AMDMIGraphX

AMDMIGraphX copied to clipboard

[Issue]: Running Argmax op using MiGraphX EP (ONNX Runtime)

2

### Problem Description Getting a Core Dump trying to running Argmax operation using MigraphX EP in ONNX Runtime. Following is a short error trace: ``` [W:onnxruntime:Default, migraphx_execution_provider.cc:1298 compile_program] Model Compile:...

smohta-all3

The following fails verification: ```python p = migraphx.program() m = p.get_main_module() p_x = m.add_parameter("x", migraphx.shape(type="float_type", lens=[1, 1, 2])) x_1 = m.add_instruction(migraphx.op("pooling", padding=[1,2], stride=[1], lengths=[3], dilations=[1], count_include_pad=True), [p_x]) # migraphx.shape(type="float_type", lens=[1,...

pfultz2

Issue in MIGraphX execution provider for onnxruntime.

2

Hi, I was trying to use [the rust bindings for onnxruntime (ort crate)](https://github.com/pykeio/ort) with with the [MIGraphX](https://github.com/ROCm/AMDMIGraphX/tree/develop) execution provider. I compiled onnxruntime library to support this execution provider using the...

aravindhv10

Switch to c++23

6

This upgrades the version to c++23 for our source code and for kernels. We still cannot use newer std library features because sles uses gcc 7. - Update `any_cast` of...

pfultz2

MIGraphX doesn't report supported Onnx operations in the API

1

The API allows us to parse in onnx and tensor flow operators but we don't have a way to query the currently installed MIGraphX version to determine what operators or...

TedThemistokleous

Implement flash decoding in attention fusion pass

6

- [x] Understand the feedback and requirements - [x] Revert previous implementation that mixed flash decoding into attention finder - [x] Create separate `find_flash_decoding` finder struct - [x] Make it...

Copilot

pfultz2

G+G fusion in rocMLIR supports optional elementwise fusion between gemms, and before/after either gemm. The current MIGraphX implementation has the elementwise op between gemms as required

bdevorem

Add ONNX model testing workflow

2

**Summary** This PR creates and updates the onnx-model.yaml workflow to automate model zoo and onnx model testing. **Changes** Removed explicit gh_token input (no longer required to run this workflow). Cleaned...

danieyan-amd

Disable offloading attention like kernels to rocMLIR if there is no wmma/mfma for requested dtype

1

### DOR (Definition of Ready) Ready ### Description rocMLIR doesn't support attention like kernels. eg. GEG, Attention, CEG if there is no associated wmma/mfma instruction for that dtype on the...

umangyadav