Comfy Workflow?
Can you provide the workflow for these?
Can you provide the workflow for these?
https://github.com/kijai/ComfyUI-MochiWrapper/tree/main/examples
I dont know how to put this on comfy, can someone help?
I did a git clone of https://github.com/kijai/ComfyUI-MochiWrapper in the custom-nodes folder. Then updating comfyui, and restarting, and reloading the workflow json, the boxes all went the right colour - now trying to see if I can find a config that works.
It downloads the models itself, but you can predownload them just so you have a clearer idea of how long it would take, they go into ComfyAI\ComfyUI_windows_portable\ComfyUI\models\diffusion_models\mochi (safetensors, 10 and 20 gig files) and ComfyAI\ComfyUI_windows_portable\ComfyUI\models\vae\mochi (mochi_preview_vae_bf16.safetensors)
can you please help nothing is working for me when trying to run this
i have a m2 mac 32gb
I am attempting to run the 16gb workflow and am getting memory errors
I then changed the vae decoder settings to 16 and batch 1 still same issues
I am attempting to split up the workflows by saving the latent to disk and starting a new flow in order to process it but not able to get the current workflow to save the latent and run it in a separate flow
On my 24GB 3090, it works with the following: Mochi model: Mochi preview dit fp8 e4m3fn; vae bf16; precision fp8; attention mode:comfy; cublas ops:False; Mochi sampler: 848x480 49 frames, 30 steps, 4.5 cfg; mochi vae tiling, tiles w and h 8; overlap 16; min block 1 , per batch 6, feed output as 8 fps.
Prompt outputs failed validation CLIPLoader: - Value not in list: clip_name: 't5\google_t5-v1_1-xxl_encoderonly-fp8_e4m3fn.safetensors' not in [] . How to solve this error?
I often get this if I'm switching between Linux (runpod) and home pc (windows) and it's just the slashes, right click fix node, or rechoose the clip to the one that's available fixes this.