train_qnli_1753094142

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: 0.0377
• 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
0.007	0.5000	11784	0.0461	5193280
0.0106	1.0000	23568	0.0445	10365728
0.0595	1.5001	35352	0.0471	15547488
0.0045	2.0001	47136	0.0377	20725792
0.0283	2.5001	58920	0.0493	25887456
0.0016	3.0001	70704	0.0753	31082368
0.002	3.5001	82488	0.0693	36266176
0.0017	4.0002	94272	0.0669	41440992
0.0	4.5002	106056	0.0962	46618176
0.025	5.0002	117840	0.0887	51803520
0.0	5.5002	129624	0.1154	56978912
0.0	6.0003	141408	0.1034	62167168
0.0	6.5003	153192	0.1117	67356288
0.0	7.0003	164976	0.1250	72532096
0.0	7.5003	176760	0.1402	77710656
0.0	8.0003	188544	0.1565	82887904
0.0	8.5004	200328	0.1776	88066400
0.0	9.0004	212112	0.1827	93248224
0.0	9.5004	223896	0.1967	98430752

Framework versions

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