train_rte_101112_1760638014

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

• Loss: 0.0625
• Num Input Tokens Seen: 6980984

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
0.0084	1.0	561	0.0786	350480
0.0693	2.0	1122	0.0625	700992
0.0546	3.0	1683	0.0631	1050848
0.0619	4.0	2244	0.1006	1400856
0.0	5.0	2805	0.1246	1749544
0.0	6.0	3366	0.1842	2099368
0.0	7.0	3927	0.1637	2447504
0.0	8.0	4488	0.1264	2794592
0.0	9.0	5049	0.2143	3145760
0.0	10.0	5610	0.1665	3495600
0.0001	11.0	6171	0.1197	3844488
0.0	12.0	6732	0.1446	4191800
0.0	13.0	7293	0.1505	4538416
0.0	14.0	7854	0.1551	4888904
0.0	15.0	8415	0.1557	5236560
0.0	16.0	8976	0.1649	5587768
0.0	17.0	9537	0.1629	5935088
0.0	18.0	10098	0.1677	6283144
0.0	19.0	10659	0.1635	6632504
0.0	20.0	11220	0.1673	6980984

Framework versions

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