qanastek/whisper-large-french-cased

Whisper Large French Cased

This model is a fine-tuned version of openai/whisper-large on the mozilla-foundation/common_voice_11_0 fr dataset. It achieves the following results on the evaluation set:

• Loss: 0.2962
• Wer: 11.9100

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: 4
• eval_batch_size: 2
• 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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3357	0.2	1000	0.3994	16.1523
0.3026	0.4	2000	0.3802	15.2403
0.2904	0.6	3000	0.3389	14.0045
0.2407	0.8	4000	0.3135	12.7947
0.2451	1.0	5000	0.2962	11.9100

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.11.0+cu102
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2