Embeddings cause the node to break

[cheald]

When an embedding is added to the conditioning text, it may be expanded to multiple tokens, which causes an error:

I added a simple print(token_id) to vector_sculptor_tokens, which produces something like:

token id 49406
token id 4431
...
token id tensor([-0.0215,  0.0080,  0.0078,  ..., -0.0245, -0.0093,  0.0171])
!!! Exception during processing !!!
Traceback (most recent call last):
  File "/srv/ComfyUI/execution.py", line 151, in recursive_execute
    output_data, output_ui = get_output_data(obj, input_data_all)
  File "/srv/ComfyUI/execution.py", line 81, in get_output_data
    return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
  File "/srv/ComfyUI/execution.py", line 74, in map_node_over_list
    results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
  File "/srv/ComfyUI/custom_nodes/Vector_Sculptor_ComfyUI/nodes.py", line 165, in exec
    sculptor_tokens = vector_sculptor_tokens(clip, text, sculptor_method, token_normalization, sculptor_intensity)
  File "/srv/ComfyUI/custom_nodes/Vector_Sculptor_ComfyUI/nodes.py", line 106, in vector_sculptor_tokens
    if token_id not in ignored_token_ids and sculptor_multiplier > 0:
RuntimeError: Boolean value of Tensor with more than one value is ambiguous

Here's a workflow which replicates the issue:

broken-with-embedding.json

The embedding is from https://civitai.com/models/148131/ultimate-text-embeddings-sdxl-pack-styles-negative but you can use any multi-vector embedding to replicate the issue.

[slackingway]

vector_sculptor_tokens expects a token id which it then converts to the corresponding embedding tensor, this can be fixed by checking if token_id is actually already an embedding with something involving isinstance(token_id, torch.Tensor)