Bastards: IDs are off

[JohnTigue]

DanC noticed that most of the bastards in the *.ipynb are off. That's turns out to be because I filtered out the hypedAFs, and then used the resulting indexes as IDs. Derp.

[JohnTigue]

There is no non-hacky simple solution. The correct answer is to go to the next level of sophistication which is adding in a metadata.tsv file. That was always the plan (esp. for #12) and now this bug has forced the issue.

[JohnTigue]

Whelp, now things are complete broke. TBProj is stalling at "fetching tensor values" and then craps out.

[JohnTigue]

My internet is pretty crappy tonight. Maybe?

[JohnTigue]

Naw, second net connection, still no workie.

[JohnTigue]

metadata.tsv is pretty simple looking. This code is TF1 unfortunately (not TF2), but gives the idea: https://github.com/normanheckscher/mnist-tensorboard-embeddings/blob/master/mnist_t-sne.py

[JohnTigue]

The TF doc, Visualizing Data using the Embedding Projector in TensorBoard, has some info:

# Save Labels separately on a line-by-line manner.
with open(os.path.join(log_dir, 'metadata.tsv'), "w") as f:
  for subwords in encoder.subwords:
    f.write("{}\n".format(subwords))
  # Fill in the rest of the labels with "unknown".
  for unknown in range(1, encoder.vocab_size - len(encoder.subwords)):
    f.write("unknown #{}\n".format(unknown))
...
embedding.metadata_path = 'metadata.tsv'

[JohnTigue]

This is preventing further publishing of Bastards analysis.