Skip to content

J
jinchat-server
Unverified 提交 e02f49a4 作者: imClumsyPanda 提交者： GitHub
操作

Merge branch 'dev' into dev

上级 300d287d 3a6a6fa8
正在显示 包含 1530 行增加 和 94 行删除
README.md
......@@ -23,13 +23,17 @@
🚩 本项目未涉及微调、训练过程，但可利用微调或训练对本项目效果进行优化。
🐳 Docker镜像：registry.cn-beijing.aliyuncs.com/isafetech/chatmydata:1.0 （感谢 @InkSong🌲 ）
💻 运行方式：docker run -d -p 80:7860 --gpus all registry.cn-beijing.aliyuncs.com/isafetech/chatmydata:1.0 
🌐 [AutoDL 镜像](https://www.codewithgpu.com/i/imClumsyPanda/langchain-ChatGLM/langchain-ChatGLM)
📓 [ModelWhale 在线运行项目](https://www.heywhale.com/mw/project/643977aa446c45f4592a1e59)
## 变更日志
参见 [变更日志](docs/CHANGELOG.md)。
参见 [版本更新日志](https://github.com/imClumsyPanda/langchain-ChatGLM/releases)。
## 硬件需求
......@@ -60,6 +64,23 @@
本项目中默认选用的 Embedding 模型 [GanymedeNil/text2vec-large-chinese](https://huggingface.co/GanymedeNil/text2vec-large-chinese/tree/main) 约占用显存 3GB，也可修改为在 CPU 中运行。
## Docker 整合包
🐳 Docker镜像地址：`registry.cn-beijing.aliyuncs.com/isafetech/chatmydata:1.0 `🌲
💻 一行命令运行：
```shell
docker run -d -p 80:7860 --gpus all registry.cn-beijing.aliyuncs.com/isafetech/chatmydata:1.0
```
- 该版本镜像大小`25.2G`，使用[v0.1.16](https://github.com/imClumsyPanda/langchain-ChatGLM/releases/tag/v0.1.16)，以`nvidia/cuda:12.1.1-cudnn8-runtime-ubuntu22.04`为基础镜像
- 该版本内置两个`embedding`模型：`m3e-base`，`text2vec-large-chinese`，内置`fastchat+chatglm-6b`
- 该版本目标为方便一键部署使用，请确保您已经在Linux发行版上安装了NVIDIA驱动程序
- 请注意，您不需要在主机系统上安装CUDA工具包，但需要安装`NVIDIA Driver`以及`NVIDIA Container Toolkit`，请参考[安装指南](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html)
- 首次拉取和启动均需要一定时间，首次启动时请参照下图使用`docker logs -f <container id>`查看日志
- 如遇到启动过程卡在`Waiting..`步骤，建议使用`docker exec -it <container id> bash`进入`/logs/`目录查看对应阶段日志
![](img/docker_logs.png)
## Docker 部署
为了能让容器使用主机GPU资源，需要在主机上安装 [NVIDIA Container Toolkit](https://github.com/NVIDIA/nvidia-container-toolkit)。具体安装步骤如下：
```shell
......@@ -198,6 +219,7 @@ Web UI 可以实现如下功能：
- [ ] 知识图谱/图数据库接入
- [ ] Agent 实现
- [x] 增加更多 LLM 模型支持
- [x] [THUDM/chatglm2-6b](https://huggingface.co/THUDM/chatglm2-6b)
- [x] [THUDM/chatglm-6b](https://huggingface.co/THUDM/chatglm-6b)
- [x] [THUDM/chatglm-6b-int8](https://huggingface.co/THUDM/chatglm-6b-int8)
- [x] [THUDM/chatglm-6b-int4](https://huggingface.co/THUDM/chatglm-6b-int4)
......@@ -221,7 +243,7 @@ Web UI 可以实现如下功能：
- [x] 选择知识库开始问答
- [x] 上传文件/文件夹至知识库
- [x] 知识库测试
- [ ] 删除知识库中文件
- [x] 删除知识库中文件
- [x] 支持搜索引擎问答
- [ ] 增加 API 支持
- [x] 利用 fastapi 实现 API 部署方式
......@@ -229,7 +251,7 @@ Web UI 可以实现如下功能：
- [x] VUE 前端
## 项目交流群
<img src="img/qr_code_32.jpg" alt="二维码" width="300" height="300" />
<img src="img/qr_code_42.jpg" alt="二维码" width="300" height="300" />
🎉 langchain-ChatGLM 项目微信交流群，如果你也对本项目感兴趣，欢迎加入群聊参与讨论交流。
api.py
......@@ -4,7 +4,7 @@ import os
import shutil
from typing import List, Optional
import urllib
import asyncio
import nltk
import pydantic
import uvicorn
......@@ -79,23 +79,37 @@ class ChatMessage(BaseModel):
}
def get_folder_path(local_doc_id: str):
return os.path.join(KB_ROOT_PATH, local_doc_id, "content")
def get_kb_path(local_doc_id: str):
return os.path.join(KB_ROOT_PATH, local_doc_id)
def get_doc_path(local_doc_id: str):
return os.path.join(get_kb_path(local_doc_id), "content")
def get_vs_path(local_doc_id: str):
return os.path.join(KB_ROOT_PATH, local_doc_id, "vector_store")
return os.path.join(get_kb_path(local_doc_id), "vector_store")
def get_file_path(local_doc_id: str, doc_name: str):
return os.path.join(KB_ROOT_PATH, local_doc_id, "content", doc_name)
return os.path.join(get_doc_path(local_doc_id), doc_name)
def validate_kb_name(knowledge_base_id: str) -> bool:
# 检查是否包含预期外的字符或路径攻击关键字
if "../" in knowledge_base_id:
return False
return True
async def upload_file(
file: UploadFile = File(description="A single binary file"),
knowledge_base_id: str = Form(..., description="Knowledge Base Name", example="kb1"),
):
saved_path = get_folder_path(knowledge_base_id)
if not validate_kb_name(knowledge_base_id):
return BaseResponse(code=403, msg="Don't attack me", data=[])
saved_path = get_doc_path(knowledge_base_id)
if not os.path.exists(saved_path):
os.makedirs(saved_path)
......@@ -125,21 +139,25 @@ async def upload_files(
],
knowledge_base_id: str = Form(..., description="Knowledge Base Name", example="kb1"),
):
saved_path = get_folder_path(knowledge_base_id)
if not validate_kb_name(knowledge_base_id):
return BaseResponse(code=403, msg="Don't attack me", data=[])
saved_path = get_doc_path(knowledge_base_id)
if not os.path.exists(saved_path):
os.makedirs(saved_path)
filelist = []
for file in files:
file_content = ''
file_path = os.path.join(saved_path, file.filename)
file_content = file.file.read()
file_content = await file.read()
if os.path.exists(file_path) and os.path.getsize(file_path) == len(file_content):
continue
with open(file_path, "ab+") as f:
with open(file_path, "wb") as f:
f.write(file_content)
filelist.append(file_path)
if filelist:
vs_path, loaded_files = local_doc_qa.init_knowledge_vector_store(filelist, get_vs_path(knowledge_base_id))
vs_path = get_vs_path(knowledge_base_id)
vs_path, loaded_files = local_doc_qa.init_knowledge_vector_store(filelist, vs_path)
if len(loaded_files):
file_status = f"documents {', '.join([os.path.split(i)[-1] for i in loaded_files])} upload success"
return BaseResponse(code=200, msg=file_status)
......@@ -163,11 +181,19 @@ async def list_kbs():
async def list_docs(
knowledge_base_id: Optional[str] = Query(default=None, description="Knowledge Base Name", example="kb1")
knowledge_base_id: str = Query(..., description="Knowledge Base Name", example="kb1")
):
local_doc_folder = get_folder_path(knowledge_base_id)
if not validate_kb_name(knowledge_base_id):
return ListDocsResponse(code=403, msg="Don't attack me", data=[])
knowledge_base_id = urllib.parse.unquote(knowledge_base_id)
kb_path = get_kb_path(knowledge_base_id)
local_doc_folder = get_doc_path(knowledge_base_id)
if not os.path.exists(kb_path):
return ListDocsResponse(code=404, msg=f"Knowledge base {knowledge_base_id} not found", data=[])
if not os.path.exists(local_doc_folder):
return {"code": 1, "msg": f"Knowledge base {knowledge_base_id} not found"}
all_doc_names = []
else:
all_doc_names = [
doc
for doc in os.listdir(local_doc_folder)
......@@ -181,11 +207,15 @@ async def delete_kb(
description="Knowledge Base Name",
example="kb1"),
):
if not validate_kb_name(knowledge_base_id):
return BaseResponse(code=403, msg="Don't attack me")
# TODO: 确认是否支持批量删除知识库
knowledge_base_id = urllib.parse.unquote(knowledge_base_id)
if not os.path.exists(get_folder_path(knowledge_base_id)):
return {"code": 1, "msg": f"Knowledge base {knowledge_base_id} not found"}
shutil.rmtree(get_folder_path(knowledge_base_id))
kb_path = get_kb_path(knowledge_base_id)
if not os.path.exists(kb_path):
return BaseResponse(code=404, msg=f"Knowledge base {knowledge_base_id} not found")
shutil.rmtree(kb_path)
return BaseResponse(code=200, msg=f"Knowledge Base {knowledge_base_id} delete success")
......@@ -194,27 +224,30 @@ async def delete_doc(
description="Knowledge Base Name",
example="kb1"),
doc_name: str = Query(
None, description="doc name", example="doc_name_1.pdf"
..., description="doc name", example="doc_name_1.pdf"
),
):
if not validate_kb_name(knowledge_base_id):
return BaseResponse(code=403, msg="Don't attack me")
knowledge_base_id = urllib.parse.unquote(knowledge_base_id)
if not os.path.exists(get_folder_path(knowledge_base_id)):
return {"code": 1, "msg": f"Knowledge base {knowledge_base_id} not found"}
if not os.path.exists(get_kb_path(knowledge_base_id)):
return BaseResponse(code=404, msg=f"Knowledge base {knowledge_base_id} not found")
doc_path = get_file_path(knowledge_base_id, doc_name)
if os.path.exists(doc_path):
os.remove(doc_path)
remain_docs = await list_docs(knowledge_base_id)
if len(remain_docs.data) == 0:
shutil.rmtree(get_folder_path(knowledge_base_id), ignore_errors=True)
shutil.rmtree(get_kb_path(knowledge_base_id), ignore_errors=True)
return BaseResponse(code=200, msg=f"document {doc_name} delete success")
else:
status = local_doc_qa.delete_file_from_vector_store(doc_path, get_vs_path(knowledge_base_id))
if "success" in status:
return BaseResponse(code=200, msg=f"document {doc_name} delete success")
else:
return BaseResponse(code=1, msg=f"document {doc_name} delete fail")
return BaseResponse(code=500, msg=f"document {doc_name} delete fail")
else:
return BaseResponse(code=1, msg=f"document {doc_name} not found")
return BaseResponse(code=404, msg=f"document {doc_name} not found")
async def update_doc(
......@@ -222,23 +255,26 @@ async def update_doc(
description="知识库名",
example="kb1"),
old_doc: str = Query(
None, description="待删除文件名，已存储在知识库中", example="doc_name_1.pdf"
..., description="待删除文件名，已存储在知识库中", example="doc_name_1.pdf"
),
new_doc: UploadFile = File(description="待上传文件"),
):
if not validate_kb_name(knowledge_base_id):
return BaseResponse(code=403, msg="Don't attack me")
knowledge_base_id = urllib.parse.unquote(knowledge_base_id)
if not os.path.exists(get_folder_path(knowledge_base_id)):
return {"code": 1, "msg": f"Knowledge base {knowledge_base_id} not found"}
if not os.path.exists(get_kb_path(knowledge_base_id)):
return BaseResponse(code=404, msg=f"Knowledge base {knowledge_base_id} not found")
doc_path = get_file_path(knowledge_base_id, old_doc)
if not os.path.exists(doc_path):
return BaseResponse(code=1, msg=f"document {old_doc} not found")
return BaseResponse(code=404, msg=f"document {old_doc} not found")
else:
os.remove(doc_path)
delete_status = local_doc_qa.delete_file_from_vector_store(doc_path, get_vs_path(knowledge_base_id))
if "fail" in delete_status:
return BaseResponse(code=1, msg=f"document {old_doc} delete failed")
return BaseResponse(code=500, msg=f"document {old_doc} delete failed")
else:
saved_path = get_folder_path(knowledge_base_id)
saved_path = get_doc_path(knowledge_base_id)
if not os.path.exists(saved_path):
os.makedirs(saved_path)
......@@ -279,7 +315,7 @@ async def local_doc_chat(
):
vs_path = get_vs_path(knowledge_base_id)
if not os.path.exists(vs_path):
# return BaseResponse(code=1, msg=f"Knowledge base {knowledge_base_id} not found")
# return BaseResponse(code=404, msg=f"Knowledge base {knowledge_base_id} not found")
return ChatMessage(
question=question,
response=f"Knowledge base {knowledge_base_id} not found",
......@@ -382,6 +418,7 @@ async def stream_chat(websocket: WebSocket):
for resp, history in local_doc_qa.get_knowledge_based_answer(
query=question, vs_path=vs_path, chat_history=history, streaming=True
):
await asyncio.sleep(0)
await websocket.send_text(resp["result"][last_print_len:])
last_print_len = len(resp["result"])
......@@ -444,7 +481,7 @@ async def document():
return RedirectResponse(url="/docs")
def api_start(host, port):
def api_start(host, port, **kwargs):
global app
global local_doc_qa
......@@ -466,7 +503,7 @@ def api_start(host, port):
# 修改了stream_chat的接口，直接通过ws://localhost:7861/local_doc_qa/stream_chat建立连接，在请求体中选择knowledge_base_id
app.websocket("/local_doc_qa/stream_chat")(stream_chat)
app.get("/", response_model=BaseResponse)(document)
app.get("/", response_model=BaseResponse, summary="swagger 文档")(document)
# 增加基于bing搜索的流式问答
# 需要说明的是，如果想测试websocket的流式问答，需要使用支持websocket的测试工具，如postman,insomnia
......@@ -474,17 +511,17 @@ def api_start(host, port):
# 在测试时选择new websocket request,并将url的协议改为ws,如ws://localhost:7861/local_doc_qa/stream_chat_bing
app.websocket("/local_doc_qa/stream_chat_bing")(stream_chat_bing)
app.post("/chat", response_model=ChatMessage)(chat)
app.post("/chat", response_model=ChatMessage, summary="与模型对话")(chat)
app.post("/local_doc_qa/upload_file", response_model=BaseResponse)(upload_file)
app.post("/local_doc_qa/upload_files", response_model=BaseResponse)(upload_files)
app.post("/local_doc_qa/local_doc_chat", response_model=ChatMessage)(local_doc_chat)
app.post("/local_doc_qa/bing_search_chat", response_model=ChatMessage)(bing_search_chat)
app.get("/local_doc_qa/list_knowledge_base", response_model=ListDocsResponse)(list_kbs)
app.get("/local_doc_qa/list_files", response_model=ListDocsResponse)(list_docs)
app.delete("/local_doc_qa/delete_knowledge_base", response_model=BaseResponse)(delete_kb)
app.delete("/local_doc_qa/delete_file", response_model=BaseResponse)(delete_doc)
app.post("/local_doc_qa/update_file", response_model=BaseResponse)(update_doc)
app.post("/local_doc_qa/upload_file", response_model=BaseResponse, summary="上传文件到知识库")(upload_file)
app.post("/local_doc_qa/upload_files", response_model=BaseResponse, summary="批量上传文件到知识库")(upload_files)
app.post("/local_doc_qa/local_doc_chat", response_model=ChatMessage, summary="与知识库对话")(local_doc_chat)
app.post("/local_doc_qa/bing_search_chat", response_model=ChatMessage, summary="与必应搜索对话")(bing_search_chat)
app.get("/local_doc_qa/list_knowledge_base", response_model=ListDocsResponse, summary="获取知识库列表")(list_kbs)
app.get("/local_doc_qa/list_files", response_model=ListDocsResponse, summary="获取知识库内的文件列表")(list_docs)
app.delete("/local_doc_qa/delete_knowledge_base", response_model=BaseResponse, summary="删除知识库")(delete_kb)
app.delete("/local_doc_qa/delete_file", response_model=BaseResponse, summary="删除知识库内的文件")(delete_doc)
app.post("/local_doc_qa/update_file", response_model=BaseResponse, summary="上传文件到知识库，并删除另一个文件")(update_doc)
local_doc_qa = LocalDocQA()
local_doc_qa.init_cfg(
......@@ -493,15 +530,21 @@ def api_start(host, port):
embedding_device=EMBEDDING_DEVICE,
top_k=VECTOR_SEARCH_TOP_K,
)
if kwargs.get("ssl_keyfile") and kwargs.get("ssl_certfile"):
uvicorn.run(app, host=host, port=port, ssl_keyfile=kwargs.get("ssl_keyfile"),
ssl_certfile=kwargs.get("ssl_certfile"))
else:
uvicorn.run(app, host=host, port=port)
if __name__ == "__main__":
parser.add_argument("--host", type=str, default="0.0.0.0")
parser.add_argument("--port", type=int, default=7861)
parser.add_argument("--ssl_keyfile", type=str)
parser.add_argument("--ssl_certfile", type=str)
# 初始化消息
args = None
args = parser.parse_args()
args_dict = vars(args)
shared.loaderCheckPoint = LoaderCheckPoint(args_dict)
api_start(args.host, args.port)
api_start(args.host, args.port, ssl_keyfile=args.ssl_keyfile, ssl_certfile=args.ssl_certfile)
chains/local_doc_qa.py
......@@ -17,6 +17,7 @@ import models.shared as shared
from agent import bing_search
from langchain.docstore.document import Document
from functools import lru_cache
from textsplitter.zh_title_enhance import zh_title_enhance
# patch HuggingFaceEmbeddings to make it hashable
......@@ -56,7 +57,7 @@ def tree(filepath, ignore_dir_names=None, ignore_file_names=None):
return ret_list, [os.path.basename(p) for p in ret_list]
def load_file(filepath, sentence_size=SENTENCE_SIZE):
def load_file(filepath, sentence_size=SENTENCE_SIZE, using_zh_title_enhance=ZH_TITLE_ENHANCE):
if filepath.lower().endswith(".md"):
loader = UnstructuredFileLoader(filepath, mode="elements")
docs = loader.load()
......@@ -79,6 +80,8 @@ def load_file(filepath, sentence_size=SENTENCE_SIZE):
loader = UnstructuredFileLoader(filepath, mode="elements")
textsplitter = ChineseTextSplitter(pdf=False, sentence_size=sentence_size)
docs = loader.load_and_split(text_splitter=textsplitter)
if using_zh_title_enhance:
docs = zh_title_enhance(docs)
write_check_file(filepath, docs)
return docs
......
cli.py
......@@ -42,7 +42,9 @@ def start():
@start.command(name="api", context_settings=dict(help_option_names=['-h', '--help']))
@click.option('-i', '--ip', default='0.0.0.0', show_default=True, type=str, help='api_server listen address.')
@click.option('-p', '--port', default=7861, show_default=True, type=int, help='api_server listen port.')
def start_api(ip, port):
@click.option('-k', '--ssl_keyfile', type=int, help='enable api https/wss service, specify the ssl keyfile path.')
@click.option('-c', '--ssl_certfile', type=int, help='enable api https/wss service, specify the ssl certificate file path.')
def start_api(ip, port, **kwargs):
# 调用api_start之前需要先loadCheckPoint,并传入加载检查点的参数，
# 理论上可以用click包进行包装，但过于繁琐，改动较大，
# 此处仍用parser包，并以models.loader.args.DEFAULT_ARGS的参数为默认参数
......@@ -51,7 +53,7 @@ def start_api(ip, port):
from models.loader import LoaderCheckPoint
from models.loader.args import DEFAULT_ARGS
shared.loaderCheckPoint = LoaderCheckPoint(DEFAULT_ARGS)
api_start(host=ip, port=port)
api_start(host=ip, port=port, **kwargs)
# # 通过cli.py调用cli_demo时需要在cli.py里初始化模型，否则会报错：
# langchain-ChatGLM: error: unrecognized arguments: start cli
......
configs/model_config.py
......@@ -27,7 +27,6 @@ EMBEDDING_MODEL = "text2vec"
# Embedding running device
EMBEDDING_DEVICE = "cuda" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu"
# supported LLM models
# llm_model_dict 处理了loader的一些预设行为，如加载位置，模型名称，模型处理器实例
# 在以下字典中修改属性值，以指定本地 LLM 模型存储位置
......@@ -58,12 +57,29 @@ llm_model_dict = {
"local_model_path": None,
"provides": "ChatGLM"
},
"chatglm2-6b": {
"name": "chatglm2-6b",
"pretrained_model_name": "THUDM/chatglm2-6b",
"local_model_path": None,
"provides": "ChatGLM"
},
"chatglm2-6b-int4": {
"name": "chatglm2-6b-int4",
"pretrained_model_name": "THUDM/chatglm2-6b-int4",
"local_model_path": None,
"provides": "ChatGLM"
},
"chatglm2-6b-int8": {
"name": "chatglm2-6b-int8",
"pretrained_model_name": "THUDM/chatglm2-6b-int8",
"local_model_path": None,
"provides": "ChatGLM"
},
"chatyuan": {
"name": "chatyuan",
"pretrained_model_name": "ClueAI/ChatYuan-large-v2",
"local_model_path": None,
"provides": None
"provides": "MOSSLLM"
},
"moss": {
"name": "moss",
......@@ -77,6 +93,46 @@ llm_model_dict = {
"local_model_path": None,
"provides": "LLamaLLM"
},
# 直接调用返回requests.exceptions.ConnectionError错误，需要通过huggingface_hub包里的snapshot_download函数
# 下载模型，如果snapshot_download还是返回网络错误，多试几次，一般是可以的，
# 如果仍然不行，则应该是网络加了防火墙(在服务器上这种情况比较常见)，基本只能从别的设备上下载，
# 然后转移到目标设备了.
"bloomz-7b1": {
"name": "bloomz-7b1",
"pretrained_model_name": "bigscience/bloomz-7b1",
"local_model_path": None,
"provides": "MOSSLLM"
},
# 实测加载bigscience/bloom-3b需要170秒左右，暂不清楚为什么这么慢
# 应与它要加载专有token有关
"bloom-3b": {
"name": "bloom-3b",
"pretrained_model_name": "bigscience/bloom-3b",
"local_model_path": None,
"provides": "MOSSLLM"
},
"baichuan-7b": {
"name": "baichuan-7b",
"pretrained_model_name": "baichuan-inc/baichuan-7B",
"local_model_path": None,
"provides": "MOSSLLM"
},
# llama-cpp模型的兼容性问题参考https://github.com/abetlen/llama-cpp-python/issues/204
"ggml-vicuna-13b-1.1-q5": {
"name": "ggml-vicuna-13b-1.1-q5",
"pretrained_model_name": "lmsys/vicuna-13b-delta-v1.1",
# 这里需要下载好模型的路径,如果下载模型是默认路径则它会下载到用户工作区的
# /.cache/huggingface/hub/models--vicuna--ggml-vicuna-13b-1.1/
# 还有就是由于本项目加载模型的方式设置的比较严格，下载完成后仍需手动修改模型的文件名
# 将其设置为与Huggface Hub一致的文件名
# 此外不同时期的ggml格式并不兼容，因此不同时期的ggml需要安装不同的llama-cpp-python库，且实测pip install 不好使
# 需要手动从https://github.com/abetlen/llama-cpp-python/releases/tag/下载对应的wheel安装
# 实测v0.1.63与本模型的vicuna/ggml-vicuna-13b-1.1/ggml-vic13b-q5_1.bin可以兼容
"local_model_path": f'''{"/".join(os.path.abspath(__file__).split("/")[:3])}/.cache/huggingface/hub/models--vicuna--ggml-vicuna-13b-1.1/blobs/''',
"provides": "LLamaLLM"
},
# 通过 fastchat 调用的模型请参考如下格式
"fastchat-chatglm-6b": {
......@@ -84,6 +140,14 @@ llm_model_dict = {
"pretrained_model_name": "chatglm-6b",
"local_model_path": None,
"provides": "FastChatOpenAILLM", # 使用fastchat api时，需保证"provides"为"FastChatOpenAILLM"
"api_base_url": "http://localhost:8000/v1", # "name"修改为fastchat服务中的"api_base_url"
"api_key": "EMPTY"
},
"fastchat-chatglm2-6b": {
"name": "chatglm2-6b", # "name"修改为fastchat服务中的"model_name"
"pretrained_model_name": "chatglm2-6b",
"local_model_path": None,
"provides": "FastChatOpenAILLM", # 使用fastchat api时，需保证"provides"为"FastChatOpenAILLM"
"api_base_url": "http://localhost:8000/v1" # "name"修改为fastchat服务中的"api_base_url"
},
......@@ -93,8 +157,18 @@ llm_model_dict = {
"pretrained_model_name": "vicuna-13b-hf",
"local_model_path": None,
"provides": "FastChatOpenAILLM", # 使用fastchat api时，需保证"provides"为"FastChatOpenAILLM"
"api_base_url": "http://localhost:8000/v1" # "name"修改为fastchat服务中的"api_base_url"
"api_base_url": "http://localhost:8000/v1", # "name"修改为fastchat服务中的"api_base_url"
"api_key": "EMPTY"
},
"openai-chatgpt-3.5": {
"name": "gpt-3.5-turbo",
"pretrained_model_name": "gpt-3.5-turbo",
"provides": "FastChatOpenAILLM",
"local_model_path": None,
"api_base_url": "https://api.openapi.com/v1",
"api_key": ""
},
}
# LLM 名称
......@@ -128,7 +202,7 @@ PROMPT_TEMPLATE = """已知信息：
根据上述已知信息，简洁和专业的来回答用户的问题。如果无法从中得到答案，请说 “根据已知信息无法回答该问题” 或 “没有提供足够的相关信息”，不允许在答案中添加编造成分，答案请使用中文。 问题是：{question}"""
# 缓存知识库数量
# 缓存知识库数量,如果是ChatGLM2,ChatGLM2-int4,ChatGLM2-int8模型若检索效果不好可以调成’10’
CACHED_VS_NUM = 1
# 文本分句长度
......@@ -174,3 +248,8 @@ BING_SEARCH_URL = "https://api.bing.microsoft.com/v7.0/search"
# 此外，如果是在服务器上，报Failed to establish a new connection: [Errno 110] Connection timed out
# 是因为服务器加了防火墙，需要联系管理员加白名单，如果公司的服务器的话，就别想了GG
BING_SUBSCRIPTION_KEY = ""
# 是否开启中文标题加强，以及标题增强的相关配置
# 通过增加标题判断，判断哪些文本为标题，并在metadata中进行标记；
# 然后将文本与往上一级的标题进行拼合，实现文本信息的增强。
ZH_TITLE_ENHANCE = False
docs/INSTALL.md
......@@ -44,4 +44,12 @@ $ pip install -r requirements.txt
$ python loader/image_loader.py
```
注：使用 `langchain.document_loaders.UnstructuredFileLoader` 进行非结构化文件接入时，可能需要依据文档进行其他依赖包的安装，请参考 [langchain 文档](https://python.langchain.com/en/latest/modules/indexes/document_loaders/examples/unstructured_file.html)。
## llama-cpp模型调用的说明
1. 首先从huggingface hub中下载对应的模型，如https://huggingface.co/vicuna/ggml-vicuna-13b-1.1/的[ggml-vic13b-q5_1.bin](https://huggingface.co/vicuna/ggml-vicuna-13b-1.1/blob/main/ggml-vic13b-q5_1.bin)，建议使用huggingface_hub库的snapshot_download下载。
2. 将下载的模型重命名。通过huggingface_hub下载的模型会被重命名为随机序列，因此需要重命名为原始文件名，如[ggml-vic13b-q5_1.bin](https://huggingface.co/vicuna/ggml-vicuna-13b-1.1/blob/main/ggml-vic13b-q5_1.bin)。
3. 基于下载模型的ggml的加载时间，推测对应的llama-cpp版本，下载对应的llama-cpp-python库的wheel文件，实测[ggml-vic13b-q5_1.bin](https://huggingface.co/vicuna/ggml-vicuna-13b-1.1/blob/main/ggml-vic13b-q5_1.bin)与llama-cpp-python库兼容,然后手动安装wheel文件。
4. 将下载的模型信息写入configs/model_config.py文件里 `llm_model_dict`中，注意保证参数的兼容性，一些参数组合可能会报错.
