Jan Vesely

> Hey @jvesely, any idea why this test I marked with `pytest.mark.pytorch` is not getting skipped in CI when pytorch isn't available? Does the test show when you do: ```...

> Yeah, on my local machine it seems to me found. The test is called . It is is the only test in test_control that is marked PyTorch now, before...

Using exceptions is quite fragile (there might be other reasons for RuntimeError) and the exception path has a pretty big performance hit torch provides `torch.is_tensor()` helper so all the newly...

> Hey, yeah, I agree, those are good suggestions. I'll do that. I am running locally pytest 7.2.0 can you retest with pytest-8.0.*, and pytest-8.1.* ? It sounds like a...

cc @kmantel

I've tracked this to the following code in TransferWithCost: ``` # Initialize intensity attributes if self.enabled_cost_functions: # Default cost params if self.owner: if self.owner.context.initialization_status != ContextFlags.DEFERRED_INIT: self.intensity_cost = self.intensity_cost_fct(self.owner.defaults.variable) else:...

Try newer LLVM and libclc. I'm able to run Pyrit on my carrizo iGPU.

Should this go straight to master? I think it's better to drop 871f960 ("Update optuna pin to latest.") from this PR. The constraint is not a pin, just an upper...

One more nit. Will the existing two tests that use instanced samplers issue warnings? If so, should the test check that the warnings are emitted in those two cases?

the bug still reproduces. after https://github.com/PrincetonUniversity/PsyNeuLink/pull/2685 the invocation line is: ``` MNET=PyTorch python Scripts/Debug/Yotam\ LCA\ Model\ LLVM.py ```