[Bugfix][Frontend][Model] Per-model serving defaults, DeepSeek-OCR resolution modes, and SWA async-eviction fix (SM70)

[kkobold]

What

Serving-quality improvements for long-generation OCR workloads, plus a
correctness fix for prefix-anchored SWA:

1. Bugfix (prefix-anchored SWA): out-of-window gap blocks were freed
   on the scheduled-token basis; under async scheduling this can free
   blocks the worker has not yet computed past. Free on the
   processed-token basis instead. Regression tests cover the async gap.
2. Per-model serving defaults with request-override bounds: an
   architecture-keyed registry applied at the chat endpoint. A model
   declares its recommended sampling/extra-args recipe; requests may
   override within declared bounds. DeepSeek-OCR registers its checkpoint
   README recipe (n-gram blocker 30/90 with whitelist tokens, temperature
   0, max_tokens 8192) so an unconfigured client gets the model's intended
   behavior via plain chat completions.
3. Architecture-keyed chat template fallbacks: DeepSeek-OCR's
   model_type previously fell through to the DeepSeek-VL2 template, which
   injects role markers the OCR checkpoint was not trained with; the
   fallback now pins the raw-concatenation prompt (golden test included).
4. Repetition-detection default for OCR serving: a bounded
   exact-pattern detector as a safety net for degenerate table pages,
   request-overridable.
5. Per-request DeepSeek-OCR resolution modes: the checkpoint's five
   official modes (tiny/small/base/large/gundam) selectable via
   mm_processor_kwargs: {"image_mode": ...}; multi-image serving
   profile; max_crops plumbing. The serving default remains the
   checkpoint's gundam configuration.

Why

On document OCR, the model's own recipe is load-bearing: without the
n-gram blocker and the correct raw prompt, long pages degenerate into
loops. These changes make the endpoint serve the checkpoint's intended
recipe by default while keeping every knob per-request-overridable within
bounds. The SWA fix closes an async-scheduling correctness gap in that
capability (#281).

Validation

All device work on Tesla V100-SXM2 32 GB (sm_70), fp16, on current main.

• 195 CPU tests green, including the prefix-anchored SWA
  admission/mask/manager suites, the new eviction-basis regression tests,
  resolution-mode suite, template golden, defaults/bounds and
  repetition-detection suites.
• Endpoint (serve) on V100: deterministic 3x byte-identical generations;
  resolution-mode discrimination (default = gundam as identity control;
  base/large produce distinct image-token counts); document-page scoring
  against known ground truth per mode.

Serving DeepSeek-OCR (validated configuration)

Launch line used for every endpoint validation in this PR (Tesla
V100-SXM2 32 GB, fp16):

vllm serve /path/to/DeepSeek-OCR \
  --dtype float16 \
  --max-model-len 8192 \
  --gpu-memory-utilization 0.85 \
  --enforce-eager \
  --no-enable-prefix-caching \
  --mm-processor-cache-gb 0 \
  --logits-processors vllm.model_executor.models.deepseek_ocr:NGramPerReqLogitsProcessor

• --logits-processors loads the n-gram blocker class; the serving-defaults
  registry supplies its per-request parameters and fails startup with the
  exact flag text if the class is missing (never silent looping).
• --no-enable-prefix-caching / --mm-processor-cache-gb 0 are the
  checkpoint recipe's engine flags.
• For multi-image requests add --limit-mm-per-prompt '{"image":8}'
  (the SM70 baseline defaults to 1 image per prompt).
• --enforce-eager is the validated configuration; CUDA graphs measured
  ~4x faster decode at ~1% window cost and can be enabled after a
  graph-mode validation pass.
• Tensor parallelism: the checkpoint has 10 attention heads, so
  --tensor-parallel-size must divide 10 (1 or 2 on a 4-GPU host); 4
  fails at startup with the head-divisibility error.

Merge base

Branch applies and merges cleanly onto current main at 34403018d
([Kernel][SM70] Optimize E4M3 page800 long decode, #290); all validation in
this PR ran on that base.

[yangzhuxinyzx]

已取消归档并解锁，但原作者在归档前删除了外部 head 分支，GitHub 因而无法原编号重开。私有备份中的原始 head SHA 已无改动恢复为公开 Draft 延续 PR：https://github.com/1CatAI/1Cat-vLLM/pull/352。#292 本身保持可见、未锁定和 Closed，以保留原作者关闭记录。