Whisper-Tiny-ASR ( Burmese )
This model is a fine-tuned version of openai/whisper-tiny for Automatic Speech Recognition (ASR) in Burmese .
Model Overview
- Architecture: Whisper (Tiny)
- Parameters: ~39M
- Task: Automatic Speech Recognition (ASR)
- Language: Burmese (
my) - Sampling Rate: 16,000 Hz (Mono)
Dataset Details
The model was trained on a standardized, high-quality Burmese speech corpus:
- Total Duration: ~22 Hours of audio
- Total Utterances: 24,560 WAV files (16 kHz, 16-bit PCM, Mono)
- Total Speakers: 13 Synthetic Burmese Speakers
- Speakers: ပီယ ၊ ဝါစာ ၊ သီရိ ၊ ဒီပ ၊ အက္ခရာ ၊ သဒ္ဒါ ၊ သရ ၊ တာရာ ၊ ကဝိ ၊ သုတ ၊ ပသာဒ ၊ ဂီတ ၊ နန္ဒ
- Training & Validation: 11 speakers (20,699 train utterances, 2,300 validation utterances)
- Test Set (Held-Out): 2 unseen speakers ( ဂီတ & နန္ဒ , 1,561 utterances )
- Text Preprocessing: Standardized Myanmar Unicode (
\u1000-\u109F) with unified word segmentation and punctuation removal. The held-out test split evaluates zero-shot acoustic generalization across completely unseen voices.
Training Configuration
| Parameter | Value |
|---|---|
| Effective Batch Size | 64 (16 per device x 4 gradient accumulation) |
| Peak Learning Rate | 1.5e-4 |
| Learning Rate Scheduler | Cosine |
| Warmup Steps | 150 |
| Total Steps | 1,500 (~9.26 Epochs) |
| Mixed Precision | BF16 |
| Weight Decay | 0.01 |
| Gradient Checkpointing | Enabled |
| Augmentation | None |
Evaluation Results
| Stage / Evaluation | Step | Train Loss | Val Loss | WER (%) | CER (%) | SER (%) | DER (%) | IER (%) | chrF |
|---|---|---|---|---|---|---|---|---|---|
| Baseline (Untrained) | 0 | - | - | 104.62 | 130.83 | 100.00 | 41.54 | 4.62 | 0.00 |
| Validation Set | 250 | 0.9590 | 0.1098 | 60.79 | 19.93 | 97.61 | 10.94 | 5.76 | 69.02 |
| Validation Set | 500 | 0.3188 | 0.0551 | 38.17 | 8.25 | 84.78 | 8.45 | 2.90 | 86.15 |
| Validation Set | 750 | 0.1719 | 0.0428 | 30.01 | 5.98 | 75.13 | 5.39 | 3.62 | 90.26 |
| Validation Set | 1000 | 0.0447 | 0.0442 | 26.92 | 5.17 | 69.13 | 5.18 | 3.08 | 91.85 |
| Validation Set | 1250 | 0.0119 | 0.0492 | 25.66 | 4.57 | 66.83 | 4.57 | 3.45 | 92.78 |
| Validation Set | 1500 | 0.0026 | 0.0508 | 25.21 | 4.47 | 65.52 | 4.47 | 3.42 | 92.91 |
| UNSEEN TEST (Final) | Final | - | - | 42.29 | 11.69 | 94.17 | 7.99 | 3.47 | 79.82 |
Usage
import torch
from transformers import pipeline
pipe = pipeline(
"automatic-speech-recognition",
model="thantzinphyo/Whisper-Tiny-ASR",
device="cuda:0" if torch.cuda.is_available() else "cpu",
)
output = pipe("audio.wav", generate_kwargs={"language": "burmese", "task": "transcribe"})
print(output["text"])
License
Apache-2.0