docs/启动API服务.md 0 → 100644
# 启动API服务
# 启动API服务
## 通过py文件启动
可以通过直接执行`api.py`文件启动API服务，默认以ip:0.0.0.0和port:7861启动http和ws服务。
```shell
python api.py
```
同时，启动时支持StartOption所列的模型加载参数，同时还支持IP和端口设置。
```shell
python api.py --model-name chatglm-6b-int8 --port 7862
```
## 通过cli.bat/cli.sh启动
也可以通过命令行控制文件继续启动。
```shell
cli.sh api --help
```
其他可设置参数和上述py文件启动方式相同。
# 以https、wss启动API服务
## 本地创建ssl相关证书文件
如果没有正式签发的CA证书，可以[安装mkcert](https://github.com/FiloSottile/mkcert#installation)工具， 然后用如下指令生成本地CA证书：
```shell
mkcert -install
mkcert api.example.com 47.123.123.123 localhost 127.0.0.1 ::1
```
默认回车保存在当前目录下，会有以生成指令第一个域名命名为前缀命名的两个pem文件。
附带两个文件参数启动即可。
````shell
python api --port 7862 --ssl_keyfile api.example.com+4-key.pem --ssl_certfile api.example.com+4.pem
./cli.sh api --port 7862 --ssl_keyfile api.example.com+4-key.pem --ssl_certfile api.example.com+4.pem
````
此外可以通过前置Nginx转发实现类似效果，可另行查阅相关资料。
\ No newline at end of file
knowledge_base/samples/isssues_merge/langchain-ChatGLM_closed.csv 0 → 100644
,title,file,url,detail,id
0,加油~以及一些建议,2023-03-31.0002,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/2,加油，我认为你的方向是对的。,0
1,当前的运行环境是什么，windows还是Linux,2023-04-01.0003,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/3,当前的运行环境是什么，windows还是Linux，python是什么版本？,1
2,请问这是在CLM基础上运行吗？,2023-04-01.0004,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/4,请问是不是需要本地安装好clm并正常运行的情况下，再按文中的步骤执行才能运行起来？,2
3,[复现问题] 构造 prompt 时从知识库中提取的文字乱码,2023-04-01.0005,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/5,hi，我在尝试复现 README 中的效果，也使用了 ChatGLM-6B 的 README 作为输入文本，但发现从知识库中提取的文字是乱码，导致构造的 prompt 不可用。想了解如何解决这个问题。,3
4,后面能否加入上下文对话功能？,2023-04-02.0006,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/6,目前的get_wiki_agent_answer函数中已经实现了历史消息传递的功能，后面我再确认一下是否有langchain中model调用过程中是否传递了chat_history。,4
5,请问：纯cpu可以吗？,2023-04-03.0007,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/7,很酷的实现，极大地开拓了我的眼界！很顺利的在gpu机器上运行了,5
6,运行报错：AttributeError: 'NoneType' object has no attribute 'message_types_by_name',2023-04-03.0008,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/8,报错：,6
7,运行环境：GPU需要多大的？,2023-04-03.0009,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/9,如果按照THUDM/ChatGLM-6B的说法，使用的GPU大小应该在13GB左右，但运行脚本后，占用了24GB还不够。,7
8,请问本地知识的格式是什么？,2023-04-03.0010,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/10,已测试格式包括docx、md文件中的文本信息，具体格式可以参考 [langchain文档](https://python.langchain.com/en/latest/modules/indexes/document_loaders/examples/unstructured_file.html?highlight=pdf#),8
9,24G的显存还是爆掉了，是否支持双卡运行,2023-04-03.0011,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/11,RuntimeError: CUDA out of memory. Tried to allocate 96.00 MiB (GPU 0; 23.70 GiB total capacity; 22.18 GiB already allocated; 12.75 MiB free; 22.18 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF,9
10,你怎么知道embeddings方式和模型训练时候的方式是一样的?,2023-04-03.0012,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/12,embedding和LLM的方式不用一致，embedding能够解决语义检索的需求就行。这个项目里用到embedding是在对本地知识建立索引和对问句转换成向量的过程。,10
11,是否能提供本地知识文件的格式？,2023-04-04.0013,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/13,是否能提供本地知识文件的格式？,11
12,是否可以像清华原版跑在8G一以下的卡？,2023-04-04.0016,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/16,是否可以像清华原版跑在8G一以下的卡？我的8G卡爆显存了🤣🤣🤣,12
13,请教一下langchain协调使用向量库和chatGLM工作的,2023-04-05.0018,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/18,代码里面这段是创建问答模型的，会接入ChatGLM和本地语料的向量库，langchain回答的时候是怎么个优先顺序？先搜向量库，没有再找chatglm么？ 还是什么机制？,13
14,在mac m2max上抛出了ValueError: 150001 is not in list这个异常,2023-04-05.0019,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/19,我把chatglm_llm.py加载模型的代码改成如下,14
15,程序运行后一直卡住,2023-04-05.0020,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/20,感谢作者的付出，不过本人在运行时出现了问题，请大家帮助。,15
16,问一下chat_history的逻辑,2023-04-06.0022,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/22,感谢开源。,16
17,为什么每次运行都会loading checkpoint,2023-04-06.0023,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/23,我把这个embeding模型下载到本地后，无法正常启动。,17
18,本地知识文件能否上传一些示例？,2023-04-06.0025,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/25,如题，怎么构造知识文件，效果更好？能否提供一个样例,18
19,What version of you are using?,2023-04-06.0026,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/26,"Hi Panda, I saw the `pip install -r requirements` command in README, and want to confirm you are using python2 or python3? because my pip and pip3 version are all is 22.3.",19
20,有兴趣交流本项目应用的朋友可以加一下微信群,2023-04-07.0027,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/27,![IMG_1630](https://user-images.githubusercontent.com/5668498/230533162-8b9bfcdd-249c-4efe-b066-4f9ba2ce9f23.jpeg),20
21,本地知识越多，回答时检索的时间是否会越长,2023-04-07.0029,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/29,是的 因为需要进行向量匹配检索,21
22,爲啥最後還是報錯 哭。。,2023-04-07.0030,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/30,Failed to import transformers.models.t5.configuration_t5 because of the following error (look up to see,22
23,对话到第二次的时候就报错UnicodeDecodeError: 'utf-8' codec can't decode,2023-04-07.0031,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/31,对话第一次是没问题的，模型返回输出后又给到请输入你的问题，我再输入问题就报错,23
24,用的in4的量化版本，推理的时候显示需要申请10Gb的显存,2023-04-07.0033,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/33,"File ""/root/.cache/huggingface/modules/transformers_modules/chatglm-6b-int4-qe/modeling_chatglm.py"", line 581, in forward",24
25,使用colab运行，python3.9，提示包导入有问题,2023-04-07.0034,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/34,"from ._util import is_directory, is_path",25
26,运行失败，Loading checkpoint未达到100%被kill了，请问下是什么原因？,2023-04-07.0035,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/35,日志如下：,26
27,弄了个交流群，自己弄好多细节不会，大家技术讨论 加connection-image 我来拉你,2023-04-08.0036,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/36,自己搞好多不清楚的，一起来弄吧。。准备搞个部署问题的解决文档出来,27
28,Error using the new version with langchain,2023-04-09.0043,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/43,Error with the new changes:,28
29,程序报错torch.cuda.OutOfMemoryError如何解决？,2023-04-10.0044,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/44,报错详细信息如下：,29
30,qa的训练数据格式是如何设置的,2023-04-10.0045,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/45,本项目不是使用微调的方式，所以并不涉及到训练过程。,30
31,The FileType.UNK file type is not supported in partition. 解决办法,2023-04-10.0046,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/46,ValueError: Invalid file /home/yawu/Documents/langchain-ChatGLM-master/data. The FileType.UNK file type is not supported in partition.,31
32,如何读取多个txt文档？,2023-04-10.0047,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/47,如题，请教一下如何读取多个txt文档？示例代码中只给了读一个文档的案例，这个input我换成string之后也只能指定一个文档，无法用通配符指定多个文档，也无法传入多个文件路径的列表。,32
33,nltk package unable to either download or load local nltk_data folder,2023-04-10.0049,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/49,I'm running this project on an offline Windows Server environment so I download the Punkt and averaged_perceptron_tagger tokenizer in this directory:,33
34,requirements.txt中需要指定langchain版本,2023-04-11.0055,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/55,langchain版本0.116下无法引入RetrievalQA，需要指定更高版本（0.136版本下无问题）,34
35,Demo演示无法给出输出内容,2023-04-12.0059,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/59,你好，测试了项目自带新闻稿示例和自行上传的一个文本，可以加载进去，但是无法给出答案，请问属于什么情况，如何解决，谢谢。PS: 1、今天早上刚下载全部代码；2、硬件服务器满足要求；3、按操作说明正常操作。,35
36,群人数过多无法进群，求帮忙拉进群,2023-04-12.0061,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/61,您好，您的群人数超过了200人，目前无法通过二维码加群，请问您方便加我微信拉我进群吗？万分感谢,36
37,群人数已满，求大佬拉入群,2023-04-12.0062,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/62,已在README中更新拉群二维码,37
38,requirements中langchain版本错误,2023-04-12.0065,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/65,langchain版本应该是0.0.12而不是0.0.120,38
39,Linux : Searchd in,2023-04-13.0068,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/68,import nltk,39
40,No sentence-transformers model found,2023-04-13.0069,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/69,加载不了这个模型，错误原因是找不到这个模型，但是路径是配置好了的,40
41,Error loading punkt: <urlopen error [Errno 111] Connection,2023-04-13.0070,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/70,运行knowledge_based_chatglm.py，出现nltk报错，具体情况如下：,41
42,[不懂就问] ptuning数据集格式,2023-04-13.0072,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/72,大家好请教 微调数据集的格式有什么玄机吗？我看 ChatGLM-6B/ptuning/readme.md的demo数据集ADGEN里content为啥都写成 类型#裙*风格#简约 这种格式的？这里面有啥玄机的？ 特此请教,42
43,Embedding model请教,2023-04-13.0074,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/74,您好，我看到项目里的embedding模型用的是：GanymedeNil/text2vec-large-chinese，请问这个项目里的embedding模型可以直接用ChatGLM嘛？,43
44,Macbook M1 运行 webui.py 时报错，请问是否可支持M系列芯片,2023-04-13.0080,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/80,```,44
45,new feature: 添加对P-tunningv2微调后的模型支持,2023-04-14.0099,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/99,能否添加新的功能，对使用[P-tunningv2](https://github.com/THUDM/ChatGLM-6B/tree/main/ptuning)微调chatglm后的模型提供加载支持,45
46,txt文件加载成功，但读取报错,2023-04-15.0106,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/106,最新版的代码。比较诡异的是我的电脑是没有D盘的，报错信息里怎么有个D盘出来了...,46
47,模型加载成功？文件无法导入。,2023-04-15.0107,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/107,所有模型均在本地。,47
48,请问用的什么操作系统呢？,2023-04-16.0110,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/110,ubuntu、centos还是windows？,48
49,报错ModuleNotFoundError: No module named 'configs.model_config',2023-04-17.0112,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/112,更新代码后，运行webui.py，报错ModuleNotFoundError: No module named 'configs.model_config'。未查得解决方法。,49
50,问特定问题会出现爆显存,2023-04-17.0116,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/116,正常提问没问题。,50
51,loading进不去？,2023-04-18.0127,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/127,在linux系统上python webui.py之后打开网页，一直在loading，是不是跟我没装detectron2有关呢？,51
52,本地知识内容数量限制？,2023-04-18.0129,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/129,本地知识文件类型是txt，超过5条以上的数据，提问的时候就爆显存了。,52
53,我本来也计划做一个类似的产品，看来不用从头开始做了,2023-04-18.0130,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/130,文本切割，还有优化空间吗？微信群已经加不进去了。,53
54,load model failed. 加载模型失败,2023-04-18.0132,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/132,```,54
55,如何在webui里回答时同时返回引用的本地数据内容？,2023-04-18.0133,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/133,如题,55
56,交流群满200人加不了了，能不能给个负责人的联系方式拉我进群？,2023-04-20.0143,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/143,同求,56
57,No sentence-transformers model found with name ‘/text2vec/‘，但是再路径下面确实有模型文件,2023-04-20.0145,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/145,另外：The dtype of attention mask (torch.int64) is not bool,57
58,请问加载模型的路径在哪里修改，默认好像前面会带上transformers_modules.,2023-04-20.0148,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/148,"<img width=""1181"" alt=""1681977897052"" src=""https://user-images.githubusercontent.com/30926001/233301106-3846680a-d842-41d2-874e-5b6514d732c4.png"">",58
59,为啥放到方法调用会出错，这个怎么处理？,2023-04-20.0150,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/150,```python,59
60,No sentence-transformers model found with name C:\Users\Administrator/.cache\torch\sentence_transformers\GanymedeNil_text2vec-large-chinese. Creating a new one with MEAN pooling.,2023-04-21.0154,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/154,卡在这块很久是正常现象吗,60
61,微信群需要邀请才能加入,2023-04-21.0155,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/155,RT，给个个人联系方式白,61
62,No sentence-transformers model found with name GanymedeNil/text2vec-large-chinese. Creating a new one with MEAN pooling,2023-04-21.0156,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/156,ls GanymedeNil/text2vec-large-chinese,62
63,embedding会加载两次,2023-04-23.0159,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/159,你好，为什么要这样设置呢，这样会加载两次呀。,63
64,扫二维码加的那个群，群成员满了进不去了,2023-04-23.0160,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/160,如题,64
65,执行python3 cli_demo.py 报错AttributeError: 'NoneType' object has no attribute 'chat',2023-04-24.0163,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/163,"刚开始怀疑是内存不足问题，换成int4,int4-qe也不行，有人知道是什么原因吗",65
66,匹配得分,2023-04-24.0167,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/167,在示例cli_demo.py中返回的匹配文本没有对应的score，可以加上这个feature吗,66
67,大佬有计划往web_ui.py加入打字机功能吗,2023-04-25.0170,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/170,目前在载入了知识库后，单张V100 32G在回答垂直领域的问题时也需要20S以上，没有打字机逐字输出的使用体验还是比较煎熬的....,67
68,Is it possible to use a verctorDB for the embedings?,2023-04-25.0171,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/171,"when I play, I have to load the local data again and again when to start. I wonder if it is possible to use",68
69,请问通过lora训练官方模型得到的微调模型文件该如何加载？,2023-04-25.0173,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/173,通过lora训练的方式得到以下文件:,69
70,from langchain.chains import RetrievalQA的代码在哪里？,2023-04-25.0174,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/174,local_doc_qa.py,70
71,哪里有knowledge_based_chatglm.py文件？怎么找不到了？？是被替换成cli_demo.py文件了吗？,2023-04-26.0175,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/175,哪里有knowledge_based_chatglm.py文件？怎么找不到了？？是被替换成cli_demo.py文件了吗？,71
72,AttributeError: 'Chatbot' object has no attribute 'value',2023-04-26.0177,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/177,Traceback (most recent call last):,72
73,控制台调api.py报警告,2023-04-26.0178,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/178,"you must pass the application as an import string to enable ""reload"" or ""workers""",73
74,如何加入群聊,2023-04-27.0183,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/183,微信群超过200人了，需要邀请，如何加入呢？,74
75,如何将Chatglm和本地知识相结合,2023-04-27.0185,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/185,您好，我想请教一下怎么才能让知识库匹配到的文本和chatglm生成的相结合，而不是说如果没搜索到，就说根据已知信息无法回答该问题，谢谢,75
76,一点建议,2023-04-27.0189,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/189,1.weiui的get_vector_store方法里面添加一个判断以兼容gradio版本导致的上传异常,76
77,windows环境下，按照教程，配置好conda环境，git完项目，修改完模型路径相关内容后，运行demo报错缺少,2023-04-28.0194,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/194,报错代码如下：,77
78,ValueError: too many values to unpack (expected 2),2023-04-28.0198,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/198,"When i tried to use the non-streaming, `ValueError: too many values to unpack (expected 2)` error came out.",78
79,加载doc后覆盖原本知识,2023-04-28.0201,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/201,加载较大量级的私有知识库后，原本的知识会被覆盖,79
80,自定义知识库回答效果很差,2023-04-28.0203,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/203,"请问加了自定义知识库知识库,回答效果很差，是因为数据量太小的原因么",80
81,python310下，安装pycocotools失败，提示低版本cython，实际已安装高版本,2023-04-29.0208,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/208,RT，纯离线环境安装，依赖安装的十分艰难，最后碰到pycocotools，始终无法安装上，求教方法！,81
82,[FEATURE] 支持 RWKV 模型（目前已有 pip package & rwkv.cpp 等等）,2023-05-01.0216,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/216,您好，我是 RWKV 的作者，介绍见：https://zhuanlan.zhihu.com/p/626083366,82
83,[BUG] 为啥主机/服务器不联网不能正常启动服务？,2023-05-02.0220,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/220,**问题描述 / Problem Description**,83
84,[BUG] 简洁阐述问题 / Concise description of the issue,2023-05-03.0222,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/222,**local variable 'torch' referenced before assignment**,84
85,不支持txt文件的中文输入,2023-05-04.0235,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/235,"vs_path, _ = local_doc_qa.init_knowledge_vector_store(filepath)",85
86,文件均未成功加载，请检查依赖包或替换为其他文件再次上传。 文件未成功加载，请重新上传文件,2023-05-05.0237,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/237,请大佬帮忙解决，谢谢！,86
87,[BUG] 使用多卡时chatglm模型加载两次,2023-05-05.0241,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/241,chatglm_llm.py文件下第129行先加载了一次chatglm模型，第143行又加载了一次,87
88,[BUG] similarity_search_with_score_by_vector函数返回多个doc时的score结果错误,2023-05-06.0252,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/252,**问题描述 / Problem Description**,88
89,可以再建一个交流群吗，这个群满了进不去。,2023-05-06.0255,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/255,上午应该已经在readme里更新过了，如果不能添加可能是网页缓存问题，可以试试看直接扫描img/qr_code_12.jpg,89
90,请问这是什么错误哇？KeyError: 'serialized_input',2023-05-06.0257,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/257,运行“python webui.py” 后这是什么错误？怎么解决啊？,90
91,修改哪里的代码，可以再cpu上跑？,2023-05-06.0258,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/258,**问题描述 / Problem Description**,91
92,ModuleNotFoundError: No module named 'modelscope',2023-05-07.0266,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/266,安装这个,92
93,加载lora微调模型时，lora参数加载成功，但显示模型未成功加载？,2023-05-08.0270,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/270,什么原因呀？,93
94,[BUG] 运行webui.py报错：name 'EMBEDDING_DEVICE' is not defined,2023-05-08.0274,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/274,解决了，我修改model_config时候把这个变量改错了,94
95,基于ptuning训练完成，新老模型都进行了加载，但是只有新的,2023-05-08.0280,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/280,licitly passing a `revision` is encouraged when loading a model with custom code to ensure no malicious code has been contributed in a newer revision.,95
96,[BUG] 使用chatyuan模型时，对话Error，has no attribute 'stream_chat',2023-05-08.0282,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/282,**问题描述 / Problem Description**,96
97,chaglm调用过程中 _call提示有一个 stop,2023-05-09.0286,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/286,**功能描述 / Feature Description**,97
98,Logger._log() got an unexpected keyword argument 'end',2023-05-10.0295,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/295,使用cli_demo的时候，加载一个普通txt文件，输入问题后，报错：“TypeError: Logger._log() got an unexpected keyword argument 'end'”,98
99,[BUG] 请问可以解释下这个FAISS.similarity_search_with_score_by_vector = similarity_search_with_score_by_vector的目的吗,2023-05-10.0296,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/296,我不太明白这个库自己写的similarity_search_with_score_by_vector方法做的事情，因为langchain原版的similarity_search_with_score_by_vector只是search faiss之后把返回的topk句子组合起来。我觉得原版理解起来没什么问题，但是这个库里自己写的我就没太看明白多做了什么其他的事情，因为没有注释。,99
100,[BUG] Windows下上传中文文件名文件，faiss无法生成向量数据库文件,2023-05-11.0318,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/318,**问题描述 / Problem Description**,100
101,cli_demo中的流式输出能否接着前一答案输出?,2023-05-11.0320,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/320,现有流式输出结果样式为:,101
102,内网部署时网页无法加载，能否增加离线静态资源,2023-05-12.0326,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/326,内网部署时网页无法加载，能否增加离线静态资源,102
103,我想把文件字符的编码格式改为encoding='utf-8'在哪修改呢，因为会有ascii codec can't decode byte报错,2023-05-14.0360,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/360,上传中文的txt文件时报错，编码格式为utf-8,103
104,Batches的进度条是在哪里设置的?能否关闭显示?,2023-05-15.0366,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/366,"使用cli_demo.py进行命令行测试时,每句回答前都有个Batches的进度条",104
105,ImportError: dlopen: cannot load any more object with static TLS or Segmentation fault,2023-05-15.0368,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/368,**问题描述 / Problem Description**,105
106,读取PDF时报错,2023-05-16.0373,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/373,在Colab上执行cli_demo.py时，在路径文件夹里放了pdf文件，在加载的过程中会显示错误，然后无法加载PDF文件,106
107,[BUG] webui报错 InvalidURL,2023-05-16.0375,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/375,python 版本：3.8.16,107
108,[FEATURE] 如果让回答不包含出处，应该怎么处理,2023-05-16.0380,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/380,**功能描述 / Feature Description**,108
109,加载PDF文件时，出现 unsupported colorspace for 'png',2023-05-16.0381,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/381,**问题描述 / Problem Description**,109
110,'ascii' codec can't encode characters in position 14-44: ordinal not in range(128) 经典bug,2023-05-16.0382,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/382,添加了知识库之后进行对话，之后再新增知识库就会出现这个问题。,110
111,微信群人数超过200了，扫码进不去了，群主可以再创建一个新群吗,2023-05-17.0391,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/391,**功能描述 / Feature Description**,111
112,TypeError: 'ListDocsResponse' object is not subscriptable,2023-05-17.0393,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/393,应该是用remain_docs.code和remain_docs.data吧？吗？,112
113,[BUG] 加载chatglm模型报错：'NoneType' object has no attribute 'message_types_by_name',2023-05-17.0398,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/398,**问题描述 / Problem Description**,113
114,[BUG] 执行 python webui.py 没有报错，但是ui界面提示 Something went wrong Expecting value: line 1 column 1 (char 0,2023-05-18.0399,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/399,**环境配置**,114
115,启动后调用api接口正常，过一会就不断的爆出 Since the angle classifier is not initialized,2023-05-18.0404,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/404,**问题描述 / Problem Description**,115
116,[BUG] write_check_file方法中，open函数未指定编码,2023-05-18.0408,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/408,"def write_check_file(filepath, docs):",116
117,导入的PDF中存在图片，有大概率出现 “unsupported colorspace for 'png'”异常,2023-05-18.0409,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/409,"pix = fitz.Pixmap(doc, img[0])",117
118,请问流程图是用什么软件画的,2023-05-18.0410,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/410,draw.io,118
119,mac 加载模型失败,2023-05-19.0417,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/417,Explicitly passing a `revision` is encouraged when loading a model with custom code to ensure no malicious code has been contributed in a newer revision.,119
120,使用GPU本地运行知识库问答，提问第一个问题出现异常。,2023-05-20.0419,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/419,配置文件model_config.py为：,120
121,想加入讨论群,2023-05-20.0420,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/420,OK,121
122,有没有直接调用LLM的API，目前只有知识库的API？,2023-05-22.0426,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/426,-------------------------------------------------------------------------------,122
123,上传文件后出现 ERROR __init__() got an unexpected keyword argument 'autodetect_encoding',2023-05-22.0428,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/428,"上传文件后出现这个问题：ERROR 2023-05-22 11:46:19,568-1d: __init__() got an unexpected keyword argument 'autodetect_encoding'",123
124,想问下README中用到的流程图用什么软件画的,2023-05-22.0431,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/431,**功能描述 / Feature Description**,124
125,No matching distribution found for langchain==0.0.174,2023-05-23.0436,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/436,ERROR: Could not find a version that satisfies the requirement langchain==0.0.174 ,125
126,[FEATURE] bing是必须的么？,2023-05-23.0437,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/437,从这个[脚步](https://github.com/imClumsyPanda/langchain-ChatGLM/blob/master/configs/model_config.py#L129)里面发现需要申请bing api，如果不申请，纯用模型推理不可吗？,126
127,同一台环境下部署了5.22号更新的langchain-chatglm v0.1.13和之前的版本，回复速度明显变慢,2023-05-23.0442,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/442,新langchain-chatglm v0.1.13版本速度很慢,127
128,Error reported during startup,2023-05-23.0443,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/443,Traceback (most recent call last):,128
129,"ValueError: not enough values to unpack (expected 2, got 1)on of the issue",2023-05-24.0449,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/449,"File "".cache\huggingface\modules\transformers_modules\chatglm-6b-int4\modeling_chatglm.py"", line 1280, in chat",129
130,[BUG] API部署，流式输出的函数，少了个question,2023-05-24.0451,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/451,**问题描述 / Problem Description**,130
131,项目结构的简洁性保持,2023-05-24.0454,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/454,**功能描述 / Feature Description**,131
132,项目群扫码进不去了,2023-05-24.0455,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/455,项目群扫码进不去了，是否可以加一下微信拉我进群，谢谢！微信号：daniel-0527,132
133,请求拉我入群讨论，海硕一枚，专注于LLM等相关技术,2023-05-24.0461,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/461,**功能描述 / Feature Description**,133
134,[BUG] chatglm-6b模型报错OSError: Error no file named pytorch_model.bin found in directory /chatGLM/model/model-6b,2023-05-26.0474,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/474,**1、简述：**,134
135,现在本项目交流群二维码扫描不进去了，需要群主通过,2023-05-27.0478,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/478,现在本项目交流群二维码扫描不进去了，需要群主通过,135
136,RuntimeError: Only Tensors of floating point and complex dtype can require gradients,2023-05-28.0483,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/483,刚更新了最新版本：,136
137,"RuntimeError: ""LayerNormKernelImpl"" not implemented for 'Half'",2023-05-28.0484,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/484,"已经解决了 params 只用两个参数 {'trust_remote_code': True, 'torch_dtype': torch.float16}",137
138,[BUG] 文件未成功加载，请重新上传文件,2023-05-31.0504,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/504,webui.py,138
139,[BUG] bug 17 ，pdf和pdf为啥还不一样呢？为啥有的pdf能识别？有的pdf识别不了呢？,2023-05-31.0506,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/506,bug 17 ，pdf和pdf为啥还不一样呢？为啥有的pdf能识别？有的pdf识别不了呢？,139
140,[FEATURE] 简洁阐述功能 / Concise description of the feature,2023-05-31.0513,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/513,**功能描述 / Feature Description**,140
141,[BUG] webui.py 加载chatglm-6b-int4 失败,2023-06-02.0524,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/524,**问题描述 / Problem Description**,141
142,[BUG] webui.py 加载chatglm-6b模型异常,2023-06-02.0525,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/525,**问题描述 / Problem Description**,142
143,增加对chatgpt的embedding和api调用的支持,2023-06-02.0531,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/531,能否支持openai的embedding api和对话的api？,143
144,[FEATURE] 调整模型下载的位置,2023-06-02.0537,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/537,模型默认下载到 $HOME/.cache/huggingface/，当 C 盘空间不足时无法完成模型的下载。configs/model_config.py 中也没有调整模型位置的参数。,144
145,[BUG] langchain=0.0.174 出错,2023-06-04.0543,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/543,**问题描述 / Problem Description**,145
146,[BUG] 更新后加载本地模型路径不正确,2023-06-05.0545,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/545,**问题描述 / Problem Description**,146
147,SystemError: 8bit 模型需要 CUDA 支持，或者改用量化后模型！,2023-06-06.0550,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/550,"docker 部署后,启动docker,过会儿容器会自动退出,logs报错 SystemError: 8bit 模型需要 CUDA 支持，或者改用量化后模型！ [NVIDIA Container Toolkit](https://github.com/NVIDIA/nvidia-container-toolkit) 也已经安装了",147
148,[BUG] 上传知识库超过1M报错,2023-06-06.0556,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/556,**问题描述 / Problem Description**,148
149,打开跨域访问后仍然报错，不能请求,2023-06-06.0560,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/560,报错信息：,149
150,dialogue_answering 里面的代码是不是没有用到？，没有看到调用,2023-06-07.0571,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/571,dialogue_answering 是干啥的,150
151,[BUG] 响应速度极慢，应从哪里入手优化？48C/128G/8卡,2023-06-07.0573,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/573,运行环境：ubuntu20.04,151
152,纯CPU环境下运行cli_demo时报错，提示找不到nvcuda.dll,2023-06-08.0576,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/576,本地部署环境是纯CPU，之前的版本在纯CPU环境下能正常运行，但上传本地知识库经常出现encode问题。今天重新git项目后，运行时出现如下问题，请问该如何解决。,152
153,如何加载本地的embedding模型（text2vec-large-chinese模型文件）,2023-06-08.0582,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/582,"因为需要离线部署，所以要把模型放到本地，我修改了chains/local_doc_qa.py中的HuggingFaceEmbeddings()，在其中加了一个cache_folder的参数，保证下载的文件在cache_folder中，model_name是text2vec-large-chinese。如cache_folder='/home/xx/model/text2vec-large-chinese', model_name='text2vec-large-chinese'，这样仍然需要联网下载报错，请问大佬如何解决该问题？",153
154,ChatGLM-6B 在另外服务器安装好了，请问如何修改model.cofnig.py 来使用它的接口呢？？,2023-06-09.0588,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/588,我本来想在这加一个api base url 但是运行web.py 发现 还是会去连huggingface 下载模型,154
155,[BUG] raise partially initialized module 'charset_normalizer' has no attribute 'md__mypyc' when call interface `upload_file`,2023-06-10.0591,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/591,**问题描述 / Problem Description**,155
156,[BUG] raise OSError: [Errno 101] Network is unreachable when call interface upload_file and upload .pdf files,2023-06-10.0592,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/592,**问题描述 / Problem Description**,156
157,如果直接用vicuna作为基座大模型，需要修改的地方有哪些？,2023-06-12.0596,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/596,vicuna模型有直接转换好的没有？也就是llama转换之后的vicuna。,157
158,[BUG] 通过cli.py调用api时抛出AttributeError: 'NoneType' object has no attribute 'get'错误,2023-06-12.0598,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/598,通过`python cli.py start api --ip localhost --port 8001` 命令调用api时，抛出：,158
159,[BUG] 通过cli.py调用api时直接报错`langchain-ChatGLM: error: unrecognized arguments: start cli`,2023-06-12.0601,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/601,通过python cli.py start cli启动cli_demo时，报错：,159
160,[BUG] error: unrecognized arguments: --model-dir conf/models/,2023-06-12.0602,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/602,关键字参数修改了吗？有没有文档啊？大佬,160
161,[BUG] 上传文件全部失败,2023-06-12.0603,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/603,ERROR: Exception in ASGI application,161
162,[BUG] config 使用 chatyuan 无法启动,2023-06-12.0604,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/604,"""chatyuan"": {",162
163,使用fashchat api之后，后台报错APIError 如图所示,2023-06-12.0606,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/606,我按照https://github.com/imClumsyPanda/langchain-ChatGLM/blob/master/docs/fastchat.md,163
164,[BUG] 启用上下文关联，每次embedding搜索到的内容都会比前一次多一段,2023-06-13.0613,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/613,**问题描述 / Problem Description**,164
165,local_doc_qa.py中MyFAISS.from_documents() 这个语句看不太懂。MyFAISS类中没有这个方法，其父类FAISS和VectorStore中也只有from_texts方法[BUG] 简洁阐述问题 / Concise description of the issue,2023-06-14.0619,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/619,local_doc_qa.py中MyFAISS.from_documents() 这个语句看不太懂。MyFAISS类中没有这个方法，其父类FAISS和VectorStore中也只有from_texts方法,165
166,[BUG] TypeError: similarity_search_with_score_by_vector() got an unexpected keyword argument 'filter',2023-06-14.0624,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/624,**问题描述 / Problem Description**,166
167,please delete this issue,2023-06-15.0633,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/633,"sorry, incorrect submission. Please remove this issue!",167
168,[BUG] vue前端镜像构建失败,2023-06-15.0635,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/635,**问题描述 / Problem Description**,168
169,ChatGLM-6B模型能否回答英文问题？,2023-06-15.0640,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/640,大佬，请问一下，如果本地知识文档是英文，ChatGLM-6B模型能否回答英文问题？不能的话，有没有替代的模型推荐，期待你的回复，谢谢,169
170,[BUG] 简洁阐述问题 / Concise description of the issue,2023-06-16.0644,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/644,**问题描述 / Problem Description**,170
171,KeyError: 3224,2023-06-16.0645,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/645,```,171
knowledge_base/samples/isssues_merge/langchain-ChatGLM_closed.jsonl 0 → 100644
{"title": "加油~以及一些建议", "file": "2023-03-31.0002", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/2", "detail": "加油，我认为你的方向是对的。", "id": 0}
{"title": "当前的运行环境是什么，windows还是Linux", "file": "2023-04-01.0003", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/3", "detail": "当前的运行环境是什么，windows还是Linux，python是什么版本？", "id": 1}
