metadata

language:
  - id
license: apache-2.0
tags:
  - whisper-event
  - generated_from_trainer
datasets:
  - mozilla-foundation/common_voice_11_0
  - magic_data
  - TITML
metrics:
  - wer
model-index:
  - name: Whisper Medium Indonesian
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: mozilla-foundation/common_voice_11_0 id
          type: mozilla-foundation/common_voice_11_0
          config: id
          split: test
        metrics:
          - name: Wer
            type: wer
            value: 3.8273540533062804
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: google/fleurs id_id
          type: google/fleurs
          config: id_id
          split: test
        metrics:
          - name: Wer
            type: wer
            value: 9.74

Whisper Medium Indonesian

This model is a fine-tuned version of openai/whisper-medium on the Indonesian mozilla-foundation/common_voice_11_0, magic_data, titml and google/fleurs dataset. It achieves the following results:

CV11 test split:

Loss: 0.0698
Wer: 3.8274

Google/fleurs test split:

Wer: 9.74

Usage

from transformers import pipeline
transcriber = pipeline(
  "automatic-speech-recognition", 
  model="cahya/whisper-medium-id"
)
transcriber.model.config.forced_decoder_ids = (
  transcriber.tokenizer.get_decoder_prompt_ids(
    language="id" 
    task="transcribe"
  )
)
transcription = transcriber("my_audio_file.mp3")

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-06
train_batch_size: 16
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: 10000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0427	0.33	1000	0.0664	4.3807
0.042	0.66	2000	0.0658	3.9426
0.0265	0.99	3000	0.0657	3.8274
0.0211	1.32	4000	0.0679	3.8366
0.0212	1.66	5000	0.0682	3.8412
0.0206	1.99	6000	0.0683	3.8689
0.0166	2.32	7000	0.0711	3.9657
0.0095	2.65	8000	0.0717	3.9980
0.0122	2.98	9000	0.0714	3.9795
0.0049	3.31	10000	0.0720	3.9887

Evaluation

We evaluated the model using the test split of two datasets, the Common Voice 11 and the Google Fleurs. As Whisper can transcribe casing and punctuation, we also evaluate its performance using raw and normalized text. (lowercase + removal of punctuations). The results are as follows:

Common Voice 11

	WER
cahya/whisper-medium-id	3.83
openai/whisper-medium	12.62

Google/Fleurs

	WER
cahya/whisper-medium-id	9.74
cahya/whisper-medium-id + text normalization	tbc
openai/whisper-medium	10.2
openai/whisper-medium + text normalization	tbc

Framework versions

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