| schema_version: 1.0.0 | |
| report_type: speech-asr | |
| task_type: automatic-speech-recognition | |
| title: Whisper Small INT8 — ExecuTorch + XNNPACK | |
| description: >- | |
| A compact INT8-quantized speech recognition model for transcribing 16 kHz audio, using an | |
| encoder-decoder Transformer that converts log-mel spectrograms into text tokens. It combines dynamic | |
| INT8 linear layers with selective FP32 components to reduce size and improve inference efficiency | |
| while preserving transcription quality. | |
| id: Arm/whisper-small-int8-xnnpack-executorch | |
| filename: whisper-small-int8-executorch.pte | |
| base_model_id: openai/whisper-small | |
| profile: Arm-Optimized | |
| weight_dtype: int8 | |
| quantization: | |
| method: PTQ-dynamic | |
| weight_bits: 8 | |
| activation_bits: 8 | |
| symmetric: true | |
| mode: dynamic | |
| weight_granularity: per-channel | |
| variant: 8da8w + embed_tokens weight-only int8 | |
| fp32_layers: | |
| - embed_positions | |
| model_size_mb: 395.046 | |
| parameter_count: 241734912 | |
| format: pte | |