{"title": "请问这是在CLM基础上运行吗？", "file": "2023-04-01.0004", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/4", "detail": "请问是不是需要本地安装好clm并正常运行的情况下，再按文中的步骤执行才能运行起来？", "id": 2}
{"title": "[复现问题] 构造 prompt 时从知识库中提取的文字乱码", "file": "2023-04-01.0005", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/5", "detail": "hi，我在尝试复现 README 中的效果，也使用了 ChatGLM-6B 的 README 作为输入文本，但发现从知识库中提取的文字是乱码，导致构造的 prompt 不可用。想了解如何解决这个问题。", "id": 3}
{"title": "后面能否加入上下文对话功能？", "file": "2023-04-02.0006", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/6", "detail": "目前的get_wiki_agent_answer函数中已经实现了历史消息传递的功能，后面我再确认一下是否有langchain中model调用过程中是否传递了chat_history。", "id": 4}
{"title": "请问：纯cpu可以吗？", "file": "2023-04-03.0007", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/7", "detail": "很酷的实现，极大地开拓了我的眼界！很顺利的在gpu机器上运行了", "id": 5}
{"title": "运行报错：AttributeError: 'NoneType' object has no attribute 'message_types_by_name'", "file": "2023-04-03.0008", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/8", "detail": "报错：", "id": 6}
{"title": "运行环境：GPU需要多大的？", "file": "2023-04-03.0009", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/9", "detail": "如果按照THUDM/ChatGLM-6B的说法，使用的GPU大小应该在13GB左右，但运行脚本后，占用了24GB还不够。", "id": 7}
{"title": "请问本地知识的格式是什么？", "file": "2023-04-03.0010", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/10", "detail": "已测试格式包括docx、md文件中的文本信息，具体格式可以参考 [langchain文档](https://python.langchain.com/en/latest/modules/indexes/document_loaders/examples/unstructured_file.html?highlight=pdf#)", "id": 8}
{"title": "24G的显存还是爆掉了，是否支持双卡运行", "file": "2023-04-03.0011", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/11", "detail": "RuntimeError: CUDA out of memory. Tried to allocate 96.00 MiB (GPU 0; 23.70 GiB total capacity; 22.18 GiB already allocated; 12.75 MiB free; 22.18 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF", "id": 9}
{"title": "你怎么知道embeddings方式和模型训练时候的方式是一样的?", "file": "2023-04-03.0012", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/12", "detail": "embedding和LLM的方式不用一致，embedding能够解决语义检索的需求就行。这个项目里用到embedding是在对本地知识建立索引和对问句转换成向量的过程。", "id": 10}
{"title": "是否能提供本地知识文件的格式？", "file": "2023-04-04.0013", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/13", "detail": "是否能提供本地知识文件的格式？", "id": 11}
{"title": "是否可以像清华原版跑在8G一以下的卡？", "file": "2023-04-04.0016", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/16", "detail": "是否可以像清华原版跑在8G一以下的卡？我的8G卡爆显存了🤣🤣🤣", "id": 12}
{"title": "请教一下langchain协调使用向量库和chatGLM工作的", "file": "2023-04-05.0018", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/18", "detail": "代码里面这段是创建问答模型的，会接入ChatGLM和本地语料的向量库，langchain回答的时候是怎么个优先顺序？先搜向量库，没有再找chatglm么？ 还是什么机制？", "id": 13}
{"title": "在mac m2max上抛出了ValueError: 150001 is not in list这个异常", "file": "2023-04-05.0019", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/19", "detail": "我把chatglm_llm.py加载模型的代码改成如下", "id": 14}
{"title": "程序运行后一直卡住", "file": "2023-04-05.0020", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/20", "detail": "感谢作者的付出，不过本人在运行时出现了问题，请大家帮助。", "id": 15}
{"title": "问一下chat_history的逻辑", "file": "2023-04-06.0022", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/22", "detail": "感谢开源。", "id": 16}
{"title": "为什么每次运行都会loading checkpoint", "file": "2023-04-06.0023", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/23", "detail": "我把这个embeding模型下载到本地后，无法正常启动。", "id": 17}
{"title": "本地知识文件能否上传一些示例？", "file": "2023-04-06.0025", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/25", "detail": "如题，怎么构造知识文件，效果更好？能否提供一个样例", "id": 18}
{"title": "What version of you are using?", "file": "2023-04-06.0026", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/26", "detail": "Hi Panda, I saw the `pip install -r requirements` command in README, and want to confirm you are using python2 or python3? because my pip and pip3 version are all is 22.3.", "id": 19}
{"title": "有兴趣交流本项目应用的朋友可以加一下微信群", "file": "2023-04-07.0027", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/27", "detail": "![IMG_1630](https://user-images.githubusercontent.com/5668498/230533162-8b9bfcdd-249c-4efe-b066-4f9ba2ce9f23.jpeg)", "id": 20}
{"title": "本地知识越多，回答时检索的时间是否会越长", "file": "2023-04-07.0029", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/29", "detail": "是的 因为需要进行向量匹配检索", "id": 21}
{"title": "爲啥最後還是報錯 哭。。", "file": "2023-04-07.0030", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/30", "detail": "Failed to import transformers.models.t5.configuration_t5 because of the following error (look up to see", "id": 22}
{"title": "对话到第二次的时候就报错UnicodeDecodeError: 'utf-8' codec can't decode", "file": "2023-04-07.0031", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/31", "detail": "对话第一次是没问题的，模型返回输出后又给到请输入你的问题，我再输入问题就报错", "id": 23}
{"title": "用的in4的量化版本，推理的时候显示需要申请10Gb的显存", "file": "2023-04-07.0033", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/33", "detail": "File \"/root/.cache/huggingface/modules/transformers_modules/chatglm-6b-int4-qe/modeling_chatglm.py\", line 581, in forward", "id": 24}
{"title": "使用colab运行，python3.9，提示包导入有问题", "file": "2023-04-07.0034", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/34", "detail": "from ._util import is_directory, is_path", "id": 25}
{"title": "运行失败，Loading checkpoint未达到100%被kill了，请问下是什么原因？", "file": "2023-04-07.0035", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/35", "detail": "日志如下：", "id": 26}
{"title": "弄了个交流群，自己弄好多细节不会，大家技术讨论 加connection-image 我来拉你", "file": "2023-04-08.0036", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/36", "detail": "自己搞好多不清楚的，一起来弄吧。。准备搞个部署问题的解决文档出来", "id": 27}
{"title": "Error using the new version with langchain", "file": "2023-04-09.0043", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/43", "detail": "Error with the new changes:", "id": 28}
{"title": "程序报错torch.cuda.OutOfMemoryError如何解决？", "file": "2023-04-10.0044", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/44", "detail": "报错详细信息如下：", "id": 29}
{"title": "qa的训练数据格式是如何设置的", "file": "2023-04-10.0045", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/45", "detail": "本项目不是使用微调的方式，所以并不涉及到训练过程。", "id": 30}
{"title": "The FileType.UNK file type is not supported in partition. 解决办法", "file": "2023-04-10.0046", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/46", "detail": "ValueError: Invalid file /home/yawu/Documents/langchain-ChatGLM-master/data. The FileType.UNK file type is not supported in partition.", "id": 31}
{"title": "如何读取多个txt文档？", "file": "2023-04-10.0047", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/47", "detail": "如题，请教一下如何读取多个txt文档？示例代码中只给了读一个文档的案例，这个input我换成string之后也只能指定一个文档，无法用通配符指定多个文档，也无法传入多个文件路径的列表。", "id": 32}
{"title": "nltk package unable to either download or load local nltk_data folder", "file": "2023-04-10.0049", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/49", "detail": "I'm running this project on an offline Windows Server environment so I download the Punkt and averaged_perceptron_tagger tokenizer in this directory:", "id": 33}
{"title": "requirements.txt中需要指定langchain版本", "file": "2023-04-11.0055", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/55", "detail": "langchain版本0.116下无法引入RetrievalQA，需要指定更高版本（0.136版本下无问题）", "id": 34}
{"title": "Demo演示无法给出输出内容", "file": "2023-04-12.0059", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/59", "detail": "你好，测试了项目自带新闻稿示例和自行上传的一个文本，可以加载进去，但是无法给出答案，请问属于什么情况，如何解决，谢谢。PS: 1、今天早上刚下载全部代码；2、硬件服务器满足要求；3、按操作说明正常操作。", "id": 35}
{"title": "群人数过多无法进群，求帮忙拉进群", "file": "2023-04-12.0061", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/61", "detail": "您好，您的群人数超过了200人，目前无法通过二维码加群，请问您方便加我微信拉我进群吗？万分感谢", "id": 36}
{"title": "群人数已满，求大佬拉入群", "file": "2023-04-12.0062", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/62", "detail": "已在README中更新拉群二维码", "id": 37}
{"title": "requirements中langchain版本错误", "file": "2023-04-12.0065", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/65", "detail": "langchain版本应该是0.0.12而不是0.0.120", "id": 38}
{"title": "Linux : Searchd in", "file": "2023-04-13.0068", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/68", "detail": "import nltk", "id": 39}
{"title": "No sentence-transformers model found", "file": "2023-04-13.0069", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/69", "detail": "加载不了这个模型，错误原因是找不到这个模型，但是路径是配置好了的", "id": 40}
{"title": "Error loading punkt: <urlopen error [Errno 111] Connection", "file": "2023-04-13.0070", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/70", "detail": "运行knowledge_based_chatglm.py，出现nltk报错，具体情况如下：", "id": 41}
{"title": "[不懂就问] ptuning数据集格式", "file": "2023-04-13.0072", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/72", "detail": "大家好请教 微调数据集的格式有什么玄机吗？我看 ChatGLM-6B/ptuning/readme.md的demo数据集ADGEN里content为啥都写成 类型#裙*风格#简约 这种格式的？这里面有啥玄机的？ 特此请教", "id": 42}
{"title": "Embedding model请教", "file": "2023-04-13.0074", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/74", "detail": "您好，我看到项目里的embedding模型用的是：GanymedeNil/text2vec-large-chinese，请问这个项目里的embedding模型可以直接用ChatGLM嘛？", "id": 43}
{"title": "Macbook M1 运行 webui.py 时报错，请问是否可支持M系列芯片", "file": "2023-04-13.0080", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/80", "detail": "```", "id": 44}
{"title": "new feature: 添加对P-tunningv2微调后的模型支持", "file": "2023-04-14.0099", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/99", "detail": "能否添加新的功能，对使用[P-tunningv2](https://github.com/THUDM/ChatGLM-6B/tree/main/ptuning)微调chatglm后的模型提供加载支持", "id": 45}
{"title": "txt文件加载成功，但读取报错", "file": "2023-04-15.0106", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/106", "detail": "最新版的代码。比较诡异的是我的电脑是没有D盘的，报错信息里怎么有个D盘出来了...", "id": 46}
{"title": "模型加载成功？文件无法导入。", "file": "2023-04-15.0107", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/107", "detail": "所有模型均在本地。", "id": 47}
{"title": "请问用的什么操作系统呢？", "file": "2023-04-16.0110", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/110", "detail": "ubuntu、centos还是windows？", "id": 48}
{"title": "报错ModuleNotFoundError: No module named 'configs.model_config'", "file": "2023-04-17.0112", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/112", "detail": "更新代码后，运行webui.py，报错ModuleNotFoundError: No module named 'configs.model_config'。未查得解决方法。", "id": 49}
{"title": "问特定问题会出现爆显存", "file": "2023-04-17.0116", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/116", "detail": "正常提问没问题。", "id": 50}
{"title": "loading进不去？", "file": "2023-04-18.0127", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/127", "detail": "在linux系统上python webui.py之后打开网页，一直在loading，是不是跟我没装detectron2有关呢？", "id": 51}
{"title": "本地知识内容数量限制？", "file": "2023-04-18.0129", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/129", "detail": "本地知识文件类型是txt，超过5条以上的数据，提问的时候就爆显存了。", "id": 52}
{"title": "我本来也计划做一个类似的产品，看来不用从头开始做了", "file": "2023-04-18.0130", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/130", "detail": "文本切割，还有优化空间吗？微信群已经加不进去了。", "id": 53}
{"title": "load model failed. 加载模型失败", "file": "2023-04-18.0132", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/132", "detail": "```", "id": 54}
{"title": "如何在webui里回答时同时返回引用的本地数据内容？", "file": "2023-04-18.0133", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/133", "detail": "如题", "id": 55}
{"title": "交流群满200人加不了了，能不能给个负责人的联系方式拉我进群？", "file": "2023-04-20.0143", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/143", "detail": "同求", "id": 56}
{"title": "No sentence-transformers model found with name ‘/text2vec/‘，但是再路径下面确实有模型文件", "file": "2023-04-20.0145", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/145", "detail": "另外：The dtype of attention mask (torch.int64) is not bool", "id": 57}
{"title": "请问加载模型的路径在哪里修改，默认好像前面会带上transformers_modules.", "file": "2023-04-20.0148", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/148", "detail": "<img width=\"1181\" alt=\"1681977897052\" src=\"https://user-images.githubusercontent.com/30926001/233301106-3846680a-d842-41d2-874e-5b6514d732c4.png\">", "id": 58}
{"title": "为啥放到方法调用会出错，这个怎么处理？", "file": "2023-04-20.0150", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/150", "detail": "```python", "id": 59}
{"title": "No sentence-transformers model found with name C:\\Users\\Administrator/.cache\\torch\\sentence_transformers\\GanymedeNil_text2vec-large-chinese. Creating a new one with MEAN pooling.", "file": "2023-04-21.0154", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/154", "detail": "卡在这块很久是正常现象吗", "id": 60}
{"title": "微信群需要邀请才能加入", "file": "2023-04-21.0155", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/155", "detail": "RT，给个个人联系方式白", "id": 61}
{"title": "No sentence-transformers model found with name GanymedeNil/text2vec-large-chinese. Creating a new one with MEAN pooling", "file": "2023-04-21.0156", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/156", "detail": "ls GanymedeNil/text2vec-large-chinese", "id": 62}
{"title": "embedding会加载两次", "file": "2023-04-23.0159", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/159", "detail": "你好，为什么要这样设置呢，这样会加载两次呀。", "id": 63}
{"title": "扫二维码加的那个群，群成员满了进不去了", "file": "2023-04-23.0160", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/160", "detail": "如题", "id": 64}
{"title": "执行python3 cli_demo.py 报错AttributeError: 'NoneType' object has no attribute 'chat'", "file": "2023-04-24.0163", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/163", "detail": "刚开始怀疑是内存不足问题，换成int4,int4-qe也不行，有人知道是什么原因吗", "id": 65}
{"title": "匹配得分", "file": "2023-04-24.0167", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/167", "detail": "在示例cli_demo.py中返回的匹配文本没有对应的score，可以加上这个feature吗", "id": 66}
{"title": "大佬有计划往web_ui.py加入打字机功能吗", "file": "2023-04-25.0170", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/170", "detail": "目前在载入了知识库后，单张V100 32G在回答垂直领域的问题时也需要20S以上，没有打字机逐字输出的使用体验还是比较煎熬的....", "id": 67}
{"title": "Is it possible to use a verctorDB for the embedings?", "file": "2023-04-25.0171", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/171", "detail": "when I play, I have to load the local data again and again when to start. I wonder if it is possible to use", "id": 68}
{"title": "请问通过lora训练官方模型得到的微调模型文件该如何加载？", "file": "2023-04-25.0173", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/173", "detail": "通过lora训练的方式得到以下文件:", "id": 69}
{"title": "from langchain.chains import RetrievalQA的代码在哪里？", "file": "2023-04-25.0174", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/174", "detail": "local_doc_qa.py", "id": 70}
{"title": "哪里有knowledge_based_chatglm.py文件？怎么找不到了？？是被替换成cli_demo.py文件了吗？", "file": "2023-04-26.0175", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/175", "detail": "哪里有knowledge_based_chatglm.py文件？怎么找不到了？？是被替换成cli_demo.py文件了吗？", "id": 71}
{"title": "AttributeError: 'Chatbot' object has no attribute 'value'", "file": "2023-04-26.0177", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/177", "detail": "Traceback (most recent call last):", "id": 72}
{"title": "控制台调api.py报警告", "file": "2023-04-26.0178", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/178", "detail": "you must pass the application as an import string to enable \"reload\" or \"workers\"", "id": 73}
{"title": "如何加入群聊", "file": "2023-04-27.0183", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/183", "detail": "微信群超过200人了，需要邀请，如何加入呢？", "id": 74}
{"title": "如何将Chatglm和本地知识相结合", "file": "2023-04-27.0185", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/185", "detail": "您好，我想请教一下怎么才能让知识库匹配到的文本和chatglm生成的相结合，而不是说如果没搜索到，就说根据已知信息无法回答该问题，谢谢", "id": 75}
{"title": "一点建议", "file": "2023-04-27.0189", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/189", "detail": "1.weiui的get_vector_store方法里面添加一个判断以兼容gradio版本导致的上传异常", "id": 76}
{"title": "windows环境下，按照教程，配置好conda环境，git完项目，修改完模型路径相关内容后，运行demo报错缺少", "file": "2023-04-28.0194", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/194", "detail": "报错代码如下：", "id": 77}
{"title": "ValueError: too many values to unpack (expected 2)", "file": "2023-04-28.0198", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/198", "detail": "When i tried to use the non-streaming, `ValueError: too many values to unpack (expected 2)` error came out.", "id": 78}
{"title": "加载doc后覆盖原本知识", "file": "2023-04-28.0201", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/201", "detail": "加载较大量级的私有知识库后，原本的知识会被覆盖", "id": 79}
{"title": "自定义知识库回答效果很差", "file": "2023-04-28.0203", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/203", "detail": "请问加了自定义知识库知识库,回答效果很差，是因为数据量太小的原因么", "id": 80}
{"title": "python310下，安装pycocotools失败，提示低版本cython，实际已安装高版本", "file": "2023-04-29.0208", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/208", "detail": "RT，纯离线环境安装，依赖安装的十分艰难，最后碰到pycocotools，始终无法安装上，求教方法！", "id": 81}
{"title": "[FEATURE] 支持 RWKV 模型（目前已有 pip package & rwkv.cpp 等等）", "file": "2023-05-01.0216", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/216", "detail": "您好，我是 RWKV 的作者，介绍见：https://zhuanlan.zhihu.com/p/626083366", "id": 82}
{"title": "[BUG] 为啥主机/服务器不联网不能正常启动服务？", "file": "2023-05-02.0220", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/220", "detail": "**问题描述 / Problem Description**", "id": 83}
{"title": "[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-05-03.0222", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/222", "detail": "**local variable 'torch' referenced before assignment**", "id": 84}
{"title": "不支持txt文件的中文输入", "file": "2023-05-04.0235", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/235", "detail": "vs_path, _ = local_doc_qa.init_knowledge_vector_store(filepath)", "id": 85}
{"title": "文件均未成功加载，请检查依赖包或替换为其他文件再次上传。 文件未成功加载，请重新上传文件", "file": "2023-05-05.0237", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/237", "detail": "请大佬帮忙解决，谢谢！", "id": 86}
{"title": "[BUG] 使用多卡时chatglm模型加载两次", "file": "2023-05-05.0241", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/241", "detail": "chatglm_llm.py文件下第129行先加载了一次chatglm模型，第143行又加载了一次", "id": 87}
{"title": "[BUG] similarity_search_with_score_by_vector函数返回多个doc时的score结果错误", "file": "2023-05-06.0252", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/252", "detail": "**问题描述 / Problem Description**", "id": 88}
{"title": "可以再建一个交流群吗，这个群满了进不去。", "file": "2023-05-06.0255", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/255", "detail": "上午应该已经在readme里更新过了，如果不能添加可能是网页缓存问题，可以试试看直接扫描img/qr_code_12.jpg", "id": 89}
{"title": "请问这是什么错误哇？KeyError: 'serialized_input'", "file": "2023-05-06.0257", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/257", "detail": "运行“python webui.py” 后这是什么错误？怎么解决啊？", "id": 90}
{"title": "修改哪里的代码，可以再cpu上跑？", "file": "2023-05-06.0258", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/258", "detail": "**问题描述 / Problem Description**", "id": 91}
{"title": "ModuleNotFoundError: No module named 'modelscope'", "file": "2023-05-07.0266", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/266", "detail": "安装这个", "id": 92}
{"title": "加载lora微调模型时，lora参数加载成功，但显示模型未成功加载？", "file": "2023-05-08.0270", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/270", "detail": "什么原因呀？", "id": 93}
{"title": "[BUG] 运行webui.py报错：name 'EMBEDDING_DEVICE' is not defined", "file": "2023-05-08.0274", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/274", "detail": "解决了，我修改model_config时候把这个变量改错了", "id": 94}
{"title": "基于ptuning训练完成，新老模型都进行了加载，但是只有新的", "file": "2023-05-08.0280", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/280", "detail": "licitly passing a `revision` is encouraged when loading a model with custom code to ensure no malicious code has been contributed in a newer revision.", "id": 95}
{"title": "[BUG] 使用chatyuan模型时，对话Error，has no attribute 'stream_chat'", "file": "2023-05-08.0282", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/282", "detail": "**问题描述 / Problem Description**", "id": 96}
{"title": "chaglm调用过程中 _call提示有一个 stop", "file": "2023-05-09.0286", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/286", "detail": "**功能描述 / Feature Description**", "id": 97}
{"title": "Logger._log() got an unexpected keyword argument 'end'", "file": "2023-05-10.0295", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/295", "detail": "使用cli_demo的时候，加载一个普通txt文件，输入问题后，报错：“TypeError: Logger._log() got an unexpected keyword argument 'end'”", "id": 98}
{"title": "[BUG] 请问可以解释下这个FAISS.similarity_search_with_score_by_vector = similarity_search_with_score_by_vector的目的吗", "file": "2023-05-10.0296", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/296", "detail": "我不太明白这个库自己写的similarity_search_with_score_by_vector方法做的事情，因为langchain原版的similarity_search_with_score_by_vector只是search faiss之后把返回的topk句子组合起来。我觉得原版理解起来没什么问题，但是这个库里自己写的我就没太看明白多做了什么其他的事情，因为没有注释。", "id": 99}
{"title": "[BUG] Windows下上传中文文件名文件，faiss无法生成向量数据库文件", "file": "2023-05-11.0318", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/318", "detail": "**问题描述 / Problem Description**", "id": 100}
{"title": "cli_demo中的流式输出能否接着前一答案输出?", "file": "2023-05-11.0320", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/320", "detail": "现有流式输出结果样式为:", "id": 101}
{"title": "内网部署时网页无法加载，能否增加离线静态资源", "file": "2023-05-12.0326", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/326", "detail": "内网部署时网页无法加载，能否增加离线静态资源", "id": 102}
{"title": "我想把文件字符的编码格式改为encoding='utf-8'在哪修改呢，因为会有ascii codec can't decode byte报错", "file": "2023-05-14.0360", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/360", "detail": "上传中文的txt文件时报错，编码格式为utf-8", "id": 103}
{"title": "Batches的进度条是在哪里设置的?能否关闭显示?", "file": "2023-05-15.0366", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/366", "detail": "使用cli_demo.py进行命令行测试时,每句回答前都有个Batches的进度条", "id": 104}
{"title": "ImportError: dlopen: cannot load any more object with static TLS or Segmentation fault", "file": "2023-05-15.0368", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/368", "detail": "**问题描述 / Problem Description**", "id": 105}
{"title": "读取PDF时报错", "file": "2023-05-16.0373", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/373", "detail": "在Colab上执行cli_demo.py时，在路径文件夹里放了pdf文件，在加载的过程中会显示错误，然后无法加载PDF文件", "id": 106}
{"title": "[BUG] webui报错 InvalidURL", "file": "2023-05-16.0375", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/375", "detail": "python 版本：3.8.16", "id": 107}
{"title": "[FEATURE] 如果让回答不包含出处，应该怎么处理", "file": "2023-05-16.0380", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/380", "detail": "**功能描述 / Feature Description**", "id": 108}
{"title": "加载PDF文件时，出现 unsupported colorspace for 'png'", "file": "2023-05-16.0381", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/381", "detail": "**问题描述 / Problem Description**", "id": 109}
{"title": "'ascii' codec can't encode characters in position 14-44: ordinal not in range(128) 经典bug", "file": "2023-05-16.0382", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/382", "detail": "添加了知识库之后进行对话，之后再新增知识库就会出现这个问题。", "id": 110}
{"title": "微信群人数超过200了，扫码进不去了，群主可以再创建一个新群吗", "file": "2023-05-17.0391", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/391", "detail": "**功能描述 / Feature Description**", "id": 111}
{"title": "TypeError: 'ListDocsResponse' object is not subscriptable", "file": "2023-05-17.0393", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/393", "detail": "应该是用remain_docs.code和remain_docs.data吧？吗？", "id": 112}
{"title": "[BUG] 加载chatglm模型报错：'NoneType' object has no attribute 'message_types_by_name'", "file": "2023-05-17.0398", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/398", "detail": "**问题描述 / Problem Description**", "id": 113}
{"title": "[BUG] 执行 python webui.py 没有报错，但是ui界面提示 Something went wrong Expecting value: line 1 column 1 (char 0", "file": "2023-05-18.0399", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/399", "detail": "**环境配置**", "id": 114}
{"title": "启动后调用api接口正常，过一会就不断的爆出 Since the angle classifier is not initialized", "file": "2023-05-18.0404", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/404", "detail": "**问题描述 / Problem Description**", "id": 115}
{"title": "[BUG] write_check_file方法中，open函数未指定编码", "file": "2023-05-18.0408", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/408", "detail": "def write_check_file(filepath, docs):", "id": 116}
{"title": "导入的PDF中存在图片，有大概率出现 “unsupported colorspace for 'png'”异常", "file": "2023-05-18.0409", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/409", "detail": "pix = fitz.Pixmap(doc, img[0])", "id": 117}
{"title": "请问流程图是用什么软件画的", "file": "2023-05-18.0410", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/410", "detail": "draw.io", "id": 118}
{"title": "mac 加载模型失败", "file": "2023-05-19.0417", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/417", "detail": "Explicitly passing a `revision` is encouraged when loading a model with custom code to ensure no malicious code has been contributed in a newer revision.", "id": 119}
{"title": "使用GPU本地运行知识库问答，提问第一个问题出现异常。", "file": "2023-05-20.0419", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/419", "detail": "配置文件model_config.py为：", "id": 120}
{"title": "想加入讨论群", "file": "2023-05-20.0420", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/420", "detail": "OK", "id": 121}
{"title": "有没有直接调用LLM的API，目前只有知识库的API？", "file": "2023-05-22.0426", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/426", "detail": "-------------------------------------------------------------------------------", "id": 122}
{"title": "上传文件后出现 ERROR __init__() got an unexpected keyword argument 'autodetect_encoding'", "file": "2023-05-22.0428", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/428", "detail": "上传文件后出现这个问题：ERROR 2023-05-22 11:46:19,568-1d: __init__() got an unexpected keyword argument 'autodetect_encoding'", "id": 123}
{"title": "想问下README中用到的流程图用什么软件画的", "file": "2023-05-22.0431", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/431", "detail": "**功能描述 / Feature Description**", "id": 124}
{"title": "No matching distribution found for langchain==0.0.174", "file": "2023-05-23.0436", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/436", "detail": "ERROR: Could not find a version that satisfies the requirement langchain==0.0.174 ", "id": 125}
{"title": "[FEATURE] bing是必须的么？", "file": "2023-05-23.0437", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/437", "detail": "从这个[脚步](https://github.com/imClumsyPanda/langchain-ChatGLM/blob/master/configs/model_config.py#L129)里面发现需要申请bing api，如果不申请，纯用模型推理不可吗？", "id": 126}
{"title": "同一台环境下部署了5.22号更新的langchain-chatglm v0.1.13和之前的版本，回复速度明显变慢", "file": "2023-05-23.0442", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/442", "detail": "新langchain-chatglm v0.1.13版本速度很慢", "id": 127}
{"title": "Error reported during startup", "file": "2023-05-23.0443", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/443", "detail": "Traceback (most recent call last):", "id": 128}
{"title": "ValueError: not enough values to unpack (expected 2, got 1)on of the issue", "file": "2023-05-24.0449", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/449", "detail": "File \".cache\\huggingface\\modules\\transformers_modules\\chatglm-6b-int4\\modeling_chatglm.py\", line 1280, in chat", "id": 129}
{"title": "[BUG] API部署，流式输出的函数，少了个question", "file": "2023-05-24.0451", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/451", "detail": "**问题描述 / Problem Description**", "id": 130}
{"title": "项目结构的简洁性保持", "file": "2023-05-24.0454", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/454", "detail": "**功能描述 / Feature Description**", "id": 131}
{"title": "项目群扫码进不去了", "file": "2023-05-24.0455", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/455", "detail": "项目群扫码进不去了，是否可以加一下微信拉我进群，谢谢！微信号：daniel-0527", "id": 132}
{"title": "请求拉我入群讨论，海硕一枚，专注于LLM等相关技术", "file": "2023-05-24.0461", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/461", "detail": "**功能描述 / Feature Description**", "id": 133}
{"title": "[BUG] chatglm-6b模型报错OSError: Error no file named pytorch_model.bin found in directory /chatGLM/model/model-6b", "file": "2023-05-26.0474", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/474", "detail": "**1、简述：**", "id": 134}
{"title": "现在本项目交流群二维码扫描不进去了，需要群主通过", "file": "2023-05-27.0478", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/478", "detail": "现在本项目交流群二维码扫描不进去了，需要群主通过", "id": 135}
{"title": "RuntimeError: Only Tensors of floating point and complex dtype can require gradients", "file": "2023-05-28.0483", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/483", "detail": "刚更新了最新版本：", "id": 136}
{"title": "RuntimeError: \"LayerNormKernelImpl\" not implemented for 'Half'", "file": "2023-05-28.0484", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/484", "detail": "已经解决了 params 只用两个参数 {'trust_remote_code': True, 'torch_dtype': torch.float16}", "id": 137}
{"title": "[BUG] 文件未成功加载，请重新上传文件", "file": "2023-05-31.0504", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/504", "detail": "webui.py", "id": 138}
{"title": "[BUG] bug 17 ，pdf和pdf为啥还不一样呢？为啥有的pdf能识别？有的pdf识别不了呢？", "file": "2023-05-31.0506", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/506", "detail": "bug 17 ，pdf和pdf为啥还不一样呢？为啥有的pdf能识别？有的pdf识别不了呢？", "id": 139}
{"title": "[FEATURE] 简洁阐述功能 / Concise description of the feature", "file": "2023-05-31.0513", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/513", "detail": "**功能描述 / Feature Description**", "id": 140}
{"title": "[BUG] webui.py 加载chatglm-6b-int4 失败", "file": "2023-06-02.0524", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/524", "detail": "**问题描述 / Problem Description**", "id": 141}
{"title": "[BUG] webui.py 加载chatglm-6b模型异常", "file": "2023-06-02.0525", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/525", "detail": "**问题描述 / Problem Description**", "id": 142}
{"title": "增加对chatgpt的embedding和api调用的支持", "file": "2023-06-02.0531", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/531", "detail": "能否支持openai的embedding api和对话的api？", "id": 143}
{"title": "[FEATURE] 调整模型下载的位置", "file": "2023-06-02.0537", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/537", "detail": "模型默认下载到 $HOME/.cache/huggingface/，当 C 盘空间不足时无法完成模型的下载。configs/model_config.py 中也没有调整模型位置的参数。", "id": 144}
{"title": "[BUG] langchain=0.0.174 出错", "file": "2023-06-04.0543", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/543", "detail": "**问题描述 / Problem Description**", "id": 145}
{"title": "[BUG] 更新后加载本地模型路径不正确", "file": "2023-06-05.0545", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/545", "detail": "**问题描述 / Problem Description**", "id": 146}
{"title": "SystemError: 8bit 模型需要 CUDA 支持，或者改用量化后模型！", "file": "2023-06-06.0550", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/550", "detail": "docker 部署后,启动docker,过会儿容器会自动退出,logs报错 SystemError: 8bit 模型需要 CUDA 支持，或者改用量化后模型！ [NVIDIA Container Toolkit](https://github.com/NVIDIA/nvidia-container-toolkit) 也已经安装了", "id": 147}
