train_math_qa_101112_1760638066

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: 0.6663
• Num Input Tokens Seen: 77914328

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: 101112
• 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
0.6321	1.0	6714	0.7315	3894384
0.6864	2.0	13428	0.7008	7788792
0.9075	3.0	20142	0.6857	11683344
0.7834	4.0	26856	0.6755	15578064
0.8392	5.0	33570	0.6742	19479304
0.5749	6.0	40284	0.6706	23378352
0.5827	7.0	46998	0.6673	27274568
0.7153	8.0	53712	0.6682	31172664
0.6194	9.0	60426	0.6663	35068368
0.7731	10.0	67140	0.6681	38966392
0.6639	11.0	73854	0.6683	42861936
0.575	12.0	80568	0.6695	46756048
0.4167	13.0	87282	0.6692	50652416
0.5908	14.0	93996	0.6714	54546936
0.5584	15.0	100710	0.6727	58442960
0.8154	16.0	107424	0.6729	62338944
0.5358	17.0	114138	0.6727	66231336
0.4849	18.0	120852	0.6730	70127040
0.6368	19.0	127566	0.6740	74021072
0.455	20.0	134280	0.6737	77914328

Framework versions

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