train_hellaswag_42_1760637628

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

• Loss: 0.7116
• Num Input Tokens Seen: 218263888

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: 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
0.8732	1.0	8979	1.0233	10917120
0.7955	2.0	17958	0.7851	21836032
0.7456	3.0	26937	0.7186	32746560
0.766	4.0	35916	0.7116	43661424
0.6636	5.0	44895	0.7137	54578912
0.6306	6.0	53874	0.7167	65488016
0.8581	7.0	62853	0.7175	76410304
0.7264	8.0	71832	0.7148	87327296
0.7082	9.0	80811	0.7139	98229232
0.5871	10.0	89790	0.7167	109127968
0.6326	11.0	98769	0.7145	120042688
0.8029	12.0	107748	0.7155	130954720
0.7298	13.0	116727	0.7155	141874656
0.6079	14.0	125706	0.7155	152783392
0.5391	15.0	134685	0.7155	163694096
0.7002	16.0	143664	0.7155	174604544
0.5754	17.0	152643	0.7155	185523328
0.7789	18.0	161622	0.7155	196433472
0.5864	19.0	170601	0.7155	207345200
0.7214	20.0	179580	0.7155	218263888

Framework versions

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