[Feature] Development Roadmap 2025 Q4/2026 Q1

[SolitaryThinker]

Motivation

Contritbutions are welcome!

Focus

• Diverse model support with training and inference
• Launch RL training infra and an effective training recipe
• New Distillation Recipes
• Quantization Support
• Improve training performance and infra
• CI/CD

Diverse model support with training and inference

• [ ] Longcat
  • https://github.com/hao-ai-lab/FastVideo/pull/883
• [ ] Cosmos2.5
  • https://github.com/hao-ai-lab/FastVideo/pull/897
• [ ] Hunyuanvideo Training Pipeline
  • https://github.com/hao-ai-lab/FastVideo/pull/754
• [ ] Hunyuanvideo-v1.5
• [ ] Lucy-Edit-Dev
• [ ] LTX
  • https://github.com/hao-ai-lab/FastVideo/pull/783
  • https://github.com/hao-ai-lab/FastVideo/pull/876

New Distillation Recipes

• [ ] Self-Forcing for Wan2.2 MoE
  • https://github.com/hao-ai-lab/FastVideo/pull/880
• [ ] Rolling-Forcing support

Launch RL training infra and an effective training recipe

• [ ] FlowGRPO

Preprocessing & Dataset

• [ ] Enable training with no precomputed latents

Training Infra

• [ ] Activation offloading
• [ ] Multi-resolution and variable length bucketing for long video input compatibility
  • https://github.com/hao-ai-lab/FastVideo/pull/894
• [ ] LoRA extraction, verification, and comparison scripts
  • https://github.com/hao-ai-lab/FastVideo/pull/865
• [ ] USP for Self-Forcing
• [ ] Ring Attention

CI/CD

• [ ] Add MFU into CI
  • https://github.com/hao-ai-lab/FastVideo/pull/871
• [ ] Add FVD into CI

Related resources

No response