CLIPtionBeamSearch. Expected image features to have 1024 dimensions but got 1280. Please ensure you are using CLIP L.

[Maelstrom2014]

CLIPtionBeamSearch
Expected image features to have 1024 dimensions but got 1280. Please ensure you are using CLIP L.

ComfyUI Error Report

Error Details

• Node ID: 226
• Node Type: CLIPtionBeamSearch
• Exception Type: ValueError
• Exception Message: Expected image features to have 1024 dimensions but got 1280. Please ensure you are using CLIP L.

Stack Trace

  File "C:\ai\comfyui\ComfyUI\execution.py", line 324, in execute
    output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
                                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

  File "C:\ai\comfyui\ComfyUI\execution.py", line 199, in get_output_data
    return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

  File "C:\ai\comfyui\ComfyUI\execution.py", line 170, in _map_node_over_list
    process_inputs(input_dict, i)

  File "C:\ai\comfyui\ComfyUI\execution.py", line 159, in process_inputs
    results.append(getattr(obj, func)(**inputs))
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^

  File "C:\ai\comfyui\ComfyUI\custom_nodes\comfy-cliption\__init__.py", line 552, in caption
    captions = model.generate_beam(image, beam_width, ramble)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

  File "C:\ai\comfyui\ComfyUI\custom_nodes\comfy-cliption\__init__.py", line 214, in generate_beam
    raise ValueError(

System Information

• ComfyUI Version: v0.3.7-6-gac2f052

[patientx]

Tried "CLIP-ViT-bigG-14-laion2B-39B-b160k.safetensors" , got :
Expected image features to have 1024 dimensions but got 1664. Please ensure you are using CLIP L.

Tried "CLIP-ViT-H-14-laion2B-s32B-b79K.safetensors" got :
Expected image features to have 1024 dimensions but got 1280. Please ensure you are using CLIP L.

Tried "clip_l.safetensors" (240mb) from sd3.5 got :
'NoneType' object has no attribute 'model'

Tried "clip_g.safetensors" (1.29gb) from sd3.5 got :
'NoneType' object has no attribute 'model'

I give up.

[pharmapsychotic]

Sorry for the confusion if you have ComfyUI Manager installed you can go to its Model Manager type "clip vision large" into the search and click Install on 139 from OpenAI. After restarting ComfyUI it will then show up as option for the Load CLIP Vision node.
I'll update the readme with some info on the CLIP models in installation step.

[pharmapsychotic]

I updated the readme. Let me know if you're able to get it working now. Thanks for filing the issue and trying it out!

[patientx]

I updated the readme. Let me know if you're able to get it working now. Thanks for filing the issue and trying it out!

when it loads that clipvision model (at least no errors when loading) , comfy exits completely without any errors even in verbose mode.

Requested to load CLIPVisionModelProjection
loaded completely 9.5367431640625e+25 581.267578125 True / normal message as usual

then it kills the app , meaning closes the commandline , I run by hand and saw these last two lines there ...

[billwuhao]

Sorry for the confusion if you have ComfyUI Manager installed you can go to its Model Manager type "clip vision large" into the search and click Install on 139 from OpenAI. After restarting ComfyUI it will then show up as option for the Load CLIP Vision node. I'll update the readme with some info on the CLIP models in installation step.

thank you, 'clip vision large' is working

[sirus20x6]

Error while deserializing header: HeaderTooLarge