Ecosyste.ms: OpenCollective
An open API service for software projects hosted on Open Collective.
github.com/llvm/torch-mlir
The Torch-MLIR project aims to provide first class support from the PyTorch ecosystem to the MLIR ecosystem.
https://github.com/llvm/torch-mlir
a0ccfcf581158d4a4aec2fd62a23015221991b05 authored about 3 years ago by Sean Silva <[email protected]>
3dc9b4ee2f399427991f049dfc6c56198434083c authored about 3 years ago by Sean Silva <[email protected]>
It's not used in CI and doesn't seem to be up to date.
12dc2fa93310f290108f9edd07e29d1f50b7e9d4 authored about 3 years ago by Sean Silva <[email protected]>
A few remain in examples/docs that will be naturally be updated in due
time.
This regresses the...
1a0b953ea7edfaf26fd3881d102c64fe0c1aab29 authored about 3 years ago by Sean Silva <[email protected]>We now use RefBackend/refbackend consistently.
8779d920b24ea3a53304998094ff39dcef8fe5b9 authored about 3 years ago by Sean Silva <[email protected]>It is superceded by the new one.
a25163fbfa1ef2769b688a522b47d8185e5d8beb authored about 3 years ago by Sean Silva <[email protected]>`tools/torchscript_e2e_test.sh` is all green.
This needs a few passes I put into torch-mlir/lib...
f9c48d0b89282da2007aa12a779926c9e4025689 authored about 3 years ago by Sean Silva <[email protected]>
Implements a python package for taking a `torch.fx.GraphModule`
and turning it into MLIR in the...
19feb58ef876bb854321f5100a4cda0b0defee54 authored about 3 years ago by Ramiro Leal-Cavazos <[email protected]>
Fixes https://github.com/llvm/mlir-npcomp/issues/311
The key change is that TorchPlugin is fold...
6d8e7f1bb1dc7b7b798584f15aff992794d36b87 authored about 3 years ago by Sean Silva <[email protected]>ecc334123ca745497f422952f0041abbca1b7ec4 authored about 3 years ago by George Petterson <[email protected]>
Should fix issue noted in #304
c24ca5d6398f246a82c25e8cd4ca461a58b29182 authored about 3 years ago by Stella Laurenzo <[email protected]>3e3459690c36360d59381f758d6b9f496cd56b40 authored about 3 years ago by Sean Silva <[email protected]>
6f710bbc47af7cec6cc87246faa17cca09981154 authored about 3 years ago by Sean Silva <[email protected]>
This is old code that barely worked, and this approach just won't scale.
TorchFX seems like the ...
It didn't make sense in the main registration file.
9e2442d6b01150d56cd7abca0296bb9243fd38ac authored about 3 years ago by Sean Silva <[email protected]>8d27c41f21e5f0cbd8129efd742f7b823d70cb10 authored about 3 years ago by Sean Silva <[email protected]>
It was an unnecessary layer of indirection -- there was nothing outside
of it, and it was just h...
After this change, there are now just two subdirectories in the
`python_packages` directory in o...
Our new dependency management solution relies:
- on the C++ side with the public iree-dialects p...
It just contained the e2e testing framework. We now fold it into the
main project to reduce comp...
2e63f4b1e11dfac6b43fbf3fcc87e91a944f7c37 authored about 3 years ago by Sean Silva <[email protected]>
44d615ac1ec57c213f0feed3083cd91a9d2f40b3 authored about 3 years ago by Sean Silva <[email protected]>
This brings back `check-npcomp-all` and the refbackend e2e tests
coverage.
This moves the bulk of the Python code (including the Torch interop)
from `frontends/pytorch` in...
This creates the `external/torch-mlir` directory as an
LLVM_EXTERNAL_PROJECTS-compatible project...
Last commit was only the last step of that.
28762699b3d797f43e1533450122db7993cf70d5 authored about 3 years ago by Sean Silva <[email protected]>It will be re-enabled after the torch-mlir excision is completed.
0d8af195508b6dcc1ffb677856e4f16f7564db32 authored about 3 years ago by Sean Silva <[email protected]>This plumbs through a vertical slice of support for lists.
The main chunk of new code here is A...
a7252f9a06c98de1508f0c26e6ffc15ea514270f authored about 3 years ago by Sean Silva <[email protected]>
This contains the following changes:
- Fix optional knowledge propagation. The initial knowledg...
We were not filling the `outs` with the neutral element of the
reduction, which resulted in read...
