Brian Pickrell

find_split_concat fix

3

Fix a bug that caused an exception in torch-migraphx benchmarks. Closes Issue https://github.com/ROCm/AMDMIGraphX/issues/3009

Infinite recursion in module::sort(), speech_transformer

2

Model `speech_transformer` in the DLM pytorch performance test causes an exception which is traced to an infinite recursion in the MIGRAPHX method `module::sort()`. The error is only seen in branch...

dimensions match in find_pointwise_reduce

3

Updates the `find_pointwise_reduce` matcher in the `fuse_reduce` compiler pass, to not attempt to fuse a `multibroadcast` instruction if the shape has an incompatible number of dimensions (rank). This cures the...

MIGraphX execution provider for Triton Inference Server

65

Can this be done by leveraging the onnxruntime work we already have as a back end? As a preliminary step, learn to add a Cuda back end, then change it...

adjust stride ordering rules for standard shape: stride can be anythi…

9

…ng in a dimension of size 1. This is for issue https://github.com/ROCm/AMDMIGraphX/issues/3117

Roialign fix and half_pixel mode support

3

Fix bugs in the implementation of ROIAlign operation which were found when attempting to run it with the `half_pixel` coordinate conversion mode, to include more thorough tests. Some bugs are...

Resize Onnx parsing

5

The old logic stepped through the args list in order, then assumed the first float-type input was the "Scales". This led to an error in an obscure case where the...

Low priority enhancement to [Issue 1670](https://github.com/ROCm/AMDMIGraphX/issues/1670): `make_gather_instruction()` function calls in `parse_resize.cpp` can be replaced by invoking the Resize op. as follows: /* TODO: to make Onnx resize ALWAYS parse to...