Replicator Path Tracing samples do not accumulate

When I run Replicator and set the renderer to Path Tracing, the samples don’t accumulate. I am expecting a more refined image than the outputed image. This happens with on_frame and on_time triggers.

I’m using the code from “Running Replicator Headlessly”:
https://docs.omniverse.nvidia.com/prod_extensions/prod_extensions/ext_replicator/headless_example.html

I added the following code, (after with rep.new_layer():) to implement path tracing and a consistent seed, (low quality render results without the global seed line too):

rep.set_global_seed(23)
rep.settings.set_render_pathtraced(samples_per_pixel=64)

I did some digging and found the following in the file “code-2022.2.0\extscache\omni.replicator.core-1.4.7+wx64.r.cp37\omni\replicator\core\scripts\orchestrator.py”

# Set frame counter explicitly
carb.settings.get_settings().set_int("/rtx/externalFrameCounter", self._frame)

which since it runs at every frame or subframe it will always reset and not accumulate samples.
this happens because the following is set, (in the same file):

# Setting resetPtAccumOnlyWhenExternalFrameCounterChanges ensures we control accumulation explicitly
# by simpling changing the /rtx/externalFrameCounter value
_settings.set_bool("/rtx-transient/resetPtAccumOnlyWhenExternalFrameCounterChanges", True)

the fix for me is to use the following in orchestrator.py so that it resets only when the frame has started:

if self._subframe == 1:
    carb.settings.get_settings().set_int("/rtx/externalFrameCounter", self._frame)

Do I need to change something to get Path Tracing using more than one sample to work correctly?
Is there a new version being released soon?
I appreciate any help I can get.

Windows 11
Omniverse Code 2022.2.0
omni.replicator.core-1.4.7+wx64.r.cp37
RTX 3090

Thank you,
Mark

Hello @mark148! I’ve pinged the dev team for help on this. They should reply shortly!

1 Like

Hi Mark,

Thanks for reporting this. The fix for this will be included in Replicator version 1.5.3

1 Like

Thank you !