card: Shenava-1 v1.0 family

README.md

@@ -5,7 +5,7 @@ license: cc-by-nc-4.0
 library_name: nemo
 pipeline_tag: automatic-speech-recognition
 base_model:
-- Reza2kn/Shenava-Koochik-1.0
+- Reza2kn/Shenava-Koochik-v1.0
 base_model_relation: finetune
 tags:
 - automatic-speech-recognition
@@ -33,14 +33,14 @@ datasets:
 
 # Shenava — Rizeh v1.0 (32M) · Persian streaming ASR
 
-**Rizeh** (ریزه, “tiny”) is the 32M mid‑tier of the **Shenava‑1** family — a FastConformer Hybrid RNNT/CTC model distilled (logit + feature KD) from the [Koochik 1.0 114M teacher](https://huggingface.co/Reza2kn/Shenava-Koochik-1.0). The **CTC head is the deployed/streaming head**. This repo holds the **fp32 NeMo source**; quantized deployment formats live in their own repos (below).
+**Rizeh** (ریزه, “tiny”) is the 32M mid‑tier of the **Shenava‑1** family — a FastConformer Hybrid RNNT/CTC model distilled (logit + feature KD) from the [Koochik v1.0 114M teacher](https://huggingface.co/Reza2kn/Shenava-Koochik-v1.0). CTC head deployed. fp32 NeMo source; quants in own repos below.
 
 ## The Shenava‑1 family
 
 A knowledge‑distillation cascade of on‑device Persian ASR models — one teacher distilled down to a 6.9M student. This model is one member; its siblings:
 
-- [`Reza2kn/Shenava-Koochik-1.0`](https://huggingface.co/Reza2kn/Shenava-Koochik-1.0) — **Koochik 1.0** (114M) · teacher / flagship
-- [`Reza2kn/Shenava-Rizeh-v1.0`](https://huggingface.co/Reza2kn/Shenava-Rizeh-v1.0) — **Rizeh v1.0** (32M) · mid-tier student ◀ **you are here (or its parent)**
+- [`Reza2kn/Shenava-Koochik-v1.0`](https://huggingface.co/Reza2kn/Shenava-Koochik-v1.0) — **Koochik v1.0** (114M) · teacher / flagship — on-device WER record
+- [`Reza2kn/Shenava-Rizeh-v1.0`](https://huggingface.co/Reza2kn/Shenava-Rizeh-v1.0) — **Rizeh v1.0** (32M) · mid-tier student ◀ **this model (or its parent)**
 - [`Reza2kn/Shenava-Rizeh-Pizeh-v1.0`](https://huggingface.co/Reza2kn/Shenava-Rizeh-Pizeh-v1.0) — **Rizeh Pizeh v1.0** (6.9M) · tiniest — real-time on a 2015 Cortex-A7
 
 ## Benchmark — fair WER/CER
@@ -49,15 +49,17 @@ ITN + Persian‑digit normalizer (the [double‑benchmark](https://huggingface.c
 
 | Member | golden‑6669 WER | CER | FLEURS‑fa WER | CER |
 |---|---|---|---|---|
-| Koochik 1.0 (114M) | 7.49% | 2.30% | 10.64% | 3.79% |
+| Koochik v1.0 (114M) | **7.49%** | 2.30% | **10.64%** | 3.79% |
 | Rizeh v1.0 (32M) | 12.11% | 3.94% | 14.45% | 5.10% |
 | Rizeh Pizeh v1.0 (6.9M) | 24.55% | 8.89% | 26.95% | 10.22% |
 
-## Quantized formats (own repos, children of this model)
+Koochik v1.0 is **#2 on the public double‑benchmark, behind only cloud Gemini** — the best on‑device Persian ASR, beating a 1.5B Whisper‑Persian by >2× WER at 1/13 the size.
 
-- [`Shenava-Rizeh-v1.0-ONNX-fp16`](https://huggingface.co/Reza2kn/Shenava-Rizeh-v1.0-ONNX-fp16) — fixed‑window CTC, ONNX fp16
-- [`Shenava-Rizeh-v1.0-CoreML-fp16`](https://huggingface.co/Reza2kn/Shenava-Rizeh-v1.0-CoreML-fp16) — CoreML fp16 mlprogram
+## Quantized formats (own repos)
 
-32M, d_model 256 / 16 layers, dw_striding ×8 (80 ms/frame), multi‑latency `[[70,13],[70,6],[70,1],[70,0]]`.
+- [`Shenava-Rizeh-v1.0-ONNX-fp16`](https://huggingface.co/Reza2kn/Shenava-Rizeh-v1.0-ONNX-fp16)
+- [`Shenava-Rizeh-v1.0-CoreML-fp16`](https://huggingface.co/Reza2kn/Shenava-Rizeh-v1.0-CoreML-fp16)
 
-Tokenizer: ve_tok_v4 (SentencePiece BPE‑1024 +blank, digit/punct/«»‑aware). Numbers are emitted in **spoken form**; apply ITN at display for digits. Part of [VisualEars / Shenava](https://shenava.app).
+32M, d_model 256 / 16 layers, ×8 subsampling, multi‑latency `[[70,13],[70,6],[70,1],[70,0]]`.
+
+Tokenizer: ve_tok_v4 (SentencePiece BPE‑1024 +blank, digit/punct/«»‑aware). Numbers are **spoken‑form**; apply ITN at display for digits. Part of [VisualEars / Shenava](https://shenava.app).