train_cola_1757340260

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on the cola dataset. It achieves the following results on the evaluation set:

• Loss: 0.1983
• Num Input Tokens Seen: 3663392

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: 4
• eval_batch_size: 4
• seed: 101112
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.2372	0.5	962	0.1779	183040
0.3332	1.0	1924	0.2056	366136
0.3624	1.5	2886	0.1612	548856
0.1165	2.0	3848	0.1469	732880
0.3072	2.5	4810	0.1503	916368
0.0702	3.0	5772	0.1434	1099816
0.0694	3.5	6734	0.1615	1283208
0.0432	4.0	7696	0.1529	1465464
0.1038	4.5	8658	0.1477	1648696
0.0965	5.0	9620	0.1465	1831728
0.0509	5.5	10582	0.1608	2014288
0.0672	6.0	11544	0.1458	2198176
0.0325	6.5	12506	0.1609	2382016
0.1084	7.0	13468	0.1641	2564208
0.0944	7.5	14430	0.1775	2746960
0.0155	8.0	15392	0.1814	2930240
0.1982	8.5	16354	0.2046	3113696
0.1116	9.0	17316	0.1818	3297136
0.1125	9.5	18278	0.1877	3480592
0.0085	10.0	19240	0.1878	3663392

Framework versions

• PEFT 0.15.2
• Transformers 4.51.3
• Pytorch 2.8.0+cu128
• Datasets 3.6.0
• Tokenizers 0.21.1