fastforwardlabs / train_embeddings_with_gensim

Train Gensim's Word2Vec algorithm to learn item embeddings for downstream tasks such as recommendation systems. Includes techniques for hyperparameter optimization and early stopping.
Apache License 2.0
2 stars 1 forks source link

AMP Review 2 #1

Open andrewrreed opened 3 years ago

andrewrreed commented 3 years ago

.project-metadata.yaml

train_gensim_word2vec.ipynb

README.md

Copyright source headers

These dang things again! Just need to add them to the top of these modules:

Misc.

andrewrreed commented 3 years ago

@melaniebeck This is an awesome, densely-packed notebook! Really enjoyed working through this and learned quite a bit about Gensim and Ray because of it. A few minor fixes documented above, then should be good to go. Lemme know if you have any questions - thanks!

cjwallace commented 3 years ago

On the runtime image, my understanding is this AMP strictly requires Python >=3.8, and providing a runtime spec (+ maybe some words in the README) is the only way we can convey that, absent any real dependency management system.

cjwallace commented 3 years ago

We could change https://github.com/fastforwardlabs/train_embeddings_with_gensim/blob/2c94ca2b45cff18004142b25028373061983fa86/.project-metadata.yaml#L10 to JupyterLab, if a notebook is the main entry point to the project.

(edit: but it won't work until the next release of CML)