BladeDISC
BladeDISC copied to clipboard
alibaba
BladeDISC is an end-to-end DynamIc Shape Compiler project for machine learning workloads.
Currently the `tensorflow_blade` workspace links what's in tao_bridge into `libtf_blade.so`, so to eliminate `libtao_ops.so` in tensorflow use case. But the missing part is to also build `tao_compiler_main`. When refactoring tensorflow_blade...
the CI system should deploy production Docker image with **release** version tag instead of the latest, e.g. `bladedisc/bladedisc:0.2.0-runtime-tensorflow2.8-cpu-aarch64`
Background: We are working on building Torch2MHLO passes based on [TorchMLIR](https://github.com/llvm/torch-mlir). The goal for the most recent is to switch from the old converter framework to TorchMLIR: - [ ]...
To move `torch_disc` folder which forces on training accelarting, we should do some code refactor to place these code to the correct place of `pytorch_blade` folder. For the top-level dir,...
In the past weeks, we have open-source some of the cpp and python code for tensorflow-blade, which optimize the tensorflow model for inference. The first stage of tensorflow blade open...
the warning logs are as the following: ``` text >>> import torch_disc 2022-06-08 01:43:10.626857: W src/compiler/mlir/converters/mhlo_converter_register.cpp:59] Overriding already registered converter aten::relu, unexpected behavior may occur 2022-06-08 01:43:10.626909: W src/compiler/mlir/converters/mhlo_converter_register.cpp:59] Overriding...
I am a pytorch user,and I want to use BladeDISC with DCU
I am trying to build a tf .whl on DCU. The instructions for BladeDISC are based Nvidia graphics. Thank you
