Unverified 提交 ee7285cd 作者: Zhi-guo Huang 提交者: GitHub

在args.py中增加ptuning相关的参数 (#838)

* 修复 bing_search.py的typo;更新model_config.py中Bing Subscription Key申请方式及注意事项

* 更新FAQ,增加了[Errno 110] Connection timed out的原因与解决方案

* 修改loader.py中load_in_8bit失败的原因和详细解决方案

* update loader.py

* stream_chat_bing

* 修改stream_chat的接口,在请求体中选择knowledge_base_id;增加stream_chat_bing接口

* 优化cli_demo.py的逻辑:支持 输入提示;多输入;重新输入

* update cli_demo.py

* 按照review建议进行修改

* 修改默认的多卡部署方案,基本保证针对新模型也不会失败

* 测试openai接口成功

* add ptuning-v2 dir

* 支持命令行输入ptuning路径

* 在FAQ中给出加载量化版本失败的原因和解决方案

* print error

* udpate

* Update args.py

* debug for fastchat_openai_llm

* temporarily save

* update faq for

---------

Co-authored-by: imClumsyPanda <littlepanda0716@gmail.com>
Co-authored-by: zg h <bj wang@hzg0601-acer.hundsun.com>
上级 f88bf2cb
...@@ -87,6 +87,12 @@ llm_model_dict = { ...@@ -87,6 +87,12 @@ llm_model_dict = {
"local_model_path": None, "local_model_path": None,
"provides": "MOSSLLMChain" "provides": "MOSSLLMChain"
}, },
"moss-int4": {
"name": "moss",
"pretrained_model_name": "fnlp/moss-moon-003-sft-int4",
"local_model_path": None,
"provides": "MOSSLLM"
},
"vicuna-13b-hf": { "vicuna-13b-hf": {
"name": "vicuna-13b-hf", "name": "vicuna-13b-hf",
"pretrained_model_name": "vicuna-13b-hf", "pretrained_model_name": "vicuna-13b-hf",
...@@ -149,6 +155,15 @@ llm_model_dict = { ...@@ -149,6 +155,15 @@ llm_model_dict = {
"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" "api_key": "EMPTY"
}, },
# 通过 fastchat 调用的模型请参考如下格式
"fastchat-chatglm-6b-int4": {
"name": "chatglm-6b-int4", # "name"修改为fastchat服务中的"model_name"
"pretrained_model_name": "chatglm-6b-int4",
"local_model_path": None,
"provides": "FastChatOpenAILLMChain", # 使用fastchat api时,需保证"provides"为"FastChatOpenAILLMChain"
"api_base_url": "http://localhost:8001/v1", # "name"修改为fastchat服务中的"api_base_url"
"api_key": "EMPTY"
},
"fastchat-chatglm2-6b": { "fastchat-chatglm2-6b": {
"name": "chatglm2-6b", # "name"修改为fastchat服务中的"model_name" "name": "chatglm2-6b", # "name"修改为fastchat服务中的"model_name"
"pretrained_model_name": "chatglm2-6b", "pretrained_model_name": "chatglm2-6b",
...@@ -173,7 +188,7 @@ llm_model_dict = { ...@@ -173,7 +188,7 @@ llm_model_dict = {
# 如果报出:raise NewConnectionError( # 如果报出:raise NewConnectionError(
# urllib3.exceptions.NewConnectionError: <urllib3.connection.HTTPSConnection object at 0x000001FE4BDB85E0>: # urllib3.exceptions.NewConnectionError: <urllib3.connection.HTTPSConnection object at 0x000001FE4BDB85E0>:
# Failed to establish a new connection: [WinError 10060] # Failed to establish a new connection: [WinError 10060]
# 则是因为内地和香港的IP都被OPENAI封了,需要切换为日本、新加坡等地 # 则是因为内地和香港的IP都被OPENAI封了,需要切换为日本、新加坡等地
"openai-chatgpt-3.5": { "openai-chatgpt-3.5": {
"name": "gpt-3.5-turbo", "name": "gpt-3.5-turbo",
"pretrained_model_name": "gpt-3.5-turbo", "pretrained_model_name": "gpt-3.5-turbo",
...@@ -186,7 +201,7 @@ llm_model_dict = { ...@@ -186,7 +201,7 @@ llm_model_dict = {
} }
# LLM 名称 # LLM 名称
LLM_MODEL = "chatglm-6b" LLM_MODEL = "fastchat-chatglm-6b-int4"
# 量化加载8bit 模型 # 量化加载8bit 模型
LOAD_IN_8BIT = False LOAD_IN_8BIT = False
# Load the model with bfloat16 precision. Requires NVIDIA Ampere GPU. # Load the model with bfloat16 precision. Requires NVIDIA Ampere GPU.
...@@ -203,7 +218,7 @@ STREAMING = True ...@@ -203,7 +218,7 @@ STREAMING = True
# Use p-tuning-v2 PrefixEncoder # Use p-tuning-v2 PrefixEncoder
USE_PTUNING_V2 = False USE_PTUNING_V2 = False
PTUNING_DIR='./ptuing-v2'
# LLM running device # LLM running device
LLM_DEVICE = "cuda" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu" LLM_DEVICE = "cuda" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu"
......
...@@ -177,3 +177,22 @@ download_with_progressbar(url, tmp_path) ...@@ -177,3 +177,22 @@ download_with_progressbar(url, tmp_path)
Q14 调用api中的 `bing_search_chat`接口时,报出 `Failed to establish a new connection: [Errno 110] Connection timed out` Q14 调用api中的 `bing_search_chat`接口时,报出 `Failed to establish a new connection: [Errno 110] Connection timed out`
