gpt-academic / request_llms /bridge_chatglm3.py
qingxu98's picture
version 3.6
17d0a32
raw
history blame
3.63 kB
model_name = "ChatGLM3"
cmd_to_install = "`pip install -r request_llms/requirements_chatglm.txt`"
from toolbox import get_conf, ProxyNetworkActivate
from .local_llm_class import LocalLLMHandle, get_local_llm_predict_fns
# ------------------------------------------------------------------------------------------------------------------------
# πŸ”ŒπŸ’» Local Model
# ------------------------------------------------------------------------------------------------------------------------
class GetGLM3Handle(LocalLLMHandle):
def load_model_info(self):
# πŸƒβ€β™‚οΈπŸƒβ€β™‚οΈπŸƒβ€β™‚οΈ ε­θΏ›η¨‹ζ‰§θ‘Œ
self.model_name = model_name
self.cmd_to_install = cmd_to_install
def load_model_and_tokenizer(self):
# πŸƒβ€β™‚οΈπŸƒβ€β™‚οΈπŸƒβ€β™‚οΈ ε­θΏ›η¨‹ζ‰§θ‘Œ
from transformers import AutoModel, AutoTokenizer
import os, glob
import os
import platform
LOCAL_MODEL_QUANT, device = get_conf('LOCAL_MODEL_QUANT', 'LOCAL_MODEL_DEVICE')
if LOCAL_MODEL_QUANT == "INT4": # INT4
_model_name_ = "THUDM/chatglm3-6b-int4"
elif LOCAL_MODEL_QUANT == "INT8": # INT8
_model_name_ = "THUDM/chatglm3-6b-int8"
else:
_model_name_ = "THUDM/chatglm3-6b" # FP16
with ProxyNetworkActivate('Download_LLM'):
chatglm_tokenizer = AutoTokenizer.from_pretrained(_model_name_, trust_remote_code=True)
if device=='cpu':
chatglm_model = AutoModel.from_pretrained(_model_name_, trust_remote_code=True, device='cpu').float()
else:
chatglm_model = AutoModel.from_pretrained(_model_name_, trust_remote_code=True, device='cuda')
chatglm_model = chatglm_model.eval()
self._model = chatglm_model
self._tokenizer = chatglm_tokenizer
return self._model, self._tokenizer
def llm_stream_generator(self, **kwargs):
# πŸƒβ€β™‚οΈπŸƒβ€β™‚οΈπŸƒβ€β™‚οΈ ε­θΏ›η¨‹ζ‰§θ‘Œ
def adaptor(kwargs):
query = kwargs['query']
max_length = kwargs['max_length']
top_p = kwargs['top_p']
temperature = kwargs['temperature']
history = kwargs['history']
return query, max_length, top_p, temperature, history
query, max_length, top_p, temperature, history = adaptor(kwargs)
for response, history in self._model.stream_chat(self._tokenizer,
query,
history,
max_length=max_length,
top_p=top_p,
temperature=temperature,
):
yield response
def try_to_import_special_deps(self, **kwargs):
# import something that will raise error if the user does not install requirement_*.txt
# πŸƒβ€β™‚οΈπŸƒβ€β™‚οΈπŸƒβ€β™‚οΈ δΈ»θΏ›η¨‹ζ‰§θ‘Œ
import importlib
# importlib.import_module('modelscope')
# ------------------------------------------------------------------------------------------------------------------------
# πŸ”ŒπŸ’» GPT-Academic Interface
# ------------------------------------------------------------------------------------------------------------------------
predict_no_ui_long_connection, predict = get_local_llm_predict_fns(GetGLM3Handle, model_name, history_format='chatglm3')