train_gsm8k_123_1760637710

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on the gsm8k dataset. It achieves the following results on the evaluation set:

• Loss: 4.7971
• Num Input Tokens Seen: 34679720

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: 0.001
• 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
0.5851	1.0	1682	0.5307	1738584
0.4526	2.0	3364	0.5005	3477544
0.532	3.0	5046	0.4856	5217560
0.5259	4.0	6728	0.4823	6947896
0.4339	5.0	8410	0.4744	8680664
0.3784	6.0	10092	0.4706	10414768
0.4588	7.0	11774	0.4696	12148592
0.3837	8.0	13456	0.4638	13883456
0.4609	9.0	15138	0.4626	15613432
0.3485	10.0	16820	0.4618	17345296
0.3763	11.0	18502	0.4609	19082288
0.5013	12.0	20184	0.4595	20812896
0.4102	13.0	21866	0.4611	22544592
0.4357	14.0	23548	0.4609	24280408
0.3804	15.0	25230	0.4613	26007520
0.3858	16.0	26912	0.4624	27741272
0.4541	17.0	28594	0.4627	29470344
0.4276	18.0	30276	0.4628	31206528
0.3947	19.0	31958	0.4636	32940696
0.4315	20.0	33640	0.4635	34679720

Framework versions

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