train_wsc_1754502826

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: 0.3594
• Num Input Tokens Seen: 490000

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: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Input Tokens Seen
0.5115	0.504	63	0.5330	25504
0.6246	1.008	126	0.5143	49696
0.3665	1.512	189	0.4442	74112
0.2995	2.016	252	0.3941	99136
0.3689	2.52	315	0.3874	123904
0.412	3.024	378	0.3728	148736
0.3331	3.528	441	0.3610	174432
0.3437	4.032	504	0.3902	198656
0.345	4.536	567	0.3696	224032
0.3501	5.04	630	0.3704	247424
0.3529	5.5440	693	0.3594	271232
0.3468	6.048	756	0.3655	295728
0.3808	6.552	819	0.3728	320464
0.3041	7.056	882	0.3842	345856
0.347	7.5600	945	0.3812	371040
0.3363	8.064	1008	0.3751	395216
0.3317	8.568	1071	0.3754	419184
0.384	9.072	1134	0.3719	444560
0.3151	9.576	1197	0.3720	469104

Framework versions

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