train_wsc_101112_1760347667

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

• Loss: 0.3457
• Num Input Tokens Seen: 488816

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: 0.03
• 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: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.451	0.504	63	0.5548	24608
0.349	1.008	126	0.3484	49296
0.3646	1.512	189	0.3629	74672
0.3553	2.016	252	0.3527	98816
0.4132	2.52	315	0.3512	123680
0.3561	3.024	378	0.3512	147776
0.3345	3.528	441	0.3568	173312
0.3828	4.032	504	0.3574	197728
0.3481	4.536	567	0.3481	222560
0.3524	5.04	630	0.3506	246848
0.3619	5.5440	693	0.3484	271008
0.3418	6.048	756	0.3473	295984
0.354	6.552	819	0.3495	320080
0.348	7.056	882	0.3514	345136
0.3481	7.5600	945	0.3505	370416
0.3579	8.064	1008	0.3484	394688
0.3466	8.568	1071	0.3484	418880
0.3358	9.072	1134	0.3500	444304
0.3486	9.576	1197	0.3457	469328

Framework versions

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