Updated comfy-core- nodes_easycache.py produces error about 'Tensor' and 'tuple' - Type error?

[GlamoramaAttack]

Custom Node Testing

• I have tried disabling custom nodes and the issue persists (see how to disable custom nodes if you need help)

Expected Behavior

Yesterday my LTX2 workflow worked correct with the "LazyCache" (comfy-core Beta)- node. Today, after updating Comfy, it produced the error message, excerpt in debug log.

Bypassing the node let's the sampler (LTXV Normalizing Sampler) generate again without further failures.

I think it's not related to the sampler, but as the error message says a type error in ComfyUI\comfy_extras\nodes_easycache.py .

ChatGPT thinks it should be (from line 457 on):

def update_cache_diff(self, output, x):
    if isinstance(output, tuple):
        output = output[0]
    if isinstance(x, tuple):
        x = x[0]

    self.cache_diff = output - x

But that looks very different. And ChatGPT's self-assurance should be mistrusted - writing about an "update" as if there was only one:

Because since the update, ComfyUI returns multiple values ​​for some samplers.
However, EasyCache was written for the old API, which always returned only one tensor.
The patch:
doesn't break anything
is compatible with old and new ComfyUI versions

Actual Behavior

see above

Steps to Reproduce

I recreated the LazyCache node (0.10.0), it has now an updated version number (0.12.3), but it produces the same error message.

Debug Logs

Requested to load LTXAV (...)
Exception during processing !!! unsupported operand type(s) for -: 'Tensor' and 'tuple'
Traceback (most recent call last):
  File "D:\ComfyUI\execution.py", line 527, in execute
(...)
  File "D:\ComfyUI\execution.py", line 331, in get_output_data
(...)
  File "D:\ComfyUI\comfy_extras\nodes_easycache.py", line 150, in lazycache_predict_noise_wrapper
    easycache.update_cache_diff(output, next_x_prev)
  File "D:\ComfyUI\comfy_extras\nodes_easycache.py", line 458, in update_cache_diff
    self.cache_diff = output - x
                      ~~~~~~~^~~
TypeError: unsupported operand type(s) for -: 'Tensor' and 'tuple'

See also:
Comfy-Core type error in: #12331

[kijai]

Right, so I didn't realize LazyCache actually already worked and one line of the EasyCache fix slipped in, submitted a PR to fix that.

Meanwhile you can actually use the EasyCache node with LTX2 now, if you haven't tried it yet, it should be worth it to compare it to the LazyCache.

[GlamoramaAttack]

I'll try it. Thanks a lot!