Added lowering to torch.sum into linalg
6724de7692284f63ea15a1ea86018e93adb06b15 authored about 3 years ago by Ramiro Leal-Cavazos <[email protected]>600cc6b9c7c35c372cbd396eb1976f0083c22aa6 authored about 3 years ago by Sean Silva <[email protected]>
We needed to resize the list, not just reserve capacity.
ed2afe43e757807e5605a0fc6174e9b043fb52b5 authored about 3 years ago by Sean Silva <[email protected]>They were confusing users, and most didn't even work anymore.
7a3570e88110151854b8d88fb7cccf2f848606ba authored about 3 years ago by Sean Silva <[email protected]>
- builder.getSymbolRefAttr is gone.
- OpAsmOpInterface's getAsmResultNames method needs explicit...
9cc4fdcaa8f5fe3d65f991ce27e17b22d098063c authored about 3 years ago by Sean Silva <[email protected]>
This should cover all the ops that are left in MT.
3b0e5910a8e18a202542d54573fb8ae484058331 authored about 3 years ago by Yi Zhang <[email protected]>Follows existing conventions for activation functions
d9df4bfc9591fcd4225752e24ecbb9e092fde11b authored about 3 years ago by dan <[email protected]>
It doesn't work properly with the new dialect registration framework.
This was latent and only w...
Change required to enable
./tools/torchscript_e2e_test.sh --config=iree
This reverts commit d8db41b3b6a53138c42b552131b0051050bc9bc1.
These printouts didn't interopera...
1c53424fe70e7a53f67313f172600c310ee67141 authored about 3 years ago by Sean Silva <[email protected]>
- Add `!torch.optional` knowledge tracking
- Changes to improve type propagation for branches an...
Added folders to a few binary computing ops, `TupleUnpack`,
`__contains__.str` and `__getitem__....
Helped with #287.
d8db41b3b6a53138c42b552131b0051050bc9bc1 authored about 3 years ago by Stella Laurenzo <[email protected]>* Now the parts of the MLIR API are directly exported under the npcomp module (i.e. `npcomp.ir`,...
4148f88576a683e2fa530ce146048c025368360b authored about 3 years ago by Stella Laurenzo <[email protected]>177ccdd55beecbf20b4f8796c2f615c905f36ae3 authored about 3 years ago by Stella Laurenzo <[email protected]>
* Requires patch https://reviews.llvm.org/D108527
32f56c67f4240a962103d97754c8c704dcfcac3b authored about 3 years ago by Stella Laurenzo <[email protected]>80ff744c56bb992603fc984010c6bab36f4092b1 authored about 3 years ago by Stella Laurenzo <[email protected]>
This converts a basic list op (torch.prim.ListConstruct) to the IREE
dialect.
```
def forwa...
With the following changes the compilation can continue until
RefineTypes pass:
- Add operators...
85ff8b692b75f1b3b59ef9e52dcac47499465585 authored about 3 years ago by Yi Zhang <[email protected]>78fd07da5f303408254a0ac81932eca81aa110d3 authored about 3 years ago by M4tr1xt4ng <[email protected]>
We plan on using these dialects "natively" as part of the npcomp backend
contract, and provide f...
All tests pass after https://github.com/google/iree/pull/6666 :)
37df45ded4006c27a66fc1810872dafbbce9bec2 authored about 3 years ago by Sean Silva <[email protected]>
Most of the change is in the reporting code to give error messages that
are useful, and adjustin...
This includes the following changes to import MT model into MLIR. There
are still a lot of work ...
This is no longer needed by IREE.
a3bfd115ee758b408c01c001e0da19f253f9ee8b authored about 3 years ago by Sean Silva <[email protected]>
This takes the example from torchscript_resnet18_e2e.py and puts it into
a slightly cleaned up n...
ResNet works with static shapes. (our test is not static though).
All tests blocked on https://...
f71845ea75fe489ed3f2df5291447b3436d86a07 authored about 3 years ago by Sean Silva <[email protected]>
To use, do `ninja npcomp-lsp-server`, copy `build/bin/npcomp-lsp-server`
into your PATH somewher...
- torch.aten.flatten.using_ints to linalg lowering
- torch.aten.max_pool2d to linalg lowering
- ...
This better represents its limited scope. This was causing confusion --
people were feeding it h...
719f0cd709065fffbcf18e90afb6dfdd3bb012cb authored about 3 years ago by Sean Silva <[email protected]>
Related to https://github.com/llvm/mlir-npcomp/issues/266
a153cf4ef26216590a95cad5d5600bbbd5e0dda5 authored about 3 years ago by Sean Silva <[email protected]>
The tests use the same (pure-Python) test framework as the
normal torchscript_e2e_test.sh, but t...
