amir22010's picture
Update README.md
5b0ac78
metadata
license: cc-by-nc-4.0

allenai/tk-instruct-base-def-pos model fined tuned on custom pyabsa english model from 'ATEPC_ENGLISH_CHECKPOINT' auto annotated hospital reviews [keywords,polarity] dataset.

Training Results:

Epoch|Training Loss|Validation Loss

1|0.150800|0.112104

2|0.103700|0.074400

3|0.074500|0.065034

4|0.065900|0.062676

5|0.059700|0.058017

6|0.050700|0.056635

7|0.046500|0.056589

8|0.046300|0.056924


Evaluation Results:

Train Precision: 0.9516791241183282

Train Recall: 0.9584393553859203

Train F1: 0.9550472769531456

Test Precision: 0.9533479692645445

Test Recall: 0.9540461369461736

Test F1: 0.953696925329429

UnseenTest Precision: 0.9549444119368051

UnseenTest Recall: 0.9538281706604325

UnseenTest F1: 0.9543859649122807