Chillarmo/whisper-small-armenian

Whisper-Small Fine-tuned for Armenian ASR

This model is a fine-tuned version of OpenAI's Whisper-small on the Common Voice 20 Armenian dataset for automatic speech recognition.

Training Results

The model was trained for 5.34 epochs with the following final results:

Metric	Value
Training Loss	0.101
Training Runtime	59,610 seconds (≈16.56 hours)
Training Samples/Second	1.34
Training Steps/Second	0.084
Total Training Steps	5,000
Epochs	5.34

Evaluation Results

Metric	Value
Evaluation Loss	0.165
Word Error Rate (WER)	29.43%
Character Error Rate (CER)	5.89%
Exact Match	20.05%
Average Prediction Length	7.71 tokens
Average Label Length	7.77 tokens
Length Ratio	0.992
Evaluation Runtime	9,486 seconds (≈2.63 hours)
Evaluation Samples/Second	0.62
Evaluation Steps/Second	0.077

Model Details

• Base Model: openai/whisper-small
• Language: Armenian (hy)
• Dataset: Chillarmo/common_voice_20_armenian
• License: Apache 2.0