Hugging Face's logo

deepsense-ai
/
trelbert

Fill-Mask
Transformers
PyTorch
Polish
bert
Inference Endpoints
Model card Files Community
2
dsai-artur-zygadlo commited on
Commit
dd48ac3
•
1 Parent(s): 10cf51c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +2 -2
README.md CHANGED
@@ -41,13 +41,13 @@ We fine-tuned TrelBERT to [KLEJ benchmark](klejbenchmark.com) tasks and achieved
41
  |NKJP-NER|94.4|
42
  |CDSC-E|93.9|
43
  |CDSC-R|93.6|
44
- |CBD|75.2|
45
  |PolEmo2.0-IN|89.3|
46
  |PolEmo2.0-OUT|78.1|
47
  |DYK|67.4|
48
  |PSC|95.7|
49
  |AR|86.1|
50
- |__avg__|__86.0__|
51
 
52
  For fine-tuning to KLEJ tasks we used [Polish RoBERTa](https://github.com/sdadas/polish-roberta) scripts, which we modified to use `transformers` library. In the CBD task, we set the maximum sequence length to 128 and implemented the same preprocessing procedure as in the MLM phase.
53
 
 
41
  |NKJP-NER|94.4|
42
  |CDSC-E|93.9|
43
  |CDSC-R|93.6|
44
+ |CBD|76.1|
45
  |PolEmo2.0-IN|89.3|
46
  |PolEmo2.0-OUT|78.1|
47
  |DYK|67.4|
48
  |PSC|95.7|
49
  |AR|86.1|
50
+ |__avg__|__86.1__|
51
 
52
  For fine-tuning to KLEJ tasks we used [Polish RoBERTa](https://github.com/sdadas/polish-roberta) scripts, which we modified to use `transformers` library. In the CBD task, we set the maximum sequence length to 128 and implemented the same preprocessing procedure as in the MLM phase.
53