12 Pull requests merged by 4 people

• update nexus-gen readme

  #568 merged May 13, 2025

• Step1x vram

  #556 merged Apr 28, 2025

• support step1x

  #555 merged Apr 27, 2025

• Flex

  #553 merged Apr 25, 2025

• liblib-controlnet

  #548 merged Apr 22, 2025

• Update train_wan_t2v.py

  #544 merged Apr 17, 2025

• bugfix

  #543 merged Apr 17, 2025

• Wan flf2v

  #541 merged Apr 17, 2025

• issue523

  #537 merged Apr 16, 2025

• support vace quant

  #536 merged Apr 16, 2025

• Update train_wan_t2v.py

  #526 merged Apr 16, 2025

• vace

  #533 merged Apr 15, 2025

1 Pull request opened by 1 person

• Migrate to modern Python Logger API

  #567 opened May 12, 2025

11 Issues closed by 8 people

• How diffsynth studio vs engine different??

  #564 closed May 13, 2025

• flex_text_to_image

  #559 closed May 2, 2025

• Wan 2.1 FLF model support?

  #557 closed Apr 30, 2025

• wan2.1 14B I2V lora training [model:Wan2.1-I2V-14B-720P]

  #552 closed Apr 25, 2025

• Add support for Wan FLF2V training

  #551 closed Apr 24, 2025

• 双卡运行视频推理服务卡住

  #542 closed Apr 20, 2025

• 运行wan_14b_image_to_video.py脚本报错：TypeError: DiTBlock.forward() missing 1 required positional argument: 'freqs'

  #545 closed Apr 18, 2025

• wan 14b i2v 训练lora后，输入帧保持能力丢失

  #540 closed Apr 17, 2025

• Error during inference: Input type (struct c10::BFloat16) and bias type (struct c10::Float8_e4m3fn) should be the same

  #535 closed Apr 16, 2025

• Wan I2V image encoder has bug when encoding png images. The alpha channel creates dim mismatch in matrix operation.

  #528 closed Apr 16, 2025

• Is teacache supported for Vace

  #534 closed Apr 15, 2025

20 Issues opened by 18 people

• 有关sample_shift, sample_steps取值问题

  #572 opened May 14, 2025

• Fine-tuning the Flux fill model for image editing tasks

  #571 opened May 14, 2025

• Are there plans to integrate fantasytalking?

  #570 opened May 13, 2025

• How to load pre-trained vae model and dit model? Is it supported? Any modification suggestions?

  #569 opened May 13, 2025

• [New] Model/Pipeline HiDream-E1

  #566 opened May 9, 2025

• 是否支持图生720P视频的训练任务

  #565 opened May 9, 2025

• how 14B T2V full training? on 80GB H100 gpu

  #563 opened May 1, 2025

• Timestep changes when use bfloat16

  #562 opened May 1, 2025

• Is prompt_emb[:, v:] = 0 intentional or a mistake?

  #561 opened Apr 30, 2025

• video background become dark using Wan2.1 , which is different from origin wan2.1 repo

  #560 opened Apr 29, 2025

• Wan2.1 FLF2V 首尾帧训练微调 / 多关键帧微调

  #558 opened Apr 28, 2025

• Will provide the training and fine-tuning code for the VACE algorithm?

  #554 opened Apr 27, 2025

• About FramePack

  #550 opened Apr 22, 2025

• 什么时候基于Wan2.1支持不同尺寸视频训练？

  #549 opened Apr 22, 2025

• Can we urgently get SkyReels-V2? this model is definitely SOTA right now

  #547 opened Apr 21, 2025

• [WAN 2.1 14B I2V + Lora] Is it possible to switch between different lora without re-loading base model?

  #546 opened Apr 19, 2025

• 请问有开源wan i2v 的motion control模块的计划吗？看到有刚开源了wan t2v的motion control

  #539 opened Apr 16, 2025

• 使用多卡进行 Full training，遇到 “torch.distributed.DistNetworkError: The client socket has timed out after 1800000ms while trying to connect to”的问题

  #538 opened Apr 16, 2025

• 8卡4090*48g, lora微调wan2.1 14B I2V模型OOM，怎么解决？

  #532 opened Apr 15, 2025

• How to choose the resolution of training data in Wan2.1

  #531 opened Apr 15, 2025

15 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• 1.3B I2V training issue

  #524 commented on Apr 16, 2025 • 0 new comments

• 使用wan的时候，不同的分辨率需要使用不同的tiled_size吗

  #525 commented on Apr 16, 2025 • 0 new comments

• Can we implement OptimalSteps to Wan 2.1 pipe? they claim 20 steps is equal to default 100 steps

  #529 commented on Apr 16, 2025 • 0 new comments

• fp8 vs bf16

  #530 commented on Apr 16, 2025 • 0 new comments

• Run the wan14B text-to-image model to generate a 21-frame video, with the first few frames being blurry and the later frames being normal.

  #519 commented on Apr 16, 2025 • 0 new comments

• want gguf for wan video

  #512 commented on Apr 16, 2025 • 0 new comments

• Upload examples via this issue

  #1 commented on Apr 17, 2025 • 0 new comments

• wan_video_dit issue with flash_attn 3

  #523 commented on Apr 17, 2025 • 0 new comments

• How to use torch.float8_e4m3fn and lora together?

  #527 commented on Apr 17, 2025 • 0 new comments

• Sage Attention vs Flash Attention Speed Comparison with Wan 2.1 - Sage Attention is 37% Faster - 720p - 14b model - tested on Windows Python VENV - no WSL

  #484 commented on Apr 18, 2025 • 0 new comments

• We cannot detect the model type. No models are loaded.

  #494 commented on Apr 22, 2025 • 0 new comments

• Does anyone train a working Wan t2v lora from the t2v script of this repo?

  #367 commented on Apr 29, 2025 • 0 new comments

• support 14B full training？

  #422 commented on May 2, 2025 • 0 new comments

• Inquiry About Wan-I2V Training/Inference Performance on A6000 GPUs

  #503 commented on May 2, 2025 • 0 new comments

• about batch size in Wan I2V training

  #501 commented on May 7, 2025 • 0 new comments