{"title": "[BUG] 上传知识库超过1M报错", "file": "2023-06-06.0556", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/556", "detail": "**问题描述 / Problem Description**", "id": 148}
{"title": "打开跨域访问后仍然报错，不能请求", "file": "2023-06-06.0560", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/560", "detail": "报错信息：", "id": 149}
{"title": "dialogue_answering 里面的代码是不是没有用到？，没有看到调用", "file": "2023-06-07.0571", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/571", "detail": "dialogue_answering 是干啥的", "id": 150}
{"title": "[BUG] 响应速度极慢，应从哪里入手优化？48C/128G/8卡", "file": "2023-06-07.0573", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/573", "detail": "运行环境：ubuntu20.04", "id": 151}
{"title": "纯CPU环境下运行cli_demo时报错，提示找不到nvcuda.dll", "file": "2023-06-08.0576", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/576", "detail": "本地部署环境是纯CPU，之前的版本在纯CPU环境下能正常运行，但上传本地知识库经常出现encode问题。今天重新git项目后，运行时出现如下问题，请问该如何解决。", "id": 152}
{"title": "如何加载本地的embedding模型（text2vec-large-chinese模型文件）", "file": "2023-06-08.0582", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/582", "detail": "因为需要离线部署，所以要把模型放到本地，我修改了chains/local_doc_qa.py中的HuggingFaceEmbeddings()，在其中加了一个cache_folder的参数，保证下载的文件在cache_folder中，model_name是text2vec-large-chinese。如cache_folder='/home/xx/model/text2vec-large-chinese', model_name='text2vec-large-chinese'，这样仍然需要联网下载报错，请问大佬如何解决该问题？", "id": 153}
{"title": "ChatGLM-6B 在另外服务器安装好了，请问如何修改model.cofnig.py 来使用它的接口呢？？", "file": "2023-06-09.0588", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/588", "detail": "我本来想在这加一个api base url 但是运行web.py 发现 还是会去连huggingface 下载模型", "id": 154}
{"title": "[BUG] raise partially initialized module 'charset_normalizer' has no attribute 'md__mypyc' when call interface `upload_file`", "file": "2023-06-10.0591", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/591", "detail": "**问题描述 / Problem Description**", "id": 155}
{"title": "[BUG] raise OSError: [Errno 101] Network is unreachable when call interface upload_file and upload .pdf files", "file": "2023-06-10.0592", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/592", "detail": "**问题描述 / Problem Description**", "id": 156}
{"title": "如果直接用vicuna作为基座大模型，需要修改的地方有哪些？", "file": "2023-06-12.0596", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/596", "detail": "vicuna模型有直接转换好的没有？也就是llama转换之后的vicuna。", "id": 157}
{"title": "[BUG] 通过cli.py调用api时抛出AttributeError: 'NoneType' object has no attribute 'get'错误", "file": "2023-06-12.0598", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/598", "detail": "通过`python cli.py start api --ip localhost --port 8001` 命令调用api时，抛出：", "id": 158}
{"title": "[BUG] 通过cli.py调用api时直接报错`langchain-ChatGLM: error: unrecognized arguments: start cli`", "file": "2023-06-12.0601", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/601", "detail": "通过python cli.py start cli启动cli_demo时，报错：", "id": 159}
{"title": "[BUG] error: unrecognized arguments: --model-dir conf/models/", "file": "2023-06-12.0602", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/602", "detail": "关键字参数修改了吗？有没有文档啊？大佬", "id": 160}
{"title": "[BUG] 上传文件全部失败", "file": "2023-06-12.0603", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/603", "detail": "ERROR: Exception in ASGI application", "id": 161}
{"title": "[BUG] config 使用 chatyuan 无法启动", "file": "2023-06-12.0604", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/604", "detail": "\"chatyuan\": {", "id": 162}
{"title": "使用fashchat api之后，后台报错APIError 如图所示", "file": "2023-06-12.0606", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/606", "detail": "我按照https://github.com/imClumsyPanda/langchain-ChatGLM/blob/master/docs/fastchat.md", "id": 163}
{"title": "[BUG] 启用上下文关联，每次embedding搜索到的内容都会比前一次多一段", "file": "2023-06-13.0613", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/613", "detail": "**问题描述 / Problem Description**", "id": 164}
{"title": "local_doc_qa.py中MyFAISS.from_documents() 这个语句看不太懂。MyFAISS类中没有这个方法，其父类FAISS和VectorStore中也只有from_texts方法[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-06-14.0619", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/619", "detail": "local_doc_qa.py中MyFAISS.from_documents() 这个语句看不太懂。MyFAISS类中没有这个方法，其父类FAISS和VectorStore中也只有from_texts方法", "id": 165}
{"title": "[BUG] TypeError: similarity_search_with_score_by_vector() got an unexpected keyword argument 'filter'", "file": "2023-06-14.0624", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/624", "detail": "**问题描述 / Problem Description**", "id": 166}
{"title": "please delete this issue", "file": "2023-06-15.0633", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/633", "detail": "sorry, incorrect submission. Please remove this issue!", "id": 167}
{"title": "[BUG] vue前端镜像构建失败", "file": "2023-06-15.0635", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/635", "detail": "**问题描述 / Problem Description**", "id": 168}
{"title": "ChatGLM-6B模型能否回答英文问题？", "file": "2023-06-15.0640", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/640", "detail": "大佬，请问一下，如果本地知识文档是英文，ChatGLM-6B模型能否回答英文问题？不能的话，有没有替代的模型推荐，期待你的回复，谢谢", "id": 169}
{"title": "[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-06-16.0644", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/644", "detail": "**问题描述 / Problem Description**", "id": 170}
{"title": "KeyError: 3224", "file": "2023-06-16.0645", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/645", "detail": "```", "id": 171}
knowledge_base/samples/isssues_merge/langchain-ChatGLM_open.csv 0 → 100644
,title,file,url,detail,id
0,效果如何优化,2023-04-04.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/14,如图所示，将该项目的README.md和该项目结合后，回答效果并不理想，请问可以从哪些方面进行优化,0
1,怎么让模型严格根据检索的数据进行回答，减少胡说八道的回答呢,2023-04-04.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/15,举个例子：,1
2,"When I try to run the `python knowledge_based_chatglm.py`, I got this error in macOS(M1 Max, OS 13.2)",2023-04-07.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/32,```python,2
3,萌新求教大佬怎么改成AMD显卡或者CPU？,2023-04-10.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/48,把.cuda()去掉就行,3
4,输出answer的时间很长，是否可以把文本向量化的部分提前做好存储起来？,2023-04-10.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/50,GPU：4090 24G显存,4
5,报错Use `repo_type` argument if needed.,2023-04-11.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/57,Traceback (most recent call last):,5
6,无法打开gradio的页面,2023-04-11.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/58,$ python webui.py,6
7,支持word，那word里面的图片正常显示吗？,2023-04-12.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/60,如题，刚刚从隔壁转过来的，想先了解下,7
8,detectron2 is not installed. Cannot use the hi_res partitioning strategy. Falling back to partitioning with the fast strategy.,2023-04-12.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/63,能够正常的跑起来，在加载content文件夹中的文件时，每加载一个文件都会提示：,8
9,cpu上运行webui，step3 asking时报错,2023-04-12.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/66,web运行，文件加载都正常，asking时报错,9
10,建议弄一个插件系统,2023-04-13.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/67,如题弄成stable-diffusion-webui那种能装插件，再开一个存储库给使用者或插件开发，存储或下载插件。,10
11,请教加载模型出错！？,2023-04-13.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/75,AttributeError: module 'transformers_modules.chatglm-6b.configuration_chatglm' has no attribute 'ChatGLMConfig 怎么解决呀,11
12,从本地知识检索内容的时候，是否可以设置相似度阈值，小于这个阈值的内容不返回，即使会小于设置的VECTOR_SEARCH_TOP_K参数呢？谢谢大佬,2023-04-13.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/76,比如 问一些 你好/你是谁 等一些跟本地知识库无关的问题,12
13,如何改成多卡推理？,2023-04-13.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/77,+1,13
14,能否弄个懒人包，可以一键体验？,2023-04-13.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/78,能否弄个懒人包，可以一键体验？,14
15,连续问问题会导致崩溃,2023-04-13.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/79,看上去不是爆内存的问题，连续问问题后，会出现如下报错,15
16,AttributeError: 'NoneType' object has no attribute 'as_retriever',2023-04-14.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/86,"环境：windows 11, anaconda/python 3.8",16
17,FileNotFoundError: Could not find module 'nvcuda.dll' (or one of its dependencies). Try using the full path with constructor syntax.,2023-04-14.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/87,请检查一下cuda或cudnn是否存在安装问题,17
18,加载txt文件失败？,2023-04-14.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/89,![JppHrGOWFa](https://user-images.githubusercontent.com/109277248/232009383-bf7c46d1-a01e-4e0a-9de6-5b5ed3e36158.jpg),18
19,NameError: name 'chatglm' is not defined,2023-04-14.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/90,"This share link expires in 72 hours. For free permanent hosting and GPU upgrades (NEW!), check out Spaces: https://huggingface.co/spaces",19
20,打不开地址？,2023-04-14.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/91,报错数据如下：,20
21,加载md文件出错,2023-04-14.00,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/98,运行 webui.py后能访问页面，上传一个md文件后，日志中有错误。等待后能加载完成，提示可以提问了，但提问没反应，日志中有错误。 具体日志如下。,21
22,建议增加获取在线知识的能力,2023-04-15.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/101,建议增加获取在线知识的能力,22
23,txt 未能成功加载,2023-04-15.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/103,hinese. Creating a new one with MEAN pooling.,23
24,pdf加载失败,2023-04-15.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/105,e:\a.txt加载成功了，e:\a.pdf加载就失败，pdf文件里面前面几页是图片，后面都是文字，加载失败没有报更多错误，请问该怎么排查？,24
25,一直停在文本加载处,2023-04-15.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/108,一直停在文本加载处,25
26," File ""/root/.cache/huggingface/modules/transformers_modules/chatglm-6b/modeling_chatglm.py"", line 440, in forward new_tensor_shape = mixed_raw_layer.size()[:-1] + ( TypeError: torch.Size() takes an iterable of 'int' (item 2 is 'float')",2023-04-17.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/113,按照最新的代码，发现,26
27,后续会提供前后端分离的功能吗？,2023-04-17.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/114,类似这种https://github.com/lm-sys/FastChat/tree/main/fastchat/serve,27
28,安装依赖报错,2023-04-17.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/115,(test) C:\Users\linh\Desktop\langchain-ChatGLM-master>pip install -r requirements.txt,28
29,问特定问题会出现爆显存,2023-04-17.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/117,正常提问没问题。,29
30,Expecting value: line 1 column 1 (char 0),2023-04-17.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/118,运行后 第一步加载配置一直报错：,30
31,embedding https://huggingface.co/GanymedeNil/text2vec-large-chinese/tree/main是免费的，效果比对openai的如何？,2023-04-17.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/119,-------------------------------------------------------------------------------,31
32,这是什么错误，在Colab上运行的。,2023-04-17.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/120,libcuda.so.1: cannot open shared object file: No such file or directory,32
33,只想用自己的lora微调后的模型进行对话，不想加载任何本地文档，该如何调整？,2023-04-18.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/121,能出一个单独的教程吗,33
34,"租的gpu,Running on local URL: http://0.0.0.0:7860 To create a public link, set `share=True` in `launch()`. 浏览器上访问不了？？？",2023-04-18.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/122,(chatglm20230401) root@autodl-container-e82d11963c-10ece0d7:~/autodl-tmp/chatglm/langchain-ChatGLM-20230418# python3.9 webui.py,34
35,本地部署中的报错请教,2023-04-18.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/124,"您好，在本地运行langchain-ChatGLM过程中，环境及依赖的包都已经满足条件，但是运行webui.py,报错如下（运行cli_demo.py报错类似），请问是哪里出了错呢？盼望您的回复，谢谢！",35
36,报错。The dtype of attention mask (torch.int64) is not bool,2023-04-18.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/131,The dtype of attention mask (torch.int64) is not bool,36
37,[求助] pip install -r requirements.txt 的时候出现以下报错。。。有大佬帮忙看看怎么搞么，下的release里面的包,2023-04-18.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/134,$ pip install -r requirements.txt,37
38,如何提升根据问题搜索到对应知识的准确率,2023-04-19.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/136,外链知识库最大的问题在于问题是短文本，知识是中长文本。如何根据问题精准的搜索到对应的知识是个最大的问题。这类本地化项目不像百度，由无数的网页，基本上每个问题都可以找到对应的页面。,38
39,是否可以增加向量召回的阈值设定，有些召回内容相关性太低，导致模型胡言乱语,2023-04-20.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/140,如题,39
40,输入长度问题,2023-04-20.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/141,感谢作者支持ptuning微调模型。,40
41,已有部署好的chatGLM-6b，如何通过接口接入？,2023-04-20.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/144,已有部署好的chatGLM-6b，如何通过接口接入，而不是重新加载一个模型；,41
42,执行web_demo.py后，显示Killed，就退出了，是不是配置不足呢？,2023-04-20.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/146,![图片](https://user-images.githubusercontent.com/26102866/233256425-c7aab999-11d7-4de9-867b-23ef18d519e4.png),42
43,执行python cli_demo1.py,2023-04-20.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/147,Traceback (most recent call last):,43
44,报错：ImportError: cannot import name 'GENERATION_CONFIG_NAME' from 'transformers.utils',2023-04-20.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/149,(mychatGLM) PS D:\Users\admin3\zrh\langchain-ChatGLM> python cli_demo.py,44
45,上传文件并加载知识库时，会不停地出现临时文件,2023-04-21.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/153,环境：ubuntu 18.04,45
46,向知识库中添加文件后点击”上传文件并加载知识库“后Segmentation fault报错。,2023-04-23.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/161,运行服务后的提示如下：,46
47,langchain-serve 集成,2023-04-24.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/162,Hey 我是来自 [langchain-serve](https://github.com/jina-ai/langchain-serve) 的dev！,47
48,大佬们，wsl的ubuntu怎么配置用cuda加速，装了运行后发现是cpu在跑,2023-04-24.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/164,大佬们，wsl的ubuntu怎么配置用cuda加速，装了运行后发现是cpu在跑,48
49,在github codespaces docker运行出错,2023-04-24.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/165,docker run -d --restart=always --name chatglm -p 7860:7860 -v /www/wwwroot/code/langchain-ChatGLM:/chatGLM chatglm,49
50,有计划接入Moss模型嘛,2023-04-24.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/166,后续会开展测试，目前主要在优化langchain部分效果，如果有兴趣也欢迎提PR,50
51,怎么实现 API 部署？,2023-04-24.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/168,利用 fastapi 实现 API 部署方式，具体怎么实现，有方法说明吗？,51
52, 'NoneType' object has no attribute 'message_types_by_name'报错,2023-04-24.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/169,_HISTOGRAMPROTO = DESCRIPTOR.message_types_by_name['HistogramProto'],52
53,能否指定自己训练的text2vector模型？,2023-04-25.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/172,请问大佬：,53
54,关于项目支持的模型以及quantization_bit潜在的影响的问题,2023-04-26.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/176,作者您好～,54
55,运行python3.9 api.py WARNING: You must pass the application as an import string to enable 'reload' or 'workers'.,2023-04-26.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/179,api.py文件最下面改成这样试试：,55
56,ValidationError: 1 validation error for HuggingFaceEmbeddings model_kwargs extra fields not permitted (type=value_error.extra),2023-04-26.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/180,ValidationError: 1 validation error for HuggingFaceEmbeddings,56
57,如果没有检索到相关性比较高的，回答“我不知道”,2023-04-26.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/181,如果通过设计system_template，让模型在搜索到的文档都不太相关的情况下回答“我不知道”,57
58,请问如果不能联网，6B之类的文件从本地上传需要放到哪里,2023-04-26.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/182,感谢大佬的项目，很有启发~,58
59,知识库问答--输入新的知识库名称是中文的话，会报error,2023-04-27.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/184,知识库问答--输入新的知识库名称是中文的话，会报error，选择要加载的知识库那里也不显示之前添加的知识库,59
60,现在能通过问题匹配的相似度值，来直接返回文档中的文段，而不经过模型吗？因为有些答案在文档中，模型自己回答，不能回答文档中的答案,2023-04-27.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/186,现在能通过问题匹配的相似度值，来直接返回文档中的文段，而不经过模型吗？因为有些答案在文档中，模型自己回答，不能回答文档中的答案。也就是说，提供向量检索回答+模型回答相结合的策略。如果相似度值高于一定数值，直接返回文档中的文本，没有高于就返回模型的回答或者不知道,60
61,"TypeError: The type of ChatGLM.callback_manager differs from the new default value; if you wish to change the type of this field, please use a type annotation",2023-04-27.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/188,"Mac 运行 python3 ./webui.py 报 TypeError: The type of ChatGLM.callback_manager differs from the new default value; if you wish to change the type of this field, please use a type annotation",61
62,Not Enough Memory,2023-04-27.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/190,"运行命令行程序python cli_demo.py， 已经成功加载pdf文件, 报“DefaultCPUAllocator: not enough memory: you tried to allocate 458288380900 bytes”错误，请问哪里可以配置default memory",62
63,参与开发问题,2023-04-27.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/191,1.是否需要进专门的开发群,63
64,对话框中代码片段格式需改进,2023-04-27.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/192,最好能改进下输出代码片段的格式，目前输出的格式还不友好。,64
65,请问未来有可能支持belle吗,2023-04-28.01,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/195,如题，谢谢大佬,65
66,TypeError: cannot unpack non-iterable NoneType object,2023-04-28.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/200,"When i tried to change the knowledge vector store through `init_knowledge_vector_store`, the error `TypeError: cannot unpack non-iterable NoneType object` came out.",66
67,生成结果,2023-04-28.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/202,你好，想问一下langchain+chatglm-6B，找到相似匹配的prompt，是直接返回prompt对应的答案信息，还是chatglm-6B在此基础上自己优化答案？,67
68,在win、ubuntu下都出现这个错误：attributeerror: 't5forconditionalgeneration' object has no attribute 'stream_chat',2023-04-29.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/207,在win、ubuntu。下载完模型后，没办法修改代码以执行本地模型，每次都要重新输入路径； LLM 模型、Embedding 模型支持也都在官网下的，在其他项目（wenda）下可以使用,68
69,[FEATURE] knowledge_based_chatglm.py: renamed or missing?,2023-04-30.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/210,"Not found. Was it renamed? Or, is it missing? How can I get it?",69
70,sudo apt-get install -y nvidia-container-toolkit-base执行报错,2023-05-01.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/211,**问题描述 / Problem Description**,70
71,效果不佳几乎答不上来,2023-05-01.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/212,提供了50条问答的docx文件,71
72,有没有可能新增一个基于chatglm api调用的方式构建langchain,2023-05-02.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/218,我有两台8G GPU/40G内存的服务器，一个台做成了chatglm的api ；想基于另外一台服务器部署langchain；网上好像没有类似的代码。,72
73,电脑是intel的集成显卡； 运行时告知我找不到nvcuda.dll，模型无法运行,2023-05-02.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/219,您好，我的电脑是intel的集成显卡，不过CPU是i5-11400 @ 2.60GHz ，内存64G；,73
74,根据langchain官方的文档和使用模式，是否可以改Faiss为Elasticsearch？会需要做哪些额外调整？求解,2023-05-03.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/221,本人新手小白，由于业务模式的原因（有一些自己的场景和优化），希望利用Elasticsearch做这个体系内部的检索机制，不知道是否可以替换，同时，还会涉及到哪些地方的改动？或者说可能会有哪些其他影响，希望作者和大佬们不吝赐教！,74
75,请问未来有可能支持t5吗,2023-05-04.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/224,请问可能支持基於t5的模型吗?,75
76,[BUG] 内存溢出 / torch.cuda.OutOfMemoryError:,2023-05-04.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/229,**问题描述 / Problem Description**,76
77,报错 No module named 'chatglm_llm',2023-05-04.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/230,明明已经安装了包，却在python里吊不出来,77
78,能出一个api部署的描述文档吗,2023-05-04.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/233,**功能描述 / Feature Description**,78
79,使用docs/API.md 出错,2023-05-04.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/234,使用API.md文档2种方法，出错,79
80,加载pdf文档报错？,2023-05-05.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/238,ew one with MEAN pooling.,80
81,上传的本地知识文件后再次上传不能显示，只显示成功了一个，别的上传成功后再次刷新就没了,2023-05-05.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/239,您好，项目有很大启发，感谢~,81
82,创建了新的虚拟环境，安装了相关包，并且自动下载了相关的模型，但是仍旧出现：OSError: Unable to load weights from pytorch checkpoint file for,2023-05-05.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/240,![78ac8e663fdc312d0e9d78da95925c4](https://user-images.githubusercontent.com/34124260/236378728-9ea4424f-0f7f-4013-9d33-820b723de321.png),82
83,[BUG] 数据加载不进来,2023-05-05.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/243,使用的.txt格式，utf-8编码，报以下错误,83
84,不能读取pdf,2023-05-05.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/244,请问是webui还是cli_demo,84
85,本地txt文件有500M，加载的时候很慢，如何提高速度？,2023-05-06.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/251,![yayRzxSYHP](https://user-images.githubusercontent.com/109277248/236592902-f5ab338d-c1e9-43dc-ae16-9df2cd3c1378.jpg),85
86,[BUG] gradio上传知识库后刷新之后 知识库就不见了 只有重启才能看到之前的上传的知识库,2023-05-06.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/253,gradio上传知识库后刷新之后 知识库就不见了 只有重启才能看到之前的上传的知识库,86
87,[FEATURE] 可以支持 OpenAI 的模型嘛？比如 GPT-3、GPT-3.5、GPT-4；embedding 增加 text-embedding-ada-002,2023-05-06.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/254,**功能描述 / Feature Description**,87
88,[FEATURE] 能否增加对于milvus向量数据库的支持 / Concise description of the feature,2023-05-06.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/256,**功能描述 / Feature Description**,88
89,CPU和GPU上跑，除了速度有区别，准确率效果回答上有区别吗？,2023-05-06.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/259,理论上没有区别,89
90,m1，请问在生成回答时怎么看是否使用了mps or cpu？,2023-05-06.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/260,m1，请问在生成回答时怎么看是否使用了mps or cpu？,90
91,知识库一刷新就没了,2023-05-07.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/263,知识库上传后刷新就没了,91
92,本地部署报没有模型,2023-05-07.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/267,建议在下载llm和embedding模型至本地后在configs/model_config中写入模型本地存储路径后再运行,92
93,[BUG] python3: can't open file 'webui.py': [Errno 2] No such file or directory,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/269,**问题描述 / Problem Description**,93
94,模块缺失提示,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/271,因为已有自己使用的docker环境，直接启动webui.py，提示,94
95,"运行api.py后，执行curl -X POST ""http://127.0.0.1:7861"" 报错？",2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/272,"执行curl -X POST ""http://127.0.0.1:7861"" \ -H 'Content-Type: application/json' \ -d '{""prompt"": ""你好"", ""history"": []}'，报错怎么解决",95
96,[BUG] colab安装requirements提示protobuf版本问题？,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/273,pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.,96
97,请问项目里面向量相似度使用了什么方法计算呀？,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/275,基本按照langchain里的FAISS.similarity_search_with_score_by_vector实现,97
98,[BUG] 安装detectron2后，pdf无法加载,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/276,**问题描述 / Problem Description**,98
99,[BUG] 使用ChatYuan-V2模型无法流式输出，会报错,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/277,一方面好像是ChatYuan本身不支持stream_chat，有人在clueai那边提了issue他们说还没开发，所以估计这个attribute调不起来；但是另一方面看报错好像是T5模型本身就不是decoder-only模型，所以不能流式输出吧（个人理解）,99
100,[BUG] 无法加载text2vec模型,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/278,**问题描述 / Problem Description**,100
101,请问能否增加网络搜索功能,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/281,请问能否增加网络搜索功能,101
102,[FEATURE] 结构化数据sql、excel、csv啥时会支持呐。,2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/283,**功能描述 / Feature Description**,102
103,TypeError: ChatGLM._call() got an unexpected keyword argument 'stop',2023-05-08.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/284,No sentence-transformers model found with name D:\DevProject\langchain-ChatGLM\GanymedeNil\text2vec-large-chinese. Creating a new one with MEAN pooling.,103
104,关于api.py的一些bug和设计逻辑问题？,2023-05-09.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/285,首先冒昧的问一下，这个api.py，开发者大佬们是在自己电脑上测试后确实没问题吗？,104
105,有没有租用的算力平台上，运行api.py后，浏览器http://localhost:7861/报错,2023-05-09.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/287,是不是租用的gpu平台上都会出现这个问题？？？,105
106,请问一下项目中有用到文档段落切割方法吗？,2023-05-09.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/288,text_load中的文档切割方法用上了吗？在代码中看好像没有用到？,106
107,"报错 raise ValueError(f""Knowledge base {knowledge_base_id} not found"") ValueError: Knowledge base ./vector_store not found",2023-05-09.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/289,"File ""/root/autodl-tmp/chatglm/langchain-ChatGLM-master/api.py"", line 183, in chat",107
108,能接入vicuna模型吗,2023-05-09.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/290,目前本地已经有了vicuna模型能直接接入吗？,108
109,[BUG] 提问公式相关问题大概率爆显存,2023-05-09.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/291,**问题描述 / Problem Description**,109
110,安装pycocotools失败，找了好多方法都不能解决。,2023-05-10.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/292,**问题描述 / Problem Description**,110
111,使用requirements安装，PyTorch安装的是CPU版本,2023-05-10.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/294,如题目，使用requirements安装，PyTorch安装的是CPU版本，运行程序的时候，也是使用CPU在工作。,111
112,能不能给一个毛坯服务器的部署教程,2023-05-10.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/298,“开发部署”你当成服务器的部署教程用就行了。,112
113, Error(s) in loading state_dict for ChatGLMForConditionalGeneration:,2023-05-10.02,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/299,运行中出现的问题，7860的端口页面显示不出来，求助。,113
114,ChatYuan-large-v2模型加载失败,2023-05-10.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/300,**实际结果 / Actual Result**,114
115,新增摘要功能,2023-05-10.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/303,你好，后续会考虑新增对长文本信息进行推理和语音理解功能吗？比如生成摘要,115
116,[BUG] pip install -r requirements.txt 出错,2023-05-10.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/304,pip install langchain -i https://pypi.org/simple,116
117,[BUG] 上传知识库文件报错,2023-05-10.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/305,![19621e29eaa547d01213bee53d81e6a](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/84606552/7f6ceb46-e494-4b0e-939c-23b585a6d9d8),117
118,[BUG] AssertionError: <class 'gradio.layouts.Accordion'> Component with id 41 not a valid input component.,2023-05-10.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/306,**问题描述 / Problem Description**,118
119,[BUG] CUDA out of memory with container deployment,2023-05-10.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/310,**问题描述 / Problem Description**,119
120,[FEATURE] 增加微调训练功能,2023-05-11.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/311,**功能描述 / Feature Description**,120
121,如何使用多卡部署，多个gpu,2023-05-11.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/315,"机器上有多个gpu,如何全使用了",121
122,请问这个知识库问答，和chatglm的关系是什么,2023-05-11.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/319,这个知识库问答，哪部分关联到了chatglm，是不是没有这个chatglm，知识库问答也可单单拎出来,122
123,[BUG] 运行的时候报错ImportError: libcudnn.so.8: cannot open shared object file: No such file or directory,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/324,**问题描述 / Problem Description**raceback (most recent call last):,123
124,webui启动成功，但有报错,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/325,**问题描述 / Problem Description**,124
125,切换MOSS的时候报错,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/327,danshi但是发布的源码中，,125
126,vicuna模型是否能接入？,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/328,您好！关于MOSS模型和vicuna模型，都是AutoModelForCausalLM来加载模型的，但是稍作更改（模型路径这些）会报这个错误。这个错误的造成是什么,126
127,你好，请问一下在阿里云CPU服务器上跑可以吗？可以的话比较理想的cpu配置是什么？,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/330,你好，请问一下在阿里云CPU服务器上跑可以吗？可以的话比较理想的cpu配置是什么？,127
128,你好，请问8核32g的CPU可以跑多轮对话吗？,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/331,什么样的cpu配置比较好呢？我目前想部署CPU下的多轮对话？,128
129,[BUG] 聊天内容输入超过10000个字符系统出现错误,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/332,聊天内容输入超过10000个字符系统出现错误，如下图所示：,129
130,能增加API的多用户访问接口部署吗？,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/333,默认部署程序仅支持单用户访问，多用户则需要排队访问。测试过相关的几个Github多用户工程，但是其中一些仍然不满足要求。本节将系统介绍如何实现多用户同时访问ChatGLM的部署接口，包括http、websocket（流式输出，stream）和web页面等方式，主要目录如下所示。,130
131,多卡部署,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/334,用单机多卡或多机多卡，fastapi部署模型，怎样提高并发,131
132,WEBUI能否指定知识库目录？,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/335,**功能描述 / Feature Description**,132
133,[BUG] Cannot read properties of undefined (reading 'error'),2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/336,**问题描述 / Problem Description**,133
134,[BUG] 1 validation error for HuggingFaceEmbeddings model_kwargs extra fields not permitted.,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/337,模型加载到 100% 后出现问题：,134
135,上传知识库需要重启能不能修复一下,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/338,挺严重的这个问题,135
136,[BUG] 4块v100卡爆显存，在LLM会话模式也一样,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/339,**问题描述 / Problem Description**,136
137,针对上传的文件配置不同的TextSpliter,2023-05-12.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/341,1. 目前的ChineseTextSpliter切分对英文尤其是代码文件不友好，而且限制固定长度；导致对话结果不如人意,137
138,[FEATURE] 未来可增加Bloom系列模型吗？根据甲骨易的测试，这系列中文评测效果不错,2023-05-13.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/346,**功能描述 / Feature Description**,138
139,[BUG] v0.1.12打包镜像后启动webui.py失败 / Concise description of the issue,2023-05-13.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/347,**问题描述 / Problem Description**,139
140,切换MOSS模型时报错,2023-05-13.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/349,昨天问了下，说是transformers版本不对，需要4.30.0，发现没有这个版本，今天更新到4.29.1，依旧报错，错误如下,140
141,[BUG] pdf文档加载失败,2023-05-13.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/350,**问题描述 / Problem Description**,141
142,建议可以在后期增强一波注释，这样也有助于更多人跟进提PR,2023-05-13.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/351,知道作者和团队在疯狂更新审查代码，只是建议后续稳定后可以把核心代码进行一些注释的补充，从而能帮助更多人了解各个模块作者的思路从而提出更好的优化。,142
143,[FEATURE] MOSS 量化版支援,2023-05-13.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/353,**功能描述 / Feature Description**,143
144,[BUG] moss模型无法加载,2023-05-13.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/356,**问题描述 / Problem Description**,144
145,[BUG] load_doc_qa.py 中的 load_file 函数有bug,2023-05-14.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/358,原函数为：,145
146,[FEATURE] API模式，知识库加载优化,2023-05-14.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/359,如题，当前版本，每次调用本地知识库接口，都将加载一次知识库，是否有更好的方式？,146
147,运行Python api.py脚本后端部署后，怎么使用curl命令调用？,2023-05-15.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/361,也就是说，我现在想做个对话机器人，想和公司的前后端联调？怎么与前后端相互调用呢？可私信，有偿解答！！！,147
148,上传知识库需要重启能不能修复一下,2023-05-15.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/363,上传知识库需要重启能不能修复一下,148
149,[BUG] pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple,2023-05-15.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/364,我的python是3.8.5的,149
150,pip install gradio 报错,2023-05-15.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/367,大佬帮我一下,150
151,[BUG] pip install gradio 一直卡不动,2023-05-15.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/369,![aba82742dd9d4d242181662eb5027a7](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/84606552/cd9600d9-f6e7-46b7-b1be-30ed8b99f76b),151
152,[BUG] 简洁阐述问题 / Concise description of the issue,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/370,初次加载本地知识库成功，但提问后，就无法重写加载本地知识库,152
153,[FEATURE] 简洁阐述功能 / Concise description of the feature,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/371,**功能描述 / Feature Description**,153
154,在windows上，模型文件默认会安装到哪,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/372,-------------------------------------------------------------------------------,154
155,[FEATURE] 兼顾对话管理,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/374,如何在知识库检索的情况下，兼顾对话管理？,155
156,llm device: cpu embedding device: cpu,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/376,**问题描述 / Problem Description**,156
157,[FEATURE] 简洁阐述功能 /文本文件的知识点之间使用什么分隔符可以分割？,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/377,**功能描述 / Feature Description**,157
