NegPip interference with hidream model (GGUF and possibly others)

[CrHasher]

Expected Behavior

Image is generated with NegPip on or off, HiDream model.

Actual Behavior

Image generation error with NegPip option enabled. Fine with option disabled (so not a major bug).

Steps to Reproduce

1. Install NegPipforSwarmUI extension
2. Select HiDream model (I'm using this GGUF: hidream-i1-dev-Q6_K.gguf)
3. Toggle option on 'Use NegPip'
4. Error on HiDream

Debug Logs

2025-06-11 21:47:58.824 [Debug] [ComfyUI-0/STDERR] !!! Exception during processing !!! Number of dimensions of repeat dims can not be smaller than number of dimensions of tensor
2025-06-11 21:47:58.830 [Warning] [ComfyUI-0/STDERR] Traceback (most recent call last):
2025-06-11 21:47:58.831 [Warning] [ComfyUI-0/STDERR]   File "D:\SwarmUI\dlbackend\comfy\ComfyUI\execution.py", line 349, in execute
2025-06-11 21:47:58.831 [Warning] [ComfyUI-0/STDERR]     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)
2025-06-11 21:47:58.832 [Warning] [ComfyUI-0/STDERR]                                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-06-11 21:47:58.832 [Warning] [ComfyUI-0/STDERR]   File "D:\SwarmUI\dlbackend\comfy\ComfyUI\execution.py", line 224, in get_output_data
2025-06-11 21:47:58.832 [Warning] [ComfyUI-0/STDERR]     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)
2025-06-11 21:47:58.833 [Warning] [ComfyUI-0/STDERR]                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-06-11 21:47:58.833 [Warning] [ComfyUI-0/STDERR]   File "D:\SwarmUI\dlbackend\comfy\ComfyUI\execution.py", line 196, in _map_node_over_list
2025-06-11 21:47:58.834 [Warning] [ComfyUI-0/STDERR]     process_inputs(input_dict, i)
2025-06-11 21:47:58.834 [Warning] [ComfyUI-0/STDERR]   File "D:\SwarmUI\dlbackend\comfy\ComfyUI\execution.py", line 185, in process_inputs
2025-06-11 21:47:58.835 [Warning] [ComfyUI-0/STDERR]     results.append(getattr(obj, func)(**inputs))
2025-06-11 21:47:58.835 [Warning] [ComfyUI-0/STDERR]                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-06-11 21:47:58.836 [Warning] [ComfyUI-0/STDERR]   File "D:\SwarmUI\dlbackend\comfy\ComfyUI\nodes.py", line 69, in encode
2025-06-11 21:47:58.836 [Warning] [ComfyUI-0/STDERR]     return (clip.encode_from_tokens_scheduled(tokens), )
2025-06-11 21:47:58.837 [Warning] [ComfyUI-0/STDERR]             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-06-11 21:47:58.837 [Warning] [ComfyUI-0/STDERR]   File "D:\SwarmUI\dlbackend\comfy\ComfyUI\comfy\sd.py", line 166, in encode_from_tokens_scheduled
2025-06-11 21:47:58.837 [Warning] [ComfyUI-0/STDERR]     pooled_dict = self.encode_from_tokens(tokens, return_pooled=return_pooled, return_dict=True)
2025-06-11 21:47:58.838 [Warning] [ComfyUI-0/STDERR]                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-06-11 21:47:58.838 [Warning] [ComfyUI-0/STDERR]   File "D:\SwarmUI\dlbackend\comfy\ComfyUI\comfy\sd.py", line 228, in encode_from_tokens
2025-06-11 21:47:58.839 [Warning] [ComfyUI-0/STDERR]     o = self.cond_stage_model.encode_token_weights(tokens)
2025-06-11 21:47:58.839 [Warning] [ComfyUI-0/STDERR]         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-06-11 21:47:58.839 [Warning] [ComfyUI-0/STDERR]   File "D:\SwarmUI\dlbackend\comfy\ComfyUI\comfy\text_encoders\hidream.py", line 116, in encode_token_weights
2025-06-11 21:47:58.840 [Warning] [ComfyUI-0/STDERR]     ll_output = self.llama.encode_token_weights(token_weight_pairs_llama)
2025-06-11 21:47:58.840 [Warning] [ComfyUI-0/STDERR]                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-06-11 21:47:58.840 [Warning] [ComfyUI-0/STDERR]   File "D:\SwarmUI\src\BuiltinExtensions\ComfyUIBackend\DLNodes\ComfyUI-ppm\nodes_ppm\clip_negpip.py", line 78, in encode_token_weights_negpip
2025-06-11 21:47:58.841 [Warning] [ComfyUI-0/STDERR]     z = torch.zeros_like(zk).repeat(1, 2, 1)
2025-06-11 21:47:58.841 [Warning] [ComfyUI-0/STDERR]         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-06-11 21:47:58.841 [Warning] [ComfyUI-0/STDERR] RuntimeError: Number of dimensions of repeat dims can not be smaller than number of dimensions of tensor
2025-06-11 21:47:58.842 [Warning] [ComfyUI-0/STDERR] 
2025-06-11 21:47:58.844 [Debug] [ComfyUI-0/STDERR] Prompt executed in 33.49 seconds

Other

No response

[mcmonkey4eva]

Yeah I imagine negpip just inherently isn't compatible, since it does clip things and hidream has a different textenc entirely.

For anything further, post on the negpip github rather than here

[CrHasher]

Ok will post there, thanks for the info. So, in essence there is need to check if textenc is compatible in the addon.