Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cuda:1! (when checking argument for argument tensors in method wrapper_CUDA_cat)

[Sxrjs2010]

I am using kaggle with 2 T4 GPUs

Error occurred when executing Joy_caption:

Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cuda:1! (when checking argument for argument tensors in method wrapper_CUDA_cat)

File "/kaggle/working/ComfyUI/execution.py", line 317, 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 "/kaggle/working/ComfyUI/execution.py", line 192, 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 "/kaggle/working/ComfyUI/execution.py", line 169, in _map_node_over_list process_inputs(input_dict, i) File "/kaggle/working/ComfyUI/execution.py", line 158, in process_inputs results.append(getattr(obj, func)(**inputs)) File "/kaggle/working/ComfyUI/custom_nodes/Comfyui_CXH_joy_caption/Joy_caption_node.py", line 175, in gen inputs_embeds = torch.cat([

Help please. Thanks!

[q5sys]

I ran into this and I patched the python to load everything onto one CUDA device. https://github.com/StartHua/Comfyui_CXH_joy_caption/pull/23 You can pull the edited file from my repo and see if this works for you as well.

[Sxrjs2010]

@q5sys It works. Thanks a lot!