158,[BUG] 上传文件失败：PermissionError: [WinError 32] 另一个程序正在使用此文件，进程无法访问。,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/379,**问题描述 / Problem Description**,158
159,[BUG] 执行python api.py 报错,2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/383,错误信息,159
160,model_kwargs extra fields not permitted (type=value_error.extra),2023-05-16.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/384,"大家好，请问这个有遇到的么,？",160
161,[BUG] 简洁阐述问题 / Concise description of the issue,2023-05-17.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/385,执行的时候出现了ls1 = [ls[0]],161
162,[FEATURE] 性能优化,2023-05-17.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/388,**功能描述 / Feature Description**,162
163,"[BUG] Moss模型问答，RuntimeError: probability tensor contains either inf, nan or element < 0",2023-05-17.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/390,**问题描述 / Problem Description**,163
164,有没有人知道v100GPU的32G显存，会报错吗？支持V100GPU吗？,2023-05-17.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/392,**问题描述 / Problem Description**,164
165,针对于编码问题比如'gbk' codec can't encode character '\xab' in position 14: illegal multibyte sequence粗浅的解决方法,2023-05-17.03,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/397,**功能描述 / Feature Description**,165
166,Could not import sentence_transformers python package. Please install it with `pip install sentence_transformers`.,2023-05-18.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/400,**问题描述 / Problem Description**,166
167,支持模型问答与检索问答,2023-05-18.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/401,不同的query，根据意图不一致，回答也应该不一样。,167
168,文本分割的时候，能不能按照txt文件的每行进行分割，也就是按照换行符号\n进行分割？？？,2023-05-18.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/403,下面的代码应该怎么修改？,168
169,local_doc_qa/local_doc_chat 接口响应是串行,2023-05-18.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/405,**问题描述 / Problem Description**,169
170,"为什么找到出处了,但是还是无法回答该问题?",2023-05-18.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/406,![图片](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/3349611/1fc81d61-2409-4330-9065-fdda1a27c86a),170
171,"请问下:知识库测试中的:添加单条内容,如果换成文本导入是是怎样的格式?我发现添加单条内容测试效果很好.",2023-05-18.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/412,"我发现在知识库测试中`添加单条内容`,并且勾选`禁止内容分句入库`,即使 `不开启上下文关联`的测试效果都非常好.",171
172,[BUG] 无法配置知识库,2023-05-18.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/413,**问题描述 / Problem Description**,172
173,[BUG] 部署在阿里PAI平台的EAS上访问页面是白屏,2023-05-19.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/414,**问题描述 / Problem Description**,173
174,API部署后调用/local_doc_qa/local_doc_chat 返回Knowledge base samples not found,2023-05-19.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/416,入参,174
175,[FEATURE] 上传word另存为的txt文件报 'ascii' codec can't decode byte 0xb9 in position 6: ordinal not in range(128),2023-05-20.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/421,上传word另存为的txt文件报,175
176,创建保存的知识库刷新后没有出来，这个知识库是永久保存的吗？可以连外部的 向量知识库吗？,2023-05-21.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/422,创建保存的知识库刷新后没有出来，这个知识库是永久保存的吗？可以连外部的 向量知识库吗？,176
177,[BUG] 用colab运行，无法加载模型，报错：'NoneType' object has no attribute 'message_types_by_name',2023-05-21.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/423,**问题描述 / Problem Description**,177
178,请问是否需要用到向量数据库？以及什么时候需要用到向量数据库？,2023-05-21.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/424,目前用的是 text2vec ， 请问是否需要用到向量数据库？以及什么时候需要用到向量数据库？,178
179,huggingface模型引用问题,2023-05-22.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/427,它最近似乎变成了一个Error？,179
180,你好，加载本地txt文件出现这个killed错误，TXT文件有100M左右大小。原因是？谢谢。,2023-05-22.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/429,"<img width=""677"" alt=""929aca3b22b8cd74e997a87b61d241b"" src=""https://github.com/imClumsyPanda/langchain-ChatGLM/assets/109277248/24024522-c884-4170-b5cf-a498491bd8bc"">",180
181,想请问一下，关于对本地知识的管理是如何管理？例如：通过http API接口添加数据 或者 删除某条数据,2023-05-22.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/430,例如：通过http API接口添加、删除、修改 某条数据。,181
182,[FEATURE] 双栏pdf识别问题,2023-05-22.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/432,试了一下模型，感觉对单栏pdf识别的准确性较高，但是由于使用的基本是ocr的技术，对一些双栏pdf论文识别出来有很多问题，请问有什么办法改善吗？,182
183,部署启动小问题，小弟初学求大佬解答,2023-05-22.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/433,1.python loader/image_loader.py时，提示ModuleNotFoundError: No module named 'configs'，但是跑python webui.py还是还能跑,183
184,能否支持检测到目录下文档有增加而去增量加载文档，不影响前台对话，其实就是支持读写分离。如果能支持查询哪些文档向量化了，删除过时文档等就更好了，谢谢。,2023-05-22.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/434,**功能描述 / Feature Description**,184
185,[BUG] 简洁阐述问题 / windows 下cuda错误，请用https://github.com/Keith-Hon/bitsandbytes-windows.git,2023-05-22.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/435,pip install git+https://github.com/Keith-Hon/bitsandbytes-windows.git,185
186,"[BUG] from commit 33bbb47, Required library version not found: libbitsandbytes_cuda121_nocublaslt.so. Maybe you need to compile it from source?",2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/438,**问题描述 / Problem Description**,186
187,[BUG] 简洁阐述问题 / Concise description of the issue上传60m的txt文件报错，显示超时，请问这个能上传的文件大小有限制吗,2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/439,"ERROR 2023-05-23 11:13:09,627-1d: Timeout reached while detecting encoding for ./docs/GLM模型格式数据.txt",187
188,[BUG] TypeError: issubclass() arg 1 must be a class,2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/440,**问题描述**,188
189,"执行python3 webui.py后，一直提示”模型未成功加载，请到页面左上角""模型配置""选项卡中重新选择后点击""加载模型""按钮“",2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/441,**问题描述 / Problem Description**,189
190,是否能提供网页文档得导入支持,2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/444,现在很多都是在线文档作为协作得工具，所以通过URL导入在线文档需求更大,190
191,[BUG] history 索引问题,2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/445,在比较对话框的history和模型chat function 中的history时， 发现并不匹配，在传入 llm._call 时，history用的索引是不是有点问题，导致上一轮对话的内容并不输入给模型。,191
192,[BUG] moss_llm没有实现,2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/447,有些方法没支持，如history_len,192
193,请问langchain-ChatGLM如何删除一条本地知识库的数据？,2023-05-23.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/448,例如：用户刚刚提交了一条错误的数据到本地知识库中了，现在如何在本地知识库从找到，并且对此删除。,193
194,[BUG] 简洁阐述问题 / UnboundLocalError: local variable 'resp' referenced before assignment,2023-05-24.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/450,"在最新一版的代码中， 运行api.py 出现了以上错误（UnboundLocalError: local variable 'resp' referenced before assignment）， 通过debug的方式观察到local_doc_qa.llm.generatorAnswer(prompt=question, history=history,streaming=True)可能不返回任何值。",194
195,请问有没有 PROMPT_TEMPLATE 能让模型不回答敏感问题,2023-05-24.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/452,## PROMPT_TEMPLATE问题,195
196,[BUG] 测试环境 Python 版本有误,2023-05-24.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/456,**问题描述 / Problem Description**,196
197,[BUG] webui 部署后样式不正确,2023-05-24.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/458,**问题描述 / Problem Description**,197
198,配置默认LLM模型的问题,2023-05-24.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/459,**问题描述 / Problem Description**,198
199,[FEATURE]是时候更新一下autoDL的镜像了,2023-05-24.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/460,如题，跑了下autoDL的镜像，发现是4.27号的，git pull新版本的代码功能+老的依赖环境，各种奇奇怪怪的问题。,199
200,[BUG] tag:0.1.13 以cpu模式下，想使用本地模型无法跑起来，各种路径参数问题,2023-05-24.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/462,-------------------------------------------------------------------------------,200
201,[BUG] 有没有同学遇到过这个错！！！加载本地txt文件出现这个killed错误，TXT文件有100M左右大小。,2023-05-25.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/463,运行cli_demo.py。是本地的txt文件太大了吗？100M左右。,201
202,API版本能否提供WEBSOCKET的流式接口,2023-05-25.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/464,webui 版本中，采用了WS的流式输出，整体感知反应很快,202
203,[BUG] 安装bug记录,2023-05-25.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/465,按照[install文档](https://github.com/imClumsyPanda/langchain-ChatGLM/blob/master/docs/INSTALL.md)安装的，,203
204,VUE的pnmp i执行失败的修复-用npm i命令即可,2023-05-25.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/466,感谢作者！非常棒的应用，用的很开心。,204
205,请教个问题，有没有人知道cuda11.4是否支持？？？,2023-05-25.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/467,请教个问题，有没有人知道cuda11.4是否支持？？？,205
206,请问有实现多轮问答中基于问题的搜索上下文关联么,2023-05-25.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/468,在基于知识库的多轮问答中，第一个问题讲述了一个主题，后续的问题描述没有包含这个主题的关键词，但又存在上下文的关联。如果用后续问题去搜索知识库有可能会搜索出无关的信息，从而导致大模型无法正确回答问题。请问这个项目要考虑这种情况吗？,206
207,[BUG] 内存不足的问题,2023-05-26.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/470,我用了本地的chatglm-6b-int4模型，然后显示了内存不足（win10+32G内存+1080ti11G），一般需要多少内存才足够？这个bug应该如何解决？,207
208,[BUG] 纯内网环境安装pycocotools失败,2023-05-26.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/472,**问题描述 / Problem Description**,208
209,[BUG] webui.py 重新加载模型会导致 KeyError,2023-05-26.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/473,**问题描述 / Problem Description**,209
210,chatyuan无法使用,2023-05-26.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/475,**问题描述 / Problem Description**,210
211,[BUG] 文本分割模型AliTextSplitter存在bug，会把“.”作为分割符,2023-05-26.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/476,"阿里达摩院的语义分割模型存在bug，默认会把"".”作为分割符进行分割而不管上下文语义。是否还有其他分割符则未知。建议的修改方案：把“.”统一替换为其他字符，分割后再替换回来。或者添加其他分割模型。",211
212,[BUG] RuntimeError: Error in faiss::FileIOReader::FileIOReader(const char*) a,2023-05-27.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/479,**问题描述 / Problem Description**,212
213,[FEATURE] 安装，为什么conda create要额外指定路径 用-p ，而不是默认的/envs下面,2023-05-28.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/481,##**功能描述 / Feature Description**,213
214,[小白求助] 通过Anaconda执行webui.py后，无法打开web链接,2023-05-28.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/485,在执行webui.py命令后，http://0.0.0.0:7860复制到浏览器后无法打开，显示“无法访问此网站”。,214
215,[BUG] 使用 p-tuningv2后的模型，重新加载报错,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/486,把p-tunningv2训练完后的相关文件放到了p-tunningv2文件夹下，勾选使用p-tuningv2点重新加载模型，控制台输错错误信息：,215
216,[小白求助] 服务器上执行webui.py后，在本地无法打开web链接,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/487,此项目执行在xxx.xx.xxx.xxx服务器上，我在webui.py上的代码为 (demo,216
217,[FEATURE] 能不能支持VisualGLM-6B,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/488,**功能描述 / Feature Description**,217
218,你好，问一下各位，后端api部署的时候，支持多用户同时问答吗？？？,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/489,支持多用户的话，最多支持多少用户问答？根据硬件而定吧？,218
219,V100GPU显存占满，而利用率却为0，这是为什么？,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/491,"<img width=""731"" alt=""de45fe2b6cb76fa091b6e8f76a3de60"" src=""https://github.com/imClumsyPanda/langchain-ChatGLM/assets/109277248/c32efd52-7dbf-4e9b-bd4d-0944d73d0b8b"">",219
220,[求助] 如果在公司内部搭建产品知识库，使用INT-4模型，200人规模需要配置多少显存的服务器？,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/492,如题，计划给公司搭一个在线知识库。,220
221,你好，请教个问题，目前问答回复需要20秒左右，如何提高速度？V10032G服务器。,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/493,**问题描述 / Problem Description**,221
222,[FEATURE] 如何实现只匹配下文，而不要上文的结果,2023-05-29.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/494,在构建自己的知识库时，主要采用问答对的形式，那么也就是我需要的回答是在我的问题下面的内容，但是目前设置了chunk_size的值以后匹配的是上下文的内容，但我实际并不需要上文的。为了实现更完整的展示下面的答案，我只能调大chunk_size的值，但实际上上文的一半内容都是我不需要的。也就是扔了一半没用的东西给prompt，在faiss.py中我也没找到这块的一些描述，请问该如何进行修改呢？,222
223,你好，问一下，我调用api.py部署，为什么用ip加端口可以使用postman调用，而改为域名使用postman无法调用？,2023-05-30.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/497,![5ufBSWxLyF](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/109277248/70e2fbac-5699-48d0-b0d1-3dc84fd042c2),223
224,调用api.py中的stream_chat，返回source_documents中出现中文乱码。,2023-05-30.04,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/498,-------------------------------------------------------------------------------,224
225,[BUG] 捉个虫，api.py中的stream_chat解析json问题,2023-05-30.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/501,**问题描述 / Problem Description**,225
226,windows本地部署遇到了omp错误,2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/502,**问题描述 / Problem Description**,226
227,"[BUG] bug14 ,""POST /local_doc_qa/upload_file HTTP/1.1"" 422 Unprocessable Entity",2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/503,上传的文件报错，返回错误，api.py,227
228,你好，请教个问题，api.py部署的时候，如何改为多线程调用？谢谢,2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/505,目前的api.py脚本不支持多线程,228
229,你好，请教一下。api.py部署的时候，能不能提供给后端流失返回结果。,2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/507,curl -X 'POST' \,229
230,流式输出，流式接口，使用server-sent events技术。,2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/508,想这样一样，https://blog.csdn.net/weixin_43228814/article/details/130063010,230
231,计划增加流式输出功能吗？ChatGLM模型通过api方式调用响应时间慢怎么破，Fastapi流式接口来解惑，能快速提升响应速度,2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/509,**问题描述 / Problem Description**,231
232,[BUG] 知识库上传时发生ERROR (could not open xxx for reading: No such file or directory),2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/510,**问题描述 / Problem Description**,232
233,api.py脚本打算增加SSE流式输出吗？,2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/511,curl调用的时候可以检测第一个字，从而提升回复的体验,233
234,[BUG] 使用tornado实现webSocket，可以多个客户端同时连接，并且实现流式回复，但是多个客户端同时使用，答案就很乱，是模型不支持多线程吗,2023-05-31.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/512,import asyncio,234
235,支持 chinese_alpaca_plus_lora 吗 基于llama的,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/514,支持 chinese_alpaca_plus_lora 吗 基于llama的，https://github.com/ymcui/Chinese-LLaMA-Alpaca这个项目的,235
236,[BUG] 现在能读图片的pdf了，但是文字的pdf反而读不了了，什么情况？？？,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/515,**问题描述 / Problem Description**,236
237,在推理的过程中卡住不动，进程无法正常结束,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/516,**问题描述 / Problem Description**,237
238,curl调用的时候，从第二轮开始，curl如何传参可以实现多轮对话？,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/517,第一轮调用：,238
239,建议添加api.py部署后的日志管理功能？,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/518,-------------------------------------------------------------------------------,239
240,有大佬知道，怎么多线程部署api.py脚本吗？,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/519,api.py部署后，使用下面的请求，时间较慢，好像是单线程，如何改为多线程部署api.py：,240
241,[BUG] 上传文件到知识库 任何格式与内容都永远失败,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/520,上传知识库的时候，传txt无法解析，就算是穿content/sample里的样例txt也无法解析，上传md、pdf等都无法加载，会持续性等待，等到了超过30分钟也不行。,241
242,关于prompt_template的问题,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/521,请问这段prompt_template是什么意思，要怎么使用？可以给一个具体模板参考下吗？,242
243,[BUG] 简洁阐述问题 / Concise description of the issue,2023-06-01.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/522,**问题描述 / Problem Description**,243
244,"中文分词句号处理（关于表达金额之间的"".""）",2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/523,建议处理12.6亿元的这样的分词，最好别分成12 和6亿这样的，需要放到一起,244
245,ImportError: cannot import name 'inference' from 'paddle',2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/526,在网上找了一圈，有说升级paddle的，我做了还是没有用，有说安装paddlepaddle的，我找了豆瓣的镜像源，但安装报错cannot detect archive format,245
246,[BUG] webscoket 接口串行问题（/local_doc_qa/stream-chat/{knowledge_base_id}）,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/527,**问题描述 / Problem Description**,246
247,[FEATURE] 刷新页面更新知识库列表,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/528,**功能描述以及改进方案**,247
248,[BUG] 使用ptuning微调模型后，问答效果并不好,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/530,### 未调用ptuning,248
249,[BUG] 多轮对话效果不佳,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/532,在进行多轮对话的时候，无论设置的history_len是多少，效果都不好。事实上我将其设置成了最大值10，但在对话中，仍然无法实现多轮对话：,249
250,"RuntimeError: MPS backend out of memory (MPS allocated: 18.00 GB, other allocations: 4.87 MB, max allowed: 18.13 GB)",2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/533,**问题描述**,250
251, 请大家重视这个issue！真正使用肯定是多用户并发问答，希望增加此功能！！！,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/534,这得看你有多少显卡,251
252,在启动项目的时候如何使用到多张gpu啊？,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/535,**在启动项目的时候如何使用到多张gpu啊？**,252
253, 使用流式输出的时候，curl调用的格式是什么？,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/536,"app.websocket(""/local_doc_qa/stream-chat/{knowledge_base_id}"")(stream_chat)中的knowledge_base_id应该填什么？？？",253
254,使用本地 vicuna-7b模型启动错误,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/538,环境: ubuntu 22.04 cuda 12.1 没有安装nccl，使用rtx2080与m60显卡并行计算,254
255,为什么会不调用GPU直接调用CPU呢,2023-06-02.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/539,我的阿里云配置是16G显存，用默认代码跑webui.py时提示,255
256,上传多个文件时会互相覆盖,2023-06-03.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/541,1、在同一个知识库中上传多个文件时会互相覆盖，无法结合多个文档的知识，有大佬知道怎么解决吗？,256
257,[BUG] ‘gcc’不是内部或外部命令/LLM对话只能持续一轮,2023-06-03.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/542,No compiled kernel found.,257
258,以API模式启动项目却没有知识库的接口列表？,2023-06-04.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/544,请问如何获取知识库的接口列表？如果没有需要自行编写的话，可不可以提供相关的获取方式，感谢,258
259,程序以API模式启动的时候，如何才能让接口以stream模式被调用呢？,2023-06-05.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/546,作者您好，我在以API模式进行程序启动后，我发现接口响应时间很长，怎么样才能让接口以stream模式被调用呢？我想实现像webui模式的回答那样,259
260,关于原文中表格转为文本后数据相关度问题。,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/547,原文中表格数据转换为文本，以 （X-Y：值；...） 的格式每一行组织成一句话，但这样做后发现相关度较低，效果很差，有何好的方案吗？,260
261,启动后LLM和知识库问答模式均只有最后一轮记录,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/548,拉取最新代码，问答时，每次页面只显示最后一次问答记录，需要修改什么参数才可以保留历史记录？,261
262,提供system message配置，以便于让回答不要超出知识库范围,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/549,**功能描述 / Feature Description**,262
263,[BUG] 使用p-tunningv2报错,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/551,按照readme的指示把p-tunningv2训练完后的文件放到了p-tunningv2文件夹下，勾选使用p-tuningv2点重新加载模型，控制台提示错误信息：,263
264,[BUG] 智障，这么多问题，也好意思放出来，浪费时间,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/553,。。。,264
265,[FEATURE] 我看代码文件中有一个ali_text_splitter.py，为什么不用他这个文本分割器了？,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/554,我看代码文件中有一个ali_text_splitter.py，为什么不用他这个文本分割器了？,265
266,加载文档函数报错,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/557,"def load_file(filepath, sentence_size=SENTENCE_SIZE):",266
267,参考指引安装docker后，运行cli_demo.py，提示killed,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/558,root@b3d1bd08095c:/chatGLM# python3 cli_demo.py,267
268,注意：如果安装错误，注意这两个包的版本 wandb==0.11.0 protobuf==3.18.3,2023-06-06.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/559,Error1: 如果启动异常报错 `protobuf` 需要更新到 `protobuf==3.18.3 `,268
269,知识库对长文的知识相关度匹配不太理想有何优化方向,2023-06-07.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/563,我们可能录入一个文章有 1W 字，里面涉及这个文章主题的很多角度问题，我们针对他提问，他相关度匹配的内容和实际我们需要的答案相差很大怎么办。,269
270,使用stream-chat函数进行流式输出的时候，能使用curl调用吗？,2023-06-07.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/565,为什么下面这样调用会报错？？？,270
271,有大佬实践过 并行 或者 多线程 的部署方案吗？,2023-06-07.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/566,+1,271
272,多线程部署遇到问题？,2023-06-07.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/567,"<img width=""615"" alt=""3d87bf74f0cf1a4820cc9e46b245859"" src=""https://github.com/imClumsyPanda/langchain-ChatGLM/assets/109277248/8787570d-88bd-434e-aaa4-cb9276d1aa50"">",272
273,[BUG] 用fastchat加载vicuna-13b模型进行知识库的问答有token的限制错误,2023-06-07.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/569,当我开启fastchat的vicuna-13b的api服务，然后config那里配置好(api本地测试过可以返回结果)，然后知识库加载好之后(知识库大概有1000多个文档，用chatGLM可以正常推理)，进行问答时出现token超过限制，就问了一句hello；,273
274,现在的添加知识库，文件多了总是报错，也不知道自己加载了哪些文件，报错后也不知道是全部失败还是一部分成功；希望能有个加载指定文件夹作为知识库的功能,2023-06-07.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/574,**功能描述 / Feature Description**,274
275,[BUG] moss模型本地加载报错,2023-06-08.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/577,moss模型本地加载报错：,275
276,加载本地moss模型报错Can't instantiate abstract class MOSSLLM with abstract methods _history_len,2023-06-08.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/578,(vicuna) ps@ps[13:56:20]:/data/chat/langchain-ChatGLM2/langchain-ChatGLM-0.1.13$ python webui.py --model-dir local_models --model moss --no-remote-model,276
277,[FEATURE] 能增加在前端页面控制prompt_template吗？或是能支持前端页面选择使用哪个prompt？,2023-06-08.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/579,目前只能在config里修改一个prompt，想在多个不同场景切换比较麻烦,277
278,[BUG] streamlit ui的bug，在增加知识库时会报错,2023-06-08.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/580,**问题描述 / Problem Description**,278
279,[FEATURE] webui/webui_st可以支持history吗？目前仅能一次对话,2023-06-08.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/581,试了下webui和webui_st都不支持历史对话啊，只能对话一次，不能默认开启所有history吗？,279
280,启动python cli_demo.py --model chatglm-6b-int4-qe报错,2023-06-09.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/585,下载好模型，和相关依赖环境，之间运行`python cli_demo.py --model chatglm-6b-int4-qe`报错了：,280
281,重新构建知识库报错,2023-06-09.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/586,**问题描述 / Problem Description**,281
282,[FEATURE] 能否屏蔽paddle，我不需要OCR，效果差依赖环境还很复杂,2023-06-09.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/587,希望能不依赖paddle,282
283,question ：文档向量化这个可以自己手动实现么？,2023-06-09.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/589,现有公司级数据500G+，需要使用这个功能，请问如何手动实现这个向量化，然后并加载,283
284,view前端能进行流式的返回吗？？,2023-06-09.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/590,view前端能进行流式的返回吗？？,284
285,"[BUG] Load parallel cpu kernel failed, using default cpu kernel code",2023-06-11.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/594,**问题描述 / Problem Description**,285
286,[BUG] 简洁阐述问题 / Concise description of the issue,2023-06-11.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/595,**问题描述 / Problem Description**,286
287,我在上传本地知识库时提示KeyError: 'name'错误，本地知识库都是.txt文件，文件数量大约是2000+。,2023-06-12.05,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/597,"<img width=""649"" alt=""KError"" src=""https://github.com/imClumsyPanda/langchain-ChatGLM/assets/59411575/1ecc8182-aeee-4a0a-bbc3-74c2f1373f2d"">",287
288,model_config.py中有vicuna-13b-hf模型的配置信息，但是好像还是不可用？,2023-06-12.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/600,@dongyihua543,288
289,"ImportError: Using SOCKS proxy, but the 'socksio' package is not installed. Make sure to install httpx using `pip install httpx[socks]`.",2023-06-12.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/605,应该代理问题，但是尝试了好多方法都解决不了，,289
290,[BUG] similarity_search_with_score_by_vector在找不到匹配的情况下出错,2023-06-12.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/607,在设置匹配阈值 VECTOR_SEARCH_SCORE_THRESHOLD 的情况下，vectorstore会返回空，此时上述处理函数会出错,290
291,[FEATURE] 请问如何搭建英文知识库呢,2023-06-12.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/609,**功能描述 / Feature Description**,291
292,谁有vicuna权重？llama转换之后的,2023-06-13.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/611,**问题描述 / Problem Description**,292
293,[FEATURE] API能实现上传文件夹的功能么？,2023-06-13.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/612,用户懒得全选所有的文件，就想上传个文件夹，请问下API能实现这个功能么？,293
294,请问在多卡部署后，上传单个文件作为知识库，用的是单卡在生成向量还是多卡？,2023-06-13.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/614,目前我检测我本地多卡部署的，好像生成知识库向量的时候用的还是单卡,294
295,[BUG] python webui.py提示非法指令,2023-06-13.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/615,(/data/conda-langchain [root@chatglm langchain-ChatGLM]# python webui.py,295
296,知识库文件跨行切分问题,2023-06-13.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/616,我的知识库文件txt文件，是一行一条知识，用\n分行。,296
297,[FEATURE] bing搜索问答有流式的API么？,2023-06-13.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/617,web端是有这个bing搜索回答，但api接口没有发现，大佬能给个提示么？,297
298,希望出一个macos m2的安装教程,2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/620,mac m2安装，模型加载成功了，知识库文件也上传成功了，但是一问答就会报错，报错内容如下,298
299,为【出处】提供高亮显示,2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/621,具体出处里面，对相关的内容高亮显示，不包含前后文。,299
300,[BUG] CPU运行cli_demo.py，不回答，hang住,2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/622,没有GPU；32G内存的ubuntu机器。,300
301,关于删除知识库里面的文档后，LLM知识库对话的时候还是会返回该被删除文档的内容,2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/623,如题，在vue前端成功执行删除知识库里面文档A.txt后，未能也在faiss索引中也删除该文档，LLM还是会返回这个A.txt的内容，并且以A.txt为出处，未能达到删除的效果,301
302,"[BUG] 调用知识库进行问答,显存会一直叠加",2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/625,"14G的显存,调用的chatglm-6b-int8模型,进行知识库问答时,最多问答四次就会爆显存了,观察了一下显存使用情况,每一次使用就会增加一次显存,请问这样是正常的吗?是否有什么配置需要开启可以解决这个问题?例如进行一次知识库问答清空上次问题的显存?",302
303,[BUG] web页面 重新构建数据库 失败，导致 原来的上传的数据库都没了,2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/626,web页面 重新构建数据库 失败，导致 原来的上传的数据库都没了,303
304,在CPU上运行webui.py报错Tensor on device cpu is not on the expected device meta!,2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/627,在CPU上运行python webui.py能启动，但最后有：RuntimeError: Tensor on device cpu is not on the expected device meta!,304
305,"OSError: [WinError 1114] 动态链接库(DLL)初始化例程失败。 Error loading ""E:\xxx\envs\langchain\lib\site-packages\torch\lib\caffe2_nvrtc.dll"" or one of its dependencies.哪位大佬知道如何解决吗？",2023-06-14.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/629,**问题描述 / Problem Description**,305
306,[BUG] WEBUI删除知识库文档，会导致知识库问答失败,2023-06-15.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/632,如题，从知识库已有文件中选择要删除的文件，点击删除后，在问答框输入内容回车报错,306
307,更新后的版本中，删除知识库中的文件，再提问出现error错误,2023-06-15.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/634,针对更新版本，识别到一个问题，过程如下：,307
308,我配置好了环境，想要实现本地知识库的问答？可是它返回给我的,2023-06-15.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/637,没有总结，只有相关度的回复，但是我看演示里面表现的，回复是可以实现总结的，我去查询代码,308
309,[BUG] NPM run dev can not successfully start the VUE frontend,2023-06-15.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/638,**问题描述 / Problem Description**,309
310,[BUG] 简洁阐述问题 / Concise description of the issue,2023-06-15.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/639,**问题描述 / Problem Description**,310
311,提一个模型加载的bug，我在截图中修复了，你们有空可以看一下。,2023-06-15.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/642,![model_load_bug](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/59411575/4432adc4-ccdd-45d9-aafc-5f2d1963403b),311
312,[求助]关于设置embedding model路径的问题,2023-06-16.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/643,如题，我之前成功跑起来过一次，但因环境丢失重新配置 再运行webui就总是报错,312
313,Lora微调后的模型可以直接使用吗,2023-06-16.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/646,看model_config.py里是有USE_LORA这个参数的，但是在cli_demo.py和webui.py这两个里面都没有用到，实际测试下来模型没有微调的效果，想问问现在这个功能实现了吗,313
314,write_check_file在tmp_files目录下生成的load_file.txt是否需要一直保留，占用空间很大，在建完索引后能否删除,2023-06-16.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/647,**功能描述 / Feature Description**,314
315,[BUG] /local_doc_qa/list_files?knowledge_base_id=test删除知识库bug,2023-06-16.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/649,1.新建test知识库并上传文件（在vue前端完成并检查后端发现确实生成了test文件夹以及下面的content和vec_store,315
316,[BUG] vue webui无法加载知识库,2023-06-16.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/650,拉取了最新的代码，分别运行了后端api和前端web，点击知识库，始终只能显示simple，无法加载知识库,316
317,不能本地加载moss模型吗？,2023-06-16.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/652,手动下载模型设置local_model_path路径依旧提示缺少文件，该如何正确配置？,317
318,macos m2 pro docker 安装失败,2023-06-17.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/654,macos m2 pro docker 安装失败,318
319, [BUG] mac m1 pro 运行提示 zsh: segmentation fault,2023-06-17.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/655,运行： python webui.py,319
320,安装 requirements 报错,2023-06-17.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/656,(langchainchatglm) D:\github\langchain-ChatGLM>pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple/,320
321,[BUG] AssertionError,2023-06-17.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/658,**问题描述 / Problem Description**,321
322,[FEATURE] 支持AMD win10 本地部署吗？,2023-06-18.06,https://github.com/imClumsyPanda/langchain-ChatGLM/issues/660,**功能描述 / Feature Description**,322
knowledge_base/samples/isssues_merge/langchain-ChatGLM_open.jsonl 0 → 100644
{"title": "效果如何优化", "file": "2023-04-04.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/14", "detail": "如图所示，将该项目的README.md和该项目结合后，回答效果并不理想，请问可以从哪些方面进行优化", "id": 0}
{"title": "怎么让模型严格根据检索的数据进行回答，减少胡说八道的回答呢", "file": "2023-04-04.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/15", "detail": "举个例子：", "id": 1}
{"title": "When I try to run the `python knowledge_based_chatglm.py`, I got this error in macOS(M1 Max, OS 13.2)", "file": "2023-04-07.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/32", "detail": "```python", "id": 2}
{"title": "萌新求教大佬怎么改成AMD显卡或者CPU？", "file": "2023-04-10.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/48", "detail": "把.cuda()去掉就行", "id": 3}
{"title": "输出answer的时间很长，是否可以把文本向量化的部分提前做好存储起来？", "file": "2023-04-10.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/50", "detail": "GPU：4090 24G显存", "id": 4}
{"title": "报错Use `repo_type` argument if needed.", "file": "2023-04-11.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/57", "detail": "Traceback (most recent call last):", "id": 5}
{"title": "无法打开gradio的页面", "file": "2023-04-11.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/58", "detail": "$ python webui.py", "id": 6}
{"title": "支持word，那word里面的图片正常显示吗？", "file": "2023-04-12.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/60", "detail": "如题，刚刚从隔壁转过来的，想先了解下", "id": 7}
{"title": "detectron2 is not installed. Cannot use the hi_res partitioning strategy. Falling back to partitioning with the fast strategy.", "file": "2023-04-12.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/63", "detail": "能够正常的跑起来，在加载content文件夹中的文件时，每加载一个文件都会提示：", "id": 8}
{"title": "cpu上运行webui，step3 asking时报错", "file": "2023-04-12.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/66", "detail": "web运行，文件加载都正常，asking时报错", "id": 9}
{"title": "建议弄一个插件系统", "file": "2023-04-13.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/67", "detail": "如题弄成stable-diffusion-webui那种能装插件，再开一个存储库给使用者或插件开发，存储或下载插件。", "id": 10}
{"title": "请教加载模型出错！？", "file": "2023-04-13.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/75", "detail": "AttributeError: module 'transformers_modules.chatglm-6b.configuration_chatglm' has no attribute 'ChatGLMConfig 怎么解决呀", "id": 11}
{"title": "从本地知识检索内容的时候，是否可以设置相似度阈值，小于这个阈值的内容不返回，即使会小于设置的VECTOR_SEARCH_TOP_K参数呢？谢谢大佬", "file": "2023-04-13.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/76", "detail": "比如 问一些 你好/你是谁 等一些跟本地知识库无关的问题", "id": 12}
{"title": "如何改成多卡推理？", "file": "2023-04-13.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/77", "detail": "+1", "id": 13}
{"title": "能否弄个懒人包，可以一键体验？", "file": "2023-04-13.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/78", "detail": "能否弄个懒人包，可以一键体验？", "id": 14}
{"title": "连续问问题会导致崩溃", "file": "2023-04-13.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/79", "detail": "看上去不是爆内存的问题，连续问问题后，会出现如下报错", "id": 15}
{"title": "AttributeError: 'NoneType' object has no attribute 'as_retriever'", "file": "2023-04-14.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/86", "detail": "环境：windows 11, anaconda/python 3.8", "id": 16}
{"title": "FileNotFoundError: Could not find module 'nvcuda.dll' (or one of its dependencies). Try using the full path with constructor syntax.", "file": "2023-04-14.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/87", "detail": "请检查一下cuda或cudnn是否存在安装问题", "id": 17}
