commit | e309cb38072ec909f43bc816b9fb64005d13fecd | [log] [tgz] |
---|---|---|
author | Ben Vanik <ben.vanik@gmail.com> | Wed Feb 01 11:59:39 2023 -0800 |
committer | GitHub <noreply@github.com> | Wed Feb 01 19:59:39 2023 +0000 |
tree | 07bc39833aa84f154caa5213718c831fd8dffe3e | |
parent | f65c5cb767f8d6a7d83dc3c9a8033f3c350e9b2e [diff] |
Add `--output=` flag to iree-run-module/iree-run-mlir. (#12016) This allows for ignoring outputs, routing them to stdout, or writing them to numpy .npy files that can be easily loaded in python. Example writing two outputs to an npy file: `iree-run-module ... --output=@file.npy --output=+file.npy` ```py with open('file.npy', 'rb') as f: print(numpy.load(f)) print(numpy.load(f)) ``` The `@` sigil will create/overwrite a file while the `+` sigil will create/append. This allows for multiple outputs to route to the same file even across runs (so you could run several times and accumulate the results in a single npy to process later). Shorthand for writing all outputs to an npy would be useful but is not yet implemented. We can extend this in many ways in the future by supporting binary payload writing (ala `--input=2xf32=@file.bin`), merging functionality with the `--expected_output=` flags, etc. For now this is better than users trying to string parse stdout, though :) Progress on #9462 (still need to rename/clean up vm_util).
IREE (Intermediate Representation Execution Environment, pronounced as “eerie”) is an MLIR-based end-to-end compiler and runtime that lowers Machine Learning (ML) models to a unified IR that scales up to meet the needs of the datacenter and down to satisfy the constraints and special considerations of mobile and edge deployments.
See our website for project details, user guides, and instructions on building from source.
IREE is still in its early phase. We have settled down on the overarching infrastructure and are actively improving various software components as well as project logistics. It is still quite far from ready for everyday use and is made available without any support at the moment. With that said, we welcome any kind of feedback on any communication channels!
See our website for more information.
IREE is licensed under the terms of the Apache 2.0 License with LLVM Exceptions. See LICENSE for more information.