Upload README.md with huggingface_hub

README.md

@@ -12,18 +12,60 @@ language:
 pipeline_tag: automatic-speech-recognition
 ---
 
-# whisper-small — QHexRT NPU bundle (Hexagon v79)
+# whisper-small — QHexRT NPU bundle (Hexagon v79 + v81)
 
-Precompiled **Whisper-small** ASR for the **QHexRT** runtime on Qualcomm Hexagon **v79**
-(Snapdragon 8 Elite / SM8750, e.g. Galaxy S25). More accurate than
-[whisper-base](https://huggingface.co/runanywhere/whisper_base_HNPU) at ~3× the size and ~1.3×
-slower decode (still comfortably real-time). Encoder + decoder are Qualcomm AI Hub
-`qnn_context_binary` (float/fp16) graphs; the host pipeline is QHexRT's own. Device-validated:
-transcription matches the HF `openai/whisper-small` reference exactly.
+Precompiled **Whisper-small** ASR (12 layers, d_model 768) for the **QHexRT** runtime on the Qualcomm Hexagon
+NPU. Two arch-pinned bundles ship as sibling dirs:
 
-Measured on S25/v79: 718 ms for 2.67 s of audio (≈3.7× real-time) vs base's 546 ms (≈5×).
+| dir | Hexagon arch | device | encoder path | device-validated |
+|-----|--------------|--------|--------------|------------------|
+| `v79/` | v79 (SM8750, Galaxy S25) | Snapdragon 8 Elite | AI-Hub **in-graph** conv | ✅ transcript == HF |
+| `v81/` | v81 (SM8850, soc 87) | Snapdragon 8 Elite Gen 5 | **forge host-conv** (`encoder_features`) | ✅ **WER 0** |
 
-## Contents (`v79/`)
+A context binary won't load on another arch (`soc_model` + `dsp_arch` baked in). The same `qhx_asr` runs both
+bundles (it branches on the encoder input). The `asr_transcribe` orchestrator is size-agnostic — base/small
+differ only in these bins + the manifest dims.
+
+---
+
+## `v81/` — Hexagon v81 (SM8850), compiled with **forge** (QAIRT 2.47)
+
+Same host-conv decomposition as whisper-base (the HTP-hostile conv stem runs host-side from
+`whisper_conv_stem.bin`; the encoder graph starts at post-conv `encoder_features [1500,768]`). fp16 encoder +
+decoder. This bundle is a **new forge port** — the spec (`recipes/whisper-small`) is generated from the same
+exporter that produced the validated whisper-base v81, then device-gated on hardware.
+
+### Files (`v81/`)
+| file | role |
+|------|------|
+| `whisper-small.json` | QHexRT manifest (`asr_transcribe` plan; 12 layers / d768 / 12 heads) |
+| `whispersmall_enc_f16.bin` | encoder (post-conv `encoder_features` → per-layer cross-attn K/V) |
+| `whispersmall_dec_f16.bin` | autoregressive decoder (self+cross attn, in-graph int32 embed + tied lm-head) |
+| `whisper_conv_stem.bin` | host-side conv-stem weights `[c1w,c1b,c2w,c2b,pos]` |
+| `whisper_small_mel_filters.bin` | HF 80-mel filter bank (host log-mel) |
+| `tokenizer.json` | Whisper tokenizer (vocab 51865) |
+
+### Run (`v81/`)
+```bash
+huggingface-cli download runanywhere/whisper_small_HNPU --local-dir whisper_small_HNPU
+adb push whisper_small_HNPU/v81 /data/local/tmp/wq/whisper_small   # QNN libs + v81 HTP skel from the QAIRT SDK
+adb shell "cd /data/local/tmp/wq && LD_LIBRARY_PATH=. ADSP_LIBRARY_PATH=. \
+  ./qhx_asr whisper_small/whisper-small.json libQnnHtp.so libQnnSystem.so whisper_small whisper_small/<audio16k>.wav"
+```
+
+### Measured (SM8850 / v81, soc 87, QAIRT 2.47)
+- **WER = 0** vs HF `openai/whisper-small` on a clean LibriSpeech clip *and* a 23 s clip (decode runs the full
+  length). Encoder cross-KV cosine **1.00000**, decoder 12/12 teacher-forced + greedy chain MATCH (export gate).
+
+---
+
+## `v79/` — Hexagon v79 (SM8750, Galaxy S25), Qualcomm AI Hub graphs
+
+Encoder + decoder are Qualcomm AI Hub `qnn_context_binary` (float/fp16) graphs; the host pipeline is QHexRT's
+own. Device-validated: transcription matches the HF `openai/whisper-small` reference exactly. Measured on
+S25/v79: 718 ms for 2.67 s of audio (≈3.7× real-time).
+
+### Files (`v79/`)
 | file | what |
 |------|------|
 | `whisper-small.json` | QHexRT manifest (ASR family, `asr_transcribe` plan; 12 layers, d_model 768) |
@@ -32,18 +74,7 @@ Measured on S25/v79: 718 ms for 2.67 s of audio (≈3.7× real-time) vs base's 5
 | `whisper_mel_filters.bin` | HF 80-mel filter bank `[201,80]` f32 (shared across whisper sizes) |
 | `tokenizer.json` | Whisper multilingual tokenizer (vocab 51865, shared) |
 
-## Run (QHexRT CLI)
-```bash
-huggingface-cli download runanywhere/whisper_small_HNPU --local-dir whisper_small_HNPU
-# QNN libs come from the QAIRT SDK (lib/aarch64-android) + the v79 HTP skel; push them next to qhx_asr.
-adb push whisper_small_HNPU/v79 /data/local/tmp/wq/whisper_small
-adb shell "cd /data/local/tmp/wq && LD_LIBRARY_PATH=. ADSP_LIBRARY_PATH=. \
-  ./qhx_asr whisper_small/whisper-small.json libQnnHtp.so libQnnSystem.so whisper_small whisper_small/<audio16k>.wav"
-```
-Audio: mono WAV (PCM16 or float32); resampled to 16 kHz host-side. Clips ≤ 30 s.
+Audio: mono WAV (PCM16 or float32); resampled to 16 kHz host-side. Clips ≤ 30 s. No custom op-package needed.
+Source model: `openai/whisper-small`.
 
-## Notes
-- Arch: **v79** only (context binaries are dsp-arch-pinned).
-- No custom op-package needed — pure-native AI Hub graphs.
-- The QHexRT `asr_transcribe` orchestrator is size-agnostic; base/small/tiny differ only in these bins + manifest dims.
-- Source model: `openai/whisper-small`, compiled via Qualcomm AI Hub for `qualcomm-snapdragon-8-elite-for-galaxy`.
+_v81 bundle built + device-validated with [QHexRT](https://github.com/RunanywhereAI/QHexRT) `forge` — `recipes/whisper-small`._