jinhongyii commented on code in PR #15742:
URL: https://github.com/apache/tvm/pull/15742#discussion_r1326594671
##
src/runtime/disco/nccl/nccl.cc:
##
@@ -91,26 +90,34 @@ void AllReduce(NDArray send, ReduceKind reduce_kind,
NDArray recv) {
NCCLThreadLocalContext* ctx =
jinhongyii commented on PR #15742:
URL: https://github.com/apache/tvm/pull/15742#issuecomment-1719661144
null stream (or default stream equivalently) is used for compute. The stream
created in init_ccl is for communication
--
This is an automated message from the Apache Git Service.
To
Thrsu commented on code in PR #15734:
URL: https://github.com/apache/tvm/pull/15734#discussion_r1326228756
##
tests/python/unittest/test_tir_transform_lower_thread_all_reduce.py:
##
Review Comment:
I have rebased it, please recheck.
--
This is an automated message from
Lunderberg opened a new pull request, #15749:
URL: https://github.com/apache/tvm/pull/15749
As a first step to addressing the Metal codegen errors that required the
reversion in https://github.com/apache/tvm/pull/15725, parametrizing the unit
tests for `allreduce`. While these tests are
Lunderberg commented on PR #15749:
URL: https://github.com/apache/tvm/pull/15749#issuecomment-1719764142
@junrushao @MasterJH5574
--
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
tlopex commented on PR #15746:
URL: https://github.com/apache/tvm/pull/15746#issuecomment-1720409267
I wonder why my new commits will be added to original successful PR. That
means I have to create new PR after it is merged?
--
This is an automated message from the Apache Git Service.
To
Hzfengsy merged PR #15749:
URL: https://github.com/apache/tvm/pull/15749
--
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:
tlopex closed pull request #15746: [TFLite][Frontend] Support quantized less
URL: https://github.com/apache/tvm/pull/15746
--
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
jikechao opened a new pull request, #15752:
URL: https://github.com/apache/tvm/pull/15752
The original implementation of Flip in PyTorch converter mistaken the type
of attribute `axis` in the Flip operator as an integer. Thus, It only parses
the first element of the `axis` and will give a
junrushao merged PR #15742:
URL: https://github.com/apache/tvm/pull/15742
--
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:
Cydia2018 opened a new pull request, #15753:
URL: https://github.com/apache/tvm/pull/15753
Purpose: This PR introduces RocBLAS and RocBLAS tuning support in Unity.
Motivation: The industry is asking for diversity of computing power. We at
Kwai AIPCG noticed that, as an alternative to
DongBaiYue opened a new pull request, #105:
URL: https://github.com/apache/tvm-rfcs/pull/105
This RFC is to add a new backend language——SYCL. Previously I also created
an RFC topic in[ the
forum](https://discuss.tvm.apache.org/t/rfc-sycl-sycl-backend-for-tvm/15678).
--
This is an
echuraev commented on code in PR #15752:
URL: https://github.com/apache/tvm/pull/15752#discussion_r1327008892
##
python/tvm/relay/frontend/pytorch.py:
##
@@ -2977,7 +2977,12 @@ def nll_loss(self, inputs, input_types):
def flip(self, inputs, input_types):
data =
junrushao merged PR #15720:
URL: https://github.com/apache/tvm/pull/15720
--
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:
abhikran-quic commented on code in PR #15678:
URL: https://github.com/apache/tvm/pull/15678#discussion_r1322336789
##
python/tvm/relax/transform/optimize_layout_transform.py:
##
@@ -0,0 +1,75 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more
abhikran-quic commented on code in PR #15678:
URL: https://github.com/apache/tvm/pull/15678#discussion_r1322336789
##
python/tvm/relax/transform/optimize_layout_transform.py:
##
@@ -0,0 +1,75 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more
csullivan commented on PR #15725:
URL: https://github.com/apache/tvm/pull/15725#issuecomment-1714981323
Prior to merging I would request that we please include a test or otherwise
link to a test that fails so that there is something semi-minimal for which can
be used to support the desired
junrushao commented on PR #15725:
URL: https://github.com/apache/tvm/pull/15725#issuecomment-1714983668
Agreed @csullivan. Given it's a codegen test, it's a bit challenging though
to have a minimal test if we don't have unittest infra that could potentially
access Metal-capable devices. As
Lucien0 opened a new issue, #15726:
URL: https://github.com/apache/tvm/issues/15726
Hi, I constructed the following case:
```
a = tir.Var("a", "int32")
b = tir.Var("b", "int32")
c = tir.Var("c", "int32")
d = tir.Var("d", "int32")
ana =
cbalint13 commented on code in PR #15685:
URL: https://github.com/apache/tvm/pull/15685#discussion_r1322321004
##
python/tvm/target/x86.py:
