lmdeploy · Issues· 597 open
Open on GitHubLocally synced open issues (discussions stay on GitHub)
- #4958
[Feature] 现在lmdeploy能支持模型GLM-5.3-Flash-W4A16-MTP运行吗？
Updated Sep 17, 2026 - #4962
[Bug]
awaiting responseUpdated Sep 15, 2026 - #4972
[Bug] Proxy connection warmup can accumulate unbounded PD connection wait tasks and OOM
Updated Sep 15, 2026 - #4967
[Bug] DistServe Proxy requests can leak Prefill scheduler metadata and OOM the Prefill engine
Updated Sep 15, 2026 - #4965
[Bug] Untrusted `migration_request` can terminate the DistServe Decode EngineLoop and cause persistent denial of service
Updated Sep 15, 2026 - #4971
[Bug] Mooncake async migration can hang forever and make Decode unavailable
Updated Sep 15, 2026 - #4698
[Bug] /update_weights deserializes request-controlled pickle data before validation
Updated Sep 14, 2026 - #4953
[Bug] Slower decode speed in v0.17 when compared to v0.14
Updated Sep 14, 2026 - #4960
[Enhance] Control token should not be treated as special token in user messages
Updated Sep 14, 2026 - #4933
[Bug] input_embeddings are omitted from prefix-cache identity, causing wrong KV reuse
Updated Sep 13, 2026 - #4905
[Feature] 咱们最新版本的lmdeploy可以加载这个 Qwen3.8-Flash吗？
Updated Sep 11, 2026 - #4944
[Bug] TurboMind SIGSEGV crash-loop triggered by session_len truncation path (deterministic libc offset across 3 crashes)
Updated Sep 8, 2026 - #4930
[Bug] Mooncake external KV keys omit KV format and weights lineage, with no default tenant isolation
Updated Sep 3, 2026 - #4824
[Feature] DeepSeek-V4 MoE is Blackwell-only via DeepGEMM — is a non-Blackwell path wanted?
Updated Sep 1, 2026 - #4530
[Feature] support DFlash: Block Diffusion for Flash Speculative Decoding
planned featureUpdated Sep 1, 2026