Daniël de Kok

Add `SparseLinear_v2`, fixing indexing issues

10

Introduce `SparseLinear_v2` to fix indexing issues `SparseLinear` does not correctly index the gradient/weight matrix (#752). This change fixes the indexing, so that the full matrix is used. To retain compatibility...

Reported in https://github.com/explosion/spaCy/discussions/11379 The weight matrix is allocated as follows: ``` model.set_param("W", model.ops.alloc((nO * length,), dtype="f")) ``` But indexed like this: ``` scores[clas] += weights[idx1 + clas] * value scores[clas]...

The output of a transformer is passed through in two different ways: - Prediction: the data is passed through the `Doc._.trf_data` attribute. - Training: the data is broadcast directly to...

Go to folder

3

Thanks for the fantastic plugin! It would be great if it supported GMail's 'gl' as well to search and open a folder. I guess that 'g' would be a fine...

NER: Ensure zero-cost sequence with sentence split in entity

2

## Description If we use a sentence splitter as one of the annotating components during training, an entity can become split in the predicted `Doc`. Before this change, training would...

## Description Fix an off-by-one in `TransitionModel.forward`, where we always did one move more than the maximum number of moves. This explosed another issue: when creating cut states, we skipped...

## Description Merge `master` into `v9`. **Warning:** do not squash. ### Types of change Maintenance ## Checklist - [x] I confirm that I have the right to submit this contribution...

This issue documents the current state of support for Apple M-series (M1/M2) support in Thinc and spacy-transformers. Subscribe to this issue if you'd like to receive status updates on support...

While at it: - Add CI tests for Python 3.12. - Remove 3.8 support in preparation of spaCy v4.

Distillation data is usually very large, add support for reading compressed text. Related: #107