Search results

[Backport][NVFP4] ds4-sm120-* hardcodes Mxfp4MoEMethod — DeepSeek-V4-Flash-NVFP4 fails to load on SM120 (works via ModelOpt→Marlin)

Status: Open.

#24 In jasl/vllm;

· danielwoz opened on Jun 30, 2026

[Bug]: "2+2等于几“回答None

bugSomething isn't workingSomething isn't working

Status: Open.

#22 In jasl/vllm;

· janboeye opened on Jun 22, 2026

[Bug]: DSA persistent_topk crashes on GB10/sm121 at long context -- FilteredTopK fallback needs >=128KB smem but sharedMemPerBlockOptin is 101376B; fix by honoring optin smem on the cooperative path

bugSomething isn't workingSomething isn't working

Status: Open.

#21 In jasl/vllm;

· SignoreUnito opened on Jun 20, 2026

GLM-5.1 NVFP4 (REAP): DSA lightning-indexer weight load fails — input_global_scale key has no destination (indexer built unquantized)

Status: Open.

#20 In jasl/vllm;

· SeedSource opened on Jun 13, 2026

[Build][spinloop] USE_SABI 3.11 breaks build on Python 3.10 — Py_buffer not in stable ABI until 3.11

Status: Open.

#11 In jasl/vllm;

· pasta-paul opened on May 24, 2026

[Installation]: DeepGEMM will not compile: unsupported CUDA architecture 12.0a

Status: Open.

#9 In jasl/vllm;

· sscckk opened on May 22, 2026

[Performance]: Does it support the 2 x Pro 6000 x 8 architecture for running DeepSeek v4 Pro

Status: Open.

#5 In jasl/vllm;

· ywtll opened on May 11, 2026

[Performance]: Does it support the SM120 architecture of the RTX5090?

Status: Open.

#4 In jasl/vllm;

· wuxiaohui0 opened on May 9, 2026

[Feature]: Integrate sm8x/sm11x support into #40991

Status: Open.

#3 In jasl/vllm;

· ehfd opened on May 6, 2026