train_hellaswag_1754507492

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: 0.0550
• Num Input Tokens Seen: 108930064

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: 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: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.0101	0.5001	4490	0.1009	5450816
0.1216	1.0001	8980	0.0550	10899840
0.0028	1.5002	13470	0.0601	16338976
0.0021	2.0002	17960	0.0608	21789168
0.0003	2.5003	22450	0.0735	27236592
0.0003	3.0003	26940	0.0747	32696128
0.0002	3.5004	31430	0.1229	38137920
0.0	4.0004	35920	0.1146	43579472
0.0005	4.5005	40410	0.1121	49022960
0.0001	5.0006	44900	0.1085	54468496
0.0	5.5006	49390	0.1544	59917136
0.0	6.0007	53880	0.1369	65358976
0.0	6.5007	58370	0.1261	70806016
0.0	7.0008	62860	0.1143	76259312
0.0	7.5008	67350	0.1545	81705616
0.0	8.0009	71840	0.1650	87153488
0.0	8.5009	76330	0.1804	92602480
0.0	9.0010	80820	0.1840	98051504
0.0	9.5011	85310	0.1884	103491728

Framework versions

• PEFT 0.15.2
• Transformers 4.51.3
• Pytorch 2.8.0+cu128
• Datasets 3.6.0
• Tokenizers 0.21.1