whisper-large-v3-genbed-combined

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: 0.5126
• Wer: 32.9275

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
0.5858	0.3300	250	0.6038	53.7168
0.5721	0.6601	500	0.5527	45.6851
0.54	0.9901	750	0.4955	41.3062
0.3576	1.3201	1000	0.5173	44.8467
0.3602	1.6502	1250	0.4491	38.1074
0.3477	1.9802	1500	0.4491	35.1437
0.2164	2.3102	1750	0.4674	34.2679
0.2119	2.6403	2000	0.4611	35.6243
0.2253	2.9703	2250	0.4442	33.5416
0.1314	3.3003	2500	0.4764	32.8367
0.1323	3.6304	2750	0.4733	33.4295
0.1407	3.9604	3000	0.4695	32.6231
0.0821	4.2904	3250	0.5085	33.3921
0.0816	4.6205	3500	0.5025	32.9649
0.0914	4.9505	3750	0.5126	32.9275

Framework versions

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