train_hellaswag_1754652169

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: 1.7472
• 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
0.1386	0.5001	4490	0.1300	5450816
0.151	1.0001	8980	0.0868	10899840
0.0125	1.5002	13470	0.0775	16338976
0.1252	2.0002	17960	0.0812	21789168
0.0373	2.5003	22450	0.0732	27236592
0.0227	3.0003	26940	0.0626	32696128
0.1278	3.5004	31430	0.0628	38137920
0.0083	4.0004	35920	0.0662	43579472
0.0526	4.5005	40410	0.0681	49022960
0.124	5.0006	44900	0.0629	54468496
0.0057	5.5006	49390	0.0646	59917136
0.1004	6.0007	53880	0.0625	65358976
0.0211	6.5007	58370	0.0690	70806016
0.0727	7.0008	62860	0.0656	76259312
0.0806	7.5008	67350	0.0788	81705616
0.0411	8.0009	71840	0.0713	87153488
0.0328	8.5009	76330	0.0820	92602480
0.0011	9.0010	80820	0.0804	98051504
0.0826	9.5011	85310	0.0826	103491728

Framework versions

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