train_qnli_1753094141

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

• Loss: 0.0436
• Num Input Tokens Seen: 103607072

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
0.0555	0.5000	11784	0.0754	5193280
0.041	1.0000	23568	0.0613	10365728
0.1024	1.5001	35352	0.0544	15547488
0.0455	2.0001	47136	0.0528	20725792
0.0237	2.5001	58920	0.0490	25887456
0.0264	3.0001	70704	0.0509	31082368
0.0692	3.5001	82488	0.0467	36266176
0.0935	4.0002	94272	0.0453	41440992
0.0368	4.5002	106056	0.0451	46618176
0.0472	5.0002	117840	0.0449	51803520
0.0268	5.5002	129624	0.0443	56978912
0.0233	6.0003	141408	0.0441	62167168
0.0673	6.5003	153192	0.0438	67356288
0.0405	7.0003	164976	0.0446	72532096
0.023	7.5003	176760	0.0442	77710656
0.0175	8.0003	188544	0.0436	82887904
0.0049	8.5004	200328	0.0438	88066400
0.038	9.0004	212112	0.0436	93248224
0.0997	9.5004	223896	0.0436	98430752

Framework versions

• PEFT 0.15.2
• Transformers 4.51.3
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.21.1