whisper-large-v3-genbed-m-model

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

• Loss: 0.7479
• Wer: 36.9425

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: 1.75e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 30000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.4385	0.6596	250	0.7026	57.3435
0.578	1.3193	500	0.6312	47.4271
0.499	1.9789	750	0.5735	43.2676
0.2829	2.6385	1000	0.5949	41.0913
0.2304	3.2982	1250	0.6149	40.5660
0.1672	3.9578	1500	0.5645	38.5399
0.1019	4.6174	1750	0.6265	42.0026
0.0911	5.2770	2000	0.6534	38.5399
0.0713	5.9367	2250	0.6533	38.1754
0.0545	6.5963	2500	0.6577	37.7466
0.0497	7.2559	2750	0.6626	39.3117
0.0425	7.9156	3000	0.6901	37.2642
0.0374	8.5752	3250	0.6919	38.6256
0.0312	9.2348	3500	0.7093	37.2856
0.0302	9.8945	3750	0.7260	35.7740
0.0233	10.5541	4000	0.7181	36.5780
0.0262	11.2137	4250	0.7352	35.5703
0.0241	11.8734	4500	0.7340	36.4172
0.0198	12.5330	4750	0.7463	36.8461
0.0201	13.1926	5000	0.7479	36.9425

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.4.1+cu121
• Datasets 3.0.0
• Tokenizers 0.19.1