'LocalFileSystem' object has no attribute 'makedirs'

[nkalavak]

🐛 Bug

To Reproduce

Steps to reproduce the behavior:

While working with the add_embedding method for TensorBoard (for the first time), there seems to be an issue with one of the lines in the writer.py files that is invoked. The simplest way to reproduce would be to test the Adding a "Projector" to TensorBoard section of the TensorBoard tutorial (Link: https://pytorch.org/tutorials/intermediate/tensorboard_tutorial.html#adding-a-projector-to-tensorboard)

warning: Embedding dir exists, did you set global_step for add_embedding()?
  File "training_visualization.py", line 87, in <module>
    writer.add_embedding(features, metadata=class_labels, label_img=images.unsqueeze(1))
  File "/home/nivii/Desktop/python_environments/pytorch_py36/lib/python3.6/site-packages/torch/utils/tensorboard/writer.py", line 794, in add_embedding
    make_tsv(metadata, save_path, metadata_header=metadata_header)
  File "/home/nivii/Desktop/python_environments/pytorch_py36/lib/python3.6/site-packages/torch/utils/tensorboard/_embedding.py", line 20, in make_tsv
    fs.write(fs.join(save_path, 'metadata.tsv'), metadata_bytes, binary_mode=True)
AttributeError: 'LocalFileSystem' object has no attribute 'write'

Expected behavior

No error, I presume. I am not sure where is should be setting the global_step in the case of the tutorial, while following it sequentially.

Environment

Please copy and paste the output from our
environment collection script
(or fill out the checklist below manually).

You can get the script and run it with:

wget https://raw.githubusercontent.com/pytorch/pytorch/master/torch/utils/collect_env.py
# For security purposes, please check the contents of collect_env.py before running it.
python collect_env.py

• PyTorch Version (e.g., 1.0): 1.4.0
• OS (e.g., Linux): Linux
• How you installed PyTorch (conda, pip, source): pip
• Build command you used (if compiling from source):
• Python version: 3.6
• CUDA/cuDNN version: 10.0/7/5/0
• GPU models and configuration: NVIDIA RTX2060
• Any other relevant information: Tensorboard 1.14.0

Additional context

cc @ezyang @gchanan @zou3519

[ezyang]

Marking as high priority as it seems that basic functionality in tensorboard does not seem to work

[savadhanistablab]

This worked for me:

• Uninstalled tensorflow,
• Reinstalled Tensorboard.
  PS: Restart Kernel and tensorboard.

[peterbell10]

write was added to tensorboard's LocalFileSystem stub in tensorflow/tensorboard#2260 which is included in tensorboard 1.15 and up (September 2019).

@savadhanistablab's workaround probably just updates tensorboard.

So, should we add a version check for tensorboard, or just give a better error if not hasattr(fs, 'write')?

[peterbell10]

1.14 seems to be the enforced minimum version currently

pytorch/torch/utils/tensorboard/__init__.py

Line 5 in 58f5a89

'This should be available in 1.14 or above.')

[beikejinmiao]

tensorboard==1.14.0 is not worked for me. So, I upgrade it to the newest 2.2.0.