All pull requests

Pull requests

• loader: fail when a completed build retains meta tensors

  #317·liminfei-amd (liminfei-amd) opened Sep 22, 2026·

  ·

• fix(kernels): build on rolling-release toolchains (CUDA_PATH veto, POSIX macro clash)

  #313·ManuelFCastillo opened Sep 14, 2026·

  ·
  1

• docs: OffloadMode.NONE VRAM figure applies to the 19B LTX-2, not LTX-2.5

  #312·Keith (stayinalive181) opened Sep 10, 2026·

  ·

• Update LICENSE file to include references to both LTX-2.x and LTX-2 Community License Agreements and remove the obsolete LICENSE.md file.

  #299·Alex Artyomov (alexanderar) opened Aug 24, 2026·

  ·

• Fix dependency resolution on macOS / Apple Silicon

  #298·medyadanhasan opened Aug 24, 2026·

  ·

• Handle transposed fp8 weights in LoRA fusion

  #293·masahiroteraoka opened Aug 20, 2026·

  ·

• Reference conditioning: RoPE layouts and per-source rotary phase

  #289·NRDX (alisson-anjos) opened Aug 17, 2026·

  ·

• pipelines: forward --quantization to the multi-GPU runners

  #280·Boning (wwelder) opened Aug 13, 2026·

  ·

• Preserve fractional frame rates when encoding SDR video

  #279·Syed Osama Ali Shah (Osamaali313) opened Aug 12, 2026·

  ·
  1

• fix(trainer): restore valid ruff target-version

  #265·blue (Blueyyds) opened Aug 2, 2026·

  ·

• fix(kernels): don't fail all2all build on CUDA 12.8 deprecation warnings

  #264·blue (Blueyyds) opened Aug 2, 2026·

  ·

• fix(trainer): correct LR schedule, final checkpoint and STG under DDP

  #263·blue (Blueyyds) opened Aug 2, 2026·

  ·

• Ampere (sm_86) compatibility for fp8-cast LoRA fusion and mono-audio input in a2vid

  #253·Larry Bowden (fastflair) opened Jul 11, 2026·

  ·

• README: Update outdated url

  #238·Giuseppe Silvano (giusilvano) opened Jun 18, 2026·

  ·

• Limit seq_length to max_tokens in timestep_samplers

  #230·lsl0007 opened Jun 8, 2026·

  ·

• perf(trainer): add tile_batch_size, cut VAE calls up to 15x in tiled encoding

  #215·Vittoria Lanzo (VittoriaLanzo) opened May 14, 2026·

  ·
  1

• Add keyframe-to-video training strategy

  #209·yizhiqianbi opened May 4, 2026·

  ·

• Free Gemma text encoder after caption embedding to reduce VRAM during training

  #85·bruefire (bruefire) opened Jan 21, 2026·

  ·

• Bugfix: fix memory allocation issues during multi-GPU training

  #67·LittleNyima (LittleNyima) opened Jan 14, 2026·

  ·

• 8Gb Vram optimized, Win11 support + Web UI + optimized transformer +40% more frames on 8Gb

  #66·nalexand opened Jan 14, 2026·

  ·
  7

• feat: Add deterministic inference pipeline for reproducible video generation

  #28·q5sys (JT) (q5sys) opened Jan 8, 2026·

  ·
  1

• Small README edits

  #20·Jack Saunders (jsaunders909) opened Jan 7, 2026·

  ·