{"title": "加载txt文件失败？", "file": "2023-04-14.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/89", "detail": "![JppHrGOWFa](https://user-images.githubusercontent.com/109277248/232009383-bf7c46d1-a01e-4e0a-9de6-5b5ed3e36158.jpg)", "id": 18}
{"title": "NameError: name 'chatglm' is not defined", "file": "2023-04-14.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/90", "detail": "This share link expires in 72 hours. For free permanent hosting and GPU upgrades (NEW!), check out Spaces: https://huggingface.co/spaces", "id": 19}
{"title": "打不开地址？", "file": "2023-04-14.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/91", "detail": "报错数据如下：", "id": 20}
{"title": "加载md文件出错", "file": "2023-04-14.00", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/98", "detail": "运行 webui.py后能访问页面，上传一个md文件后，日志中有错误。等待后能加载完成，提示可以提问了，但提问没反应，日志中有错误。 具体日志如下。", "id": 21}
{"title": "建议增加获取在线知识的能力", "file": "2023-04-15.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/101", "detail": "建议增加获取在线知识的能力", "id": 22}
{"title": "txt 未能成功加载", "file": "2023-04-15.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/103", "detail": "hinese. Creating a new one with MEAN pooling.", "id": 23}
{"title": "pdf加载失败", "file": "2023-04-15.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/105", "detail": "e:\\a.txt加载成功了，e:\\a.pdf加载就失败，pdf文件里面前面几页是图片，后面都是文字，加载失败没有报更多错误，请问该怎么排查？", "id": 24}
{"title": "一直停在文本加载处", "file": "2023-04-15.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/108", "detail": "一直停在文本加载处", "id": 25}
{"title": " File \"/root/.cache/huggingface/modules/transformers_modules/chatglm-6b/modeling_chatglm.py\", line 440, in forward new_tensor_shape = mixed_raw_layer.size()[:-1] + ( TypeError: torch.Size() takes an iterable of 'int' (item 2 is 'float')", "file": "2023-04-17.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/113", "detail": "按照最新的代码，发现", "id": 26}
{"title": "后续会提供前后端分离的功能吗？", "file": "2023-04-17.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/114", "detail": "类似这种https://github.com/lm-sys/FastChat/tree/main/fastchat/serve", "id": 27}
{"title": "安装依赖报错", "file": "2023-04-17.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/115", "detail": "(test) C:\\Users\\linh\\Desktop\\langchain-ChatGLM-master>pip install -r requirements.txt", "id": 28}
{"title": "问特定问题会出现爆显存", "file": "2023-04-17.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/117", "detail": "正常提问没问题。", "id": 29}
{"title": "Expecting value: line 1 column 1 (char 0)", "file": "2023-04-17.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/118", "detail": "运行后 第一步加载配置一直报错：", "id": 30}
{"title": "embedding https://huggingface.co/GanymedeNil/text2vec-large-chinese/tree/main是免费的，效果比对openai的如何？", "file": "2023-04-17.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/119", "detail": "-------------------------------------------------------------------------------", "id": 31}
{"title": "这是什么错误，在Colab上运行的。", "file": "2023-04-17.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/120", "detail": "libcuda.so.1: cannot open shared object file: No such file or directory", "id": 32}
{"title": "只想用自己的lora微调后的模型进行对话，不想加载任何本地文档，该如何调整？", "file": "2023-04-18.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/121", "detail": "能出一个单独的教程吗", "id": 33}
{"title": "租的gpu,Running on local URL: http://0.0.0.0:7860 To create a public link, set `share=True` in `launch()`. 浏览器上访问不了？？？", "file": "2023-04-18.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/122", "detail": "(chatglm20230401) root@autodl-container-e82d11963c-10ece0d7:~/autodl-tmp/chatglm/langchain-ChatGLM-20230418# python3.9 webui.py", "id": 34}
{"title": "本地部署中的报错请教", "file": "2023-04-18.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/124", "detail": "您好，在本地运行langchain-ChatGLM过程中，环境及依赖的包都已经满足条件，但是运行webui.py,报错如下（运行cli_demo.py报错类似），请问是哪里出了错呢？盼望您的回复，谢谢！", "id": 35}
{"title": "报错。The dtype of attention mask (torch.int64) is not bool", "file": "2023-04-18.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/131", "detail": "The dtype of attention mask (torch.int64) is not bool", "id": 36}
{"title": "[求助] pip install -r requirements.txt 的时候出现以下报错。。。有大佬帮忙看看怎么搞么，下的release里面的包", "file": "2023-04-18.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/134", "detail": "$ pip install -r requirements.txt", "id": 37}
{"title": "如何提升根据问题搜索到对应知识的准确率", "file": "2023-04-19.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/136", "detail": "外链知识库最大的问题在于问题是短文本，知识是中长文本。如何根据问题精准的搜索到对应的知识是个最大的问题。这类本地化项目不像百度，由无数的网页，基本上每个问题都可以找到对应的页面。", "id": 38}
{"title": "是否可以增加向量召回的阈值设定，有些召回内容相关性太低，导致模型胡言乱语", "file": "2023-04-20.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/140", "detail": "如题", "id": 39}
{"title": "输入长度问题", "file": "2023-04-20.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/141", "detail": "感谢作者支持ptuning微调模型。", "id": 40}
{"title": "已有部署好的chatGLM-6b，如何通过接口接入？", "file": "2023-04-20.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/144", "detail": "已有部署好的chatGLM-6b，如何通过接口接入，而不是重新加载一个模型；", "id": 41}
{"title": "执行web_demo.py后，显示Killed，就退出了，是不是配置不足呢？", "file": "2023-04-20.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/146", "detail": "![图片](https://user-images.githubusercontent.com/26102866/233256425-c7aab999-11d7-4de9-867b-23ef18d519e4.png)", "id": 42}
{"title": "执行python cli_demo1.py", "file": "2023-04-20.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/147", "detail": "Traceback (most recent call last):", "id": 43}
{"title": "报错：ImportError: cannot import name 'GENERATION_CONFIG_NAME' from 'transformers.utils'", "file": "2023-04-20.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/149", "detail": "(mychatGLM) PS D:\\Users\\admin3\\zrh\\langchain-ChatGLM> python cli_demo.py", "id": 44}
{"title": "上传文件并加载知识库时，会不停地出现临时文件", "file": "2023-04-21.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/153", "detail": "环境：ubuntu 18.04", "id": 45}
{"title": "向知识库中添加文件后点击”上传文件并加载知识库“后Segmentation fault报错。", "file": "2023-04-23.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/161", "detail": "运行服务后的提示如下：", "id": 46}
{"title": "langchain-serve 集成", "file": "2023-04-24.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/162", "detail": "Hey 我是来自 [langchain-serve](https://github.com/jina-ai/langchain-serve) 的dev！", "id": 47}
{"title": "大佬们，wsl的ubuntu怎么配置用cuda加速，装了运行后发现是cpu在跑", "file": "2023-04-24.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/164", "detail": "大佬们，wsl的ubuntu怎么配置用cuda加速，装了运行后发现是cpu在跑", "id": 48}
{"title": "在github codespaces docker运行出错", "file": "2023-04-24.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/165", "detail": "docker run -d --restart=always --name chatglm -p 7860:7860 -v /www/wwwroot/code/langchain-ChatGLM:/chatGLM chatglm", "id": 49}
{"title": "有计划接入Moss模型嘛", "file": "2023-04-24.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/166", "detail": "后续会开展测试，目前主要在优化langchain部分效果，如果有兴趣也欢迎提PR", "id": 50}
{"title": "怎么实现 API 部署？", "file": "2023-04-24.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/168", "detail": "利用 fastapi 实现 API 部署方式，具体怎么实现，有方法说明吗？", "id": 51}
{"title": " 'NoneType' object has no attribute 'message_types_by_name'报错", "file": "2023-04-24.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/169", "detail": "_HISTOGRAMPROTO = DESCRIPTOR.message_types_by_name['HistogramProto']", "id": 52}
{"title": "能否指定自己训练的text2vector模型？", "file": "2023-04-25.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/172", "detail": "请问大佬：", "id": 53}
{"title": "关于项目支持的模型以及quantization_bit潜在的影响的问题", "file": "2023-04-26.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/176", "detail": "作者您好～", "id": 54}
{"title": "运行python3.9 api.py WARNING: You must pass the application as an import string to enable 'reload' or 'workers'.", "file": "2023-04-26.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/179", "detail": "api.py文件最下面改成这样试试：", "id": 55}
{"title": "ValidationError: 1 validation error for HuggingFaceEmbeddings model_kwargs extra fields not permitted (type=value_error.extra)", "file": "2023-04-26.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/180", "detail": "ValidationError: 1 validation error for HuggingFaceEmbeddings", "id": 56}
{"title": "如果没有检索到相关性比较高的，回答“我不知道”", "file": "2023-04-26.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/181", "detail": "如果通过设计system_template，让模型在搜索到的文档都不太相关的情况下回答“我不知道”", "id": 57}
{"title": "请问如果不能联网，6B之类的文件从本地上传需要放到哪里", "file": "2023-04-26.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/182", "detail": "感谢大佬的项目，很有启发~", "id": 58}
{"title": "知识库问答--输入新的知识库名称是中文的话，会报error", "file": "2023-04-27.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/184", "detail": "知识库问答--输入新的知识库名称是中文的话，会报error，选择要加载的知识库那里也不显示之前添加的知识库", "id": 59}
{"title": "现在能通过问题匹配的相似度值，来直接返回文档中的文段，而不经过模型吗？因为有些答案在文档中，模型自己回答，不能回答文档中的答案", "file": "2023-04-27.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/186", "detail": "现在能通过问题匹配的相似度值，来直接返回文档中的文段，而不经过模型吗？因为有些答案在文档中，模型自己回答，不能回答文档中的答案。也就是说，提供向量检索回答+模型回答相结合的策略。如果相似度值高于一定数值，直接返回文档中的文本，没有高于就返回模型的回答或者不知道", "id": 60}
{"title": "TypeError: The type of ChatGLM.callback_manager differs from the new default value; if you wish to change the type of this field, please use a type annotation", "file": "2023-04-27.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/188", "detail": "Mac 运行 python3 ./webui.py 报 TypeError: The type of ChatGLM.callback_manager differs from the new default value; if you wish to change the type of this field, please use a type annotation", "id": 61}
{"title": "Not Enough Memory", "file": "2023-04-27.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/190", "detail": "运行命令行程序python cli_demo.py， 已经成功加载pdf文件, 报“DefaultCPUAllocator: not enough memory: you tried to allocate 458288380900 bytes”错误，请问哪里可以配置default memory", "id": 62}
{"title": "参与开发问题", "file": "2023-04-27.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/191", "detail": "1.是否需要进专门的开发群", "id": 63}
{"title": "对话框中代码片段格式需改进", "file": "2023-04-27.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/192", "detail": "最好能改进下输出代码片段的格式，目前输出的格式还不友好。", "id": 64}
{"title": "请问未来有可能支持belle吗", "file": "2023-04-28.01", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/195", "detail": "如题，谢谢大佬", "id": 65}
{"title": "TypeError: cannot unpack non-iterable NoneType object", "file": "2023-04-28.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/200", "detail": "When i tried to change the knowledge vector store through `init_knowledge_vector_store`, the error `TypeError: cannot unpack non-iterable NoneType object` came out.", "id": 66}
{"title": "生成结果", "file": "2023-04-28.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/202", "detail": "你好，想问一下langchain+chatglm-6B，找到相似匹配的prompt，是直接返回prompt对应的答案信息，还是chatglm-6B在此基础上自己优化答案？", "id": 67}
{"title": "在win、ubuntu下都出现这个错误：attributeerror: 't5forconditionalgeneration' object has no attribute 'stream_chat'", "file": "2023-04-29.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/207", "detail": "在win、ubuntu。下载完模型后，没办法修改代码以执行本地模型，每次都要重新输入路径； LLM 模型、Embedding 模型支持也都在官网下的，在其他项目（wenda）下可以使用", "id": 68}
{"title": "[FEATURE] knowledge_based_chatglm.py: renamed or missing?", "file": "2023-04-30.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/210", "detail": "Not found. Was it renamed? Or, is it missing? How can I get it?", "id": 69}
{"title": "sudo apt-get install -y nvidia-container-toolkit-base执行报错", "file": "2023-05-01.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/211", "detail": "**问题描述 / Problem Description**", "id": 70}
{"title": "效果不佳几乎答不上来", "file": "2023-05-01.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/212", "detail": "提供了50条问答的docx文件", "id": 71}
{"title": "有没有可能新增一个基于chatglm api调用的方式构建langchain", "file": "2023-05-02.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/218", "detail": "我有两台8G GPU/40G内存的服务器，一个台做成了chatglm的api ；想基于另外一台服务器部署langchain；网上好像没有类似的代码。", "id": 72}
{"title": "电脑是intel的集成显卡； 运行时告知我找不到nvcuda.dll，模型无法运行", "file": "2023-05-02.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/219", "detail": "您好，我的电脑是intel的集成显卡，不过CPU是i5-11400 @ 2.60GHz ，内存64G；", "id": 73}
{"title": "根据langchain官方的文档和使用模式，是否可以改Faiss为Elasticsearch？会需要做哪些额外调整？求解", "file": "2023-05-03.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/221", "detail": "本人新手小白，由于业务模式的原因（有一些自己的场景和优化），希望利用Elasticsearch做这个体系内部的检索机制，不知道是否可以替换，同时，还会涉及到哪些地方的改动？或者说可能会有哪些其他影响，希望作者和大佬们不吝赐教！", "id": 74}
{"title": "请问未来有可能支持t5吗", "file": "2023-05-04.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/224", "detail": "请问可能支持基於t5的模型吗?", "id": 75}
{"title": "[BUG] 内存溢出 / torch.cuda.OutOfMemoryError:", "file": "2023-05-04.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/229", "detail": "**问题描述 / Problem Description**", "id": 76}
{"title": "报错 No module named 'chatglm_llm'", "file": "2023-05-04.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/230", "detail": "明明已经安装了包，却在python里吊不出来", "id": 77}
{"title": "能出一个api部署的描述文档吗", "file": "2023-05-04.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/233", "detail": "**功能描述 / Feature Description**", "id": 78}
{"title": "使用docs/API.md 出错", "file": "2023-05-04.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/234", "detail": "使用API.md文档2种方法，出错", "id": 79}
{"title": "加载pdf文档报错？", "file": "2023-05-05.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/238", "detail": "ew one with MEAN pooling.", "id": 80}
{"title": "上传的本地知识文件后再次上传不能显示，只显示成功了一个，别的上传成功后再次刷新就没了", "file": "2023-05-05.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/239", "detail": "您好，项目有很大启发，感谢~", "id": 81}
{"title": "创建了新的虚拟环境，安装了相关包，并且自动下载了相关的模型，但是仍旧出现：OSError: Unable to load weights from pytorch checkpoint file for", "file": "2023-05-05.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/240", "detail": "![78ac8e663fdc312d0e9d78da95925c4](https://user-images.githubusercontent.com/34124260/236378728-9ea4424f-0f7f-4013-9d33-820b723de321.png)", "id": 82}
{"title": "[BUG] 数据加载不进来", "file": "2023-05-05.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/243", "detail": "使用的.txt格式，utf-8编码，报以下错误", "id": 83}
{"title": "不能读取pdf", "file": "2023-05-05.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/244", "detail": "请问是webui还是cli_demo", "id": 84}
{"title": "本地txt文件有500M，加载的时候很慢，如何提高速度？", "file": "2023-05-06.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/251", "detail": "![yayRzxSYHP](https://user-images.githubusercontent.com/109277248/236592902-f5ab338d-c1e9-43dc-ae16-9df2cd3c1378.jpg)", "id": 85}
{"title": "[BUG] gradio上传知识库后刷新之后 知识库就不见了 只有重启才能看到之前的上传的知识库", "file": "2023-05-06.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/253", "detail": "gradio上传知识库后刷新之后 知识库就不见了 只有重启才能看到之前的上传的知识库", "id": 86}
{"title": "[FEATURE] 可以支持 OpenAI 的模型嘛？比如 GPT-3、GPT-3.5、GPT-4；embedding 增加 text-embedding-ada-002", "file": "2023-05-06.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/254", "detail": "**功能描述 / Feature Description**", "id": 87}
{"title": "[FEATURE] 能否增加对于milvus向量数据库的支持 / Concise description of the feature", "file": "2023-05-06.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/256", "detail": "**功能描述 / Feature Description**", "id": 88}
{"title": "CPU和GPU上跑，除了速度有区别，准确率效果回答上有区别吗？", "file": "2023-05-06.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/259", "detail": "理论上没有区别", "id": 89}
{"title": "m1，请问在生成回答时怎么看是否使用了mps or cpu？", "file": "2023-05-06.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/260", "detail": "m1，请问在生成回答时怎么看是否使用了mps or cpu？", "id": 90}
{"title": "知识库一刷新就没了", "file": "2023-05-07.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/263", "detail": "知识库上传后刷新就没了", "id": 91}
{"title": "本地部署报没有模型", "file": "2023-05-07.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/267", "detail": "建议在下载llm和embedding模型至本地后在configs/model_config中写入模型本地存储路径后再运行", "id": 92}
{"title": "[BUG] python3: can't open file 'webui.py': [Errno 2] No such file or directory", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/269", "detail": "**问题描述 / Problem Description**", "id": 93}
{"title": "模块缺失提示", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/271", "detail": "因为已有自己使用的docker环境，直接启动webui.py，提示", "id": 94}
{"title": "运行api.py后，执行curl -X POST \"http://127.0.0.1:7861\" 报错？", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/272", "detail": "执行curl -X POST \"http://127.0.0.1:7861\" \\ -H 'Content-Type: application/json' \\ -d '{\"prompt\": \"你好\", \"history\": []}'，报错怎么解决", "id": 95}
{"title": "[BUG] colab安装requirements提示protobuf版本问题？", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/273", "detail": "pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.", "id": 96}
{"title": "请问项目里面向量相似度使用了什么方法计算呀？", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/275", "detail": "基本按照langchain里的FAISS.similarity_search_with_score_by_vector实现", "id": 97}
{"title": "[BUG] 安装detectron2后，pdf无法加载", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/276", "detail": "**问题描述 / Problem Description**", "id": 98}
{"title": "[BUG] 使用ChatYuan-V2模型无法流式输出，会报错", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/277", "detail": "一方面好像是ChatYuan本身不支持stream_chat，有人在clueai那边提了issue他们说还没开发，所以估计这个attribute调不起来；但是另一方面看报错好像是T5模型本身就不是decoder-only模型，所以不能流式输出吧（个人理解）", "id": 99}
{"title": "[BUG] 无法加载text2vec模型", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/278", "detail": "**问题描述 / Problem Description**", "id": 100}
{"title": "请问能否增加网络搜索功能", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/281", "detail": "请问能否增加网络搜索功能", "id": 101}
{"title": "[FEATURE] 结构化数据sql、excel、csv啥时会支持呐。", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/283", "detail": "**功能描述 / Feature Description**", "id": 102}
{"title": "TypeError: ChatGLM._call() got an unexpected keyword argument 'stop'", "file": "2023-05-08.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/284", "detail": "No sentence-transformers model found with name D:\\DevProject\\langchain-ChatGLM\\GanymedeNil\\text2vec-large-chinese. Creating a new one with MEAN pooling.", "id": 103}
{"title": "关于api.py的一些bug和设计逻辑问题？", "file": "2023-05-09.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/285", "detail": "首先冒昧的问一下，这个api.py，开发者大佬们是在自己电脑上测试后确实没问题吗？", "id": 104}
{"title": "有没有租用的算力平台上，运行api.py后，浏览器http://localhost:7861/报错", "file": "2023-05-09.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/287", "detail": "是不是租用的gpu平台上都会出现这个问题？？？", "id": 105}
{"title": "请问一下项目中有用到文档段落切割方法吗？", "file": "2023-05-09.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/288", "detail": "text_load中的文档切割方法用上了吗？在代码中看好像没有用到？", "id": 106}
{"title": "报错 raise ValueError(f\"Knowledge base {knowledge_base_id} not found\") ValueError: Knowledge base ./vector_store not found", "file": "2023-05-09.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/289", "detail": "File \"/root/autodl-tmp/chatglm/langchain-ChatGLM-master/api.py\", line 183, in chat", "id": 107}
{"title": "能接入vicuna模型吗", "file": "2023-05-09.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/290", "detail": "目前本地已经有了vicuna模型能直接接入吗？", "id": 108}
{"title": "[BUG] 提问公式相关问题大概率爆显存", "file": "2023-05-09.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/291", "detail": "**问题描述 / Problem Description**", "id": 109}
{"title": "安装pycocotools失败，找了好多方法都不能解决。", "file": "2023-05-10.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/292", "detail": "**问题描述 / Problem Description**", "id": 110}
{"title": "使用requirements安装，PyTorch安装的是CPU版本", "file": "2023-05-10.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/294", "detail": "如题目，使用requirements安装，PyTorch安装的是CPU版本，运行程序的时候，也是使用CPU在工作。", "id": 111}
{"title": "能不能给一个毛坯服务器的部署教程", "file": "2023-05-10.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/298", "detail": "“开发部署”你当成服务器的部署教程用就行了。", "id": 112}
{"title": " Error(s) in loading state_dict for ChatGLMForConditionalGeneration:", "file": "2023-05-10.02", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/299", "detail": "运行中出现的问题，7860的端口页面显示不出来，求助。", "id": 113}
{"title": "ChatYuan-large-v2模型加载失败", "file": "2023-05-10.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/300", "detail": "**实际结果 / Actual Result**", "id": 114}
{"title": "新增摘要功能", "file": "2023-05-10.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/303", "detail": "你好，后续会考虑新增对长文本信息进行推理和语音理解功能吗？比如生成摘要", "id": 115}
{"title": "[BUG] pip install -r requirements.txt 出错", "file": "2023-05-10.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/304", "detail": "pip install langchain -i https://pypi.org/simple", "id": 116}
{"title": "[BUG] 上传知识库文件报错", "file": "2023-05-10.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/305", "detail": "![19621e29eaa547d01213bee53d81e6a](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/84606552/7f6ceb46-e494-4b0e-939c-23b585a6d9d8)", "id": 117}
{"title": "[BUG] AssertionError: <class 'gradio.layouts.Accordion'> Component with id 41 not a valid input component.", "file": "2023-05-10.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/306", "detail": "**问题描述 / Problem Description**", "id": 118}
{"title": "[BUG] CUDA out of memory with container deployment", "file": "2023-05-10.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/310", "detail": "**问题描述 / Problem Description**", "id": 119}
{"title": "[FEATURE] 增加微调训练功能", "file": "2023-05-11.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/311", "detail": "**功能描述 / Feature Description**", "id": 120}
{"title": "如何使用多卡部署，多个gpu", "file": "2023-05-11.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/315", "detail": "机器上有多个gpu,如何全使用了", "id": 121}
{"title": "请问这个知识库问答，和chatglm的关系是什么", "file": "2023-05-11.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/319", "detail": "这个知识库问答，哪部分关联到了chatglm，是不是没有这个chatglm，知识库问答也可单单拎出来", "id": 122}
{"title": "[BUG] 运行的时候报错ImportError: libcudnn.so.8: cannot open shared object file: No such file or directory", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/324", "detail": "**问题描述 / Problem Description**raceback (most recent call last):", "id": 123}
{"title": "webui启动成功，但有报错", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/325", "detail": "**问题描述 / Problem Description**", "id": 124}
{"title": "切换MOSS的时候报错", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/327", "detail": "danshi但是发布的源码中，", "id": 125}
{"title": "vicuna模型是否能接入？", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/328", "detail": "您好！关于MOSS模型和vicuna模型，都是AutoModelForCausalLM来加载模型的，但是稍作更改（模型路径这些）会报这个错误。这个错误的造成是什么", "id": 126}
{"title": "你好，请问一下在阿里云CPU服务器上跑可以吗？可以的话比较理想的cpu配置是什么？", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/330", "detail": "你好，请问一下在阿里云CPU服务器上跑可以吗？可以的话比较理想的cpu配置是什么？", "id": 127}
{"title": "你好，请问8核32g的CPU可以跑多轮对话吗？", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/331", "detail": "什么样的cpu配置比较好呢？我目前想部署CPU下的多轮对话？", "id": 128}
{"title": "[BUG] 聊天内容输入超过10000个字符系统出现错误", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/332", "detail": "聊天内容输入超过10000个字符系统出现错误，如下图所示：", "id": 129}
{"title": "能增加API的多用户访问接口部署吗？", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/333", "detail": "默认部署程序仅支持单用户访问，多用户则需要排队访问。测试过相关的几个Github多用户工程，但是其中一些仍然不满足要求。本节将系统介绍如何实现多用户同时访问ChatGLM的部署接口，包括http、websocket（流式输出，stream）和web页面等方式，主要目录如下所示。", "id": 130}
{"title": "多卡部署", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/334", "detail": "用单机多卡或多机多卡，fastapi部署模型，怎样提高并发", "id": 131}
{"title": "WEBUI能否指定知识库目录？", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/335", "detail": "**功能描述 / Feature Description**", "id": 132}
{"title": "[BUG] Cannot read properties of undefined (reading 'error')", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/336", "detail": "**问题描述 / Problem Description**", "id": 133}
{"title": "[BUG] 1 validation error for HuggingFaceEmbeddings model_kwargs extra fields not permitted.", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/337", "detail": "模型加载到 100% 后出现问题：", "id": 134}
{"title": "上传知识库需要重启能不能修复一下", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/338", "detail": "挺严重的这个问题", "id": 135}
{"title": "[BUG] 4块v100卡爆显存，在LLM会话模式也一样", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/339", "detail": "**问题描述 / Problem Description**", "id": 136}
{"title": "针对上传的文件配置不同的TextSpliter", "file": "2023-05-12.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/341", "detail": "1. 目前的ChineseTextSpliter切分对英文尤其是代码文件不友好，而且限制固定长度；导致对话结果不如人意", "id": 137}
{"title": "[FEATURE] 未来可增加Bloom系列模型吗？根据甲骨易的测试，这系列中文评测效果不错", "file": "2023-05-13.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/346", "detail": "**功能描述 / Feature Description**", "id": 138}
{"title": "[BUG] v0.1.12打包镜像后启动webui.py失败 / Concise description of the issue", "file": "2023-05-13.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/347", "detail": "**问题描述 / Problem Description**", "id": 139}
{"title": "切换MOSS模型时报错", "file": "2023-05-13.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/349", "detail": "昨天问了下，说是transformers版本不对，需要4.30.0，发现没有这个版本，今天更新到4.29.1，依旧报错，错误如下", "id": 140}
{"title": "[BUG] pdf文档加载失败", "file": "2023-05-13.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/350", "detail": "**问题描述 / Problem Description**", "id": 141}
{"title": "建议可以在后期增强一波注释，这样也有助于更多人跟进提PR", "file": "2023-05-13.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/351", "detail": "知道作者和团队在疯狂更新审查代码，只是建议后续稳定后可以把核心代码进行一些注释的补充，从而能帮助更多人了解各个模块作者的思路从而提出更好的优化。", "id": 142}
{"title": "[FEATURE] MOSS 量化版支援", "file": "2023-05-13.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/353", "detail": "**功能描述 / Feature Description**", "id": 143}
{"title": "[BUG] moss模型无法加载", "file": "2023-05-13.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/356", "detail": "**问题描述 / Problem Description**", "id": 144}
{"title": "[BUG] load_doc_qa.py 中的 load_file 函数有bug", "file": "2023-05-14.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/358", "detail": "原函数为：", "id": 145}
{"title": "[FEATURE] API模式，知识库加载优化", "file": "2023-05-14.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/359", "detail": "如题，当前版本，每次调用本地知识库接口，都将加载一次知识库，是否有更好的方式？", "id": 146}
{"title": "运行Python api.py脚本后端部署后，怎么使用curl命令调用？", "file": "2023-05-15.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/361", "detail": "也就是说，我现在想做个对话机器人，想和公司的前后端联调？怎么与前后端相互调用呢？可私信，有偿解答！！！", "id": 147}
{"title": "上传知识库需要重启能不能修复一下", "file": "2023-05-15.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/363", "detail": "上传知识库需要重启能不能修复一下", "id": 148}
{"title": "[BUG] pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple", "file": "2023-05-15.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/364", "detail": "我的python是3.8.5的", "id": 149}
{"title": "pip install gradio 报错", "file": "2023-05-15.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/367", "detail": "大佬帮我一下", "id": 150}
{"title": "[BUG] pip install gradio 一直卡不动", "file": "2023-05-15.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/369", "detail": "![aba82742dd9d4d242181662eb5027a7](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/84606552/cd9600d9-f6e7-46b7-b1be-30ed8b99f76b)", "id": 151}
{"title": "[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/370", "detail": "初次加载本地知识库成功，但提问后，就无法重写加载本地知识库", "id": 152}
{"title": "[FEATURE] 简洁阐述功能 / Concise description of the feature", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/371", "detail": "**功能描述 / Feature Description**", "id": 153}
{"title": "在windows上，模型文件默认会安装到哪", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/372", "detail": "-------------------------------------------------------------------------------", "id": 154}
{"title": "[FEATURE] 兼顾对话管理", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/374", "detail": "如何在知识库检索的情况下，兼顾对话管理？", "id": 155}
{"title": "llm device: cpu embedding device: cpu", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/376", "detail": "**问题描述 / Problem Description**", "id": 156}
{"title": "[FEATURE] 简洁阐述功能 /文本文件的知识点之间使用什么分隔符可以分割？", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/377", "detail": "**功能描述 / Feature Description**", "id": 157}
{"title": "[BUG] 上传文件失败：PermissionError: [WinError 32] 另一个程序正在使用此文件，进程无法访问。", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/379", "detail": "**问题描述 / Problem Description**", "id": 158}
{"title": "[BUG] 执行python api.py 报错", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/383", "detail": "错误信息", "id": 159}
{"title": "model_kwargs extra fields not permitted (type=value_error.extra)", "file": "2023-05-16.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/384", "detail": "大家好，请问这个有遇到的么,？", "id": 160}
{"title": "[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-05-17.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/385", "detail": "执行的时候出现了ls1 = [ls[0]]", "id": 161}
{"title": "[FEATURE] 性能优化", "file": "2023-05-17.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/388", "detail": "**功能描述 / Feature Description**", "id": 162}
{"title": "[BUG] Moss模型问答，RuntimeError: probability tensor contains either inf, nan or element < 0", "file": "2023-05-17.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/390", "detail": "**问题描述 / Problem Description**", "id": 163}
{"title": "有没有人知道v100GPU的32G显存，会报错吗？支持V100GPU吗？", "file": "2023-05-17.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/392", "detail": "**问题描述 / Problem Description**", "id": 164}
{"title": "针对于编码问题比如'gbk' codec can't encode character '\\xab' in position 14: illegal multibyte sequence粗浅的解决方法", "file": "2023-05-17.03", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/397", "detail": "**功能描述 / Feature Description**", "id": 165}
{"title": "Could not import sentence_transformers python package. Please install it with `pip install sentence_transformers`.", "file": "2023-05-18.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/400", "detail": "**问题描述 / Problem Description**", "id": 166}
{"title": "支持模型问答与检索问答", "file": "2023-05-18.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/401", "detail": "不同的query，根据意图不一致，回答也应该不一样。", "id": 167}
{"title": "文本分割的时候，能不能按照txt文件的每行进行分割，也就是按照换行符号\\n进行分割？？？", "file": "2023-05-18.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/403", "detail": "下面的代码应该怎么修改？", "id": 168}
{"title": "local_doc_qa/local_doc_chat 接口响应是串行", "file": "2023-05-18.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/405", "detail": "**问题描述 / Problem Description**", "id": 169}
{"title": "为什么找到出处了,但是还是无法回答该问题?", "file": "2023-05-18.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/406", "detail": "![图片](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/3349611/1fc81d61-2409-4330-9065-fdda1a27c86a)", "id": 170}
{"title": "请问下:知识库测试中的:添加单条内容,如果换成文本导入是是怎样的格式?我发现添加单条内容测试效果很好.", "file": "2023-05-18.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/412", "detail": "我发现在知识库测试中`添加单条内容`,并且勾选`禁止内容分句入库`,即使 `不开启上下文关联`的测试效果都非常好.", "id": 171}
