train_openbookqa_101112_1760638028

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

• Loss: 0.5169
• Num Input Tokens Seen: 8474968

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: 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
0.4802	1.0	1116	0.5300	424568
0.4469	2.0	2232	0.5230	848552
0.569	3.0	3348	0.5213	1271776
0.5901	4.0	4464	0.5203	1694896
0.5406	5.0	5580	0.5215	2118456
0.9049	6.0	6696	0.5220	2542752
0.5261	7.0	7812	0.5169	2966512
0.2801	8.0	8928	0.5187	3390048
0.5042	9.0	10044	0.5191	3814704
0.4908	10.0	11160	0.5244	4238440
0.5505	11.0	12276	0.5205	4662136
0.4383	12.0	13392	0.5199	5086336
0.5399	13.0	14508	0.5230	5510768
0.4704	14.0	15624	0.5228	5933936
0.7452	15.0	16740	0.5216	6357536
0.5127	16.0	17856	0.5226	6779872
0.3553	17.0	18972	0.5202	7203216
0.5657	18.0	20088	0.5203	7626944
0.6928	19.0	21204	0.5183	8051216
0.8785	20.0	22320	0.5183	8474968

Framework versions

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