lubobill1990/whisper-large-v3-yixing-dialect

This model is a fine-tuned version of openai/whisper-large-v3 on the generator dataset. It achieves the following results on the evaluation set:

• Loss: 1.0375

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 3407
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 5
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.6523	0.0346	50	1.8361
1.1887	0.0692	100	1.4349
1.1177	0.1037	150	1.3058
1.2017	0.1383	200	1.2580
1.2097	0.1729	250	1.2214
0.8582	0.2075	300	1.1897
0.8613	0.2420	350	1.1691
0.9025	0.2766	400	1.1537
1.0347	0.3112	450	1.1366
0.8058	0.3458	500	1.1220
0.8358	0.3804	550	1.1087
0.9403	0.4149	600	1.1042
1.0258	0.4495	650	1.0950
0.976	0.4841	700	1.0891
0.632	0.5187	750	1.0800
0.7036	0.5533	800	1.0700
0.7675	0.5878	850	1.0653
1.0105	0.6224	900	1.0629
0.6001	0.6570	950	1.0558
0.7627	0.6916	1000	1.0524
0.7371	0.7261	1050	1.0507
0.6548	0.7607	1100	1.0513
0.8578	0.7953	1150	1.0455
0.6846	0.8299	1200	1.0432
0.785	0.8645	1250	1.0401
0.819	0.8990	1300	1.0388
0.9689	0.9336	1350	1.0380
0.8372	0.9682	1400	1.0375

Framework versions

• PEFT 0.15.2
• Transformers 4.51.3
• Pytorch 2.7.0+cu128
• Datasets 3.5.1
• Tokenizers 0.21.1