Search code examples
pythontensorflowword2vectensorboard

Tensorboard Embeddings: "parsing metadata" hangs

I am trying to visualize embeddings using tensorboard but the embedding tab seems to hang on "parsing metadata".

I checked the code, metadata tsv file, and the projector_config.ptxt against the tensorboard embedding visualization tutorial ; everything seems to be correct and tensorboard is not giving me any messages in the terminal.

The code I am using to generate the embeddings and visualization can be found here.

I am running tensorflow 1.2.1 for python 2.7 with gpu support.

Solution

  • It turns out that the Tensorboard Embedding Visualization Tutorial is incorrect; the metadata_path property in projector_config.pbtxt needs to be set relative to the log directory. Otherwise, TensorBoard (the embedding projector frontend) will misleadingly halt at the "Parsing metadata" step as it searches for the metadata file in a path that does not exist.

    For more on this, see the corresponding issue on github.