whisper-large-v2-finetuning-3

This model is a fine-tuned version of guilhermebastos96/whisper-large-v2-finetuning-2 on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2959
• Wer: 7.9254

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0365	0.5089	1000	0.2219	12.8233
0.0154	1.0178	2000	0.2462	9.3545
0.0255	1.5267	3000	0.2492	9.2442
0.0178	2.0356	4000	0.2386	9.3401
0.0121	2.5445	5000	0.2447	8.9741
0.0051	3.0534	6000	0.2619	8.8478
0.0034	3.5623	7000	0.2634	8.3427
0.0014	4.0712	8000	0.2776	8.0597
0.001	4.5802	9000	0.2961	8.0022
0.0006	5.0891	10000	0.2959	7.9254

Framework versions

• Transformers 4.42.3
• Pytorch 2.2.1
• Datasets 2.20.0
• Tokenizers 0.19.1