These were legacy concepts that are now superceded by direct Torch to
linalg-on-tensors lowering...
7c788dbfec010df14f4d225491ab209d3ce181ee authored about 3 years ago by Sean Silva <[email protected]>
Show an example of classifying image from
https://commons.wikimedia.org/wiki/File:YellowLabrador...
8494455282674866cff5ce63529dcf0a06eea2b2 authored about 3 years ago by Stella Laurenzo <[email protected]>
`Conv2dNoPaddingModule_basic` and `Conv2dWithPaddingModule_basic` start
failing because of resul...
* Adds a minimal setup.py for frontends/pytorch
* Makes npcomp-core export its headers and libra...
cd44a35177cc9ae4edd4e3e32a8a22fab7e3a56f authored about 3 years ago by Stella Laurenzo <[email protected]>
dc305c5101a5115baca14a379130e2f1bb7546f4 authored about 3 years ago by Stella Laurenzo <[email protected]>
77e8ecaff8cb7a9b491faad7f3f37bdc36e32d0e authored about 3 years ago by Stella Laurenzo <[email protected]>
Add scripts for generating .env and update instructions in README.
6fbf94f0b2dcdeb8b27a56daa8b5dd898d59312b authored about 3 years ago by Yi Zhang <[email protected]>* Change aligned_alloc -> malloc. It can fail (and does on MacOS) and is a bit over-aggressive o...
ec611c1e6f44eb5b49c658fd98740000935a1058 authored about 3 years ago by Stella Laurenzo <[email protected]>* Adapt to python build system updates.
* Bump llvm to 310c9496d80961188e8d8f8ad306cdf44bd754...
2dbab50444e9c30eabbd3355a47545c0650fd100 authored about 3 years ago by Stella Laurenzo <[email protected]>
* Added additional *ToLLVM conversion patterns (they were disaggregated from standard).
* Misc ...
1. Add m_TorchConstantIntList
2. Lowering for aten.conv2d
3. Lowering aten.AdaptiveAvgPool2d
5f1b2ba3232181551cbe3c31b72c409de0d64bf4 authored over 3 years ago by Yi Zhang <[email protected]>
- Build adjustments for `.cpp.inc` dialect files.
- Renaming of `memref.dim` to `tensor.dim` for...
This includes IREE and RefBackend.
This includes a fixup to torchscript_e2e_test.sh for handlin...
ef118eb1e12797ea55ea3f89a9411ad44d4f848b authored over 3 years ago by Sean Silva <[email protected]>I'm seeing the following error:
```
CMake Error in frontends/pytorch/csrc/CMakeLists.txt:
Imp...
Use the new `tools/torchscript_e2e_test.sh`.
Also, fix a few whitespace/comment issues.
c289d83407951e6d8ec654ebbafb3e4c6e4202f9 authored over 3 years ago by Sean Silva <[email protected]>
- Add support for "expected failures" in test reporting. The new error
reports look like
[th...
We plumb through e2e a fair number of interesting cases:
- unary, binary, ternary elementwise op...
The underlying issue seems to be resolved now:
https://github.com/pytorch/pytorch/issues/60485
49b5b7272bbdc46801826714e57dec5f984fd722 authored over 3 years ago by Sean Silva <[email protected]>
Changes:
- Change to operand ordering of `linalg.fill`.
This printing is chosen to be similar to how MLIR prints the values by
default.
This allows it to fold away in trivial cases.
60a947b4a7cc0edd21f00e0cc10e6e3bacaf3299 authored over 3 years ago by Sean Silva <[email protected]>
1. Add TorchToSCF pass.
2. Convert prim.If and prim.If.yield.
- Fold more for aten.gt.int, aten.ne.int and Aten__Getitem__TOp
- Some format cleaning up
This adds a pattern to MaximizeValueSemantics which does a simple
abstract interpretation within...
1. Added a simplified version of torch.aten.batch_norm which only handles
inference and assumes ...
For some reason, pytorch_nightly was being installed for the LLVM build,
and so the wrong line g...
This temporarily works around the CMake error:
```
CMake Error in frontends/pytorch/csrc/CMakeL...
e6adecac834dccd6ed25b54e2b9142aca0b34673 authored over 3 years ago by Yi Zhang <[email protected]>
This removes most of the trivial cases that MaximizeValueSemantics needs
to handle, making it ea...
This brings in a change to canonicalization of `dim` ops, which we need
to adjust our pass pipel...