mirror of
https://github.com/deepset-ai/haystack.git
synced 2025-12-09 13:56:58 +00:00
f3dc9edd26
* initial ExtractiveReader implementation * initial ExtractiveReader implementation * fix mypy * remove unused import * Use AutoTokenizer * rename reader to model * combine no-answer logit * support document slicing with proper probabilities * add variable stride * validate model * fix typo * make postprocessing easier to understand * remove debug code * set default reader * add ExtractiveReader to __init__ * remove validation * use new answer class * add batching * use v2 lazy imports * move reader * fix type hints * add doc strings * add nucleus sampling * fix types * fix doc string * add no_answer parameter * remove print statement * fix gpu support * turn into binary classification task * change dataclass so document does not need to be provided for no answer * add simple tests * add unit tests * rename reader folder to readers * add integration tests * fix type hints * add release notes * remove accidentally included test file * remove unnecessary __init__ file * revert __init__ file to main * rename test script by adding test_ prefix * undo accidentally moving of test script after renaming it * remove use of bisect * rename _flatten and _unflatten * make variable name more intuitive * remove type: ignore * fix mypy issue * refactor long tuple * add doc strings * explain HF test * remove unnecessary top_k check --------- Co-authored-by: ZanSara <sara.zanzottera@deepset.ai>
8 lines
306 B
YAML
8 lines
306 B
YAML
---
|
|
preview:
|
|
- |
|
|
This adds an ExtractiveReader for v2. It should be a replacement where
|
|
FARMReader would have been used before for inference.
|
|
The confidence scores are calculated differently from FARMReader because
|
|
each span is considered to be an independent binary classification task.
|