train_cb_1757340191

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: 0.1360
• Num Input Tokens Seen: 367864

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
0.8637	0.5088	29	0.5433	20064
0.2282	1.0175	58	0.2346	37832
0.1669	1.5263	87	0.2458	57288
0.0698	2.0351	116	0.1708	74520
0.6367	2.5439	145	0.1745	93080
0.1197	3.0526	174	0.1617	111928
0.2465	3.5614	203	0.1476	131160
0.1638	4.0702	232	0.1464	150056
0.0107	4.5789	261	0.1360	167208
0.0856	5.0877	290	0.1542	186160
0.0062	5.5965	319	0.1584	206000
0.1789	6.1053	348	0.1645	224064
0.0373	6.6140	377	0.1678	243840
0.0026	7.1228	406	0.1604	261504
0.1333	7.6316	435	0.1662	280352
0.0041	8.1404	464	0.1668	299344
0.2603	8.6491	493	0.1665	318672
0.0191	9.1579	522	0.1650	337480
0.0748	9.6667	551	0.1643	356456

Framework versions

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