Commit 82cd38ce authored by suily's avatar suily
Browse files

Update README.md

parent b1270ba3
......@@ -81,7 +81,7 @@ deepspeed0.14.2只能使用transformers4.31.0,此版本不支持k100ai使用bf
```
pip show pip #查看依赖库安装地址site-packages
1、site-packages/transformers/utils/import_utils.py,修改def is_torch_bf16_gpu_available():
1、site-packages/transformers/utils/import_utils.py,修改def is_torch_bf16_gpu_available()277行后:
...
#TODO:if torch.cuda.is_available() and torch.version.cuda is not None:
if torch.cuda.is_available():
......@@ -93,8 +93,9 @@ if torch.cuda.is_available():
return False
else:
return False
return True
2、site-packages/transformers/utils/import_utils.py,修改def is_torch_tf32_available():
2、site-packages/transformers/utils/import_utils.py,修改def is_torch_tf32_available()322行后:
...
#TODO:if not torch.cuda.is_available() or torch.version.cuda is None:
if not torch.cuda.is_available():
......@@ -181,7 +182,7 @@ huggingface-cli download --resume-download THUDM/chatglm3-6b --local-dir checkpo
```
以基于VTimeLLM-7B的VTimeLLM为例,模型目录结构如下:
```
VTimeLLM:
VTimeLLM/checkpoints:
── clip
│   └── ViT-L-14.pt
── vicuna-7b-v1.5
......@@ -190,6 +191,7 @@ VTimeLLM:
以基于Vicuna v1.5的VTimeLLM为例,训练运行代码:
```
# 若使用小数据集需修改sh文件,例如per_device_train_batch_size=gradient_accumulation_steps=2【per_device_train_batch_size*n_gpus<10不能超过小数据集的size】
cd VTimeLLM
wandb off
sh scripts/stage1.sh
......@@ -210,7 +212,7 @@ tar -xzvf vtimellm-vicuna-v1-5-7b.tar.gz
```
以基于Vicuna v1.5的VTimeLLM为例,模型目录结构如下:
```
VTimeLLM:
VTimeLLM/checkpoints:
── clip
│   └── ViT-L-14.pt
── vtimellm-vicuna-v1-5-7b-stage1
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment