CLIP Text Encode (Advanced) Suddenly Throwing Errors
Kinglord opened this issue · 4 comments
I'm not sure what I did but suddenly one of my advanced clip nodes started throwing an error in the console and killing the job. I think it has something to do with the length of my prompt but I'm not sure, there's a lot of what I think is useless spam but the final line has what I think is causing the problem.
RuntimeError: Sizes of tensors must match except in dimension 0. Expected size 2464 but got size 308 for tensor number 1 in the list.
I'm going to swap back over to the default CLIP node since it works fine with the same prompt for now but hopefully I can get things sorted.
What settings and prompt did you use?
there's a lot of what I think is useless spam but the final line has what I think is causing the problem.
that useless spam is probably also very helpful for me to track down the issue
Sure thing, here's the full error log. I'll post my prompt under it that broke it, I had the same thing happen to another prompt as soon as I added some weights to something so I think it's related to the number of terms I have weighted?
Traceback (most recent call last): File "X:\BOS\WD_Gen\ComfyUI\ComfyUI\execution.py", line 184, in execute executed += recursive_execute(self.server, prompt, self.outputs, x, extra_data) File "X:\BOS\WD_Gen\ComfyUI\ComfyUI\execution.py", line 60, in recursive_execute executed += recursive_execute(server, prompt, outputs, input_unique_id, extra_data) File "X:\BOS\WD_Gen\ComfyUI\ComfyUI\execution.py", line 60, in recursive_execute executed += recursive_execute(server, prompt, outputs, input_unique_id, extra_data) File "X:\BOS\WD_Gen\ComfyUI\ComfyUI\execution.py", line 69, in recursive_execute outputs[unique_id] = getattr(obj, obj.FUNCTION)(**input_data_all) File "X:\BOS\WD_Gen\ComfyUI\ComfyUI\custom_nodes\ComfyUI_ADV_CLIP_emb\nodes.py", line 19, in encode embeddings_final = advanced_encode(clip, text, token_normalization, weight_interpretation, w_max=1.0) File "X:\BOS\WD_Gen\ComfyUI\ComfyUI\custom_nodes\ComfyUI_ADV_CLIP_emb\adv_encode.py", line 193, in advanced_encode return advanced_encode_from_tokens(clip, tokenized, token_normalization, weight_interpretation, w_max) File "X:\BOS\WD_Gen\ComfyUI\ComfyUI\custom_nodes\ComfyUI_ADV_CLIP_emb\adv_encode.py", line 183, in advanced_encode_from_tokens weighted_emb = from_masked(unweighted_tokens, weights, word_ids, weighted_emb, clip) File "X:\BOS\WD_Gen\ComfyUI\ComfyUI\custom_nodes\ComfyUI_ADV_CLIP_emb\adv_encode.py", line 87, in from_masked embs = batched_clip_encode(masked_tokens, clip, len(tokens)) File "X:\BOS\WD_Gen\ComfyUI\ComfyUI\custom_nodes\ComfyUI_ADV_CLIP_emb\adv_encode.py", line 54, in batched_clip_encode embs = torch.cat(embs) RuntimeError: Sizes of tensors must match except in dimension 0. Expected size 2464 but got size 154 for tensor number 1 in the list.
Prompt:
1girl, (detailed face), (detailed eyes), best quality, hyper detailed, masterpiece, (mature Swedish woman), Freya Adelsköld, techwear, (mecha musume:.75), (himedere), aloof, short hair, multicolored hair, (platinum blonde hair), white inner hair, (grey eyes:1.35), thick lips, upturned nose, eyelashes, looking at viewer, highest details, luminous eyes, tall, medium breasts, tough, (pale skin), oval face, regal armor, light rays, (high contrast), (colorful), simple background, trending on artstation, detailed, award winning illustration, hd, 4k, digital art, cinematic lighting,
I tried all the various settings and they all seemed to give the same error. The one I'm using most though is length+mean and comfy++
Let me know if I can help send anything else.
was able to reproduce, should be fixed now when you pull