train_cb_1757340217

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: 0.3687
• Num Input Tokens Seen: 359688

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: 456
• 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
0.4646	0.5088	29	0.2152	18048
0.4224	1.0175	58	0.2754	36328
0.1384	1.5263	87	0.1717	56168
0.9708	2.0351	116	0.1764	73792
0.187	2.5439	145	0.1948	92768
0.1273	3.0526	174	0.1582	110064
0.0853	3.5614	203	0.1560	129808
0.5225	4.0702	232	0.2043	147240
0.0862	4.5789	261	0.2014	164744
0.0088	5.0877	290	0.1787	184440
0.0051	5.5965	319	0.1657	202456
0.001	6.1053	348	0.1849	220288
0.1069	6.6140	377	0.1918	239200
0.0181	7.1228	406	0.1944	256296
0.0004	7.6316	435	0.1970	275688
0.0886	8.1404	464	0.1968	294608
0.1007	8.6491	493	0.2060	312144
0.0083	9.1579	522	0.2111	330152
0.0012	9.6667	551	0.2125	347976

Framework versions

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