train_math_qa_1754652176

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

• Loss: 0.8026
• Num Input Tokens Seen: 38732208

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: 123
• 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
0.9385	0.5	3357	0.8721	1938656
0.8336	1.0	6714	0.8212	3870688
0.8288	1.5	10071	0.8165	5801696
0.8502	2.0	13428	0.8096	7741288
0.7843	2.5	16785	0.8062	9678632
0.8009	3.0	20142	0.8065	11611120
0.8214	3.5	23499	0.8087	13550032
0.7798	4.0	26856	0.8033	15489040
0.7908	4.5	30213	0.8038	17420720
0.813	5.0	33570	0.8034	19360624
0.7837	5.5	36927	0.8044	21295472
0.8133	6.0	40284	0.8047	23230504
0.7966	6.5	43641	0.8026	25166536
0.8272	7.0	46998	0.8034	27107328
0.7807	7.5	50355	0.8030	29044256
0.7946	8.0	53712	0.8029	30985288
0.8	8.5	57069	0.8026	32925544
0.8165	9.0	60426	0.8028	34856680
0.7526	9.5	63783	0.8028	36790952
0.7855	10.0	67140	0.8029	38732208

Framework versions

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