train_rte_1754652145

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

• Loss: 0.1820
• Num Input Tokens Seen: 3481336

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
6.7786	0.5009	281	6.3206	176032
2.0606	1.0018	562	2.2780	349200
0.8595	1.5027	843	0.5880	524208
0.4327	2.0036	1124	0.3796	699264
0.239	2.5045	1405	0.2867	873600
0.196	3.0053	1686	0.2456	1048184
0.2079	3.5062	1967	0.2240	1223864
0.2358	4.0071	2248	0.2165	1397624
0.1778	4.5080	2529	0.2119	1570936
0.2055	5.0089	2810	0.1969	1746384
0.2039	5.5098	3091	0.1901	1922384
0.1752	6.0107	3372	0.1904	2092320
0.2109	6.5116	3653	0.1882	2267520
0.1554	7.0125	3934	0.1853	2441688
0.1665	7.5134	4215	0.1854	2614936
0.1686	8.0143	4496	0.1831	2790832
0.1655	8.5152	4777	0.1827	2963888
0.1594	9.0160	5058	0.1820	3137352
0.1618	9.5169	5339	0.1834	3312648

Framework versions

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