Sync from GitHub via hub-sync

CLAUDE.md

@@ -88,7 +88,7 @@ Legend: ✅ production-ready · ⚠️ works only with a required pinned image
 | `dots-ocr-1.5.py` | ✅ | vLLM 0.17.1 | l4x1 | see gotcha (`content_format`, mirror, bbox space) |
 | `glm-ocr.py` | ✅ | vLLM (nightly) | l4x1 | `VLLM_USE_DEEP_GEMM=0`; no pyarrow cap |
 | `ovis-ocr2.py` | ✅ | vLLM (stable ≥0.22.1) | l4x1 / a10g-small | `gdn_prefill_backend="triton"` (card); card-exact prompt via `enable_thinking=False` template + `llm.generate`; `<img>` region tags filtered by default (`--keep-image-tags`) |
-| `xiaomi-ocr-0-saturate.py` | 🧪 | vLLM or SGLang + saturate | a10g-small / rtx-pro-6000 | single-stage only; OTSL→HTML + math delimiters vendored from upstream; `max_tokens` 12000 to fit the 16384 context; default `Auto()` window |
+| `xiaomi-ocr-0-saturate.py` | ✅ | vLLM or SGLang + saturate | a10g-small / rtx-pro-6000 | single-stage only; OTSL→HTML + math delimiters vendored from upstream; `max_tokens` 12000 to fit the 16384 context; default `Auto()` window |
 | `glm-ocr-v2.py` | 🧪 | vLLM (nightly) | l4x1 | CommitScheduler incremental — on hold (see Deferred) |
 | `nanonets-ocr.py` | ✅ | vLLM | a10g-small | `--max-model-len` 32768 (`--max-tokens` 15000) |
 | `nanonets-ocr2.py` | ⚠️+image | vLLM `:v0.10.2` | a10g-small | Qwen2.5-VL ≥0.11 regression → pure `!` |
@@ -344,6 +344,7 @@ ARM wheels) — if a nightly-recipe install fails on resolution, wait and retry
 
 ## Change log
 
+- **2026-11-02** — `xiaomi-ocr-0-saturate.py` moved to `core`. NASA demo: 7/7 in 31 s; checked pages read well, but a table lost one item and some cells lost separators. 16 multilingual historical newspaper pages: 16/16 with no error rows, but repetition loops on 2 small-print German Fraktur pages (lines about 10–13 px after the 2048 px fit), a partial loop on a third, and Chinese characters inserted on 3 pages. `finish_reason` stayed `stop` on all of them.
 - **2026-10-30** — added `xiaomi-ocr-0-saturate.py` for `xiaomi-research/Xiaomi-OCR-0` (0.8B, Qwen3.5 arch), on
   saturate `==0.3.0`. Do not set `max_limit` by copying older saturate recipes: on `rtx-pro-6000`, `Auto(max_limit=48)`
   gave 4.7 pages/s and the default `Auto()` with `prepare_workers` gave 9.7 (1,200 pages, 95% GPU, 0 failures).

