train_math_qa_1754507502

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: 0.6648
• 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
0.6914	0.5	3357	0.7456	1938656
0.775	1.0	6714	0.7076	3870688
0.6317	1.5	10071	0.6910	5801696
0.7525	2.0	13428	0.6778	7741288
0.6871	2.5	16785	0.6760	9678632
0.7372	3.0	20142	0.6749	11611120
0.7415	3.5	23499	0.6681	13550032
0.7147	4.0	26856	0.6683	15489040
0.7027	4.5	30213	0.6750	17420720
0.7274	5.0	33570	0.6648	19360624
0.5673	5.5	36927	0.6664	21295472
0.6829	6.0	40284	0.6708	23230504
0.7997	6.5	43641	0.6713	25166536
0.7792	7.0	46998	0.6682	27107328
0.5814	7.5	50355	0.6710	29044256
0.6262	8.0	53712	0.6701	30985288
0.6886	8.5	57069	0.6694	32925544
0.7695	9.0	60426	0.6695	34856680
0.6357	9.5	63783	0.6699	36790952
0.8185	10.0	67140	0.6699	38732208

Framework versions

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