Scott Lundberg

Fix bad WOE scores.

3

I found that the WOEEncoder was not giving the right scores, and it was because of the the agg stats not being correct from pandas. The stats did work if...

The code is currently based all around files, which makes working in memory tricky. This adds a sklearn-style API wrapper for easy use in things like notebooks.

It does not look like this package supports sample weighting. But am I missing something? Seems like it would be useful to clarify this to prevent others from digging around...

[BUG] layer_past is ignored by DeepSpeedSelfAttention's compute_attention

5

**Describe the bug** The past_key_values provided to the model are ignored by DeepSpeedSelfAttention. **To Reproduce** Call a deep speed inference model with past_key_values and note that at the following line...

Now using pyparsing and supporting infix notation!

1

This is a complete rewrite of the underlying parser of the Guidance language. It now uses pyparsing which is more feature rich and widely used compared to parsimonious. It is...

the title says it all. Right now we need to convert things to lists...

This is a bigger issue centered on how to do Alias objects with sparse representations, but right now SHAP explanations often need per-row feature names and output names, but since...

eachline() reports extra line in GZip file but not in unzipped file

10

I found an issue where eachline() was returning an extra empty line "" after the end of a gz file I was reading. The file ends in a single newline,...

This removes the need for the user to set the `stateless=True` flag for pure grammar functions. Now we attempt to treat all functions like pure grammar functions using tracing, and...