malixian opened a new issue, #15987:
URL: https://github.com/apache/tvm/issues/15987
### Expected behavior
I try to use MetaScheduler to tuning matmul, and the dimensions of the
matrix are m=8192, n=14336, k=8192.
When n=8192, everything is ok, but once m or n is equal to 14336, an err
This is an automated email from the ASF dual-hosted git repository.
github-bot pushed a change to branch nightly
in repository https://gitbox.apache.org/repos/asf/tvm.git
from 3c4ee86d97 [CMake] Fix order of GNUInstallDirs module (#15966)
add 885fc27390 [TVMScript][TIR] Pretty print TIR
JackWeiw commented on PR #15986:
URL: https://github.com/apache/tvm/pull/15986#issuecomment-1780389209
CC @Lunderberg @wrongtest-intellif
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the speci
JackWeiw opened a new pull request, #15986:
URL: https://github.com/apache/tvm/pull/15986
`compact_buffer_region` PASS modify shared buffer stride[0] to
`T.int64(72) * T.min((n + T.int64(63)) // T.int64(64) * T.int64(64),
T.int64(96))` and stride[1] is `T.int64(72)`
but in LowerOpa
spectrometerHBH opened a new pull request, #15985:
URL: https://github.com/apache/tvm/pull/15985
On M2 Ultra, q4f16_1:
7B: 74 tok/s -> 89 tok/s
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go
Hzfengsy merged PR #15950:
URL: https://github.com/apache/tvm/pull/15950
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.
JackWeiw closed pull request #15984: [Fix][TIR] Symbolic strides lower
URL: https://github.com/apache/tvm/pull/15984
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscr
JackWeiw opened a new pull request, #15984:
URL: https://github.com/apache/tvm/pull/15984
compact_buffer_region PASS modify shared buffer stride[0] to
`T.int64(72) * T.min((n + T.int64(63)) // T.int64(64) * T.int64(64),
T.int64(96))` and stride[1] is `T.int64(72)`
but in `LowerOpaq
JackWeiw commented on PR #15961:
URL: https://github.com/apache/tvm/pull/15961#issuecomment-1780341103
> It looks like this PR isn't unity-specific. Can the PR be applied to the
`main` branch instead, so we get the bugfix on both branches?
![1](https://github.com/apache/tvm/assets/126441
This is an automated email from the ASF dual-hosted git repository.
wuwei pushed a commit to branch unity
in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/unity by this push:
new ef39f37f49 [Unity][BYOC] Support variable-length atten
vinx13 merged PR #15959:
URL: https://github.com/apache/tvm/pull/15959
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.or
vinx13 merged PR #15977:
URL: https://github.com/apache/tvm/pull/15977
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.or
This is an automated email from the ASF dual-hosted git repository.
wuwei pushed a commit to branch unity
in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/unity by this push:
new 04c6863a25 [Unity][BYOC] Fix cuBLAS BYOC compatibilty
slyubomirsky commented on PR #15916:
URL: https://github.com/apache/tvm/pull/15916#issuecomment-1780155887
See also [this
issue](https://discuss.tvm.apache.org/t/validity-of-common-subexpression-elimination-pass-that-simplifies-call-tir-args/15832).
Allowing `call_tir` to be more general wo
tqchen commented on PR #15916:
URL: https://github.com/apache/tvm/pull/15916#issuecomment-1780084913
Thanks for the PR, I know this is indeed a generalization and there are some
tradeoffs to be considered here. Specifically, we should consider the following
alternative:
- C0: We enfo
tqchen commented on code in PR #15916:
URL: https://github.com/apache/tvm/pull/15916#discussion_r1372345395
##
include/tvm/relax/expr_functor.h:
##
@@ -278,6 +278,37 @@ class ExprVisitor : public ExprFunctor {
virtual void VisitSpan(const Span& span);
virtual void VisitPri
Lunderberg commented on code in PR #15916:
URL: https://github.com/apache/tvm/pull/15916#discussion_r1372341781
##
src/relax/transform/call_tir_rewrite.cc:
##
@@ -111,41 +111,69 @@ class CallTIRMutator : public ExprMutator {
<< expr->struct_info_;
}
Lunderberg commented on PR #15855:
URL: https://github.com/apache/tvm/pull/15855#issuecomment-1780074018
> ah i see, one way to get around is define the callback as a global test
function and call that with call_packed.
That's what I ended up doing, with a global definition which can
Lunderberg opened a new pull request, #15983:
URL: https://github.com/apache/tvm/pull/15983
TVM containers, such as tvm::runtime::Array, require the contained objects
to inherit from `ObjectRef`. As a result, the wrapper types `IntImm`,
`FloatImm`, and `StringImm` are often used to allow n
This is an automated email from the ASF dual-hosted git repository.
masahi pushed a commit to branch unity
in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/unity by this push:
new 7ef36ebb5d [Unity] Support symbolic PrimValue argumen
masahi merged PR #15980:
URL: https://github.com/apache/tvm/pull/15980
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.or
tqchen commented on PR #15855:
URL: https://github.com/apache/tvm/pull/15855#issuecomment-1780006889
ah i see, one way to get around is define the callback as a global test
function and call that with call_packed. e.g. `test.vm.assert_notnull`
https://github.com/apache/tvm/blob/unity
slyubomirsky commented on code in PR #15916:
URL: https://github.com/apache/tvm/pull/15916#discussion_r1372276835
##
python/tvm/relax/op/base.py:
##
@@ -97,7 +97,11 @@ def call_tir(
ret: Call
A call node for the call_tir operator.
"""
-if isinstance(args,
slyubomirsky closed pull request #15971: [Unity][Op] Allow the argument to
`call_tir` to be a var bound to a tuple, not a tuple literal
URL: https://github.com/apache/tvm/pull/15971
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub
slyubomirsky commented on PR #15971:
URL: https://github.com/apache/tvm/pull/15971#issuecomment-1779980410
Didn't see it, yep, they're duplicates. There is one case that the other PR
misses so hopefully that can be updated.
--
This is an automated message from the Apache Git Service.
To r
slyubomirsky commented on code in PR #15916:
URL: https://github.com/apache/tvm/pull/15916#discussion_r1372273991
##
src/relax/transform/call_tir_rewrite.cc:
##
@@ -111,41 +111,69 @@ class CallTIRMutator : public ExprMutator {
<< expr->struct_info_;
}
slyubomirsky commented on PR #15916:
URL: https://github.com/apache/tvm/pull/15916#issuecomment-1779975180
I think you may need to update the StructInfo inference for
`call_tir_inplace` like in #15971, since (without modification) that assumes
the argument is a tuple literal. The test cases
This is an automated email from the ASF dual-hosted git repository.
github-bot pushed a change to branch dependabot/pip/apps/microtvm/werkzeug-3.0.1
in repository https://gitbox.apache.org/repos/asf/tvm.git
at 48d7096e14 Bump werkzeug from 2.2.3 to 3.0.1 in /apps/microtvm
No new revisions
dependabot[bot] opened a new pull request, #15982:
URL: https://github.com/apache/tvm/pull/15982
Bumps [werkzeug](https://github.com/pallets/werkzeug) from 2.2.3 to 3.0.1.
Release notes
Sourced from https://github.com/pallets/werkzeug/releases";>werkzeug's
releases.
3.0.1
This is an automated email from the ASF dual-hosted git repository.
wuwei pushed a commit to branch unity
in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/unity by this push:
new 5808cea9af [Unity][BYOC] CoreML Scaffolding (#15556)
5
vinx13 merged PR #15556:
URL: https://github.com/apache/tvm/pull/15556
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.or
JackWeiw commented on code in PR #15961:
URL: https://github.com/apache/tvm/pull/15961#discussion_r1372017992
##
src/tir/transforms/inject_ptx_async_copy.cc:
##
@@ -113,9 +116,11 @@ class PTXAsyncCopyInjector : public StmtMutator {
return PrimExpr();
}();
Lunderberg opened a new pull request, #15981:
URL: https://github.com/apache/tvm/pull/15981
- Removed unused `import os`
- Used `tvm.testing.main()` inside `if __name__=="__main__"`
- Added parametrized fixture `exec_mode` instead of marking all tests.
- Replace `@pytest.
Lunderberg opened a new pull request, #15980:
URL: https://github.com/apache/tvm/pull/15980
Prior this this commit, all symbolic variables needed to be defined either
by tensor shapes, or by an explicit `tvm.runtime.ShapeTuple` argument. This
commit allows arguments `arg: R.Prim(value="n")
lhutton1 commented on PR #104:
URL: https://github.com/apache/tvm-rfcs/pull/104#issuecomment-1779468180
Regarding the changes required to support scalability in the data type, I've
been prototyping adding a new `scalable_` attribute to `DataType` that wraps
`DLDataType`.
However, I'v
Lunderberg commented on PR #15961:
URL: https://github.com/apache/tvm/pull/15961#issuecomment-1779454823
It looks like this PR isn't unity-specific. Can the PR be applied to the
`main` branch instead, so we get the bugfix on both branches?
--
This is an automated message from the Apache
quic-sanirudh commented on PR #15971:
URL: https://github.com/apache/tvm/pull/15971#issuecomment-1779451578
> This may be a duplicate of #15916, which also resolves the analogous
problem for `FuseOps`, `RewriteDataflowReshape`, and `FoldConstant`.
