All issues

Search results

[Bug]: SM80 router logits round through BF16 for M>8; backport upstream #54048

Status: Open.

#65 In wtdcode/vllm-backport;

· Morrowmake opened on Sep 8, 2026

[Performance]: Qwen3.8-Flash-Next AWQ on 4xA100: benchmark configuration needed to reproduce 936 tps

Status: Open.

#64 In wtdcode/vllm-backport;

· 77598072 opened on Sep 8, 2026

[Bug]: lmcache cant enable on latest-sm80

bugSomething isn't workingSomething isn't working

Status: Open.

#61 In wtdcode/vllm-backport;

· dogethdev opened on Sep 4, 2026

[Feature]:

Status: Open.

#60 In wtdcode/vllm-backport;

· grantchenhuarong opened on Sep 4, 2026

[Feature] DFlash2

Status: Open.

#59 In wtdcode/vllm-backport;

· wtdcode opened on Sep 3, 2026

[Feature]: GitHub Container Registry

Status: Open.

#53 In wtdcode/vllm-backport;

· halexan opened on Sep 1, 2026

[Usage]: Request for the exact CLI command and environment variables used for the 8xA6000 (TP4PP2) benchmark

Status: Open.

#52 In wtdcode/vllm-backport;

· momo5511 opened on Aug 31, 2026

[Feature]: DeepSeek-V4-Flash-Vision-Exp

Status: Open.

#48 In wtdcode/vllm-backport;

· halexan opened on Aug 31, 2026

[Feature]:Request for cu129-based Docker images like vllm/vllm-openai

Status: Open.

#46 In wtdcode/vllm-backport;

· qinggangwu opened on Aug 31, 2026

[Feature]: Add FP8 GLM-5.3-Flash support for sm89

Status: Open.

#39 In wtdcode/vllm-backport;

· Pragmatism0220 opened on Aug 29, 2026

[Feature]: glm-5.3 support in sm80 a100

Status: Open.

#38 In wtdcode/vllm-backport;

· veritynoob opened on Aug 29, 2026

[Bug]:[NIXL][PD] After days of uptime the UCX connection dies (NIXL_ERR_REMOTE_DISCONNECT), and the HMA failure path keeps serving HTTP 200 responses with corrupted tokens

bugSomething isn't workingSomething isn't working

Status: Open.

#33 In wtdcode/vllm-backport;

· me040526 opened on Aug 27, 2026