| commit | 575509fcc692d0544d6709033c64419621b343b4 | [log] [tgz] |
|---|---|---|
| author | Geoffrey Martin-Noble <gcmn@google.com> | Sat Nov 13 10:46:35 2021 -0800 |
| committer | GitHub <noreply@github.com> | Sat Nov 13 13:46:35 2021 -0500 |
| tree | b4dfcac23cda8260b4c00d07f5e4e6b4690e0291 | |
| parent | bd2001aa8405373e5a06bf431e1b8b7459a5ddc0 [diff] |
Add default-flags configurations for all TFLite benchmarks (#7580) Following up on https://github.com/google/iree/pull/7553, this adds default flag configurations for the remaining TFLite models. Now that we've got timestamps in the Buildkite runs, we can see what this does to artifact transfer time via the presubmit: https://buildkite.com/iree/iree-benchmark/builds/1447 I'm contrasting with the run at the merge base on main: https://buildkite.com/iree/iree-benchmark/builds/1445 For this PR's run, downloads took 1:35 and 4:12 for the two RPIs, compared to 2:06 and 3:30 before. That looks like it's basically in the noise. We could do more detailed analysis, but that doesn't seem worth it to me. Just spot-checking a few, it seems the RPI connected to the Pixels consistently downloads faster than the ones connected to the Samsungs. As for the benchmarks themselves, the run on the RPI took around 34-35 minutes for each phone, 6-8 min longer than before. Overall run latency, excluding waiting for agents, increased from 43:12 to 49:50. I think this is acceptable given our current limitations.
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!
| CI System | Build System | Platform | Architecture | Configuration / Component | Status |
|---|---|---|---|---|---|
| Kokoro | Bazel | Linux | x86-64 |  | |
| Kokoro | CMake & Bazel | Linux | x86-64 (swiftshader) | Integrations |  |
| Kokoro | CMake & Bazel | Linux | x86-64 (turing) | Integrations |  |
| Kokoro | CMake | Linux | x86-64 (swiftshader) |  | |
| Kokoro | CMake | Linux | x86-64 (swiftshader) | asan |  |
| Kokoro | CMake | Linux | x86-64 (turing) |  | |
| Kokoro | CMake | Android | arm64-v8a | Runtime (build only) |  |
| Kokoro | CMake | Bare Metal | risc-v-32 | Runtime |  |
| Kokoro | CMake | Linux | risc-v-64 | Runtime |  |
| Buildkite | CMake | Android | arm64-v8a | Runtime |  |
| BuildKite | CMake | Android | arm64-v8a | Runtime Benchmarks |  |
| BuildKite | CMake | Linux | x86-64 | Tracing + Standalone Runtime |  |
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.