whisper-khmer-small

This model is a fine-tuned version of openai/whisper-small on the openslr, google/fleurs and KMCS Train dataset. It achieves the following results on the evaluation set:

• Loss: 0.17
• Wer: 0.5211

Model description

This model is fine-tuned with Google FLEURS, OpenSLR (SLR42) and KMCS Train dataset.

from transformers import pipeline

pipe = pipeline(
    task="automatic-speech-recognition",
    model="dataticon-dev/whisper-khmer-small",
)

result = pipe("audio.wav",
  generate_kwargs={
    "language":"<|km|>",
    "task":"transcribe"},
    batch_size=8
)

print(result["text"])