ROE_Patent_Breeze7B_20250124_Type1

This model is a fine-tuned version of MediaTek-Research/Breeze-7B-Instruct-v1_0 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7582

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: 1
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 8
• optimizer: Use OptimizerNames.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: 0.05
• num_epochs: 17
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.0139	1.0	13	1.0349
0.9149	2.0	26	0.9424
0.8664	3.0	39	0.8711
0.7648	4.0	52	0.8391
0.7454	5.0	65	0.8082
0.7115	6.0	78	0.7896
0.6892	7.0	91	0.7794
0.6379	8.0	104	0.7733
0.6592	9.0	117	0.7650
0.6346	10.0	130	0.7655
0.6214	11.0	143	0.7607
0.6093	12.0	156	0.7601
0.6085	13.0	169	0.7590
0.6268	14.0	182	0.7585
0.609	15.0	195	0.7575
0.6206	15.7273	204	0.7582

Framework versions

• PEFT 0.14.1.dev0
• Transformers 4.48.1
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0