train_math_qa_1754507505

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: 0.6484
• 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
0.6574	0.5	3357	0.7108	1938656
0.7483	1.0	6714	0.6811	3870688
0.6417	1.5	10071	0.6659	5801696
0.7456	2.0	13428	0.6549	7741288
0.6224	2.5	16785	0.6529	9678632
0.6071	3.0	20142	0.6484	11611120
0.6323	3.5	23499	0.6511	13550032
0.6184	4.0	26856	0.6527	15489040
0.566	4.5	30213	0.6706	17420720
0.7218	5.0	33570	0.6527	19360624
0.4094	5.5	36927	0.6629	21295472
0.6057	6.0	40284	0.6692	23230504
0.6518	6.5	43641	0.6869	25166536
0.5514	7.0	46998	0.6820	27107328
0.4348	7.5	50355	0.7125	29044256
0.4093	8.0	53712	0.6995	30985288
0.4233	8.5	57069	0.7024	32925544
0.5576	9.0	60426	0.7077	34856680
0.5336	9.5	63783	0.7066	36790952
0.5528	10.0	67140	0.7073	38732208

Framework versions

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