Do we have roadmap

[nitinmukesh]

Thank you for sharing the project.

Do we have roadmap, like what features are you planning to support.
int4, GGUF, support and examples.

[FurkanGozukara]

Thank you for sharing the project.

Do we have roadmap, like what features are you planning to support. int4, GGUF, support and examples.

yes GGUF and int4 would make this project different than others

I requested it for Wan 2.1 several times on DiffSynth-Studio :(

[yingdachen]

gguf is supported with DiffSynth-Engine.

[FurkanGozukara]

gguf is supported with DiffSynth-Engine.

So we have to move from DiffSynth-Studio to DiffSynth-Engine right?

Becuase i tested and DiffSynth-Studio doesn't support

[nitinmukesh]

gguf is supported with DiffSynth-Engine.

Please do provide some examples when you get time. No rush. Thank you

[FurkanGozukara]

I am willing to test GGUF asap if the syntax similar to DiffSynth-Studio

e.g.

this model manager from DiffSynth-Studio works with single safetensors files as well but not GGUF

GGUF not working : https://huggingface.co/city96/Wan2.1-T2V-14B-gguf/tree/main

single tensor i tested it works : https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/diffusion_models

        clip_path = get_common_file(os.path.join("models", "models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth"),
                                    os.path.join("models", "Wan-AI", "Wan2.1-I2V-14B-720P", "models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth"))
        t5_path = get_common_file(os.path.join("models", "models_t5_umt5-xxl-enc-bf16.pth"),
                                  os.path.join("models", "Wan-AI", "Wan2.1-I2V-14B-720P", "models_t5_umt5-xxl-enc-bf16.pth"))
        vae_path = get_common_file(os.path.join("models", "Wan2.1_VAE.pth"),
                                  os.path.join("models", "Wan-AI", "Wan2.1-I2V-14B-720P", "Wan2.1_VAE.pth"))
        model_manager.load_models([clip_path], torch_dtype=torch.float32)
        model_manager.load_models(
            [
                [
                    os.path.join("models", "Wan-AI", "Wan2.1-I2V-14B-720P", "diffusion_pytorch_model-00001-of-00007.safetensors"),
                    os.path.join("models", "Wan-AI", "Wan2.1-I2V-14B-720P", "diffusion_pytorch_model-00002-of-00007.safetensors"),
                    os.path.join("models", "Wan-AI", "Wan2.1-I2V-14B-720P", "diffusion_pytorch_model-00003-of-00007.safetensors"),
                    os.path.join("models", "Wan-AI", "Wan2.1-I2V-14B-720P", "diffusion_pytorch_model-00004-of-00007.safetensors"),
                    os.path.join("models", "Wan-AI", "Wan2.1-I2V-14B-720P", "diffusion_pytorch_model-00005-of-00007.safetensors"),
                    os.path.join("models", "Wan-AI", "Wan2.1-I2V-14B-720P", "diffusion_pytorch_model-00006-of-00007.safetensors"),
                    os.path.join("models", "Wan-AI", "Wan2.1-I2V-14B-720P", "diffusion_pytorch_model-00007-of-00007.safetensors"),
                ],
                t5_path,
                vae_path,
            ],
            torch_dtype=torch_dtype,
        )  

[akaitsuki-ii]

For now, GGUF is only available for Flux DiT and T5 model with Diffsynth-Engine, please have a check at the example here. An known issue is the fetch_model method will download all files under the model repo even with specified path, which will be fixed in later version.

[FurkanGozukara]

For now, GGUF is only available for Flux DiT and T5 model with Diffsynth-Engine, please have a check at the example here. An known issue is the fetch_model method will download all files under the model repo even with specified path, which will be fixed in later version.

Thank. When can we expect for Wan 2.1? There are already comfyui nodes fully support perhaps it can make it easier for you. I am willing to test if you implement

[akaitsuki-ii]

@nitinmukesh We don't have a long-term roadmap for the project right now. But we did have several work items planned for the next few weeks, including optimization for video generation (e.g. sequence parallel, teacache) and better documentation. Will update the README after internal discussion.

[akaitsuki-ii]

For now, GGUF is only available for Flux DiT and T5 model with Diffsynth-Engine, please have a check at the example here. An known issue is the fetch_model method will download all files under the model repo even with specified path, which will be fixed in later version.

Thank. When can we expect for Wan 2.1? There are already comfyui nodes fully support perhaps it can make it easier for you. I am willing to test if you implement

Yes, I think it‘s not difficult to enable GGUF for Wan2.1. Let me have a try later this week.

[FurkanGozukara]

For now, GGUF is only available for Flux DiT and T5 model with Diffsynth-Engine, please have a check at the example here. An known issue is the fetch_model method will download all files under the model repo even with specified path, which will be fixed in later version.

Thank. When can we expect for Wan 2.1? There are already comfyui nodes fully support perhaps it can make it easier for you. I am willing to test if you implement

Yes, I think it‘s not difficult to enable GGUF for Wan2.1. Let me have a try later this week.

awesome looking forward too it. i hope you can first implement to DiffSynth-Studio

[akaitsuki-ii]

GGUF for Wan2.1 is available now, please have try by installing from source (we are going to make new release later) and follow the example here. You could find more wan and umt5 GGUF format models on modelscope. Ping me if any problem.

[FurkanGozukara]

GGUF for Wan2.1 is available now, please have try by installing from source (we are going to make new release later) and follow the example here. You could find more wan and umt5 GGUF format models on modelscope. Ping me if any problem.

awesome i need to check code and implement into DiffSynth-Studio and test