Sanitize public model card metadata

README.md

@@ -1,80 +1,18 @@
 ---
-license: mit
+license: "mit"
 tags:
-- qhexrt
-- hexagon
-- npu
-- qnn
-- asr
-base_model: UsefulSensors/moonshine-tiny
-pipeline_tag: automatic-speech-recognition
+- "hnpu"
+- "hexagon"
+- "npu"
+- "asr"
+base_model: "UsefulSensors/moonshine-tiny"
+pipeline_tag: "automatic-speech-recognition"
 ---
 
-# Moonshine-tiny → Qualcomm Hexagon NPU (QHexRT)
+# moonshine tiny HNPU
 
-Prebuilt **QNN context binaries** for running [`UsefulSensors/moonshine-tiny`](https://huggingface.co/UsefulSensors/moonshine-tiny)
-speech-to-text **on-device on the Qualcomm Hexagon NPU** via the [QHexRT](https://github.com/RunanywhereAI/QHexRT)
-runtime. **Arch: `v81`** (SM8850 / Snapdragon 8 Elite Gen 5). Device-validated at **WER = 0** vs HF.
+Prebuilt HNPU artifacts for [UsefulSensors/moonshine-tiny](https://huggingface.co/UsefulSensors/moonshine-tiny), a public automatic speech recognition model.
 
-A context binary is **arch-pinned** (the `dsp_arch` + `soc_model` are baked in) — these `v81/` bins won't load
-on another Hexagon arch. Other arches are sibling `<arch>/` dirs (re-converted), added to this same repo.
+For model behavior, license, intended use, and limitations, see the [upstream model card](https://huggingface.co/UsefulSensors/moonshine-tiny).
 
-## How it runs
-
-Moonshine is an encoder–decoder ASR model on the **raw 16 kHz waveform** (no mel spectrogram). The pipeline:
-- **Host** runs the 3-conv raw-audio stem (`conv1 k127/s64 + tanh → GroupNorm → conv2 k7/s3 + gelu →
-  conv3 k3/s2 + gelu`) — the conv is HTP-hostile so it stays on the CPU; the encoder graph starts at the
-  post-conv features.
-- **NPU encoder** (bidirectional, partial interleaved RoPE) → per-decoder-layer cross-attention K/V.
-- **NPU decoder** (one autoregressive step: causal self-attn + RoPE in-graph, cross-attn to the cached
-  encoder states, gated SwiGLU; tied lm-head) → tokens, detokenized on the host.
-
-Variable-length audio is handled on a fixed graph (`n_audio = 415`, ~10 s window): the host pads/truncates
-the features and masks the padding (`encoder_mask` / `cross_mask`). Precision **fp16** (encoder + decoder).
-
-## Files (`v81/`)
-
-| file | role | size |
-|---|---|---|
-| `moonshine-tiny.json` | QHexRT manifest (declarative run plan; `moonshine_transcribe` host-op) | ~1 KB |
-| `moonshinetiny_enc_f16.bin` | encoder context binary | 17.5 MB |
-| `moonshinetiny_dec_f16.bin` | decoder context binary | 56.5 MB |
-| `moonshine_conv_stem.bin` | host raw-audio conv-stem weights `[c1w,c2w,c2b,c3w,c3b,gnw,gnb]` | 6.8 MB |
-| `tokenizer.json` | SentencePiece-style BPE (byte_fallback; metaspace detok) | 3.8 MB |
-
-The QNN runtime libs (`libQnnHtp.so` / `libQnnSystem.so` + the v81 HTP skel) come from the **QAIRT SDK**, not
-this repo. The `qhx_asr` tool comes from a QHexRT build.
-
-## Run
-
-```bash
-hf download runanywhere/moonshine_tiny_HNPU --local-dir moonshine_tiny_HNPU
-# Windows: adb push from PowerShell with native paths.
-adb push moonshine_tiny_HNPU/v81 /data/local/tmp/wq/moonshine
-adb push my_audio_16k_mono.wav /data/local/tmp/wq/moonshine/
-adb shell "cd /data/local/tmp/wq && export ADSP_LIBRARY_PATH='/data/local/tmp/wq/dsp;/data/local/tmp/wq;/vendor/dsp/cdsp'; \
-  LD_LIBRARY_PATH=. ./qhx_asr moonshine/moonshine-tiny.json libQnnHtp.so libQnnSystem.so moonshine moonshine/my_audio_16k_mono.wav"
-```
-
-Tool arg order is invariant: `qhx_asr <manifest> libQnnHtp.so libQnnSystem.so <artifacts_root> <audio16k.wav>`.
-Input audio must be **16 kHz mono** (PCM16 or float32).
-
-## Measured (v81, SM8850, soc_model 87, QAIRT 2.47)
-
-| metric | value |
-|---|---|
-| Parity | **WER = 0.0000** vs HF `UsefulSensors/moonshine-tiny` (LibriSpeech sample) |
-| Latency | 26 tokens in **551 ms** for a 5.9 s clip |
-| Precision | fp16 encoder + decoder |
-
-Example: a 5.9 s clip → *"Mr. Quilter is the apostle of the middle classes, and we are glad to welcome his
-gospel."* (matches HF exactly).
-
-## Caveats
-
-- **v81 only** here (arch-pinned). **fp16** weights. Audio window ~10 s (`n_audio = 415`); longer clips are
-  truncated to the window.
-- Parity is greedy (temperature 0) vs the HF reference. WER measured on a standard LibriSpeech sample.
-
-Converted + device-validated with the [QHexRT `forge`](https://github.com/RunanywhereAI/QHexRT) pipeline
-(`recipes/moonshine-tiny`).
+Artifacts are architecture-pinned. Available artifact directories: `v75/`, `v79/`, `v81/`.