PDG (TOPs) Network Sticks on First Frame During Cook

[spelafort]

I am encountering an issue with MLOPs and PDG. Specifically, I am attempting to create a simple setup that involves reading CSV files (for prompts) and then uses @pdg_frame to interpolate between prompts, baking out the images of the results.

While I can successfully export images by using $F as the interpolation value and manually scrubbing the timeline, the issue arises when I attempt to cook the ropfetch in TOPs and have it driven by @pdg_frame. During this process, it seems to get stuck on the first frame, resulting in repeated frames instead of the desired interpolation. Important to note is that if I manually select a work item in TOPs, the scene viewer does show the correct output-- just not when I cook any of the Rop nodes in TOPs.

I tried exporting as a bgeo instead of an image, but exporting as Geo yields a similar result.

Upon further investigation, I suspect the issue may lie with the sd_solver node or how my TOPs network and localscheduler are handling the process. It's worth noting that while the ropgeo node's cook type default is set to "out of process", MLOPs nodes only seem to function properly if I cook them in process, as others have pointed out.

I will attach an example file to illustrate the issue further.

Steps to Reproduce:

1. Open the example file attached, and attempt to cook any of the rop nodes in TOPs
2. Observe that the cooked result gets stuck on the first frame instead of interpolating between frames.

Expected Behavior:
The PDG network should successfully interpolate between prompts based on the provided tops frame attribute (@pdg_frame in the channel in obj/Prompt Arithmetic/prompt_interpolation node), allowing for proper frame-by-frame processing and export.

Additional Information:
-I have already attempted swapping $F references for @pdg_frame.
-Exporting as Geo also exhibits a similar issue, although the ropgeo node properly picks up the frame number while the ropfetch node does not.
-The issue seems to be related to how MLOPs nodes interact with TOPs, particularly regarding cook types and in-process cooking.

any help appreciated.
PDG_MLOPS_Broken.zip