openai/whisper-medium-mix-es

This model is a fine-tuned version of openai/whisper-medium on the mozilla-foundation/common_voice_11_0, google/fleurs, facebook/multilingual_librispeech and facebook/voxpopuli datasets. It achieves the following results on the evaluation set:

• Loss: 0.1344
• Wer: 6.3465

Using the evaluation script provided in the Whisper Sprint the model achieves these results on the test sets (WER):

• google/fleurs: 4.0266 %
  (python run_eval_whisper_streaming.py --model_id="deepdml/whisper-medium-mix-es" --dataset="google/fleurs" --config="es_419" --device=0 --language="es")

• facebook/multilingual_librispeech: 4.6644 %
  (python run_eval_whisper_streaming.py --model_id="deepdml/whisper-medium-mix-es" --dataset="facebook/multilingual_librispeech" --config="spanish" --device=0 --language="es")

• facebook/voxpopuli: 8.3668 %
  (python run_eval_whisper_streaming.py --model_id="deepdml/whisper-medium-mix-es" --dataset="facebook/voxpopuli" --config="es" --device=0 --language="es")

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

Training data used:

• mozilla-foundation/common_voice_11_0: es, train+validation
• google/fleurs: es_419, train
• facebook/multilingual_librispeech: spanish, train
• facebook/voxpopuli: es, train

Evaluating over test split from mozilla-foundation/common_voice_11_0 dataset.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 32
• eval_batch_size: 16
• 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
0.266	0.2	1000	0.1657	8.0395
0.1394	0.4	2000	0.1539	7.3937
0.1316	0.6	3000	0.1452	6.9656
0.1165	0.8	4000	0.1392	6.5765
0.2816	1.0	5000	0.1344	6.3465

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu117
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2