publish whisper-tiny OpenASR packs

.gitattributes

@@ -0,0 +1 @@
+*.oasr filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,114 @@
+---
+license: apache-2.0
+base_model: openai/whisper-tiny
+pipeline_tag: automatic-speech-recognition
+library_name: openasr
+tags:
+  - automatic-speech-recognition
+  - speech-to-text
+  - openasr
+  - oasr
+  - whisper-tiny
+---
+
+<div align="center">
+
+# Whisper Tiny · OpenASR
+
+**The smallest multilingual Whisper for fast local transcription**
+
+[![License](https://img.shields.io/badge/license-Apache--2.0-2563eb.svg)](https://huggingface.co/openai/whisper-tiny/blob/main/README.md)
+[![Format](https://img.shields.io/badge/format-.oasr-7c3aed.svg)](https://github.com/QuintinShaw/openasr)
+[![Runtime](https://img.shields.io/badge/runtime-OpenASR-111827.svg)](https://openasr.org)
+[![Base model](https://img.shields.io/badge/base-whisper--tiny-f59e0b.svg)](https://huggingface.co/openai/whisper-tiny)
+
+Native speech-to-text in the **[OpenASR](https://github.com/QuintinShaw/openasr)** runtime —
+engineered for peak performance on CPU & GPU, **no Python at inference time**.
+
+</div>
+
+---
+
+## ✨ Highlights
+
+- 🎧 **Multilingual ASR** — transcribes many languages and can translate speech to English
+- ⚡ **39M parameters** — the smallest Whisper checkpoint, the fastest and lightest to run
+- 🌐 **Weak-supervision scale** — trained with Whisper's 680k-hour labelled speech corpus
+- 🦀 **Native in OpenASR** — `.oasr` packs run with no Python at inference, engineered for peak performance on CPU & GPU
+
+## 🚀 Quickstart
+
+```bash
+# 1. Install the OpenASR CLI  ·  https://openasr.org
+# 2. Pull a build (pick a quant — see the table below)
+openasr pull whisper-tiny:q8
+
+# 3. Transcribe
+openasr transcribe audio.wav --model whisper-tiny
+```
+
+All builds for this model:
+
+```bash
+openasr pull whisper-tiny:fp16
+openasr pull whisper-tiny:q8
+openasr pull whisper-tiny:q4
+```
+
+## 📦 Available builds
+
+| Quant | File (`.oasr`) | Size | RAM peak | RTF · M1 CPU | RTF · M1 GPU | JFK ΔWER vs fp16 |
+|:------|:---------------|-----:|---------:|-------------:|-------------:|-----------------:|
+| fp16 | `whisper-tiny-fp16.oasr` | 79 MB | 321 MB | 0.04× | 0.03× | 0.0% |
+| q8_0 | `whisper-tiny-q8_0.oasr` | 63 MB | 275 MB | 0.04× | 0.03× | 0.0% |
+| q4_k | `whisper-tiny-q4_k.oasr` | 61 MB | 274 MB | 0.04× | 0.03× | 0.0% |
+
+<sub>RTF = real-time factor on the fixed 11s JFK clip (**lower is faster**); RAM peak measured per pack
+in an isolated subprocess. JFK ΔWER compares each quantized build's JFK transcript to this model's
+fp16 JFK transcript, so it measures quantization drift rather than absolute recognition accuracy.
+**q8_0** is the recommended default — near-reference quality at a fraction of the
+footprint.</sub>
+
+## 🧠 About Whisper Tiny
+
+Whisper Tiny is OpenAI's 39M-parameter multilingual Whisper checkpoint, the smallest member of
+the Whisper family. It uses the standard Whisper encoder-decoder architecture for automatic
+speech recognition and speech translation, trained with large-scale weak supervision on 680k
+hours of labelled speech. The tiny model trades some accuracy for the lowest footprint and
+fastest inference, which suits low-resource devices and latency-sensitive use. This OpenASR
+repo repackages the original `openai/whisper-tiny` weights as `.oasr` packs that run natively
+in the OpenASR runtime with no Python at inference time. For most users the q8_0 build is the
+recommended default; q4_k is for the tightest memory budgets and fp16 is for verification or
+maximum fidelity.
+
+## ⚙️ How these packs were made
+
+Converted from [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) with the OpenASR importer:
+
+```bash
+openasr model-pack import-whisper-local <src> <out>.oasr \
+  --package-id whisper-tiny --quantization {fp16,q8-0,q4-k}
+```
+
+The `.oasr` container is GGUF-backed; packs use zero-copy mmap weight binding and graph
+buffer reuse to keep peak memory low.
+
+## ⚖️ License
+
+These packs **inherit the upstream model's license: Apache-2.0**
+([source](https://huggingface.co/openai/whisper-tiny/blob/main/README.md)). OpenASR packaging retains the upstream copyright and
+NOTICE; the only modifications are format conversion and quantization.
+
+## 🙏 Acknowledgements
+
+This pack is a redistribution of **Whisper Tiny**, released by **OpenAI**
+([openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny)).
+All credit for the original model, training recipe, and weights belongs to OpenAI. The
+upstream Hugging Face model card declares Apache-2.0 licensing; OpenASR only converts the
+weights into `.oasr` packages and adds quantized builds for local runtime use.
+
+## 🔗 Links
+
+- 🦀 **OpenASR** — <https://github.com/QuintinShaw/openasr>
+- 🌐 **Website** — <https://openasr.org>
+- 🤗 **Upstream model** — [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny)

whisper-tiny-fp16.oasr

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc3f2c4c1f3e887d8e2e7f614febcdbd2bfe4acd325b104afe2d6c206f3ce60f
+size 78793088

whisper-tiny-q4_k.oasr

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79a947c5060fa981f29e1729ff349946f06dc689bdc02a1808dc8cabe705d853
+size 60950912

whisper-tiny-q8_0.oasr

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13346d7bd47d349ae9e35be5cc8595997be0c2d2a5bbe41b4e3ef1c60a98ca17
+size 63310208