Comments (10)
@silvasean
Hi, I just tried the newest commit and both commands seem to go through. Although I am running a custom-built IREE and e2e test failed, I'll have a look at the tests myself, Thanks!
from torch-mlir.
The latest IREE has the fix for builtin.module. This is the current list I see locally (will update xfail sets soon)
PASS - "MmModule_basic"
PASS - "MmModule_chained"
PASS - "MmDagModule_basic"
PASS - "MmTanhModule_basic"
PASS - "AdaptiveAvgPool2dModule_basic"
PASS - "FlattenStaticModule_basic"
PASS - "FlattenRank0Module_basic"
XFAIL - "FlattenDynamicModule_basic"
XFAIL - "MaxPool2dModule_basic"
XPASS - "ResNet18Module_basic"
PASS - "Mlp1LayerModule_basic"
PASS - "Mlp2LayerModule_basic"
PASS - "Conv2dNoPaddingModule_basic"
FAIL - "Conv2dWithPaddingModule_basic"
XFAIL - "Conv2dWithPaddingDilationStrideModule_basic"
PASS - "BatchNorm1DModule_basic"
PASS - "BatchNorm2DModule_basic"
PASS - "BatchNorm3DModule_basic"
XFAIL - "QuantizedMLP_basic"
PASS - "ElementwiseUnaryModule_basic"
PASS - "ElementwiseBinaryModule_basic"
PASS - "ElementwiseTernaryModule_basic"
PASS - "ElementwiseAddModule_basic"
PASS - "ElementwiseUnsqueezeBroadcastModule_basic"
PASS - "ElementwiseFlattenBroadcastModule_basic"
PASS - "ElementwiseReluModule_basic"
from torch-mlir.
I just refreshed the instructions: a153cf4
Can you give them a try? If not, can you include the full logs with the errors?
from torch-mlir.
Hi @silvasean,
About the failed tests yesterday... I started a fresh docker and built npcomp, I installed iree (not my custom built one) via pip. But the tests still failed:
Maybe the newest commits caused this?
from torch-mlir.
Currently IREE backend is failing because we bumped our LLVM version to pickup some fixes and changes from the upstream but IREE is still using the old LLVM. If you add -v you will see the failure is because builtin.module
can't be recognized. The op is the result of recent MLIR refactoring and the older MLIR used by IREE is not aware of this op.
Other than the LLVM version issue, there might be some smaller issues which should be fixed shortly.
from torch-mlir.
Ha, seems ResNet18Module_basic
passed with IREE as well?
from torch-mlir.
yep :)
from torch-mlir.
@silvasean Do you know which patch in iree fix the builtin.module issue?
from torch-mlir.
It's an asm format change from upstream. It will have come in via an iree submodule bump. We don't track such unstable changes between releases.
from torch-mlir.
@stellaraccident Thanks, I'll just update upstream version directly.
from torch-mlir.
Related Issues (20)
- `make_simple_dynamo_backend` does not work for single argument models. HOT 2
- Compilation of Quantized Model Failed HOT 1
- Missing dtype support for OnnxToTorch HOT 3
- Strange issue related to `pyproject.toml` when building release wheels HOT 4
- iree-compiler failed to compile gpt2
- `NotSupportedError` error when executing Readme example "TorchScript ResNet18"
- onnx.Transpose version 1 not lowered (but it's essentially the same as version 13)
- boolean indexing ops: AtenNonzeroOp, AtenIndexTensorOp, AtenMaskedSelectOp
- QuantizedMLP currently failing in CI
- migrate to from member `x.cast<T>()` to `mlir::cast<T>(x)` HOT 1
- Update version pin to onnx 1.16 HOT 4
- Include Shape Information Into DecomposeComplexOps Patterns HOT 4
- onnx.MatMul version 1 not lowered (but it's essentially the same as version 13)
- Crash with TorchBackendToLinalgOnTensorsBackendPipeline
- link out of date
- unsupported aten.meshgrid operator
- Trying to lower elided IR from torch onnx to torch causes crash HOT 2
- [ONNX] Systematically expand ONNX functions before conversion to Torch to avoid needing bespoke conversions HOT 12
- [onnx] Attempting to lower onnx mlir with elided attributes causes crash HOT 7
- Some issues with migrating to one-shot-bufferize HOT 11
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from torch-mlir.