Grigory Sizov

Add Multinomial and Bernoulli Naive Bayes algorithms

3

Currently `linfa-bayes` crate contains Gaussian Naive Bayes algorithm. It should not be very difficult to add other kinds of Naive Bayes present in `sklearn`: - [x] Multinomial Naive Bayes -...

Enable depth_k != depth_v in local_attention_2d and masked_local_attention_2d

2

This PR resolves a [TODO in tensor2tensor/layers/common_attention_test.py](https://github.com/tensorflow/tensor2tensor/blob/0b3ee79c460ac5b0d5bb32b9e48001a933cdc5db/tensor2tensor/layers/common_attention_test.py#L520) : enable `depth_v != depth_k` for `common_attention.local_attention_2d` and `common_attention.masked_local_attention_2d` Modification is simple: one just needs to alter the shape passed to `scatter_blocks_2d` when...

Fix symbolic shape deduplication

5

Summary: When deduplicating symbolic shapes, we need iterate over shapes of tensors in the graph, as well as over shapes in input/output tensor accessors, in case some of them are...

Update profiler cache

1

Summary: Bump profiler remote cache version. Differential Revision: D47208662

Fix +/-inf in LSE returned by forward

1

Forward op was returning `+inf` in LSE for queries which have no keys to attend to, e.g. when K/V length happens to be 0. This diverges from the definition of...