train_math_qa_42_1760637604

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: 2.6924
• Num Input Tokens Seen: 69300976

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: 1e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• 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
0.7517	2.0	11936	0.7964	6927672
0.8372	4.0	23872	0.7418	13862896
0.5544	6.0	35808	0.7228	20792536
0.6566	8.0	47744	0.7486	27722384
0.7514	10.0	59680	0.7892	34648704
0.3566	12.0	71616	0.9536	41580136
0.3106	14.0	83552	1.1857	48509240
0.1969	16.0	95488	1.8086	55439312
0.1342	18.0	107424	2.4534	62374112
0.0171	20.0	119360	2.6924	69300976

Framework versions

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