Model Doesn’t Load Properly in ComfyUI — Parts Not Auto-Assembling
I’ve downloaded all 7 .safetensors parts along with the config.json and index.json for the WAN 2.1 I2V 14B model. All files are in the same folder and named correctly, but ComfyUI is not recognizing them as a single model. Instead, it shows each .safetensors part separately.
This makes the model unusable unless there’s a hidden merge step that hasn’t been documented.
Please provide:
A version with a single .safetensors file
OR
Clear step-by-step instructions for getting the multi-part setup working in ComfyUI
Btw not sure if this is for all users, but having them separate does not help, but hinders downloading. Downloading one file takes a long time - yes, but all of them separate at the same time will resulting on one of them timing and and failing to download. You end up just waiting on one to download so you can start the next.
Hi. I had a similar problem. Basically, have you tried running it not with ComfyUI, but via command line or something? Without interface. I haven't tried it yet, but I'd like to know.
I haven’t tried running it via command line yet, just through ComfyUI. I do know there are tools and scripts out there to merge the model parts, but honestly, it feels like I’m probably just missing one tiny step or doing something slightly wrong. What’s frustrating is I haven’t found anyone clearly outlining the setup process for these multi-part .safetensors models in ComfyUI — so I’m kind of in the dark and just guessing at this point.
If you end up trying it outside ComfyUI and it works, definitely let me know , though currently I'm simply using the Wan2_1-I2V-14B-720P_fp8_e5m2.safetensors file instead.
It's been three months, is anyone planning to fix this?