Skip to content
Closed
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
46 changes: 43 additions & 3 deletions docs/arch-gaps.md
Original file line number Diff line number Diff line change
Expand Up @@ -1041,7 +1041,10 @@ takes the reviewed set: such a pack prints `. FLM ... registry: none (reviewed,
Still open, and alerting on purpose: `MiniCPMV4_7ForConditionalGeneration` (needs one converter registration, not an
architecture), `MiniCPMOcrForConditionalGeneration` (own `minicpm-ocr` library), `smolvla-NPU2` (no config at all), and
the `!! NPU FLM` families whose registry mapping is HRX-only (`hy_v3`, `dflash`, `gemma4`, `hunyuan-dense`, `gemma3`) —
that last group is engine #588 (the family must be served on the NPU route), not a registry edit.
that last group is **not a registry edit**: the family has to be served on the NPU route. It used to be recorded here as
engine #588, which is **closed** (2026-10-27) and was about *watching FLM's releases*, not about serving a family on the
route — so this class currently has **no owning issue**, and an arrival that lands in it is recorded on this page and
nowhere else. See the 2026-11-10 section below for the arrivals that are sitting in it.

## Census run 2026-11-09 (`engine #1667`)

Expand Down Expand Up @@ -1072,9 +1075,11 @@ Left as they are, on purpose, under the policy above: the six `. REMOTE-CODE` cl
(`HajeriOperatorForCausalLM`, `MM200ForCausalLM`, `K2HorizonForCausalLM`, `KLMForCausalLM`,
`KimiK31SuperExpertForCausalLM`, `ThalamusReasoningQwen3_5ForConditionalGeneration`); the `? UNVERIFIABLE` gated repos,
which are retried on every run; and the `!! FLM` lines. Those keep the 2026-11-08 reasons: `MiniCPMV4_7`, `MiniCPMOcr`
and `smolvla` as above, and `!! NPU FLM` is engine #588. This run's issue window also includes the `gpt-oss`
and `smolvla` as above, and `!! NPU FLM` was recorded as engine #588 — which is **closed** (2026-10-27) and was
about watching FLM's releases, not about serving a family on the NPU route, so that class now has no owning issue
(corrected 2026-11-10; see the section at the end of this page). This run's issue window also includes the `gpt-oss`
(`GPT-OSS-20B`, `GPT-OSS-Safeguard-20b`) and `lfm2` (`LFM2-1.2B`, `LFM2-2.6B`) Q4NX packs. Their mapping is HRX-only
too, so they belong to #588, not to a registry edit.
too, so they are the same class — a route to serve, not a registry edit.

## 2026-11-10: `DeepseekV4FlashSuperExpertForCausalLM` — reviewed, and recorded here

Expand All @@ -1087,3 +1092,38 @@ the single undocumented entry out of 37.
MLA-style geometry (`kv_lora_rank` 512, `q_lora_rank` 1536, `qk_rope_head_dim` 64, `v_head_dim` 128, 32 heads /
16 KV, IM 4096). **Not an alias of `deepseek_v41`**: nothing at the pin decodes AW-SECD weights, so mapping it to
that architecture would name a backend that cannot read the file. Reported, not alerted (`"watch": "report"`).

## FLM 1.0.12's HRX build: four family roots with no row (engine#1787, 2026-11-10)

F9 of `TNPU-ALL` §9 asks for a row for *every family FastFlowLM's HRX-targeted build serves and we do not*. The build is
on the box, so the checklist was enumerated from it rather than from the release notes:

```
$ ls ~/flm-f9/xclbins | wc -l -> 54 (per-family xclbin directories)
$ ls ~/flm-f9/lib/*.so | wc -l -> 34 (per-family shared objects)
```

Mapping each of the 54 by **family root** (not by model name: `Gemma3-1B` normalises to `gemma31b`, which appears nowhere,
and matching that way reported 53 of 54 as unknown) against `registry/*.json`, `docs/registry.md` and this page:

* **48** roots have a trace somewhere — spot-checked rather than assumed: `d3` → `D3LM` in `census.json`, `gemma3` →
`Gemma3ForCausalLM` in `architectures.json`, `qwen3.5` → `Qwen3_5ForConditionalGeneration`, `lfm2` → `Lfm2ForCausalLM`.
* **6** xclbins over **4** family roots have **no trace at all**:

| xclbin (FLM 1.0.12 HRX) | family root | state |
|---|---|---|
| `LFM2.5-1.2B` | `lfm2.5` | no row, no registry entry. We map `Lfm2ForCausalLM`, and `registry/significant.json` already carries the note *"lfm2-shaped config, different tensors"*, so a 2.5 generation is exactly the arrival this page exists to catch |
| `LFM2.5-1.2B-Thinking` | `lfm2.5` | as above |
| `Medgemma-1.5-4B` | `medgemma` | no row, no registry entry |
| `Medgemma-4B` | `medgemma` | as above |
| `Nanbeige4.1-3B` | `nanbeige4.1` | no row, no registry entry |
| `Translategemma-4B-Instruct` | `translategemma` | no row, no registry entry |

**These rows are un-inspected, and say so on purpose.** Nothing above describes what any of the four families computes: the
xclbin names are evidence that *their* build carries a compiled family for it, and that is all. §1's five conditions are
what converts an entry from *listed* to *settled* — imports with zero gaps and serving on the device, a known-answer prompt
under the model's own template with three identical answers, decode tok/s and TTFT under a ticket, and the kernel set and
engine commit named. Writing an architecture description from a model name is the defect this week has been about; a row
that is honest about being un-inspected is more useful than a plausible paragraph.

The `!! NPU FLM` class above is where these belong once someone owns it. As of this run that class has no owning issue.
Loading