chore: bump inference to 0.6.6 #1563
Conversation
- bump `unstructured-inference` to `0.6.6` - specify default model name for element detection to be `detectron2_onnx` to keep current behavior - NOTE: the updated inference package by default would use yolox as element detection model; this will be evaluated and enabled in a separated PR
This pull request includes updated ingest test fixtures. Please review and merge if appropriate. Co-authored-by: badGarnet <[email protected]>
There was a problem hiding this comment.
the changes all seem to be positive: joining sentences back together where they should be
There was a problem hiding this comment.
duplicated text element removed
There was a problem hiding this comment.
better element categories (not sure why since default model is still detectron...); better text joining
There was a problem hiding this comment.
better joining of text into coherent elements; note this does look different from yolox model output we have seen before forcing default to detectron so this shows impact from just other improvements in inference
There was a problem hiding this comment.
properly recognizing list items now
|
In general changes to fixture are due to better grouping of elements and dedupe of elements containing the same/overlap text |
Occasionally the es test can fail because the index fail to be created on the first try. Experiments show adding timeout doesn't help but add retry mitigates the issue. See history of commits in branch: yao/bump-inference-to-0.6.6 #1563 --------- Co-authored-by: ryannikolaidis <[email protected]> Co-authored-by: badGarnet <[email protected]>
unstructured-inferenceto0.6.6detectron2_onnxto keep current behavior