emilios/whisper-md-hu

Whisper medium Hungarian El Greco

This model is a fine-tuned version of openai/whisper-medium on the mozilla-foundation/common_voice_11_0,google/fleurs hu,hu_hu dataset. It achieves the following results on the evaluation set:

• Loss: 0.3428
• Wer: 18.6422

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: 3e-06
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• distributed_type: multi-GPU
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 10000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0621	1.05	1000	0.2690	20.5099
0.0174	2.1	2000	0.2705	19.2292
0.006	3.15	3000	0.2954	18.9890
0.0028	4.2	4000	0.3093	18.8023
0.0016	5.25	5000	0.3240	18.9653
0.0018	6.3	6000	0.3313	18.6451
0.0014	7.35	7000	0.3330	18.9446
0.0016	8.39	8000	0.3428	18.6422
0.0015	9.44	9000	0.3508	18.9564
0.001	10.49	10000	0.3569	18.8556

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 2.0.0.dev20221216+cu116
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2