Update README.md
Browse files
README.md
CHANGED
|
@@ -19,6 +19,8 @@ I originally used <a href="https://github.com/monsoonNLP/transformers">a modifie
|
|
| 19 |
This small model has comparable results to Multilingual BERT on <a href="https://colab.research.google.com/drive/18FQxp9QGOORhMENafQilEmeAo88pqVtP">BBC Hindi news classification</a>
|
| 20 |
and on <a href="https://colab.research.google.com/drive/1UYn5Th8u7xISnPUBf72at1IZIm3LEDWN">Hindi movie reviews / sentiment analysis</a> (using SimpleTransformers)
|
| 21 |
|
|
|
|
|
|
|
| 22 |
Question-answering on MLQA dataset: https://colab.research.google.com/drive/1i6fidh2tItf_-IDkljMuaIGmEU6HT2Ar#scrollTo=IcFoAHgKCUiQ
|
| 23 |
|
| 24 |
A larger model (<a href="https://huggingface.co/monsoon-nlp/hindi-tpu-electra">Hindi-TPU-Electra</a>) using ELECTRA base size outperforms both models on Hindi movie reviews / sentiment analysis, but
|
|
|
|
| 19 |
This small model has comparable results to Multilingual BERT on <a href="https://colab.research.google.com/drive/18FQxp9QGOORhMENafQilEmeAo88pqVtP">BBC Hindi news classification</a>
|
| 20 |
and on <a href="https://colab.research.google.com/drive/1UYn5Th8u7xISnPUBf72at1IZIm3LEDWN">Hindi movie reviews / sentiment analysis</a> (using SimpleTransformers)
|
| 21 |
|
| 22 |
+
You can get higher accuracy using ktrain by adjusting learning rate (also: changing model_type in config.json - this is an open issue with ktrain): https://colab.research.google.com/drive/1mSeeSfVSOT7e-dVhPlmSsQRvpn6xC05w?usp=sharing
|
| 23 |
+
|
| 24 |
Question-answering on MLQA dataset: https://colab.research.google.com/drive/1i6fidh2tItf_-IDkljMuaIGmEU6HT2Ar#scrollTo=IcFoAHgKCUiQ
|
| 25 |
|
| 26 |
A larger model (<a href="https://huggingface.co/monsoon-nlp/hindi-tpu-electra">Hindi-TPU-Electra</a>) using ELECTRA base size outperforms both models on Hindi movie reviews / sentiment analysis, but
|