MiniMax H3 controlnet adaln width 8 does not match the base model's timestep embedding width 2688
- Dominant language
- Python
- Stars
- 133k
- Forks
- 15.7k
- Avg merge
- 1d 7h
- Merged PRs (30d)
- 158
Description
### Custom Node Testing
- [ ] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)
### Expected Behavior
no error.
### Actual Behavior
base model: minimax_h3_ref2va_bf16.safetensor
controlnet: minimax_h3_fun_controlnet_union_pruned_bf16.safetensors
comfyui official workflow template
An error has occurred: File "D:\ai\self_use_package\ComfyUI_windows_portable\ComfyUI\comfy\patcher_extension.py", line 106, in __call__
return new_executor.execute(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ai\self_use_package\ComfyUI_windows_portable\ComfyUI\comfy\patcher_extension.py", line 113, in execute
return self.original(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ai\self_use_package\ComfyUI_windows_portable\ComfyUI\comfy\ldm\minimax\model.py", line 759, in _forward
h = blocks_replace[("double_block", i)](
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ai\self_use_package\ComfyUI_windows_portable\ComfyUI\comfy_extras\nodes_minimax_h3.py", line 553, in __call__
return self.control_patch.after_block(self.block_index, args, out)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ai\self_use_package\ComfyUI_windows_portable\ComfyUI\comfy_extras\nodes_minimax_h3.py", line 502, in after_block
self.control_stream = self.model_patch.model.init_stream(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ai\self_use_package\ComfyUI_windows_portable\ComfyUI\comfy\ldm\minimax\controlnet.py", line 47, in init_stream
raise RuntimeError(
RuntimeError: MiniMax H3 controlnet adaln width 8 does not match the base model's timestep embedding width 2688: the controlnet and base checkpoint use different adaln forms (curve basis vs full), convert the controlnet to match the base model.
### Steps to Reproduce
base model: minimax_h3_ref2va_bf16.safetensor
controlnet: minimax_h3_fun_controlnet_union_pruned_bf16.safetensors
comfyui official template
### Debug Logs
```powershell
File "D:\ai\self_use_package\ComfyUI_windows_portable\ComfyUI\comfy\patcher_extension.py", line 106, in __call__
return new_executor.execute(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ai\self_use_package\ComfyUI_windows_portable\ComfyUI\comfy\patcher_extension.py", line 113, in execute
return self.original(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ai\self_use_package\ComfyUI_windows_portable\ComfyUI\comfy\ldm\minimax\model.py", line 759, in _forward
h = blocks_replace[("double_block", i)](
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ai\self_use_package\ComfyUI_windows_portable\ComfyUI\comfy_extras\nodes_minimax_h3.py", line 553, in __call__
return self.control_patch.after_block(self.block_index, args, out)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ai\self_use_package\ComfyUI_windows_portable\ComfyUI\comfy_extras\nodes_minimax_h3.py", line 502, in after_block
self.control_stream = self.model_patch.model.init_stream(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ai\self_use_package\ComfyUI_windows_portable\ComfyUI\comfy\ldm\minimax\controlnet.py", line 47, in init_stream
raise RuntimeError(
RuntimeError: MiniMax H3 controlnet adaln width 8 does not match the base model's timestep embedding width 2688: the controlnet and base checkpoint use different adaln forms (curve basis vs full), convert the controlnet to match the base model.
```
### Other
_No response_
Contributor guide
Research direction
Start with comfy/ldm/minimax/controlnet.py at init_stream, then trace the call from comfy_extras/nodes_minimax_h3.py and comfy/ldm/minimax/model.py. Reproduce with the two named checkpoints and the official workflow, then determine whether the controlnet/base combination can run without the reported width error and add or update coverage for the supported behavior.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 42/100