train_cb_1757340170

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

• Loss: 0.0902
• Num Input Tokens Seen: 361992

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: 42
• 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
1.0002	0.5088	29	0.7479	18048
0.277	1.0175	58	0.1704	36928
0.091	1.5263	87	0.1233	54176
0.3287	2.0351	116	0.1255	73136
0.1758	2.5439	145	0.1202	91216
0.076	3.0526	174	0.1193	110696
0.1815	3.5614	203	0.1115	129448
0.1655	4.0702	232	0.1046	147176
0.2647	4.5789	261	0.0995	164424
0.0395	5.0877	290	0.1002	183416
0.3311	5.5965	319	0.0946	203256
0.0178	6.1053	348	0.0960	220912
0.0569	6.6140	377	0.0902	240336
0.1175	7.1228	406	0.0930	257848
0.069	7.6316	435	0.0944	276824
0.1596	8.1404	464	0.0930	294504
0.0465	8.6491	493	0.0926	313576
0.0086	9.1579	522	0.0932	332256
0.3315	9.6667	551	0.0926	350336

Framework versions

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