train_cola_101112_1757596144

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

• Loss: 0.2487
• Num Input Tokens Seen: 6920328

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: 2
• eval_batch_size: 2
• 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
0.1588	1.0	3848	0.2635	346088
0.2684	2.0	7696	0.2537	692000
0.2971	3.0	11544	0.2506	1038152
0.2253	4.0	15392	0.2665	1384064
0.2699	5.0	19240	0.2665	1730008
0.2474	6.0	23088	0.2510	2076088
0.2821	7.0	26936	0.2515	2422232
0.2325	8.0	30784	0.2572	2768128
0.324	9.0	34632	0.2493	3114488
0.234	10.0	38480	0.2507	3460672
0.3039	11.0	42328	0.2561	3806392
0.2341	12.0	46176	0.2489	4152080
0.2633	13.0	50024	0.2489	4498120
0.2366	14.0	53872	0.2490	4844176
0.2518	15.0	57720	0.2507	5190512
0.163	16.0	61568	0.2510	5536504
0.3332	17.0	65416	0.2482	5882480
0.3443	18.0	69264	0.2485	6228440
0.2256	19.0	73112	0.2479	6574016
0.2547	20.0	76960	0.2487	6920328

Framework versions

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