Repository Issue Activity (beta feature)

#

#

#2990 [Bug]: Use SageAttention backend Wan2.2 and Hunyuan-Video Quality Crash

#2992 [Rebase][Bug] Wan 2.1 / Wan 2.2 with --enable-layerwise-offload crashes with CUDA illegal memory access during dummy warmup (surfaced on dev/vllm-align rebase)

#

#2931 [Bug]: Chat completions omit diffusion profiler metrics; BAGEL img2img breaks on GLM-style mm_processor_kwargs

#

#

#1830 [CI Failure]: Qwen3-omni，occasionally, the generated audio and text are inconsistent.

#

#2347 [Roadmap][Feature] Support Moore Threads (MUSA) GPUs

#

#2568 [RFC] TTS Concurrent Streaming: Root Cause Analysis and Optimization Plan

#

#

#

#2993 [Rebase][Bug] Qwen2.5-Omni talker CUDA graph capture fails with cudaErrorStreamCaptureUnsupported on dev/vllm-align rebase (suspected PyTorch upgrade regression)

#

#2957 [RFC]: Stabilize the lora_path contract and expose a lora_loader seam in DiffusionLoRAManager

#

#

#

#2994 [Rebase][Bug] Performance regression across Qwen-Image and Qwen-Image-Edit benchmarks on dev/vllm-align rebase (throughput below / latency above baseline)

#

#