Oh nice, this is great, thanks for th
Lunderberg commented on PR #15968:
URL: https://github.com/apache/tvm/pull/15968#issuecomment-1779442186
Looks like the CI failures are due to a check that the PR body is non-empty.
Can you add a description to the PR?
--
This is an automated message from the Apache Git Service.
To respo
Lunderberg merged PR #15970:
URL: https://github.com/apache/tvm/pull/15970
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscr...@tvm.apach
This is an automated email from the ASF dual-hosted git repository.
lunderberg pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/main by this push:
new de56d8c950 [Hotfix] Mark python-FFI handling with T
Lunderberg commented on PR #15971:
URL: https://github.com/apache/tvm/pull/15971#issuecomment-1779402991
This may be a duplicate of https://github.com/apache/tvm/pull/15916, which
also resolves the analogous problem for `FuseOps`, `RewriteDataflowReshape`,
and `FoldConstant`.
--
This is
Lunderberg commented on PR #15977:
URL: https://github.com/apache/tvm/pull/15977#issuecomment-1779395939
Current CI failures were present in unity head, but should be resolved after
PR#15941. (See [this
comment](https://github.com/apache/tvm/pull/15941#issuecomment-1779377445) for
details
github-actions[bot] commented on PR #15977:
URL: https://github.com/apache/tvm/pull/15977#issuecomment-1779395462
Failed to re-run CI in https://github.com/apache/tvm/actions/runs/6641806894
```
Traceback (most recent call last):
File "ci/scripts/github/github_tvmbot.py
Lunderberg commented on PR #15977:
URL: https://github.com/apache/tvm/pull/15977#issuecomment-1779394478
@tvm-bot re-run
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To
Lunderberg commented on PR #15977:
URL: https://github.com/apache/tvm/pull/15977#issuecomment-1779387049
@ci-bot re-run
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To u
Lunderberg commented on PR #15978:
URL: https://github.com/apache/tvm/pull/15978#issuecomment-1779386185
Rebased onto head to re-run CI. Previous failures were present in unity
head, and have been resolved. (See [this
comment](https://github.com/apache/tvm/pull/15941#issuecomment-17793774
Lunderberg commented on PR #15864:
URL: https://github.com/apache/tvm/pull/15864#issuecomment-1779381464
Sounds good, and thank you!
I'm rebasing the commit on top of unity head, as the current CI failures are
due to a breakage on `unity` head, and are resolved with
https://github.co
This is an automated email from the ASF dual-hosted git repository.
lunderberg pushed a commit to branch unity
in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/unity by this push:
new 4d19c8ab1f [Unity][Transform] Improved canonicali
Lunderberg merged PR #15941:
URL: https://github.com/apache/tvm/pull/15941
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscr...@tvm.apach
Lunderberg commented on PR #15941:
URL: https://github.com/apache/tvm/pull/15941#issuecomment-1779377445
No problem!
I'm going to merge this in, as it resolves a few test failures in unity head
that resulted from a conflict between
[PR#15791](https://github.com/apache/tvm/pull/15791)
Lunderberg commented on PR #15855:
URL: https://github.com/apache/tvm/pull/15855#issuecomment-1779338988
One unrelated question, though. In your pseudocode, you have the signature
`def test(x: Object, callback)`, but I wasn't able to pass a callback directly
into a relax function. I could
Lunderberg commented on PR #15855:
URL: https://github.com/apache/tvm/pull/15855#issuecomment-1779329489
Thank you, @tqchen, and I really like that unit test to validate the desired
behavior in the VM, and not just how that behavior impacts a use case. I've
updated the PR to include the un
liquanfeng commented on issue #15101:
URL: https://github.com/apache/tvm/issues/15101#issuecomment-1779026435
A very useful job! And I try it on
tests/python/relax/test_codegen_cudnn.py::test_conv2d_offload according to
tests/python/relax/test_vm_multi_device.py::test_multi_device as shown
HongHongHongL opened a new pull request, #15979:
URL: https://github.com/apache/tvm/pull/15979
This PR uses fast_exp to convert power op to fast but approximate
counterpart.
When y is not an integer, fast_exp(log(x) * y) is much faster than
topi.power(x, y). However, if y is not a co
ekalda commented on PR #15953:
URL: https://github.com/apache/tvm/pull/15953#issuecomment-1778785343
Thanks @cbalint13 for the great work on this!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to
This is an automated email from the ASF dual-hosted git repository.
ekalda pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/main by this push:
new 885fc27390 [TVMScript][TIR] Pretty print TIR LLVM funct
ekalda merged PR #15953:
URL: https://github.com/apache/tvm/pull/15953
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.or
57 matches
Mail list logo