train_copa_1757340277

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

• Loss: 0.0768
• Num Input Tokens Seen: 281312

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: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.081	0.5	45	0.1215	14144
0.3339	1.0	90	0.1037	28192
0.0463	1.5	135	0.0964	42208
0.082	2.0	180	0.0777	56256
0.2678	2.5	225	0.0822	70368
0.0878	3.0	270	0.0920	84320
0.1495	3.5	315	0.1532	98400
0.0075	4.0	360	0.0768	112416
0.429	4.5	405	0.1562	126496
0.0002	5.0	450	0.1207	140544
0.0092	5.5	495	0.1345	154592
0.002	6.0	540	0.1524	168768
0.0064	6.5	585	0.1678	182848
0.0449	7.0	630	0.1447	196896
0.1323	7.5	675	0.1635	210912
0.0001	8.0	720	0.2237	225024
0.0211	8.5	765	0.2088	239200
0.0121	9.0	810	0.2073	253152
0.0034	9.5	855	0.2088	267040
0.1445	10.0	900	0.2092	281312

Framework versions

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