search

Knowledge-Graph-Hub / neat-ml

Network Embedding All the Things
BSD 3-Clause "New" or "Revised" License
18 stars 1 forks source link

Updates for newest grape #82

Closed caufieldjh closed 2 years ago

caufieldjh commented 2 years ago

See linked issue (#81).

caufieldjh commented 2 years ago

Getting an ImportError:

 ImportError: cannot import name 'LinkPredictionTransformer' from 'grape' (/opt/hostedtoolcache/Python/3.8.12/x64/lib/python3.8/site-packages/grape/__init__.py)

Where are the transformers now? (They're in grape.transformers, and LinkPredictionTransformer is EdgePredictionTransformer

caufieldjh commented 2 years ago

Not entirely sure what to do with generating BERT embeddings for now - neat_schema doesn't include params for it, and not sure what is or isn't in Embiggen to support it

caufieldjh commented 2 years ago

A work in progress note:

  File "/home/harry/neat-env/lib/python3.8/site-packages/neat/cli.py", line 84, in run
    train_data, validation_data = model.make_train_valid_data(
  File "/home/harry/neat-env/lib/python3.8/site-packages/neat/link_prediction/model.py", line 93, in make_train_valid_data
    graph_args["directed"] = training_graph_args["directed"]
TypeError: 'str' object does not support item assignment

Looks like something that should be a dict isn't one - not sure what's happening yet

caufieldjh commented 2 years ago

This now appears to work as expected, with the sole caveat that the edge prediction methods have not been replaced with the grape methods - they are still our own sklearn/tf methods.

I'd like to split that part of the update into its own PR so we can complete the name change and get NEAT on Pypi.