{"title": "[BUG] 无法配置知识库", "file": "2023-05-18.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/413", "detail": "**问题描述 / Problem Description**", "id": 172}
{"title": "[BUG] 部署在阿里PAI平台的EAS上访问页面是白屏", "file": "2023-05-19.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/414", "detail": "**问题描述 / Problem Description**", "id": 173}
{"title": "API部署后调用/local_doc_qa/local_doc_chat 返回Knowledge base samples not found", "file": "2023-05-19.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/416", "detail": "入参", "id": 174}
{"title": "[FEATURE] 上传word另存为的txt文件报 'ascii' codec can't decode byte 0xb9 in position 6: ordinal not in range(128)", "file": "2023-05-20.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/421", "detail": "上传word另存为的txt文件报", "id": 175}
{"title": "创建保存的知识库刷新后没有出来，这个知识库是永久保存的吗？可以连外部的 向量知识库吗？", "file": "2023-05-21.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/422", "detail": "创建保存的知识库刷新后没有出来，这个知识库是永久保存的吗？可以连外部的 向量知识库吗？", "id": 176}
{"title": "[BUG] 用colab运行，无法加载模型，报错：'NoneType' object has no attribute 'message_types_by_name'", "file": "2023-05-21.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/423", "detail": "**问题描述 / Problem Description**", "id": 177}
{"title": "请问是否需要用到向量数据库？以及什么时候需要用到向量数据库？", "file": "2023-05-21.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/424", "detail": "目前用的是 text2vec ， 请问是否需要用到向量数据库？以及什么时候需要用到向量数据库？", "id": 178}
{"title": "huggingface模型引用问题", "file": "2023-05-22.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/427", "detail": "它最近似乎变成了一个Error？", "id": 179}
{"title": "你好，加载本地txt文件出现这个killed错误，TXT文件有100M左右大小。原因是？谢谢。", "file": "2023-05-22.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/429", "detail": "<img width=\"677\" alt=\"929aca3b22b8cd74e997a87b61d241b\" src=\"https://github.com/imClumsyPanda/langchain-ChatGLM/assets/109277248/24024522-c884-4170-b5cf-a498491bd8bc\">", "id": 180}
{"title": "想请问一下，关于对本地知识的管理是如何管理？例如：通过http API接口添加数据 或者 删除某条数据", "file": "2023-05-22.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/430", "detail": "例如：通过http API接口添加、删除、修改 某条数据。", "id": 181}
{"title": "[FEATURE] 双栏pdf识别问题", "file": "2023-05-22.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/432", "detail": "试了一下模型，感觉对单栏pdf识别的准确性较高，但是由于使用的基本是ocr的技术，对一些双栏pdf论文识别出来有很多问题，请问有什么办法改善吗？", "id": 182}
{"title": "部署启动小问题，小弟初学求大佬解答", "file": "2023-05-22.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/433", "detail": "1.python loader/image_loader.py时，提示ModuleNotFoundError: No module named 'configs'，但是跑python webui.py还是还能跑", "id": 183}
{"title": "能否支持检测到目录下文档有增加而去增量加载文档，不影响前台对话，其实就是支持读写分离。如果能支持查询哪些文档向量化了，删除过时文档等就更好了，谢谢。", "file": "2023-05-22.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/434", "detail": "**功能描述 / Feature Description**", "id": 184}
{"title": "[BUG] 简洁阐述问题 / windows 下cuda错误，请用https://github.com/Keith-Hon/bitsandbytes-windows.git", "file": "2023-05-22.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/435", "detail": "pip install git+https://github.com/Keith-Hon/bitsandbytes-windows.git", "id": 185}
{"title": "[BUG] from commit 33bbb47, Required library version not found: libbitsandbytes_cuda121_nocublaslt.so. Maybe you need to compile it from source?", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/438", "detail": "**问题描述 / Problem Description**", "id": 186}
{"title": "[BUG] 简洁阐述问题 / Concise description of the issue上传60m的txt文件报错，显示超时，请问这个能上传的文件大小有限制吗", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/439", "detail": "ERROR 2023-05-23 11:13:09,627-1d: Timeout reached while detecting encoding for ./docs/GLM模型格式数据.txt", "id": 187}
{"title": "[BUG] TypeError: issubclass() arg 1 must be a class", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/440", "detail": "**问题描述**", "id": 188}
{"title": "执行python3 webui.py后，一直提示”模型未成功加载，请到页面左上角\"模型配置\"选项卡中重新选择后点击\"加载模型\"按钮“", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/441", "detail": "**问题描述 / Problem Description**", "id": 189}
{"title": "是否能提供网页文档得导入支持", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/444", "detail": "现在很多都是在线文档作为协作得工具，所以通过URL导入在线文档需求更大", "id": 190}
{"title": "[BUG] history 索引问题", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/445", "detail": "在比较对话框的history和模型chat function 中的history时， 发现并不匹配，在传入 llm._call 时，history用的索引是不是有点问题，导致上一轮对话的内容并不输入给模型。", "id": 191}
{"title": "[BUG] moss_llm没有实现", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/447", "detail": "有些方法没支持，如history_len", "id": 192}
{"title": "请问langchain-ChatGLM如何删除一条本地知识库的数据？", "file": "2023-05-23.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/448", "detail": "例如：用户刚刚提交了一条错误的数据到本地知识库中了，现在如何在本地知识库从找到，并且对此删除。", "id": 193}
{"title": "[BUG] 简洁阐述问题 / UnboundLocalError: local variable 'resp' referenced before assignment", "file": "2023-05-24.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/450", "detail": "在最新一版的代码中， 运行api.py 出现了以上错误（UnboundLocalError: local variable 'resp' referenced before assignment）， 通过debug的方式观察到local_doc_qa.llm.generatorAnswer(prompt=question, history=history,streaming=True)可能不返回任何值。", "id": 194}
{"title": "请问有没有 PROMPT_TEMPLATE 能让模型不回答敏感问题", "file": "2023-05-24.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/452", "detail": "## PROMPT_TEMPLATE问题", "id": 195}
{"title": "[BUG] 测试环境 Python 版本有误", "file": "2023-05-24.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/456", "detail": "**问题描述 / Problem Description**", "id": 196}
{"title": "[BUG] webui 部署后样式不正确", "file": "2023-05-24.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/458", "detail": "**问题描述 / Problem Description**", "id": 197}
{"title": "配置默认LLM模型的问题", "file": "2023-05-24.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/459", "detail": "**问题描述 / Problem Description**", "id": 198}
{"title": "[FEATURE]是时候更新一下autoDL的镜像了", "file": "2023-05-24.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/460", "detail": "如题，跑了下autoDL的镜像，发现是4.27号的，git pull新版本的代码功能+老的依赖环境，各种奇奇怪怪的问题。", "id": 199}
{"title": "[BUG] tag:0.1.13 以cpu模式下，想使用本地模型无法跑起来，各种路径参数问题", "file": "2023-05-24.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/462", "detail": "-------------------------------------------------------------------------------", "id": 200}
{"title": "[BUG] 有没有同学遇到过这个错！！！加载本地txt文件出现这个killed错误，TXT文件有100M左右大小。", "file": "2023-05-25.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/463", "detail": "运行cli_demo.py。是本地的txt文件太大了吗？100M左右。", "id": 201}
{"title": "API版本能否提供WEBSOCKET的流式接口", "file": "2023-05-25.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/464", "detail": "webui 版本中，采用了WS的流式输出，整体感知反应很快", "id": 202}
{"title": "[BUG] 安装bug记录", "file": "2023-05-25.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/465", "detail": "按照[install文档](https://github.com/imClumsyPanda/langchain-ChatGLM/blob/master/docs/INSTALL.md)安装的，", "id": 203}
{"title": "VUE的pnmp i执行失败的修复-用npm i命令即可", "file": "2023-05-25.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/466", "detail": "感谢作者！非常棒的应用，用的很开心。", "id": 204}
{"title": "请教个问题，有没有人知道cuda11.4是否支持？？？", "file": "2023-05-25.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/467", "detail": "请教个问题，有没有人知道cuda11.4是否支持？？？", "id": 205}
{"title": "请问有实现多轮问答中基于问题的搜索上下文关联么", "file": "2023-05-25.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/468", "detail": "在基于知识库的多轮问答中，第一个问题讲述了一个主题，后续的问题描述没有包含这个主题的关键词，但又存在上下文的关联。如果用后续问题去搜索知识库有可能会搜索出无关的信息，从而导致大模型无法正确回答问题。请问这个项目要考虑这种情况吗？", "id": 206}
{"title": "[BUG] 内存不足的问题", "file": "2023-05-26.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/470", "detail": "我用了本地的chatglm-6b-int4模型，然后显示了内存不足（win10+32G内存+1080ti11G），一般需要多少内存才足够？这个bug应该如何解决？", "id": 207}
{"title": "[BUG] 纯内网环境安装pycocotools失败", "file": "2023-05-26.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/472", "detail": "**问题描述 / Problem Description**", "id": 208}
{"title": "[BUG] webui.py 重新加载模型会导致 KeyError", "file": "2023-05-26.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/473", "detail": "**问题描述 / Problem Description**", "id": 209}
{"title": "chatyuan无法使用", "file": "2023-05-26.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/475", "detail": "**问题描述 / Problem Description**", "id": 210}
{"title": "[BUG] 文本分割模型AliTextSplitter存在bug，会把“.”作为分割符", "file": "2023-05-26.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/476", "detail": "阿里达摩院的语义分割模型存在bug，默认会把\".”作为分割符进行分割而不管上下文语义。是否还有其他分割符则未知。建议的修改方案：把“.”统一替换为其他字符，分割后再替换回来。或者添加其他分割模型。", "id": 211}
{"title": "[BUG] RuntimeError: Error in faiss::FileIOReader::FileIOReader(const char*) a", "file": "2023-05-27.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/479", "detail": "**问题描述 / Problem Description**", "id": 212}
{"title": "[FEATURE] 安装，为什么conda create要额外指定路径 用-p ，而不是默认的/envs下面", "file": "2023-05-28.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/481", "detail": "##**功能描述 / Feature Description**", "id": 213}
{"title": "[小白求助] 通过Anaconda执行webui.py后，无法打开web链接", "file": "2023-05-28.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/485", "detail": "在执行webui.py命令后，http://0.0.0.0:7860复制到浏览器后无法打开，显示“无法访问此网站”。", "id": 214}
{"title": "[BUG] 使用 p-tuningv2后的模型，重新加载报错", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/486", "detail": "把p-tunningv2训练完后的相关文件放到了p-tunningv2文件夹下，勾选使用p-tuningv2点重新加载模型，控制台输错错误信息：", "id": 215}
{"title": "[小白求助] 服务器上执行webui.py后，在本地无法打开web链接", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/487", "detail": "此项目执行在xxx.xx.xxx.xxx服务器上，我在webui.py上的代码为 (demo", "id": 216}
{"title": "[FEATURE] 能不能支持VisualGLM-6B", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/488", "detail": "**功能描述 / Feature Description**", "id": 217}
{"title": "你好，问一下各位，后端api部署的时候，支持多用户同时问答吗？？？", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/489", "detail": "支持多用户的话，最多支持多少用户问答？根据硬件而定吧？", "id": 218}
{"title": "V100GPU显存占满，而利用率却为0，这是为什么？", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/491", "detail": "<img width=\"731\" alt=\"de45fe2b6cb76fa091b6e8f76a3de60\" src=\"https://github.com/imClumsyPanda/langchain-ChatGLM/assets/109277248/c32efd52-7dbf-4e9b-bd4d-0944d73d0b8b\">", "id": 219}
{"title": "[求助] 如果在公司内部搭建产品知识库，使用INT-4模型，200人规模需要配置多少显存的服务器？", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/492", "detail": "如题，计划给公司搭一个在线知识库。", "id": 220}
{"title": "你好，请教个问题，目前问答回复需要20秒左右，如何提高速度？V10032G服务器。", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/493", "detail": "**问题描述 / Problem Description**", "id": 221}
{"title": "[FEATURE] 如何实现只匹配下文，而不要上文的结果", "file": "2023-05-29.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/494", "detail": "在构建自己的知识库时，主要采用问答对的形式，那么也就是我需要的回答是在我的问题下面的内容，但是目前设置了chunk_size的值以后匹配的是上下文的内容，但我实际并不需要上文的。为了实现更完整的展示下面的答案，我只能调大chunk_size的值，但实际上上文的一半内容都是我不需要的。也就是扔了一半没用的东西给prompt，在faiss.py中我也没找到这块的一些描述，请问该如何进行修改呢？", "id": 222}
{"title": "你好，问一下，我调用api.py部署，为什么用ip加端口可以使用postman调用，而改为域名使用postman无法调用？", "file": "2023-05-30.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/497", "detail": "![5ufBSWxLyF](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/109277248/70e2fbac-5699-48d0-b0d1-3dc84fd042c2)", "id": 223}
{"title": "调用api.py中的stream_chat，返回source_documents中出现中文乱码。", "file": "2023-05-30.04", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/498", "detail": "-------------------------------------------------------------------------------", "id": 224}
{"title": "[BUG] 捉个虫，api.py中的stream_chat解析json问题", "file": "2023-05-30.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/501", "detail": "**问题描述 / Problem Description**", "id": 225}
{"title": "windows本地部署遇到了omp错误", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/502", "detail": "**问题描述 / Problem Description**", "id": 226}
{"title": "[BUG] bug14 ,\"POST /local_doc_qa/upload_file HTTP/1.1\" 422 Unprocessable Entity", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/503", "detail": "上传的文件报错，返回错误，api.py", "id": 227}
{"title": "你好，请教个问题，api.py部署的时候，如何改为多线程调用？谢谢", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/505", "detail": "目前的api.py脚本不支持多线程", "id": 228}
{"title": "你好，请教一下。api.py部署的时候，能不能提供给后端流失返回结果。", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/507", "detail": "curl -X 'POST' \\", "id": 229}
{"title": "流式输出，流式接口，使用server-sent events技术。", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/508", "detail": "想这样一样，https://blog.csdn.net/weixin_43228814/article/details/130063010", "id": 230}
{"title": "计划增加流式输出功能吗？ChatGLM模型通过api方式调用响应时间慢怎么破，Fastapi流式接口来解惑，能快速提升响应速度", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/509", "detail": "**问题描述 / Problem Description**", "id": 231}
{"title": "[BUG] 知识库上传时发生ERROR (could not open xxx for reading: No such file or directory)", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/510", "detail": "**问题描述 / Problem Description**", "id": 232}
{"title": "api.py脚本打算增加SSE流式输出吗？", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/511", "detail": "curl调用的时候可以检测第一个字，从而提升回复的体验", "id": 233}
{"title": "[BUG] 使用tornado实现webSocket，可以多个客户端同时连接，并且实现流式回复，但是多个客户端同时使用，答案就很乱，是模型不支持多线程吗", "file": "2023-05-31.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/512", "detail": "import asyncio", "id": 234}
{"title": "支持 chinese_alpaca_plus_lora 吗 基于llama的", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/514", "detail": "支持 chinese_alpaca_plus_lora 吗 基于llama的，https://github.com/ymcui/Chinese-LLaMA-Alpaca这个项目的", "id": 235}
{"title": "[BUG] 现在能读图片的pdf了，但是文字的pdf反而读不了了，什么情况？？？", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/515", "detail": "**问题描述 / Problem Description**", "id": 236}
{"title": "在推理的过程中卡住不动，进程无法正常结束", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/516", "detail": "**问题描述 / Problem Description**", "id": 237}
{"title": "curl调用的时候，从第二轮开始，curl如何传参可以实现多轮对话？", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/517", "detail": "第一轮调用：", "id": 238}
{"title": "建议添加api.py部署后的日志管理功能？", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/518", "detail": "-------------------------------------------------------------------------------", "id": 239}
{"title": "有大佬知道，怎么多线程部署api.py脚本吗？", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/519", "detail": "api.py部署后，使用下面的请求，时间较慢，好像是单线程，如何改为多线程部署api.py：", "id": 240}
{"title": "[BUG] 上传文件到知识库 任何格式与内容都永远失败", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/520", "detail": "上传知识库的时候，传txt无法解析，就算是穿content/sample里的样例txt也无法解析，上传md、pdf等都无法加载，会持续性等待，等到了超过30分钟也不行。", "id": 241}
{"title": "关于prompt_template的问题", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/521", "detail": "请问这段prompt_template是什么意思，要怎么使用？可以给一个具体模板参考下吗？", "id": 242}
{"title": "[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-06-01.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/522", "detail": "**问题描述 / Problem Description**", "id": 243}
{"title": "中文分词句号处理（关于表达金额之间的\".\"）", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/523", "detail": "建议处理12.6亿元的这样的分词，最好别分成12 和6亿这样的，需要放到一起", "id": 244}
{"title": "ImportError: cannot import name 'inference' from 'paddle'", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/526", "detail": "在网上找了一圈，有说升级paddle的，我做了还是没有用，有说安装paddlepaddle的，我找了豆瓣的镜像源，但安装报错cannot detect archive format", "id": 245}
{"title": "[BUG] webscoket 接口串行问题（/local_doc_qa/stream-chat/{knowledge_base_id}）", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/527", "detail": "**问题描述 / Problem Description**", "id": 246}
{"title": "[FEATURE] 刷新页面更新知识库列表", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/528", "detail": "**功能描述以及改进方案**", "id": 247}
{"title": "[BUG] 使用ptuning微调模型后，问答效果并不好", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/530", "detail": "### 未调用ptuning", "id": 248}
{"title": "[BUG] 多轮对话效果不佳", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/532", "detail": "在进行多轮对话的时候，无论设置的history_len是多少，效果都不好。事实上我将其设置成了最大值10，但在对话中，仍然无法实现多轮对话：", "id": 249}
{"title": "RuntimeError: MPS backend out of memory (MPS allocated: 18.00 GB, other allocations: 4.87 MB, max allowed: 18.13 GB)", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/533", "detail": "**问题描述**", "id": 250}
{"title": " 请大家重视这个issue！真正使用肯定是多用户并发问答，希望增加此功能！！！", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/534", "detail": "这得看你有多少显卡", "id": 251}
{"title": "在启动项目的时候如何使用到多张gpu啊？", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/535", "detail": "**在启动项目的时候如何使用到多张gpu啊？**", "id": 252}
{"title": " 使用流式输出的时候，curl调用的格式是什么？", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/536", "detail": "app.websocket(\"/local_doc_qa/stream-chat/{knowledge_base_id}\")(stream_chat)中的knowledge_base_id应该填什么？？？", "id": 253}
{"title": "使用本地 vicuna-7b模型启动错误", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/538", "detail": "环境: ubuntu 22.04 cuda 12.1 没有安装nccl，使用rtx2080与m60显卡并行计算", "id": 254}
{"title": "为什么会不调用GPU直接调用CPU呢", "file": "2023-06-02.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/539", "detail": "我的阿里云配置是16G显存，用默认代码跑webui.py时提示", "id": 255}
{"title": "上传多个文件时会互相覆盖", "file": "2023-06-03.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/541", "detail": "1、在同一个知识库中上传多个文件时会互相覆盖，无法结合多个文档的知识，有大佬知道怎么解决吗？", "id": 256}
{"title": "[BUG] ‘gcc’不是内部或外部命令/LLM对话只能持续一轮", "file": "2023-06-03.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/542", "detail": "No compiled kernel found.", "id": 257}
{"title": "以API模式启动项目却没有知识库的接口列表？", "file": "2023-06-04.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/544", "detail": "请问如何获取知识库的接口列表？如果没有需要自行编写的话，可不可以提供相关的获取方式，感谢", "id": 258}
{"title": "程序以API模式启动的时候，如何才能让接口以stream模式被调用呢？", "file": "2023-06-05.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/546", "detail": "作者您好，我在以API模式进行程序启动后，我发现接口响应时间很长，怎么样才能让接口以stream模式被调用呢？我想实现像webui模式的回答那样", "id": 259}
{"title": "关于原文中表格转为文本后数据相关度问题。", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/547", "detail": "原文中表格数据转换为文本，以 （X-Y：值；...） 的格式每一行组织成一句话，但这样做后发现相关度较低，效果很差，有何好的方案吗？", "id": 260}
{"title": "启动后LLM和知识库问答模式均只有最后一轮记录", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/548", "detail": "拉取最新代码，问答时，每次页面只显示最后一次问答记录，需要修改什么参数才可以保留历史记录？", "id": 261}
{"title": "提供system message配置，以便于让回答不要超出知识库范围", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/549", "detail": "**功能描述 / Feature Description**", "id": 262}
{"title": "[BUG] 使用p-tunningv2报错", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/551", "detail": "按照readme的指示把p-tunningv2训练完后的文件放到了p-tunningv2文件夹下，勾选使用p-tuningv2点重新加载模型，控制台提示错误信息：", "id": 263}
{"title": "[BUG] 智障，这么多问题，也好意思放出来，浪费时间", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/553", "detail": "。。。", "id": 264}
{"title": "[FEATURE] 我看代码文件中有一个ali_text_splitter.py，为什么不用他这个文本分割器了？", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/554", "detail": "我看代码文件中有一个ali_text_splitter.py，为什么不用他这个文本分割器了？", "id": 265}
{"title": "加载文档函数报错", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/557", "detail": "def load_file(filepath, sentence_size=SENTENCE_SIZE):", "id": 266}
{"title": "参考指引安装docker后，运行cli_demo.py，提示killed", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/558", "detail": "root@b3d1bd08095c:/chatGLM# python3 cli_demo.py", "id": 267}
{"title": "注意：如果安装错误，注意这两个包的版本 wandb==0.11.0 protobuf==3.18.3", "file": "2023-06-06.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/559", "detail": "Error1: 如果启动异常报错 `protobuf` 需要更新到 `protobuf==3.18.3 `", "id": 268}
{"title": "知识库对长文的知识相关度匹配不太理想有何优化方向", "file": "2023-06-07.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/563", "detail": "我们可能录入一个文章有 1W 字，里面涉及这个文章主题的很多角度问题，我们针对他提问，他相关度匹配的内容和实际我们需要的答案相差很大怎么办。", "id": 269}
{"title": "使用stream-chat函数进行流式输出的时候，能使用curl调用吗？", "file": "2023-06-07.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/565", "detail": "为什么下面这样调用会报错？？？", "id": 270}
{"title": "有大佬实践过 并行 或者 多线程 的部署方案吗？", "file": "2023-06-07.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/566", "detail": "+1", "id": 271}
{"title": "多线程部署遇到问题？", "file": "2023-06-07.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/567", "detail": "<img width=\"615\" alt=\"3d87bf74f0cf1a4820cc9e46b245859\" src=\"https://github.com/imClumsyPanda/langchain-ChatGLM/assets/109277248/8787570d-88bd-434e-aaa4-cb9276d1aa50\">", "id": 272}
{"title": "[BUG] 用fastchat加载vicuna-13b模型进行知识库的问答有token的限制错误", "file": "2023-06-07.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/569", "detail": "当我开启fastchat的vicuna-13b的api服务，然后config那里配置好(api本地测试过可以返回结果)，然后知识库加载好之后(知识库大概有1000多个文档，用chatGLM可以正常推理)，进行问答时出现token超过限制，就问了一句hello；", "id": 273}
{"title": "现在的添加知识库，文件多了总是报错，也不知道自己加载了哪些文件，报错后也不知道是全部失败还是一部分成功；希望能有个加载指定文件夹作为知识库的功能", "file": "2023-06-07.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/574", "detail": "**功能描述 / Feature Description**", "id": 274}
{"title": "[BUG] moss模型本地加载报错", "file": "2023-06-08.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/577", "detail": "moss模型本地加载报错：", "id": 275}
{"title": "加载本地moss模型报错Can't instantiate abstract class MOSSLLM with abstract methods _history_len", "file": "2023-06-08.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/578", "detail": "(vicuna) ps@ps[13:56:20]:/data/chat/langchain-ChatGLM2/langchain-ChatGLM-0.1.13$ python webui.py --model-dir local_models --model moss --no-remote-model", "id": 276}
{"title": "[FEATURE] 能增加在前端页面控制prompt_template吗？或是能支持前端页面选择使用哪个prompt？", "file": "2023-06-08.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/579", "detail": "目前只能在config里修改一个prompt，想在多个不同场景切换比较麻烦", "id": 277}
{"title": "[BUG] streamlit ui的bug，在增加知识库时会报错", "file": "2023-06-08.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/580", "detail": "**问题描述 / Problem Description**", "id": 278}
{"title": "[FEATURE] webui/webui_st可以支持history吗？目前仅能一次对话", "file": "2023-06-08.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/581", "detail": "试了下webui和webui_st都不支持历史对话啊，只能对话一次，不能默认开启所有history吗？", "id": 279}
{"title": "启动python cli_demo.py --model chatglm-6b-int4-qe报错", "file": "2023-06-09.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/585", "detail": "下载好模型，和相关依赖环境，之间运行`python cli_demo.py --model chatglm-6b-int4-qe`报错了：", "id": 280}
{"title": "重新构建知识库报错", "file": "2023-06-09.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/586", "detail": "**问题描述 / Problem Description**", "id": 281}
{"title": "[FEATURE] 能否屏蔽paddle，我不需要OCR，效果差依赖环境还很复杂", "file": "2023-06-09.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/587", "detail": "希望能不依赖paddle", "id": 282}
{"title": "question ：文档向量化这个可以自己手动实现么？", "file": "2023-06-09.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/589", "detail": "现有公司级数据500G+，需要使用这个功能，请问如何手动实现这个向量化，然后并加载", "id": 283}
{"title": "view前端能进行流式的返回吗？？", "file": "2023-06-09.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/590", "detail": "view前端能进行流式的返回吗？？", "id": 284}
{"title": "[BUG] Load parallel cpu kernel failed, using default cpu kernel code", "file": "2023-06-11.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/594", "detail": "**问题描述 / Problem Description**", "id": 285}
{"title": "[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-06-11.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/595", "detail": "**问题描述 / Problem Description**", "id": 286}
{"title": "我在上传本地知识库时提示KeyError: 'name'错误，本地知识库都是.txt文件，文件数量大约是2000+。", "file": "2023-06-12.05", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/597", "detail": "<img width=\"649\" alt=\"KError\" src=\"https://github.com/imClumsyPanda/langchain-ChatGLM/assets/59411575/1ecc8182-aeee-4a0a-bbc3-74c2f1373f2d\">", "id": 287}
{"title": "model_config.py中有vicuna-13b-hf模型的配置信息，但是好像还是不可用？", "file": "2023-06-12.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/600", "detail": "@dongyihua543", "id": 288}
{"title": "ImportError: Using SOCKS proxy, but the 'socksio' package is not installed. Make sure to install httpx using `pip install httpx[socks]`.", "file": "2023-06-12.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/605", "detail": "应该代理问题，但是尝试了好多方法都解决不了，", "id": 289}
{"title": "[BUG] similarity_search_with_score_by_vector在找不到匹配的情况下出错", "file": "2023-06-12.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/607", "detail": "在设置匹配阈值 VECTOR_SEARCH_SCORE_THRESHOLD 的情况下，vectorstore会返回空，此时上述处理函数会出错", "id": 290}
{"title": "[FEATURE] 请问如何搭建英文知识库呢", "file": "2023-06-12.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/609", "detail": "**功能描述 / Feature Description**", "id": 291}
{"title": "谁有vicuna权重？llama转换之后的", "file": "2023-06-13.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/611", "detail": "**问题描述 / Problem Description**", "id": 292}
{"title": "[FEATURE] API能实现上传文件夹的功能么？", "file": "2023-06-13.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/612", "detail": "用户懒得全选所有的文件，就想上传个文件夹，请问下API能实现这个功能么？", "id": 293}
{"title": "请问在多卡部署后，上传单个文件作为知识库，用的是单卡在生成向量还是多卡？", "file": "2023-06-13.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/614", "detail": "目前我检测我本地多卡部署的，好像生成知识库向量的时候用的还是单卡", "id": 294}
{"title": "[BUG] python webui.py提示非法指令", "file": "2023-06-13.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/615", "detail": "(/data/conda-langchain [root@chatglm langchain-ChatGLM]# python webui.py", "id": 295}
{"title": "知识库文件跨行切分问题", "file": "2023-06-13.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/616", "detail": "我的知识库文件txt文件，是一行一条知识，用\\n分行。", "id": 296}
{"title": "[FEATURE] bing搜索问答有流式的API么？", "file": "2023-06-13.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/617", "detail": "web端是有这个bing搜索回答，但api接口没有发现，大佬能给个提示么？", "id": 297}
{"title": "希望出一个macos m2的安装教程", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/620", "detail": "mac m2安装，模型加载成功了，知识库文件也上传成功了，但是一问答就会报错，报错内容如下", "id": 298}
{"title": "为【出处】提供高亮显示", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/621", "detail": "具体出处里面，对相关的内容高亮显示，不包含前后文。", "id": 299}
{"title": "[BUG] CPU运行cli_demo.py，不回答，hang住", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/622", "detail": "没有GPU；32G内存的ubuntu机器。", "id": 300}
{"title": "关于删除知识库里面的文档后，LLM知识库对话的时候还是会返回该被删除文档的内容", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/623", "detail": "如题，在vue前端成功执行删除知识库里面文档A.txt后，未能也在faiss索引中也删除该文档，LLM还是会返回这个A.txt的内容，并且以A.txt为出处，未能达到删除的效果", "id": 301}
{"title": "[BUG] 调用知识库进行问答,显存会一直叠加", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/625", "detail": "14G的显存,调用的chatglm-6b-int8模型,进行知识库问答时,最多问答四次就会爆显存了,观察了一下显存使用情况,每一次使用就会增加一次显存,请问这样是正常的吗?是否有什么配置需要开启可以解决这个问题?例如进行一次知识库问答清空上次问题的显存?", "id": 302}
{"title": "[BUG] web页面 重新构建数据库 失败，导致 原来的上传的数据库都没了", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/626", "detail": "web页面 重新构建数据库 失败，导致 原来的上传的数据库都没了", "id": 303}
{"title": "在CPU上运行webui.py报错Tensor on device cpu is not on the expected device meta!", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/627", "detail": "在CPU上运行python webui.py能启动，但最后有：RuntimeError: Tensor on device cpu is not on the expected device meta!", "id": 304}
{"title": "OSError: [WinError 1114] 动态链接库(DLL)初始化例程失败。 Error loading \"E:\\xxx\\envs\\langchain\\lib\\site-packages\\torch\\lib\\caffe2_nvrtc.dll\" or one of its dependencies.哪位大佬知道如何解决吗？", "file": "2023-06-14.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/629", "detail": "**问题描述 / Problem Description**", "id": 305}
{"title": "[BUG] WEBUI删除知识库文档，会导致知识库问答失败", "file": "2023-06-15.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/632", "detail": "如题，从知识库已有文件中选择要删除的文件，点击删除后，在问答框输入内容回车报错", "id": 306}
{"title": "更新后的版本中，删除知识库中的文件，再提问出现error错误", "file": "2023-06-15.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/634", "detail": "针对更新版本，识别到一个问题，过程如下：", "id": 307}
{"title": "我配置好了环境，想要实现本地知识库的问答？可是它返回给我的", "file": "2023-06-15.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/637", "detail": "没有总结，只有相关度的回复，但是我看演示里面表现的，回复是可以实现总结的，我去查询代码", "id": 308}
{"title": "[BUG] NPM run dev can not successfully start the VUE frontend", "file": "2023-06-15.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/638", "detail": "**问题描述 / Problem Description**", "id": 309}
{"title": "[BUG] 简洁阐述问题 / Concise description of the issue", "file": "2023-06-15.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/639", "detail": "**问题描述 / Problem Description**", "id": 310}
{"title": "提一个模型加载的bug，我在截图中修复了，你们有空可以看一下。", "file": "2023-06-15.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/642", "detail": "![model_load_bug](https://github.com/imClumsyPanda/langchain-ChatGLM/assets/59411575/4432adc4-ccdd-45d9-aafc-5f2d1963403b)", "id": 311}
{"title": "[求助]关于设置embedding model路径的问题", "file": "2023-06-16.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/643", "detail": "如题，我之前成功跑起来过一次，但因环境丢失重新配置 再运行webui就总是报错", "id": 312}
{"title": "Lora微调后的模型可以直接使用吗", "file": "2023-06-16.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/646", "detail": "看model_config.py里是有USE_LORA这个参数的，但是在cli_demo.py和webui.py这两个里面都没有用到，实际测试下来模型没有微调的效果，想问问现在这个功能实现了吗", "id": 313}
{"title": "write_check_file在tmp_files目录下生成的load_file.txt是否需要一直保留，占用空间很大，在建完索引后能否删除", "file": "2023-06-16.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/647", "detail": "**功能描述 / Feature Description**", "id": 314}
{"title": "[BUG] /local_doc_qa/list_files?knowledge_base_id=test删除知识库bug", "file": "2023-06-16.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/649", "detail": "1.新建test知识库并上传文件（在vue前端完成并检查后端发现确实生成了test文件夹以及下面的content和vec_store", "id": 315}
{"title": "[BUG] vue webui无法加载知识库", "file": "2023-06-16.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/650", "detail": "拉取了最新的代码，分别运行了后端api和前端web，点击知识库，始终只能显示simple，无法加载知识库", "id": 316}
{"title": "不能本地加载moss模型吗？", "file": "2023-06-16.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/652", "detail": "手动下载模型设置local_model_path路径依旧提示缺少文件，该如何正确配置？", "id": 317}
{"title": "macos m2 pro docker 安装失败", "file": "2023-06-17.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/654", "detail": "macos m2 pro docker 安装失败", "id": 318}
{"title": " [BUG] mac m1 pro 运行提示 zsh: segmentation fault", "file": "2023-06-17.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/655", "detail": "运行： python webui.py", "id": 319}
{"title": "安装 requirements 报错", "file": "2023-06-17.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/656", "detail": "(langchainchatglm) D:\\github\\langchain-ChatGLM>pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple/", "id": 320}
{"title": "[BUG] AssertionError", "file": "2023-06-17.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/658", "detail": "**问题描述 / Problem Description**", "id": 321}
{"title": "[FEATURE] 支持AMD win10 本地部署吗？", "file": "2023-06-18.06", "url": "https://github.com/imClumsyPanda/langchain-ChatGLM/issues/660", "detail": "**功能描述 / Feature Description**", "id": 322}
loader/image_loader.py
......@@ -5,9 +5,6 @@ from langchain.document_loaders.unstructured import UnstructuredFileLoader
from paddleocr import PaddleOCR
import os
import nltk
from configs.model_config import NLTK_DATA_PATH
nltk.data.path = [NLTK_DATA_PATH] + nltk.data.path
class UnstructuredPaddleImageLoader(UnstructuredFileLoader):
"""Loader that uses unstructured to load image files, such as PNGs and JPGs."""
