train_multirc_123_1765143191

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

• Loss: 0.1333
• Num Input Tokens Seen: 264547520

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: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.2557	1.0	6130	0.1697	13255424
0.1781	2.0	12260	0.1504	26471216
0.133	3.0	18390	0.1469	39694112
0.0205	4.0	24520	0.1373	52929744
0.1351	5.0	30650	0.1356	66152480
0.0862	6.0	36780	0.1371	79389648
0.0952	7.0	42910	0.1420	92621824
0.0184	8.0	49040	0.1333	105830544
0.0353	9.0	55170	0.1360	119047920
0.1154	10.0	61300	0.1387	132272272
0.2578	11.0	67430	0.1359	145487264
0.1963	12.0	73560	0.1357	158737232
0.0198	13.0	79690	0.1431	171979232
0.0615	14.0	85820	0.1469	185199728
0.0885	15.0	91950	0.1426	198426688
0.1168	16.0	98080	0.1411	211640976
0.0501	17.0	104210	0.1444	224870720
0.0268	18.0	110340	0.1442	238102672
0.0331	19.0	116470	0.1440	251320768
0.06	20.0	122600	0.1450	264547520

Framework versions

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