T5-base-finetuned-mrpc

This model is T5 fine-tuned on GLUE MRPC dataset. It acheives the following results on the validation set

• Accuracy: 0.8922

Model Details

T5 is an encoder-decoder model pre-trained on a multi-task mixture of unsupervised and supervised tasks and for which each task is converted into a text-to-text format.

Training procedure

Tokenization

Since, T5 is a text-to-text model, the labels of the dataset are converted as follows: For each example, a sentence as been formed as "mrpc sentence1: " + mrpc_sentence1 + "sentence 2: " + mrpc_sentence2 and fed to the tokenizer to get the input_ids and attention_mask. For each label, label is choosen as "equivalent" if label is 1, else label is "not_equivalent" and tokenized to get input_ids and attention_mask . During training, these inputs_ids having pad token are replaced with -100 so that loss is not calculated for them. Then these input ids are given as labels, and above attention_mask of labels is given as decoder attention mask.

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 3e-4
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: epsilon=1e-08
• num_epochs: 3.0

Training results

Epoch	Training Loss	Validation Accuracy
1	0.1925	0.8799
2	0.0767	0.8922
3	0.0251	0.8922