kaggle-contradictory-my-dear-watson

These are the notebooks that:

fine tune a Bert model,
fine tune a RoBERTa model,
add a Global Average Pooling to an xlm-roberta-large model output hidden layer and fine tune.

Datasets used for fine tuning are glue/mnli, and xnli.

Name		Name	Last commit message	Last commit date
Latest commit History 16 Commits
MyTrainer.ipynb		MyTrainer.ipynb
README.md		README.md
manuallyTrain_sentPairs_bert_score64%.ipynb		manuallyTrain_sentPairs_bert_score64%.ipynb
pytorch_gap_xlmroberta_multiple_datasets.ipynb		pytorch_gap_xlmroberta_multiple_datasets.ipynb
train_automatically_bert_on_multiple_datasets.ipynb		train_automatically_bert_on_multiple_datasets.ipynb
train_manually_bert_on_multiple_datasets.ipynb		train_manually_bert_on_multiple_datasets.ipynb
train_manually_roberta_on_multiple_datasets.ipynb		train_manually_roberta_on_multiple_datasets.ipynb

Provide feedback