......@@ -35,6 +32,10 @@ class UnstructuredPaddleImageLoader(UnstructuredFileLoader):
if __name__ == "__main__":
import sys
sys.path.append(os.path.dirname(os.path.dirname(__file__)))
from configs.model_config import NLTK_DATA_PATH
nltk.data.path = [NLTK_DATA_PATH] + nltk.data.path
filepath = os.path.join(os.path.dirname(os.path.dirname(__file__)), "knowledge_base", "samples", "content", "test.jpg")
loader = UnstructuredPaddleImageLoader(filepath, mode="elements")
docs = loader.load()
......
models/chatglm_llm.py
......@@ -65,6 +65,7 @@ class ChatGLM(BaseAnswer, LLM, ABC):
answer_result.history = history
answer_result.llm_output = {"answer": stream_resp}
yield answer_result
self.checkPoint.clear_torch_cache()
else:
response, _ = self.checkPoint.model.chat(
self.checkPoint.tokenizer,
......
models/fastchat_openai_llm.py
......@@ -23,6 +23,7 @@ def _build_message_template() -> Dict[str, str]:
class FastChatOpenAILLM(RemoteRpcModel, LLM, ABC):
api_base_url: str = "http://localhost:8000/v1"
model_name: str = "chatglm-6b"
max_token: int = 10000
......@@ -31,8 +32,14 @@ class FastChatOpenAILLM(RemoteRpcModel, LLM, ABC):
checkPoint: LoaderCheckPoint = None
history = []
history_len: int = 10
def __init__(self, checkPoint: LoaderCheckPoint = None):
api_key: str = ""
def __init__(self,
checkPoint: LoaderCheckPoint = None,
# api_base_url:str="http://localhost:8000/v1",
# model_name:str="chatglm-6b",
# api_key:str=""
):
super().__init__()
self.checkPoint = checkPoint
......@@ -60,7 +67,7 @@ class FastChatOpenAILLM(RemoteRpcModel, LLM, ABC):
return self.api_base_url
def set_api_key(self, api_key: str):
pass
self.api_key = api_key
def set_api_base_url(self, api_base_url: str):
self.api_base_url = api_base_url
......@@ -73,7 +80,8 @@ class FastChatOpenAILLM(RemoteRpcModel, LLM, ABC):
try:
import openai
# Not support yet
openai.api_key = "EMPTY"
# openai.api_key = "EMPTY"
openai.key = self.api_key
openai.api_base = self.api_base_url
except ImportError:
raise ValueError(
......@@ -116,7 +124,8 @@ class FastChatOpenAILLM(RemoteRpcModel, LLM, ABC):
try:
import openai
# Not support yet
openai.api_key = "EMPTY"
# openai.api_key = "EMPTY"
openai.api_key = self.api_key
openai.api_base = self.api_base_url
except ImportError:
raise ValueError(
......
models/llama_llm.py
......@@ -6,14 +6,17 @@ import torch
import transformers
from transformers.generation.logits_process import LogitsProcessor
from transformers.generation.utils import LogitsProcessorList, StoppingCriteriaList
from typing import Optional, List, Dict, Any
from typing import Optional, List, Dict, Any,Union
from models.loader import LoaderCheckPoint
from models.base import (BaseAnswer,
AnswerResult)
class InvalidScoreLogitsProcessor(LogitsProcessor):
def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor) -> torch.FloatTensor:
def __call__(self, input_ids: Union[torch.LongTensor,list], scores: Union[torch.FloatTensor,list]) -> torch.FloatTensor:
# llama-cpp模型返回的是list,为兼容性考虑，需要判断input_ids和scores的类型，将list转换为torch.Tensor
input_ids = torch.tensor(input_ids) if isinstance(input_ids,list) else input_ids
scores = torch.tensor(scores) if isinstance(scores,list) else scores
if torch.isnan(scores).any() or torch.isinf(scores).any():
scores.zero_()
scores[..., 5] = 5e4
......@@ -163,7 +166,20 @@ class LLamaLLM(BaseAnswer, LLM, ABC):
self.stopping_criteria = transformers.StoppingCriteriaList()
# 观测输出
gen_kwargs.update({'stopping_criteria': self.stopping_criteria})
# llama-cpp模型的参数与transformers的参数字段有较大差异，直接调用会返回不支持的字段错误
# 因此需要先判断模型是否是llama-cpp模型，然后取gen_kwargs与模型generate方法字段的交集
# 仅将交集字段传给模型以保证兼容性
# todo llama-cpp模型在本框架下兼容性较差，后续可以考虑重写一个llama_cpp_llm.py模块
if "llama_cpp" in self.checkPoint.model.__str__():
import inspect
common_kwargs_keys = set(inspect.getfullargspec(self.checkPoint.model.generate).args)&set(gen_kwargs.keys())
common_kwargs = {key:gen_kwargs[key] for key in common_kwargs_keys}
#? llama-cpp模型的generate方法似乎只接受.cpu类型的输入，响应很慢，慢到哭泣
#?为什么会不支持GPU呢，不应该啊？
output_ids = torch.tensor([list(self.checkPoint.model.generate(input_id_i.cpu(),**common_kwargs)) for input_id_i in input_ids])
else:
output_ids = self.checkPoint.model.generate(**gen_kwargs)
new_tokens = len(output_ids[0]) - len(input_ids[0])
reply = self.decode(output_ids[0][-new_tokens:])
......
models/loader/loader.py
......@@ -67,9 +67,11 @@ class LoaderCheckPoint:
self.load_in_8bit = params.get('load_in_8bit', False)
self.bf16 = params.get('bf16', False)
def _load_model_config(self, model_name):
if self.model_path:
self.model_path = re.sub("\s","",self.model_path)
checkpoint = Path(f'{self.model_path}')
else:
if not self.no_remote_model:
......@@ -78,10 +80,12 @@ class LoaderCheckPoint:
raise ValueError(
"本地模型local_model_path未配置路径"
)
try:
model_config = AutoConfig.from_pretrained(checkpoint, trust_remote_code=True)
return model_config
except Exception as e:
print(e)
return checkpoint
def _load_model(self, model_name):
"""
......@@ -93,6 +97,7 @@ class LoaderCheckPoint:
t0 = time.time()
if self.model_path:
self.model_path = re.sub("\s","",self.model_path)
checkpoint = Path(f'{self.model_path}')
else:
if not self.no_remote_model:
......@@ -103,7 +108,7 @@ class LoaderCheckPoint:
)
self.is_llamacpp = len(list(Path(f'{checkpoint}').glob('ggml*.bin'))) > 0
if 'chatglm' in model_name.lower():
if 'chatglm' in model_name.lower() or "chatyuan" in model_name.lower():
LoaderClass = AutoModel
else:
LoaderClass = AutoModelForCausalLM
......@@ -126,8 +131,14 @@ class LoaderCheckPoint:
.half()
.cuda()
)
# 支持自定义cuda设备
elif ":" in self.llm_device:
model = LoaderClass.from_pretrained(checkpoint,
config=self.model_config,
torch_dtype=torch.bfloat16 if self.bf16 else torch.float16,
trust_remote_code=True).half().to(self.llm_device)
else:
from accelerate import dispatch_model
from accelerate import dispatch_model,infer_auto_device_map
model = LoaderClass.from_pretrained(checkpoint,
config=self.model_config,
......@@ -151,6 +162,13 @@ class LoaderCheckPoint:
dtype=torch.float16 if not self.load_in_8bit else torch.int8,
max_memory=max_memory,
no_split_module_classes=model._no_split_modules)
# 对于chaglm和moss意外的模型应使用自动指定，而非调用chatglm的配置方式
# 其他模型定义的层类几乎不可能与chatglm和moss一致，使用chatglm_auto_configure_device_map
# 百分百会报错，使用infer_auto_device_map虽然可能导致负载不均衡，但至少不会报错
# 实测在bloom模型上如此
# self.device_map = infer_auto_device_map(model,
# dtype=torch.int8,
# no_split_module_classes=model._no_split_modules)
model = dispatch_model(model, device_map=self.device_map)
else:
......@@ -166,7 +184,7 @@ class LoaderCheckPoint:
elif self.is_llamacpp:
try:
from models.extensions.llamacpp_model_alternative import LlamaCppModel
from llama_cpp import Llama
except ImportError as exc:
raise ValueError(
......@@ -177,7 +195,16 @@ class LoaderCheckPoint:
model_file = list(checkpoint.glob('ggml*.bin'))[0]
print(f"llama.cpp weights detected: {model_file}\n")
model, tokenizer = LlamaCppModel.from_pretrained(model_file)
model = Llama(model_path=model_file._str)
# 实测llama-cpp-vicuna13b-q5_1的AutoTokenizer加载tokenizer的速度极慢，应存在优化空间
# 但需要对huggingface的AutoTokenizer进行优化
# tokenizer = model.tokenizer
# todo 此处调用AutoTokenizer的tokenizer，但后续可以测试自带tokenizer是不是兼容
#* -> 自带的tokenizer不与transoformers的tokenizer兼容,无法使用
tokenizer = AutoTokenizer.from_pretrained(self.model_name)
return model, tokenizer
elif self.load_in_8bit:
......@@ -267,10 +294,21 @@ class LoaderCheckPoint:
# 在调用chat或者stream_chat时,input_ids会被放到model.device上
# 如果transformer.word_embeddings.device和model.device不同,则会导致RuntimeError
# 因此这里将transformer.word_embeddings,transformer.final_layernorm,lm_head都放到第一张卡上
encode = ""
if 'chatglm2' in self.model_name:
device_map = {
f"{layer_prefix}.embedding.word_embeddings": 0,
f"{layer_prefix}.rotary_pos_emb": 0,
f"{layer_prefix}.output_layer": 0,
f"{layer_prefix}.encoder.final_layernorm": 0,
f"base_model.model.output_layer": 0
}
encode = ".encoder"
else:
device_map = {f'{layer_prefix}.word_embeddings': 0,
f'{layer_prefix}.final_layernorm': 0, 'lm_head': 0,
f'base_model.model.lm_head': 0, }
used = 2
gpu_target = 0
for i in range(num_trans_layers):
......@@ -278,7 +316,7 @@ class LoaderCheckPoint:
gpu_target += 1
used = 0
assert gpu_target < num_gpus
device_map[f'{layer_prefix}.layers.{i}'] = gpu_target
device_map[f'{layer_prefix}{encode}.layers.{i}'] = gpu_target
used += 1
return device_map
......@@ -395,7 +433,7 @@ class LoaderCheckPoint:
print(
"如果您使用的是 macOS 建议将 pytorch 版本升级至 2.0.0 或更高版本，以支持及时清理 torch 产生的内存占用。")
elif torch.has_cuda:
device_id = "0" if torch.cuda.is_available() else None
device_id = "0" if torch.cuda.is_available() and (":" not in self.llm_device) else None
CUDA_DEVICE = f"{self.llm_device}:{device_id}" if device_id else self.llm_device
with torch.cuda.device(CUDA_DEVICE):
torch.cuda.empty_cache()
......@@ -442,5 +480,6 @@ class LoaderCheckPoint:
self.model.transformer.prefix_encoder.float()
except Exception as e:
print("加载PrefixEncoder模型参数失败")
# llama-cpp模型（至少vicuna-13b）的eval方法就是自身，其没有eval方法
if not self.is_llamacpp:
self.model = self.model.eval()
models/moss_llm.py
......@@ -6,7 +6,7 @@ from models.base import (BaseAnswer,
AnswerResult)
import torch
# todo 建议重写instruction,在该instruction下，各模型的表现比较差
META_INSTRUCTION = \
"""You are an AI assistant whose name is MOSS.
- MOSS is a conversational language model that is developed by Fudan University. It is designed to be helpful, honest, and harmless.
......@@ -20,7 +20,7 @@ META_INSTRUCTION = \
Capabilities and tools that MOSS can possess.
"""
# todo 在MOSSLLM类下，各模型的响应速度很慢，后续要检查一下原因
class MOSSLLM(BaseAnswer, LLM, ABC):
max_token: int = 2048
temperature: float = 0.7
......@@ -42,10 +42,11 @@ class MOSSLLM(BaseAnswer, LLM, ABC):
return self.checkPoint
@property
def set_history_len(self) -> int:
def _history_len(self) -> int:
return self.history_len
def _set_history_len(self, history_len: int) -> None:
def set_history_len(self, history_len: int) -> None:
self.history_len = history_len
def _call(self, prompt: str, stop: Optional[List[str]] = None) -> str:
......@@ -59,11 +60,13 @@ class MOSSLLM(BaseAnswer, LLM, ABC):
prompt_w_history = str(history)
prompt_w_history += '<|Human|>: ' + prompt + '<eoh>'
else:
prompt_w_history = META_INSTRUCTION
prompt_w_history = META_INSTRUCTION.replace("MOSS", self.checkPoint.model_name.split("/")[-1])
prompt_w_history += '<|Human|>: ' + prompt + '<eoh>'
inputs = self.checkPoint.tokenizer(prompt_w_history, return_tensors="pt")
with torch.no_grad():
# max_length似乎可以设的小一些，而repetion_penalty应大一些，否则chatyuan,bloom等模型为满足max会重复输出
#
outputs = self.checkPoint.model.generate(
inputs.input_ids.cuda(),
attention_mask=inputs.attention_mask.cuda(),
......
models/shared.py
......@@ -44,4 +44,5 @@ def loaderLLM(llm_model: str = None, no_remote_model: bool = False, use_ptuning_
if 'FastChatOpenAILLM' in llm_model_info["provides"]:
modelInsLLM.set_api_base_url(llm_model_info['api_base_url'])
modelInsLLM.call_model_name(llm_model_info['name'])
modelInsLLM.set_api_key(llm_model_info['api_key'])
return modelInsLLM
requirements.txt
......@@ -23,9 +23,13 @@ openai
#accelerate~=0.18.0
#peft~=0.3.0
#bitsandbytes; platform_system != "Windows"
#llama-cpp-python==0.1.34; platform_system != "Windows"
#https://github.com/abetlen/llama-cpp-python/releases/download/v0.1.34/llama_cpp_python-0.1.34-cp310-cp310-win_amd64.whl; platform_system == "Windows"
# 要调用llama-cpp模型，如vicuma-13b量化模型需要安装llama-cpp-python库
# but!!! 实测pip install 不好使，需要手动从ttps://github.com/abetlen/llama-cpp-python/releases/下载
# 而且注意不同时期的ggml格式并不！兼！容!!!因此需要安装的llama-cpp-python版本也不一致，需要手动测试才能确定
# 实测ggml-vicuna-13b-1.1在llama-cpp-python 0.1.63上可正常兼容
# 不过！！！本项目模型加载的方式控制的比较严格，与llama-cpp-python的兼容性较差，很多参数设定不能使用，
# 建议如非必要还是不要使用llama-cpp
torch~=2.0.0
pydantic~=1.10.7
starlette~=0.26.1
......@@ -33,5 +37,4 @@ numpy~=1.23.5
tqdm~=4.65.0
requests~=2.28.2
tenacity~=8.2.2
# 默认下载的charset_normalizer模块版本过高会抛出，`artially initialized module 'charset_normalizer' has no attribute 'md__mypyc' (most likely due to a circular import)`
charset_normalizer==2.1.0
\ No newline at end of file
test/textsplitter/test_zh_title_enhance.py 0 → 100644
from configs.model_config import *
from langchain.embeddings.huggingface import HuggingFaceEmbeddings
import nltk
from vectorstores import MyFAISS
from chains.local_doc_qa import load_file
nltk.data.path = [NLTK_DATA_PATH] + nltk.data.path
if __name__ == "__main__":
filepath = os.path.join(os.path.dirname(os.path.dirname(os.path.dirname(__file__))),
"knowledge_base", "samples", "content", "test.txt")
embeddings = HuggingFaceEmbeddings(model_name=embedding_model_dict[EMBEDDING_MODEL],
model_kwargs={'device': EMBEDDING_DEVICE})
docs = load_file(filepath, using_zh_title_enhance=True)
vector_store = MyFAISS.from_documents(docs, embeddings)
query = "指令提示技术有什么示例"
search_result = vector_store.similarity_search(query)
print(search_result)
pass
textsplitter/__init__.py
from .chinese_text_splitter import ChineseTextSplitter
from .ali_text_splitter import AliTextSplitter
from .zh_title_enhance import zh_title_enhance
\ No newline at end of file
textsplitter/zh_title_enhance.py 0 → 100644
from langchain.docstore.document import Document
import re
def under_non_alpha_ratio(text: str, threshold: float = 0.5):
"""Checks if the proportion of non-alpha characters in the text snippet exceeds a given
threshold. This helps prevent text like "-----------BREAK---------" from being tagged
as a title or narrative text. The ratio does not count spaces.
Parameters
----------
text
The input string to test
threshold
If the proportion of non-alpha characters exceeds this threshold, the function
returns False
"""
if len(text) == 0:
return False
alpha_count = len([char for char in text if char.strip() and char.isalpha()])
total_count = len([char for char in text if char.strip()])
try:
ratio = alpha_count / total_count
return ratio < threshold
except:
return False
def is_possible_title(
text: str,
title_max_word_length: int = 20,
non_alpha_threshold: float = 0.5,
) -> bool:
"""Checks to see if the text passes all of the checks for a valid title.
Parameters
----------
text
The input text to check
title_max_word_length
The maximum number of words a title can contain
non_alpha_threshold
The minimum number of alpha characters the text needs to be considered a title
"""
# 文本长度为0的话，肯定不是title
if len(text) == 0:
print("Not a title. Text is empty.")
return False
# 文本中有标点符号，就不是title
ENDS_IN_PUNCT_PATTERN = r"[^\w\s]\Z"
ENDS_IN_PUNCT_RE = re.compile(ENDS_IN_PUNCT_PATTERN)
if ENDS_IN_PUNCT_RE.search(text) is not None:
return False
# 文本长度不能超过设定值，默认20
# NOTE(robinson) - splitting on spaces here instead of word tokenizing because it
# is less expensive and actual tokenization doesn't add much value for the length check
if len(text) > title_max_word_length:
return False
# 文本中数字的占比不能太高，否则不是title
if under_non_alpha_ratio(text, threshold=non_alpha_threshold):
return False
# NOTE(robinson) - Prevent flagging salutations like "To My Dearest Friends," as titles
if text.endswith((",", ".", "，", "。")):
return False
if text.isnumeric():
print(f"Not a title. Text is all numeric:\n\n{text}") # type: ignore
return False
# 开头的字符内应该有数字，默认5个字符内
if len(text) < 5:
text_5 = text
else:
text_5 = text[:5]
alpha_in_text_5 = sum(list(map(lambda x: x.isnumeric(), list(text_5))))
if not alpha_in_text_5:
return False
return True
def zh_title_enhance(docs: Document) -> Document:
title = None
if len(docs) > 0:
for doc in docs:
if is_possible_title(doc.page_content):
doc.metadata['category'] = 'cn_Title'
title = doc.page_content
elif title:
doc.page_content = f"下文与({title})有关。{doc.page_content}"
return docs
else:
print("文件不存在")
vectorstores/MyFAISS.py
......@@ -6,6 +6,8 @@ from langchain.docstore.base import Docstore
from langchain.docstore.document import Document
import numpy as np
import copy
import os
from configs.model_config import *
class MyFAISS(FAISS, VectorStore):
......@@ -22,6 +24,9 @@ class MyFAISS(FAISS, VectorStore):
docstore=docstore,
index_to_docstore_id=index_to_docstore_id,
normalize_L2=normalize_L2)
self.score_threshold=VECTOR_SEARCH_SCORE_THRESHOLD
self.chunk_size = CHUNK_SIZE
self.chunk_conent = False
def seperate_list(self, ls: List[int]) -> List[List[int]]:
# TODO: 增加是否属于同一文档的判断
......@@ -52,7 +57,11 @@ class MyFAISS(FAISS, VectorStore):
if i == -1 or 0 < self.score_threshold < scores[0][j]:
# This happens when not enough docs are returned.
continue
if i in self.index_to_docstore_id:
_id = self.index_to_docstore_id[i]
# 执行接下来的操作
else:
continue
doc = self.docstore.search(_id)
if (not self.chunk_conent) or ("context_expand" in doc.metadata and not doc.metadata["context_expand"]):
# 匹配出的文本如果不需要扩展上下文则执行如下代码
......@@ -113,8 +122,10 @@ class MyFAISS(FAISS, VectorStore):
try:
if isinstance(source, str):
ids = [k for k, v in self.docstore._dict.items() if v.metadata["source"] == source]
vs_path = os.path.join(os.path.split(os.path.split(source)[0])[0], "vector_store")
else:
ids = [k for k, v in self.docstore._dict.items() if v.metadata["source"] in source]
vs_path = os.path.join(os.path.split(os.path.split(source[0])[0])[0], "vector_store")
if len(ids) == 0:
return f"docs delete fail"
else:
......@@ -122,6 +133,9 @@ class MyFAISS(FAISS, VectorStore):
index = list(self.index_to_docstore_id.keys())[list(self.index_to_docstore_id.values()).index(id)]
self.index_to_docstore_id.pop(index)
self.docstore._dict.pop(id)
# TODO: 从 self.index 中删除对应id
# self.index.reset()
self.save_local(vs_path)
return f"docs delete success"
except Exception as e:
print(e)
......
views/Dockerfile
......@@ -15,7 +15,7 @@ COPY . /app
RUN pnpm run build
FROM frontend AS final
COPY --from=frontend /app/dist /app/public
......
views/src/api/chat.ts
......@@ -16,6 +16,24 @@ export const chatfile = (params: any) => {
})
}
export const getKbsList = () => {
return api({
url: '/local_doc_qa/list_knowledge_base',
method: 'get',
})
}
export const deleteKb = (knowledge_base_id: any) => {
return api({
url: '/local_doc_qa/delete_knowledge_base',
method: 'delete',
params: {
knowledge_base_id,
},
})
}
export const getfilelist = (knowledge_base_id: any) => {
return api({
url: '/local_doc_qa/list_files',
......@@ -35,8 +53,8 @@ export const bing_search = (params: any) => {
export const deletefile = (params: any) => {
return api({
url: '/local_doc_qa/delete_file',
method: 'post',
data: JSON.stringify(params),
method: 'delete',
params,
})
}
export const web_url = () => {
......@@ -45,3 +63,18 @@ export const web_url = () => {
export const setapi = () => {
return window.baseApi
}
export const getkblist = (knowledge_base_id: any) => {
return api({
url: '/local_doc_qa/list_knowledge_base',
method: 'get',
params: {},
})
}
export const deletekb = (params: any) => {
return api({
url: '/local_doc_qa/delete_knowledge_base',
method: 'post',
data: JSON.stringify(params),
})
}
views/src/views/chat/index.vue
......@@ -555,7 +555,7 @@ const options = computed(() => {
return common
})
function handleSelect(key: 'copyText' | 'delete' | 'toggleRenderType') {
function handleSelect(key: string) {
if (key == '清除会话') {
handleClear()
}
......@@ -658,7 +658,6 @@ function searchfun() {
<NDropdown
v-if="isMobile"
:trigger="isMobile ? 'click' : 'hover'"
:placement="!inversion ? 'right' : 'left'"
:options="options"
@select="handleSelect"
>
......
views/src/views/chat/layout/sider/knowledge-base/index.vue
......@@ -3,15 +3,16 @@ import { NButton, NForm, NFormItem, NInput, NPopconfirm } from 'naive-ui'
import { onMounted, ref } from 'vue'
import filelist from './filelist.vue'
import { SvgIcon } from '@/components/common'
import { deletefile, getfilelist } from '@/api/chat'
import { deleteKb, getKbsList } from '@/api/chat'
import { idStore } from '@/store/modules/knowledgebaseid/id'
const items = ref<any>([])
const choice = ref('')
const store = idStore()
onMounted(async () => {
choice.value = store.knowledgeid
const res = await getfilelist({})
const res = await getKbsList()
res.data.data.forEach((item: any) => {
items.value.push({
value: item,
......@@ -52,8 +53,8 @@ const handleClick = () => {
}
}
async function handleDelete(item: any) {
await deletefile(item.value)
const res = await getfilelist({})
await deleteKb(item.value)
const res = await getKbsList()
items.value = []
res.data.data.forEach((item: any) => {
items.value.push({
......
webui.py
......@@ -218,7 +218,12 @@ def change_chunk_conent(mode, label_conent, history):
def add_vs_name(vs_name, chatbot):
if vs_name in get_vs_list():
if vs_name is None or vs_name.strip() == "" :
vs_status = "知识库名称不能为空，请重新填写知识库名称"
chatbot = chatbot + [[None, vs_status]]
return gr.update(visible=True), gr.update(visible=True), gr.update(visible=True), gr.update(
visible=False), chatbot, gr.update(visible=False)
elif vs_name in get_vs_list():
vs_status = "与已有知识库名称冲突，请重新选择其他名称后提交"
chatbot = chatbot + [[None, vs_status]]
return gr.update(visible=True), gr.update(visible=True), gr.update(visible=True), gr.update(
......
webui_st.py
......@@ -143,7 +143,7 @@ def init_model(llm_model: str = 'chat-glm-6b', embedding_model: str = 'text2vec'
# return history + [[None, model_status]]
def get_vector_store(vs_id, files, sentence_size, history, one_conent, one_content_segmentation):
def get_vector_store(local_doc_qa, vs_id, files, sentence_size, history, one_conent, one_content_segmentation):
vs_path = os.path.join(KB_ROOT_PATH, vs_id, "vector_store")
filelist = []
if not os.path.exists(os.path.join(KB_ROOT_PATH, vs_id, "content")):
......@@ -455,6 +455,8 @@ with st.sidebar:
cols = st.columns([12, 10])
kb_name = cols[0].text_input(
'新知识库名称', placeholder='新知识库名称', label_visibility='collapsed')
if 'kb_name' not in st.session_state:
st.session_state.kb_name = kb_name
cols[1].button('新建知识库', on_click=on_new_kb)
vs_path = st.selectbox(
'选择知识库', vs_list, on_change=on_vs_change, key='vs_path')
......
Markdown 格式
0% 或
您添加了 0 人 到此讨论。请谨慎行事。
请先完成此评论的编辑！
请 注册 或者 后发表评论