All commands should be run from root directory.
Figures and saved models are saved after training in their respective folder.
### Start training | Default to BasicLSTM
python -m src.training.mainGridsearch csv files are saved in gridsearch-results/ folder.
You can modify the gridsearch parameters in gridsearch_config.yml before running this command:
### Start gridsearch
python -m src.training.gridsearch| Hyperparameters | Possible values |
|---|---|
| model_type | ['BasicLSTM', 'BiLSTM', 'HybridCNNLSTM', 'HybridLSTMCNN', 'DistillBert', 'DistillBertEmotion', 'PyramidCNN', 'TinyBert'] |
| optimizer_type | ['adam', 'adamw', 'sgd'] |
| loss_criterion | ['bceloss', 'bcelosswithlogits', 'crossentropy'] |
| lr | [float] |
| epochs | [int] |
| batch_size | [int] |
| patience_es | [int] |
| scheduler_type | ['', reduce_lr_on_plateau', 'linear_schedule_with_warmup'] |
| patience_lr | [int] |
| save_condition | ['loss', 'acc'] |
| fix_length | [null or int] |
To print the model size parameters, loss and accuracy on the test set, run this command:
### Example on our best BasicLSTM trained model
python -m src.evaluation.test_save_stats --model BasicLSTM --saved_model_path saved-models/BasicLSTM_2021-12-08_01-04-25_trained_testAcc=0.7107.pth --loss_criterion bcelosswithlogits --only_test 1