Alissonerdx

/

EditAnything

like 94

Hi! First of all, thank you for this amazing workflow.

I'm consistently getting the following error when running the workflow:

RuntimeError: cannot reshape tensor of 0 elements into shape [1, 0, 32, -1] because the unspecified dimension size -1 can be any value and is ambiguous

The exception is raised from the LTXVEditAnythingLoopingSampler node.

I've already ruled out several common causes:

Latest BFSNodes (1.21.1)
Latest KJNodes (1.4.8)
Latest RES4LYF nightly
Latest ComfyUI Manager
Tested with both:
ltx-2.3-22b-dev-Q6_K.gguf
ltx-2-3-22b-dev_transformer_only_fp8_input_scaled.safetensors
Using the recommended Distilled LoRA.
Using the matching Edit Anything .standard and .module files from the same Reference V2 build.
Disabling visual_crossattn produces exactly the same error.

The workflow starts correctly, loads all models and LoRAs, generates the guide and reference latents, and then crashes immediately at the very first diffusion step (0%).

Before the crash, I always see the following warning:

Could not force-load model before ref_attn install (out of memory). Falling back to random init.

ref_attn modules: created=False

Applied 192 ref_attn LoRA deltas

I'm not sure whether this warning is related to the crash or not.

I also found a recent ComfyUI issue describing a very similar error:

https://github.com/Comfy-Org/ComfyUI/issues/15233

The reported error is almost identical (cannot reshape tensor of 0 elements into shape [2, 0, 32, -1]) and seems to be related to LTXAVModel after ComfyUI v0.29.0. I don't know if this workflow is affected by the same underlying issue, but the similarity caught my attention.

Has anyone experienced this, or is there anything else I should check?

Thank you!