train_cb_1757340216

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: 0.2338
• 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
0.8625	0.5088	29	0.6374	18048
0.2666	1.0175	58	0.2585	36328
0.1799	1.5263	87	0.3019	56168
0.1285	2.0351	116	0.3430	73792
0.279	2.5439	145	0.2732	92768
0.0709	3.0526	174	0.2643	110064
0.0766	3.5614	203	0.2460	129808
0.5976	4.0702	232	0.2900	147240
0.1025	4.5789	261	0.3045	164744
0.0229	5.0877	290	0.2338	184440
0.0101	5.5965	319	0.2631	202456
0.002	6.1053	348	0.2788	220288
0.172	6.6140	377	0.2708	239200
0.1097	7.1228	406	0.2558	256296
0.009	7.6316	435	0.2577	275688
0.2481	8.1404	464	0.2552	294608
0.1663	8.6491	493	0.2559	312144
0.0393	9.1579	522	0.2580	330152
0.0009	9.6667	551	0.2560	347976

Framework versions

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