SimpleT5 Versions Save

SimpleT5 v0.1.4

• Added support for all the pytorch-lightning supported loggers #11
• Added support to save model only at last epoch #21
• Added dataloader_num_workers parameter to .train( ) method to specify number of worker in train/test/val dataloader #19
• fixed warnings and made compatible with latest transformers and pytorch-lightning

byt5 support, transformers upgrade, dropping ONNX support

SimpleT5 - version 0.1.0

• Supports ByT5 model - Thanks to @mapmeld for his contribution

from simplet5 import SimpleT5
model = SimpleT5()
model.from_pretrained("byt5", "google/byt5-small")

• Added precision flag to support mixed precision training

# train
model.train(train_df=train_df, # pandas dataframe with 2 columns: source_text & target_text
            eval_df=eval_df, # pandas dataframe with 2 columns: source_text & target_text
            source_max_token_len = 512, 
            target_max_token_len = 128,
            batch_size = 8,
            max_epochs = 5,
            use_gpu = True,
            outputdir = "outputs",
            early_stopping_patience_epochs = 0,
            precision = 32
            )