MODEL_PATH='<path>'
model = AutoModelForCausalLM.from_pretrained(
MODEL_PATH,
low_cpu_mem_usage=True,
trust_remote_code=True,
torch_dtype=torch.float16,
device_map="auto"
)
报错大概是这个错误栈,transformers的错误栈:
File "/home/lichengjie/workspace/inference/xinference/model/llm/pytorch/glm4v.py", line 87, in load
model = AutoModelForCausalLM.from_pretrained(
File "/home/lichengjie/miniconda3/envs/xinf/lib/python3.10/site-packages/transformers/models/auto/auto_factory.py", line 550, in from_pretrained
model_class = get_class_from_dynamic_module(
File "/home/lichengjie/miniconda3/envs/xinf/lib/python3.10/site-packages/transformers/dynamic_module_utils.py", line 501, in get_class_from_dynamic_module
final_module = get_cached_module_file(
File "/home/lichengjie/miniconda3/envs/xinf/lib/python3.10/site-packages/transformers/dynamic_module_utils.py", line 326, in get_cached_module_file
modules_needed = check_imports(resolved_module_file)
File "/home/lichengjie/miniconda3/envs/xinf/lib/python3.10/site-packages/transformers/dynamic_module_utils.py", line 181, in check_imports
raise ImportError(
ImportError: [address=[0.0.0.0:33683](http://0.0.0.0:33683/), pid=1938627] This modeling file requires the following packages that were not found in your environment: flash_attn. Run `pip install flash_attn`
System Info / 系統信息
Python 3.10 Cuda 12.1
Who can help? / 谁可以帮助到您?
@zRzRzRzRzRzRzR
Information / 问题信息
Reproduction / 复现过程
报错大概是这个错误栈,transformers的错误栈:
Expected behavior / 期待表现
没安装flash attention的库就走原来的逻辑。 这个库windows平台用不了。