roberta-base-relu-qnli

This model is a fine-tuned version of JeremiahZ/roberta-base-qnli on the GLUE QNLI dataset. It achieves the following results on the evaluation set:

• eval_loss: 0.5725
• eval_model_preparation_time: 0.0022
• eval_accuracy: 0.9264
• eval_runtime: 41.1034
• eval_samples_per_second: 132.909
• eval_steps_per_second: 33.233
• step: 0

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-5, 2e-5, 3e-5
• train_batch_size: 16
• eval_batch_size: 4
• seed: 42
• num_epochs: 10

The best model was selected based on the highest accuracy, which is the key evaluation metric for this task.

Framework versions

• Transformers 4.50.0.dev0
• Pytorch 2.6.0+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0