train_piqa_1754741701

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

• Loss: 0.1320
• Num Input Tokens Seen: 22103448

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
0.1248	0.5	1813	0.1588	1118368
0.0805	1.0	3626	0.1470	2216600
0.0542	1.5	5439	0.1503	3320792
0.087	2.0	7252	0.1482	4419000
0.1423	2.5	9065	0.1463	5525176
0.1749	3.0	10878	0.1625	6628280
0.1449	3.5	12691	0.1320	7736376
0.2093	4.0	14504	0.1529	8844408
0.2112	4.5	16317	0.1506	9951832
0.0171	5.0	18130	0.1470	11048200
0.2367	5.5	19943	0.1649	12157032
0.1281	6.0	21756	0.1827	13257624
0.0894	6.5	23569	0.1733	14360952
0.1768	7.0	25382	0.1844	15468632
0.1353	7.5	27195	0.1813	16574840
0.2293	8.0	29008	0.1806	17678024
0.2512	8.5	30821	0.1802	18780040
0.1342	9.0	32634	0.1810	19894712
0.086	9.5	34447	0.1813	21014840
0.0014	10.0	36260	0.1800	22103448

Framework versions

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