train_math_qa_123_1760637723

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: 0.6289
• 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
0.7124	1.0	6714	0.6483	3894688
0.7039	2.0	13428	0.6445	7789256
0.5342	3.0	20142	0.6289	11683856
0.501	4.0	26856	0.6441	15585872
0.3563	5.0	33570	0.7951	19482128
0.3972	6.0	40284	0.9929	23376776
0.2464	7.0	46998	1.1039	27278720
0.0013	8.0	53712	1.2798	31180904
0.4622	9.0	60426	1.3532	35077032
0.0026	10.0	67140	1.5241	38976336
0.0347	11.0	73854	1.7321	42875264
0.0002	12.0	80568	1.7844	46772480
0.0001	13.0	87282	2.4031	50673016
0.0108	14.0	93996	2.3681	54573896
0.0	15.0	100710	2.5477	58472760
0.0	16.0	107424	2.7958	62371472
0.0	17.0	114138	3.0958	66268336
0.0	18.0	120852	3.5948	70167432
0.0	19.0	127566	3.7205	74065984
0.0	20.0	134280	3.7401	77961608

Framework versions

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