Issues

• FastMetal-14B-QAD hangs before first denoise step on M4 Max (1.3B/5B fine; GPU command buffer never completes)

  installationperformanceplatformscope: attentionscope: docsscope: inferencescope: kernelscope: model
  hao-ai-lab/FastVideo#2016 · Filter by author YYC-Cube opened on Oct 24, 2026

• Local trailing-slash and HF cache snapshot paths select the wrong config/preset

  installationplatformscope: attentionscope: inferencescope: kernelscope: model
  hao-ai-lab/FastVideo#1996 · Filter by author Jackmeson1 opened on Oct 20, 2026

• CFG validation accepts negative_prompt=None with empty embeddings, then raises a bare assertion

  installationplatformscope: attentionscope: inferencescope: kernelscope: model
  hao-ai-lab/FastVideo#1995 · Filter by author Jackmeson1 opened on Oct 20, 2026

• output_video_name survives request preparation but does not select the output filename

  installationscope: attentionscope: inferencescope: kernelscope: model
  hao-ai-lab/FastVideo#1994 · Filter by author Jackmeson1 opened on Oct 20, 2026

• [Bug] SP communication warmup hardcodes device="cuda", breaking every non-CUDA accelerator

  installationplatformscope: attentionscope: distributedscope: inference
  hao-ai-lab/FastVideo#1987 · Filter by author li-lizhe opened on Oct 19, 2026

• [RFC]: Causal Wan training with sequence parallelism repeats the non-attention work on every rank

  installationperformanceplatformscope: attentionscope: distributedscope: kernelscope: modelscope: training
  hao-ai-lab/FastVideo#1950 · Filter by author zmy1260222041 opened on Oct 11, 2026
  1

• Migration plan for the new FastVideo cookbook

  installationperformanceplatformscope: attentionscope: docsscope: inferencescope: modelscope: training
  hao-ai-lab/FastVideo#1949 · Filter by author leo2ultimate opened on Oct 9, 2026
  2

  

• [Bug] tests/test_fasth3_packaging.py: stale TORCH_CUDA_ARCH_LIST assertions fail since #1833

  buginstallationplatformscope: attentionscope: kernel
  hao-ai-lab/FastVideo#1928 · Filter by author SolitaryThinker opened on Oct 7, 2026
  1

• [RFC]: Environment variable policy: typed registry enforced by a contract test

  installationscope: attentionscope: distributedscope: docsscope: inferencescope: kernelscope: modelscope: training
  hao-ai-lab/FastVideo#1893 · Filter by author Davids048 opened on Sep 29, 2026

• [Feature] Cookbook: adding OpenAI API serving recipes for the other model families

  scope: attentionscope: model
  hao-ai-lab/FastVideo#1889 · Filter by author YZJF opened on Sep 25, 2026
  3

• [Bug] MPS generation fails when multiprocessing worker returns output tensors

  installationperformanceplatformscope: attentionscope: inferencescope: model
  hao-ai-lab/FastVideo#1846 · Filter by author leo2ultimate opened on Sep 14, 2026
  5

• [Feature] Add Dreamverse mode selection for T2VA, FL2VA, and Ref2VA

  installationplatformscope: attentionscope: docsscope: inferencescope: model
  hao-ai-lab/FastVideo#1834 · Filter by author jayzou3773 opened on Sep 9, 2026
  1

  

• Support FSDP inference with precomputed quantized weights

  scope: attentionscope: distributedscope: inferencescope: training
  Task

  hao-ai-lab/FastVideo#1822 · Filter by author Davids048 opened on Sep 5, 2026
  

• Support LoRA lifecycle operations for MXFP8 and NVFP4 inference

  scope: inferencescope: training
  hao-ai-lab/FastVideo#1821 · Filter by author Davids048 opened on Sep 5, 2026
  1

• [kernel] [Feature] NVFP4 QAT: fix broken measurement plumbing and build MFU foundations (pre-work for DGX Spark QAT kernel optimization)

  installationplatformscope: docsscope: kernelscope: training
  hao-ai-lab/FastVideo#1722 · Filter by author yogyam opened on Aug 19, 2026
  1

• [Bug] Dreamverse 9:16 Rendering Artifact

  hao-ai-lab/FastVideo#1713 · Filter by author Someone45 opened on Aug 17, 2026
  2

• [Bug] MiniMax H3 text encoder needs ~97 GB resident to load a 63 GB bf16 checkpoint(FAS-494)

  installationperformanceplatformscope: attentionscope: docsscope: inferencescope: model
  hao-ai-lab/FastVideo#1709 · Filter by author KyleNeverGivesUp opened on Aug 15, 2026
  3

• [RFC]: Command "dreamverse-server --host 0.0.0.0 --port 8009": GPU 0 warmup failed

  installationscope: attentionscope: inferencescope: training
  hao-ai-lab/FastVideo#1685 · Filter by author nanwanj opened on Aug 6, 2026
  1

• [ci] [dashboard]: add explicit benchmark cohort comparison mode

  platformscope: attention
  hao-ai-lab/FastVideo#1636 · Filter by author Satyam-53 opened on Jul 21, 2026

• [ci] [dashboard]: add advanced hardware and software profile filters

  installationscope: attentionscope: docsscope: model
  hao-ai-lab/FastVideo#1635 · Filter by author Satyam-53 opened on Jul 21, 2026

• [ci] [dashboard]: replace automatic all-cohort trend rendering with a cohort overview

  performancescope: attentionscope: inference
  hao-ai-lab/FastVideo#1634 · Filter by author Satyam-53 opened on Jul 21, 2026

• [ci] [dashboard]: add exact benchmark cohort selection and clarify configuration identity

  installationscope: attentionscope: docsscope: model
  hao-ai-lab/FastVideo#1633 · Filter by author Satyam-53 opened on Jul 21, 2026
  

• [RFC]: Questions about VSA weight provenance & infer VSA_TILE_SIZE mismatch

  installationplatformscope: attentionscope: inferencescope: kernelscope: modelscope: trainingstale
  hao-ai-lab/FastVideo#1606 · Filter by author bitzyz opened on Jul 15, 2026
  2

• Dev Roadmap 2026 Q3

  scope: attentionscope: docsscope: inferencescope: kernelscope: modelscope: training
  hao-ai-lab/FastVideo#1603 · Filter by author SolitaryThinker opened on Jul 15, 2026
  2

• [CI] Investigate Vanilla training references under Modal GPU substitution

  scope: attentionscope: distributedscope: trainingstale
  hao-ai-lab/FastVideo#1592 · Filter by author macthecadillac opened on Jul 12, 2026
  3