AMDMIGraphX
AMDMIGraphX copied to clipboard
ROCmSoftwarePlatform
AMD's graph optimization engine.
* Updates parsing of `Split` ONNX operator to support dynamic shapes for the case of no `split` attribute or input. * Should be possible to extend dynamic shape support to...
This will fuse the inputs but only when using the `MIGRAPHX_ENABLE_MLIR_INPUT_FUSION` env variable.
### Problem Description Trying out torch.compile via torch_migraphx and using the example code in torch_migraphx/examples/dynamo/stable_diffusion (but compiling only the unet) does not seem to give a performance increase. Passing in...
On Windows, the symbol must be exported from the DLL for the test to see the function while dynamically loading the library.
Adding softmaxcrossentropy op in MIGraphX. This is used in Llama v2 https://github.com/onnx/onnx/blob/main/docs/Operators.md#softmaxcrossentropyloss This function is a bit complicated but is able to be built up by smaller Onnx ops I've...
Model performing worse or similar to eager pytorch. Model Location: `/mnt/nas_share/migraphx/models/torch_benchmarks/AllenaiLongformerBase`
Model performing worse or similar to eager pytorch. Model Location: `/mnt/nas_share/migraphx/models/torch_benchmarks/mobilenet_v3_large`
Model performing worse or similar to eager pytorch. Model Location: `/mnt/nas_share/migraphx/models/torch_benchmarks/vgg16`
Model performing worse or similar to eager pytorch. Model Location: `/mnt/nas_share/migraphx/models/torch_benchmarks/alexnet`