README.md

@@ -87,7 +87,7 @@ These are the maintained recipes. Each one has a tested `[tool.hf-jobs]` header,
 | [`glm-ocr.py`](https://huggingface.co/datasets/uv-scripts/ocr/blob/main/glm-ocr.py) | [GLM-OCR](https://huggingface.co/zai-org/GLM-OCR) | 0.9B | 94.62 OmniDocBench v1.5 | MIT | `a10g-small` |
 | [`paddleocr-vl-1.6.py`](https://huggingface.co/datasets/uv-scripts/ocr/blob/main/paddleocr-vl-1.6.py) | [PaddleOCR-VL-1.6](https://huggingface.co/PaddlePaddle/PaddleOCR-VL-1.6) | 0.9B | 96.33 OmniDocBench v1.6; six task modes | Apache-2.0 | `a10g-small` |
 | [`ovis-ocr2.py`](https://huggingface.co/datasets/uv-scripts/ocr/blob/main/ovis-ocr2.py) | [OvisOCR2](https://huggingface.co/ATH-MaaS/OvisOCR2) | 0.9B | 96.58 OmniDocBench v1.6; LaTeX and HTML tables | Apache-2.0 | `a10g-small` |
-| [`xiaomi-ocr-0-saturate.py`](https://huggingface.co/datasets/uv-scripts/ocr/blob/main/xiaomi-ocr-0-saturate.py) | [Xiaomi-OCR-0](https://huggingface.co/xiaomi-research/Xiaomi-OCR-0) | 0.8B | Whole-page Markdown with LaTeX and HTML tables on vLLM or SGLang; the card's benchmark scores use a two-stage pipeline this recipe does not run (experimental) | Apache-2.0 | `a10g-small` |
+| [`xiaomi-ocr-0-saturate.py`](https://huggingface.co/datasets/uv-scripts/ocr/blob/main/xiaomi-ocr-0-saturate.py) | [Xiaomi-OCR-0](https://huggingface.co/xiaomi-research/Xiaomi-OCR-0) | 0.8B | Whole-page Markdown with LaTeX and HTML tables on vLLM or SGLang; the card's benchmark scores use a two-stage pipeline this recipe does not run; it can loop or insert Chinese characters on small-print historical newspapers | Apache-2.0 | `a10g-small` |
 | [`lighton-ocr2.py`](https://huggingface.co/datasets/uv-scripts/ocr/blob/main/lighton-ocr2.py) | [LightOnOCR-2-1B](https://huggingface.co/lightonai/LightOnOCR-2-1B) | 1B | 83.2 olmOCR-Bench | Apache-2.0 | `a10g-small` |
 | [`hunyuan-ocr-1.5.py`](https://huggingface.co/datasets/uv-scripts/ocr/blob/main/hunyuan-ocr-1.5.py) | [HunyuanOCR-1.5](https://huggingface.co/tencent/HunyuanOCR) | 1B | 12 task types, including spotting, charts and translation | [Hunyuan Community](https://huggingface.co/tencent/HunyuanOCR/blob/main/LICENSE) (excludes EU, UK, South Korea) | `a10g-small` |
 | [`minicpm-ocr.py`](https://huggingface.co/datasets/uv-scripts/ocr/blob/main/minicpm-ocr.py) | [MiniCPM-OCR 1.0](https://huggingface.co/openbmb/MiniCPM-OCR-1.0) | 1.3B | OpenBMB's full pipeline: layout regions, per-region prompts, Markdown with HTML tables; Hub dataset or Bucket in/out; runs on a public [`cr.hf.co`](https://huggingface.co/containers/davanstrien/minicpm-ocr) image (experimental) | Apache-2.0 | `a10g-large` |

models.json

@@ -264,8 +264,8 @@
     "languages": {
       "evidence": "not-stated"
     },
-    "support": "experimental",
-    "support_note": "New on 2026-10-30; tested on one English scan collection only.",
+    "support": "core",
+    "support_note": "Works well on clean print. On small-print historical newspapers it can loop or insert Chinese characters, with finish_reason still stop; check tables for lost items and separators.",
     "jobs": {
       "image": "vllm/vllm-openai:v0.29.0",
       "flavor": "a10g-small",

xiaomi-ocr-0-saturate.py

@@ -66,6 +66,9 @@ Model: xiaomi-research/Xiaomi-OCR-0 (0.8B, Apache-2.0)
 - Upstream behaviour kept: a paragraph (text between blank lines) that holds an OTSL
   table tag is replaced by the converted table, so text in the same paragraph is lost.
   `raw` keeps the model's full output.
+- Limits seen in testing: on small-print historical newspapers the model can loop or
+  insert Chinese characters, and `finish_reason` stays "stop". Tables can lose an item
+  or the separators between items in a cell. Check a sample of your own pages.
 
 The SERVING dict below is the per-model tuning prior. Agents can `ast.literal_eval`
 it without running the script; the script itself consumes it.