pickle module ：EOFError Ran out of input

[WanShaw]

when I want to use the model of vit_tiny_patch16_224 from flowvison module ,it prompt this EOFError: Ran out of input.
环境就是OneFlow实训平台的3090显卡：oneflow-0.7.0+torch-1.8.1-cu11.2-cudnn8

[shangguanshiyuan]

您好，我在oneflow 0.7.0和0.8.0的版本未能复现您的问题，是否方便提供更多的信息，比如出错的代码片段等，谢谢。

[WanShaw]

稍等一下

[WanShaw]

File "of_train.py", line 10, in
model = Vit_t_16_224_pre(num_classes=10, pretrained=True)
File "/workspace/of_Vit.py", line 6, in Vit_t_16_224_pre
model = vit_tiny_patch16_224(pretrained=pretrained)
File "/usr/local/miniconda3/lib/python3.7/site-packages/flowvision/models/vision_transformer.py", line 412, in vit_tiny_patch16_224
"vit_tiny_patch16_224", pretrained=pretrained, progress=progress, **model_kwargs
File "/usr/local/miniconda3/lib/python3.7/site-packages/flowvision/models/vision_transformer.py", line 382, in _create_vision_transformer
state_dict = load_state_dict_from_url(model_urls[arch], progress=progress)
File "/usr/local/miniconda3/lib/python3.7/site-packages/flowvision/models/utils.py", line 133, in load_state_dict_from_url
state_dict = flow.load(pretrained_weight_dir)
File "/usr/local/miniconda3/lib/python3.7/site-packages/oneflow/framework/check_point_v2.py", line 322, in load
res = pickle.loads(pickle_bytes)
EOFError: Ran out of input

[WanShaw]

单独执行的话，也会出现Segmentation fault (core dumped)

[shangguanshiyuan]

您好，这个of_train.py、of_Vit.py是您自己写的还是实训平台提供的？

[WanShaw]

自己实现的，调用模型代码
def Vit_t_16_224_pre(num_classes=10, pretrained=True):
from flowvision.models import vit_tiny_patch16_224
model = vit_tiny_patch16_224(pretrained=pretrained)
model.head = nn.Linear(in_features=192, out_features=num_classes)
return model

[WanShaw]

不好意思了，在训练文件里直接调用就没事了，谢谢了

[shangguanshiyuan]

好的，未能定位到您的问题，如果您已经解决了的话，这个issue稍后将会被关闭，如果遇到了其他问题也欢迎随时反馈。