Mark Daoust

Results 254 comments of Mark Daoust

Basic Recommender Ranking: Big Optimization issue

It looks like the model is undertrained, and just learning the average. It you train it longer, it learns a similar distribution to the test labels. It's easy to overfit....

Basic Recommender Ranking: Big Optimization issue

Yes. Broader is exactly what I'd expect since this is being approached as a regression. MSE loss doesn't encourage sharpness, it tends to give a blurry result. It doesn't meant...

Basic Recommender Ranking: Big Optimization issue

Thanks for the PR! LGTM. Don't forget to follow the CLA instructions the bot posted: https://github.com/tensorflow/recommenders/pull/593#issuecomment-1352734325 We can't merge the PR unless you do.

Basic Ranking - Predictions Histogram Distribution

@maciejkula, I like this change. Can you take a look and merge?

fixed errors - converted train_features and test_features to tensor, …

Hi, thanks for pointing this out and taking the time to send a PR. I always prefer referencing columns by name instead of index, how about we do this: https://github.com/gitshiva/tensorflow-docs/pull/2

fix: update g3doc links

Thanks!

fix initial_epoch in fine tuning step

Thanks!

Fix to version numbering so that it is in accordance with PEP 404

Thanks Mark. I'll fix it in the merge.

Fixing deprecated keyword lr by new learning_rate

Thanks!

Ensure async clients work over REST transport

Also reported here: https://github.com/google-gemini/generative-ai-python/issues/499