Skip to content

docs(README): remove degenerate DFlash perf row from #85 perf table #88

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
solderzzc merged 1 commit into from
Apr 26, 2026
Merged

docs(README): remove degenerate DFlash perf row from #85 perf table #88

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
3 changes: 1 addition & 2 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,13 +80,12 @@ Benchmark results for full-RAM (no SSD streaming) MoE inference on M1 Ultra. The
| Configuration | Short (~126 tok) | Medium (~400 tok) | Long (~800 tok) |
|---|---|---|---|
| **Vanilla full-GPU** | **61.7 tok/s** | **62.3 tok/s** | **62.1 tok/s** |
| `--dflash` (block_size=16) † | 52.3 tok/s | **70.3 tok/s** (+13%) | **69.9 tok/s** (+13%) |

> *Hardware:* Apple M1 Ultra, 64 GB unified memory, macOS 26.x. Model ~20 GB on disk, ~21.6 GB resident weight + ~2.1 GB KV at runtime.
> *Flags:* `--repeat-penalty 1.1 --max-tokens 2000`, `temperature: 0.6`, single-stream `/v1/chat/completions`.
> *Vanilla baseline before* `needsMoeFlush` *gate (for reference):* 19.2 / 18.1 / 18.3 tok/s — see #84.

† DFlash uses [`z-lab/Qwen3.6-35B-A3B-DFlash`](https://huggingface.co/z-lab/Qwen3.6-35B-A3B-DFlash) (~948 MB) as the block-diffusion draft model. DFlash gives a clean +13% on medium/long generations but regresses short prompts (block overhead doesn't amortize at low token counts) and changes stop-condition behavior (`finish_reason=null` vs `stop`/`length`). Recommend a quality eval before using as default.
> ⚠️ **DFlash on this model is currently unsuitable for production.** DFlash uses pure greedy (`argMax`) decoding regardless of `temperature`, which on Qwen3.6-35B-A3B + the [`z-lab/Qwen3.6-35B-A3B-DFlash`](https://huggingface.co/z-lab/Qwen3.6-35B-A3B-DFlash) draft locks into low-entropy attractors (`"and and and..."`, `"**UMA** **UMA**..."`). Earlier 70 tok/s DFlash numbers were degenerate output that scored high acceptance because draft and target both committed to the same locked-in token. Repetition-penalty mitigation works on some prompts but tanks acceptance on others — the proper fix is stochastic posterior sampling with rejection-based accept ([Leviathan/Chen](https://arxiv.org/abs/2211.17192) formulation), which is a DFlash architecture change tracked at [z-lab/dflash#91](https://github.com/z-lab/dflash/issues/91).

### DeepSeek-V4-Flash (126 GB, Q3-mixed-gs128-affine) — M5 Pro 64 GB

Expand Down
Loading