train_winogrande_101112_1760638067

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

• Loss: 0.2432
• Num Input Tokens Seen: 34100624

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: 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
0.2238	2.0	16160	0.2324	3410576
0.2434	4.0	32320	0.2204	6820640
0.1744	6.0	48480	0.1129	10230096
0.0037	8.0	64640	0.0676	13639760
0.0006	10.0	80800	0.0786	17048912
0.063	12.0	96960	0.1006	20458720
0.0001	14.0	113120	0.1434	23869008
0.0	16.0	129280	0.1929	27280464
0.0	18.0	145440	0.2320	30690528
0.0	20.0	161600	0.2432	34100624

Framework versions

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