使用CodeShell-7B-Chat-int4报错

[safehumeng]

代码如下

import time import torch from transformers import AutoModelForCausalLM, AutoTokenizer from transformers import GenerationConfig

model_dir = "./CodeShell-7B-Chat-int4" device = torch.device('cpu') model = AutoModelForCausalLM.from_pretrained(model_dir, trust_remote_code=True).to(device) tokenizer = AutoTokenizer.from_pretrained(model_dir)

history = [] query = """""" response = model.chat(query, history, tokenizer) print(response)

报错：

[2023-12-06 16:26:08,981] [INFO] [real_accelerator.py:158:get_accelerator] Setting ds_accelerator to cuda (auto detect) Traceback (most recent call last): File "/data-ext/codeshell/demo.py", line 8, in model = AutoModelForCausalLM.from_pretrained(model_dir, trust_remote_code=True).to(device) File "/data-ext/anaconda3/lib/python3.9/site-packages/transformers/modeling_utils.py", line 2271, in to return super().to(*args, *kwargs) File "/data-ext/anaconda3/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1145, in to return self._apply(convert) File "/data-ext/anaconda3/lib/python3.9/site-packages/torch/nn/modules/module.py", line 797, in _apply module._apply(fn) File "/data-ext/anaconda3/lib/python3.9/site-packages/torch/nn/modules/module.py", line 797, in _apply module._apply(fn) File "/data-ext/anaconda3/lib/python3.9/site-packages/torch/nn/modules/module.py", line 797, in _apply module._apply(fn) [Previous line repeated 2 more times] File "/data-ext/anaconda3/lib/python3.9/site-packages/torch/nn/modules/module.py", line 820, in _apply param_applied = fn(param) File "/data-ext/anaconda3/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1143, in convert return t.to(device, dtype if t.is_floating_point() or t.is_complex() else None, non_blocking) File "/root/.cache/huggingface/modules/transformers_modules/CodeShell-7B-Chat-int4/quantizer.py", line 40, in Params4bitTo new_param = Params4bit(self.to(device=device, dtype=dtype, non_blocking=non_blocking), File "/root/.cache/huggingface/modules/transformers_modules/CodeShell-7B-Chat-int4/quantizer.py", line 40, in Params4bitTo new_param = Params4bit(self.to(device=device, dtype=dtype, non_blocking=non_blocking), File "/root/.cache/huggingface/modules/transformers_modules/CodeShell-7B-Chat-int4/quantizer.py", line 40, in Params4bitTo new_param = Params4bit(self.to(device=device, dtype=dtype, non_blocking=non_blocking), [Previous line repeated 986 more times] File "/root/.cache/huggingface/modules/transformers_modules/CodeShell-7B-Chat-int4/quantizer.py", line 31, in Params4bitTo device, dtype, non_blocking, convert_to_format = torch._C._nn._parse_to(args, **kwargs) RecursionError: maximum recursion depth exceeded while calling a Python object