这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--! 这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--!
---
Q15 加载chatglm-6b-int8或chatglm-6b-int4抛出 `RuntimeError: Only Tensors of floating point andcomplex dtype can require gradients`
疑为chatglm的quantization的问题或torch版本差异问题,针对已经变为Parameter的torch.zeros矩阵也执行Parameter操作,从而抛出 `RuntimeError: Only Tensors of floating point andcomplex dtype can require gradients`。解决办法是在chatglm-项目的原始文件中的quantization.py文件374行改为:
```
try:
self.weight =Parameter(self.weight.to(kwargs["device"]), requires_grad=False)
except Exception as e:
pass
```
如果上述方式不起作用,则在.cache/hugggingface/modules/目录下针对chatglm项目的原始文件中的quantization.py文件执行上述操作,若软链接不止一个,按照错误提示选择正确的路径。
注:虽然模型可以顺利加载但在cpu上仍存在推理失败的可能:即针对每个问题,模型一直输出gugugugu。
因此,最好不要试图用cpu加载量化模型,原因可能是目前python主流量化包的量化操作是在gpu上执行的,会天然地存在gap。
...@@ -245,7 +245,7 @@ if __name__ == "__main__": ...@@ -245,7 +245,7 @@ if __name__ == "__main__":
chain = FastChatOpenAILLMChain() chain = FastChatOpenAILLMChain()
chain.set_api_key("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") chain.set_api_key("EMPTY")
# chain.set_api_base_url("https://api.openai.com/v1") # chain.set_api_base_url("https://api.openai.com/v1")
# chain.call_model_name("gpt-3.5-turbo") # chain.call_model_name("gpt-3.5-turbo")
......
import argparse import argparse
import os import os
from configs.model_config import * from configs.model_config import *
...@@ -43,7 +44,8 @@ parser.add_argument('--no-remote-model', action='store_true', help='remote in th ...@@ -43,7 +44,8 @@ parser.add_argument('--no-remote-model', action='store_true', help='remote in th
parser.add_argument('--model-name', type=str, default=LLM_MODEL, help='Name of the model to load by default.') parser.add_argument('--model-name', type=str, default=LLM_MODEL, help='Name of the model to load by default.')
parser.add_argument('--lora', type=str, help='Name of the LoRA to apply to the model by default.') parser.add_argument('--lora', type=str, help='Name of the LoRA to apply to the model by default.')
parser.add_argument("--lora-dir", type=str, default=LORA_DIR, help="Path to directory with all the loras") parser.add_argument("--lora-dir", type=str, default=LORA_DIR, help="Path to directory with all the loras")
parser.add_argument('--use-ptuning-v2',type=str,default=USE_PTUNING_V2,help="whether use ptuning-v2 checkpoint")
parser.add_argument("--ptuning-dir",type=str,default=PTUNING_DIR,help="the dir of ptuning-v2 checkpoint")
# Accelerate/transformers # Accelerate/transformers
parser.add_argument('--load-in-8bit', action='store_true', default=LOAD_IN_8BIT, parser.add_argument('--load-in-8bit', action='store_true', default=LOAD_IN_8BIT,
help='Load the model with 8-bit precision.') help='Load the model with 8-bit precision.')
......
...@@ -454,6 +454,7 @@ class LoaderCheckPoint: ...@@ -454,6 +454,7 @@ class LoaderCheckPoint:
self.model_config.pre_seq_len = prefix_encoder_config['pre_seq_len'] self.model_config.pre_seq_len = prefix_encoder_config['pre_seq_len']
self.model_config.prefix_projection = prefix_encoder_config['prefix_projection'] self.model_config.prefix_projection = prefix_encoder_config['prefix_projection']
except Exception as e: except Exception as e:
print(e)
print("加载PrefixEncoder config.json失败") print("加载PrefixEncoder config.json失败")
self.model, self.tokenizer = self._load_model() self.model, self.tokenizer = self._load_model()
...@@ -471,6 +472,7 @@ class LoaderCheckPoint: ...@@ -471,6 +472,7 @@ class LoaderCheckPoint:
self.model.transformer.prefix_encoder.load_state_dict(new_prefix_state_dict) self.model.transformer.prefix_encoder.load_state_dict(new_prefix_state_dict)
self.model.transformer.prefix_encoder.float() self.model.transformer.prefix_encoder.float()
except Exception as e: except Exception as e:
print(e)
print("加载PrefixEncoder模型参数失败") 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 # 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: if not self.is_llamacpp:
......
Markdown 格式
0%
您添加了 0 到此讨论。请谨慎行事。
请先完成此评论的编辑!
注册 或者 后发表评论