metadata

language:
  - ca
license: apache-2.0
tags:
  - whisper-event
  - generated_from_trainer
datasets:
  - mozilla-foundation/common_voice_11_0
  - google/fleurs
  - openslr
  - collectivat/tv3_parla
  - projecte-aina/parlament_parla
metrics:
  - wer
model-index:
  - name: Whisper Medium Ca
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Common Voice 11.0
          type: mozilla-foundation/common_voice_11_0
          config: ca
          split: test
          args: ca
        metrics:
          - name: Wer
            type: wer
            value: 10.0031

Whisper Medium Ca

This model is a fine-tuned version of openai/whisper-medium on the Common Voice 11.0, the Fleurs, the SLR69, the tb3_parla and the parlament_parla datasets. It achieves the following results on the evaluation set:

eval_loss: 0.1905
eval_wer: 10.0031
eval_runtime: 10456.4485
eval_samples_per_second: 1.563
eval_steps_per_second: 0.195
epoch: 0.2
step: 2000

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: 32
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 64
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 1000
training_steps: 10000
mixed_precision_training: Native AMP

Framework versions

Transformers 4.26.0.dev0
Pytorch 1.13.1+cu117
Datasets 2.7.1.dev0
Tokenizers 0.13.2