Jacob Hinkle
Jacob Hinkle
Nope this issue is still open. However, I've done this in https://github.com/jacobhinkle/diffeomorphic_autoencoders/blob/master/run_all.py Note that that code is not using the lagomorph CLI, because I wrote that first and later put...
This PR was motivated by cases like the following. Consider this basic fusion: https://github.com/csarofeen/pytorch/blob/a06224f1976e2896d27f199984bd6b3f98707424/third_party/nvfuser/test/test_gpu_match_frontend.cpp#L197-L205 Using manual and automatic scheduling give identical `fusion_ir` printouts, but looking at the generated kernels, they...
With the printout above of the Fusion operation log, I was able to exactly match the automatic scheduler exactly for this problem, and learned a few things about how the...
This is where the canned CLI tool is going to limit you if you want to extend the basic training loop. There's no realistic way to support all the things...
Note that one direction works just fine already; given a displacement field `u`, you can compose it with an affine transform via `lm.affine_interp(u, A, T)`. So this issue really only...
Once this capability exists, I think it should be demonstrated in a jupyter notebook where we do affine atlas building, then use the affine transforms (without optimizing them) to precompose...
I looked at this example using `PYTORCH_NVFUSER_DUMP=segmenter_logging` as well as a little snooping with gdb. It seems that the segmenter hits the second view and has to segment. The reason...
I don't understand why the line after doesn't get put in the first segment. See the "NOTE" in this copy of Kevin's Fusion ```python def nvfuser_fusion_id0(fd: FusionDefinition, insert_views: bool =...
Studying how views affect segmentation with this sequential fusion: ```python def simple_fusion(fd: FusionDefinition) -> None: T = fd.define_tensor( symbolic_sizes=[-1, -1], contiguous=[True, True], dtype=DataType.Float ) current_shape = [8192, 3072] other_shape =...
This is a good test case. I think I know where the heuristic fails. This is probably related to https://github.com/csarofeen/pytorch/pull/2455