bert_uncased_L-4_H-128_A-2_mnli

This model is a fine-tuned version of google/bert_uncased_L-4_H-128_A-2 on the GLUE MNLI dataset. It achieves the following results on the evaluation set:

• Loss: 0.6425
• Accuracy: 0.7331

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 256
• eval_batch_size: 256
• seed: 10
• optimizer: Use 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: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.8752	1.0	1534	0.7743	0.6558
0.7714	2.0	3068	0.7263	0.6857
0.7255	3.0	4602	0.6946	0.7020
0.6927	4.0	6136	0.6789	0.7087
0.6662	5.0	7670	0.6657	0.7205
0.6441	6.0	9204	0.6691	0.7229
0.625	7.0	10738	0.6622	0.7258
0.607	8.0	12272	0.6531	0.7314
0.5894	9.0	13806	0.6613	0.7308
0.5754	10.0	15340	0.6591	0.7293
0.5615	11.0	16874	0.6635	0.7287
0.5477	12.0	18408	0.6701	0.7344
0.5343	13.0	19942	0.6699	0.7343

Framework versions

• Transformers 4.46.3
• Pytorch 2.2.1+cu118
• Datasets 2.17.0
• Tokenizers 0.20.3