train_math_qa_123_1760637725

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: 1.0601
• Num Input Tokens Seen: 77961608

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: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
1.1789	1.0	6714	1.1266	3894688
1.2559	2.0	13428	1.0804	7789256
0.8631	3.0	20142	1.0672	11683856
1.0857	4.0	26856	1.0640	15585872
0.9917	5.0	33570	1.0624	19482128
1.1152	6.0	40284	1.0682	23376776
1.3225	7.0	46998	1.0651	27278720
1.0668	8.0	53712	1.0638	31180904
1.0528	9.0	60426	1.0663	35077032
1.3899	10.0	67140	1.0647	38976336
0.9245	11.0	73854	1.0665	42875264
1.1201	12.0	80568	1.0715	46772480
0.8035	13.0	87282	1.0635	50673016
0.8389	14.0	93996	1.0636	54573896
0.906	15.0	100710	1.0643	58472760
0.8204	16.0	107424	1.0663	62371472
1.0922	17.0	114138	1.0643	66268336
0.9974	18.0	120852	1.0601	70167432
1.2872	19.0	127566	1.0669	74065984
1.2264	20.0	134280	1.0669	77961608

Framework versions

• PEFT 0.17.1
• Transformers 4.51.3
• Pytorch 2.9.0+cu128
• Datasets 4.0.0
• Tokenizers 0.21.4