Jim Dowling

We would love to be able to run Tensorboard with logs in HDFS. Currently, we have to copy logs from hdfs to local disk and then run Tensorboard.

I can confirm the same issue is on Ubuntu 16.04 with nodejs v10.15.0

The guys at pepperdata said they would contribute this back, but no sign of it yet.

Hi galuszak. Thanks for this PR - It looks good.! You just need to sign the CLA and mail it to [email protected] and we can merge this: https://docs.google.com/document/d/1V8Sx9miu3-Z8KPA08eWnYXa8LRnqwvusG9LtkuQ1pBk/edit?usp=sharing

Feel free to cherry pick. 2nd commit is file formats. Original blog is from me here (see image there): https://www.logicalclocks.com/blog/guide-to-file-formats-for-machine-learning-with-feature-store

We hit this issue for a FUSE mounted client. The problem is that the FUSE mounted client didn't support mmap, which this library uses to read the .safetensors file.

This would be a really useful PR for other file systems like HDFS.

Create your API key in Hopsworks. Then, create an environment variable in Huggingface spaces called HOPSWORKS_API_KEY https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/hub/secrets.png https://huggingface.co/docs/hub/spaces-overview