train_hellaswag_1754507490

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: 0.1269
• 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
0.1504	0.5001	4490	0.2625	5450816
0.2005	1.0001	8980	0.1893	10899840
0.0142	1.5002	13470	0.1743	16338976
0.2198	2.0002	17960	0.1841	21789168
0.0654	2.5003	22450	0.1486	27236592
0.0142	3.0003	26940	0.1269	32696128
0.2299	3.5004	31430	0.1411	38137920
0.002	4.0004	35920	0.1463	43579472
0.2379	4.5005	40410	0.1384	49022960
0.2197	5.0006	44900	0.1399	54468496
0.1109	5.5006	49390	0.1304	59917136
0.1593	6.0007	53880	0.1340	65358976
0.0458	6.5007	58370	0.1402	70806016
0.1249	7.0008	62860	0.1337	76259312
0.0229	7.5008	67350	0.1371	81705616
0.0872	8.0009	71840	0.1364	87153488
0.0598	8.5009	76330	0.1364	92602480
0.0016	9.0010	80820	0.1359	98051504
0.1272	9.5011	85310	0.1360	103491728

Framework versions

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