train_cb_101112_1760637983

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: 0.0086
• Num Input Tokens Seen: 723584

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: 101112
• 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: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.1849	1.0	57	0.2210	36112
0.2067	2.0	114	0.2628	71552
0.2322	3.0	171	0.0987	108088
0.1315	4.0	228	0.1422	144720
0.0016	5.0	285	0.0857	181120
0.0007	6.0	342	0.0523	217128
0.0	7.0	399	0.0471	253536
0.0001	8.0	456	0.0086	290112
0.0	9.0	513	0.0798	325872
0.0	10.0	570	0.0866	361920
0.0	11.0	627	0.0805	398432
0.0	12.0	684	0.0890	435536
0.0	13.0	741	0.0862	471520
0.0	14.0	798	0.0856	507256
0.0	15.0	855	0.0890	543064
0.0	16.0	912	0.0900	579704
0.0	17.0	969	0.0913	615960
0.0	18.0	1026	0.0912	652368
0.0	19.0	1083	0.0852	687976
0.0	20.0	1140	0.0901	723584

Framework versions

• PEFT 0.17.1
• Transformers 4.51.3
• Pytorch 2.9.0+cu128
• Datasets 4.0.0
• Tokenizers 0.21.4