gpt1_sst2_right

This model is a fine-tuned version of openai-gpt on sst2 dataset of GLUE benchmark. It achieves the following results on the evaluation set:

• Loss: 0.4216
• Accuracy: 0.9255
• Recall: 0.9369
• Precision: 0.9183

For testing, the model is loaded as a pipeline, and used for the prediction of each sample in test split. The samples and their predictions are recorded in test_preds.csv file. Access to Repository for finetuning.

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

For batched training, <pad> token is added to the tokenizer and the following padding-truncation options are adapted:

• Padding Side: "right"
• Truncation Side: "right"

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 4

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Recall	Precision
0.2	1.0	4210	0.2958	0.9037	0.8649	0.9412
0.1455	2.0	8420	0.3172	0.9186	0.9505	0.8960
0.0892	3.0	12630	0.3637	0.9278	0.9257	0.9320
0.0584	4.0	16840	0.4216	0.9255	0.9369	0.9183

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0