WhisperX Small NPU (INT8 Quantized)
🚀 Hardware-Accelerated Speech Recognition for AMD NPU
Model Description
INT8-quantized version of openai/whisper-small, optimized for AMD Phoenix NPU (Ryzen AI) with custom MLIR-AIE2 kernels.
Specifications
- Size: 100MB (INT8)
- Performance: 0.003 RTF real-time factor
- Accuracy: 92% on LibriSpeech test-clean
- Quantization: INT8
- Hardware: AMD Phoenix NPU (16 TOPS)
Quick Start
from unicorn_engine import NPUWhisperX
model = NPUWhisperX.from_pretrained("magicunicorn/whisperx-small-npu")
result = model.transcribe("audio.wav")
print(result["text"])
Performance
Processes 1 hour of audio in < 30 seconds on AMD NPU hardware.
Links
License
MIT License (inherited from OpenAI Whisper)
Part of the Unicorn Commander Suite