How to make node run *every* generation?

[WASasquatch]

I have added NSP to new CLIPTextEncode node, but it seems to actually run only once, and then it seems to be using a cached result?

Basically, it parses all noodles in the prompt with random words of a term. However, it seems to only run once and tell me what the parsed from, after that, it doesn't tell me a new prompt, and the gens are always of the same "random" word... so, not random.

Example from the console with two generations (with HR-Fix) the second one, the CLIPTextEncode doesn't print the new prompt, which it does before returning.

Parsed Prompt: London, England
100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:04<00:00,  4.30it/s]
making attention of type 'vanilla-xformers' with 512 in_channels
building MemoryEfficientAttnBlock with 512 in_channels...
Working with z of shape (1, 4, 32, 32) = 4096 dimensions.
making attention of type 'vanilla-xformers' with 512 in_channels
building MemoryEfficientAttnBlock with 512 in_channels...
Restored from C:\ComfyUI_windows_portable\ComfyUI\models\vae\kl-f8-anime2.vae.pt
  0%|                                                                                           | 0/20 [00:00<?, ?it/s]C:\ComfyUI_windows_portable\python_embeded\lib\site-packages\torchsde\_brownian\brownian_interval.py:594: UserWarning: Should have tb<=t1 but got tb=1.2392144203186035 and t1=1.239214.
  warnings.warn(f"Should have {tb_name}<=t1 but got {tb_name}={tb} and t1={self._end}.")
100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:16<00:00,  1.24it/s]
Downloading and loading MiDaS Model...
Using cache found in C:\Users\jorda/.cache\torch\hub\intel-isl_MiDaS_master
MiDaS is using device: cuda
Using cache found in C:\Users\jorda/.cache\torch\hub\intel-isl_MiDaS_master
Approximating depth from image...
100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:04<00:00,  4.89it/s]
100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:17<00:00,  1.17it/s]
Downloading and loading MiDaS Model...
Using cache found in C:\Users\jorda/.cache\torch\hub\intel-isl_MiDaS_master
MiDaS is using device: cuda
Using cache found in C:\Users\jorda/.cache\torch\hub\intel-isl_MiDaS_master
Approximating depth from image...

[comfyanonymous]

https://github.com/comfyanonymous/ComfyUI/blob/master/nodes.py#L817

If this IS_CHANGED function exist and returns a different value than the previous time the node (and everything depending on it) will be executed again.

[WASasquatch]

https://github.com/comfyanonymous/ComfyUI/blob/master/nodes.py#L817

If this IS_CHANGED function exist and returns a different value than the previous time the node (and everything depending on it) will be executed again.

I'm not sure how to get this to work. I don't have a image involved, so I tried just returning the text, since it should be different. Then I tried emulating what you have by getting hash of the string (didn't look if that was possible, but thought to try) but I keep getting the same error

Exception in thread Thread-1 (prompt_worker):
Traceback (most recent call last):
  File "threading.py", line 1016, in _bootstrap_inner
  File "threading.py", line 953, in run
  File "C:\ComfyUI_windows_portable\ComfyUI\main.py", line 40, in prompt_worker
    e.execute(item[-2], item[-1])
  File "C:\ComfyUI_windows_portable\ComfyUI\execution.py", line 148, in execute
    recursive_output_delete_if_changed(prompt, self.old_prompt, self.outputs, x)
  File "C:\ComfyUI_windows_portable\ComfyUI\execution.py", line 97, in recursive_output_delete_if_changed
    input_data_all = get_input_data(inputs, class_def)
  File "C:\ComfyUI_windows_portable\ComfyUI\execution.py", line 21, in get_input_data
    obj = outputs[input_unique_id][output_index]
KeyError: '4'

Tried

    @classmethod
    def IS_CHANGED(s, clip, text, noodle_key = '__'):
        m = hashlib.sha256()
        m.update(text)
        return m.digest().hex()

and

    @classmethod
    def IS_CHANGED(s, clip, text, noodle_key = '__'):
        return text

also

    @classmethod
    def IS_CHANGED(s, clip, text, noodle_key = '__'):
        return random.randint(-9999999999, 9999999999)

[WASasquatch]

Ended up just adding a seed, since I can add that with random.seed() method which is useful for calling back a certain set of random terms anyway. That seems to do the trick.