训练时报错invalid file name: model_data/.pt

[chensisi0730]

训练总步数: 1562 每步耗时: 0.4430051675679284 最新每步耗时: 0.44242509608079467 最新每步loss 3.1989 Traceback (most recent call last): File "/data1/code/github_code/chatbot/execute.py", line 165, in train() File "/data1/code/github_code/chatbot/execute.py", line 114, in train torch.save({'modelA_state_dict': encoder.state_dict(), File "/home/tt/anaconda3/envs/py38_yolov5_5x/lib/python3.8/site-packages/torch/serialization.py", line 422, in save with _open_zipfile_writer(f) as opened_zipfile: File "/home/tt/anaconda3/envs/py38_yolov5_5x/lib/python3.8/site-packages/torch/serialization.py", line 309, in _open_zipfile_writer return container(name_or_buffer) File "/home/tt/anaconda3/envs/py38_yolov5_5x/lib/python3.8/site-packages/torch/serialization.py", line 287, in init super(_open_zipfile_writer_file, self).init(torch._C.PyTorchFileWriter(str(name))) RuntimeError: [enforce fail at inline_container.cc:354] . invalid file name: model_data/.pt

[LiMuwenan]

execute.py中checkpoint_prefix = os.path.join(checkpoint_dir, "model_data.pt")拼接的路径有问题。我创建了和config同级的model_data目录，然后在这个目录里存模型解决

[Agnes4m]

新建文件夹model_date

[kianna-ops]

可以将checkpoint_prefix = os.path.join(checkpoint_dir, "model_data.pt")改成checkpoint_prefix = checkpoint_dir + ".pt"