##
@@ -16,127 +16,19 @@
# under the License.
"""Common x86 related utilities"""
from .._ffi import register_func
-from .target import
abhikran-quic commented on code in PR #15678:
URL: https://github.com/apache/tvm/pull/15678#discussion_r1322336789
##
python/tvm/relax/transform/optimize_layout_transform.py:
##
@@ -0,0 +1,75 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more
csullivan commented on PR #15694:
URL: https://github.com/apache/tvm/pull/15694#issuecomment-1718370549
Thanks @Lunderberg @ganler @masahi, this is merged.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL
csullivan merged PR #15694:
URL: https://github.com/apache/tvm/pull/15694
--
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:
junrushao commented on issue #15716:
URL: https://github.com/apache/tvm/issues/15716#issuecomment-1718314358
I'm able to confirm on my end that this bug exists in both main and unity
branch. @ysh329 if it doesn't bother you too much, would you mind doing a `git
bisect` to find out which
csullivan merged PR #15672:
URL: https://github.com/apache/tvm/pull/15672
--
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:
junrushao merged PR #15662:
URL: https://github.com/apache/tvm/pull/15662
--
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:
quic-sanirudh opened a new pull request, #15664:
URL: https://github.com/apache/tvm/pull/15664
This PR adds a small change to verify equality of `tvm.ir.Range` as a
structural equal. This assumes that in most cases, comparing two `Range`s means
to compare its `min` and `extent` as opposed
wrongtest-intellif opened a new pull request, #15665:
URL: https://github.com/apache/tvm/pull/15665
Previously, the optimization from `floordiv(floormod(..))` to
`floormod(floordiv(..))` do not check the divisibility. Which may create wrong
iteration form. The change add a default
toyowata commented on issue #15643:
URL: https://github.com/apache/tvm/issues/15643#issuecomment-1701257673
Hi @lhutton1
Yes, I will make a PR.
--
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
toyowata opened a new pull request, #15649:
URL: https://github.com/apache/tvm/pull/15649
Fixed output_data_sec section was missing in the corstone300.ld for Ethos-U.
See more detail: https://github.com/apache/tvm/issues/15643
--
This is an automated message from the Apache Git
tqchen merged PR #15652:
URL: https://github.com/apache/tvm/pull/15652
--
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:
tqchen merged PR #15659:
URL: https://github.com/apache/tvm/pull/15659
--
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:
tqchen merged PR #15653:
URL: https://github.com/apache/tvm/pull/15653
--
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:
junrushao opened a new pull request, #15659:
URL: https://github.com/apache/tvm/pull/15659
This PR refactors the NDArrayCache support with the following changes:
- Support loading metadata from a string rather than a concrete JSON file on
disc;
- Remove dependency to
junrushao merged PR #15658:
URL: https://github.com/apache/tvm/pull/15658
--
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:
junrushao commented on PR #15659:
URL: https://github.com/apache/tvm/pull/15659#issuecomment-1703734529
CC: @kparzysz-quic @tqchen
--
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
junrushao merged PR #15455:
URL: https://github.com/apache/tvm/pull/15455
--
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:
multiverstack-intellif commented on PR #15638:
URL: https://github.com/apache/tvm/pull/15638#issuecomment-1698588609
> > Thanks for the bugfixes! Definitely not an expert here, and do you think
we need to add a test for changes introduced in presurger_set.cc?
>
> @junrushao ,
>
junrushao commented on PR #15638:
URL: https://github.com/apache/tvm/pull/15638#issuecomment-1698593649
Sounds great, and it seems that we all agree with the changes being made!
Will merge it in as soon as the CI is green.
--
This is an automated message from the Apache Git Service.
To
quic-sanirudh commented on PR #15599:
URL: https://github.com/apache/tvm/pull/15599#issuecomment-1698598177
@tvm-bot rerun
--
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.
kparzysz-quic opened a new pull request, #15666:
URL: https://github.com/apache/tvm/pull/15666
When a module with imported modules is exported into a shared library, the
imported modules are serialized and embedded inside of that library. This is
done by generating a raw binary from the
junrushao opened a new pull request, #15668:
URL: https://github.com/apache/tvm/pull/15668
This PR adds a new flag `--cpus` in `./docker/bash.sh`, which is passed to
docker command that allows limiting the number of CPU cores of a docker
container.
Related materials:
junrushao commented on PR #15668:
URL: https://github.com/apache/tvm/pull/15668#issuecomment-1705762320
CC: @tqchen
--
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
haoyang9804 commented on PR #15683:
URL: https://github.com/apache/tvm/pull/15683#issuecomment-1708513363
cc @vvchernov @echuraev
--
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
ashutosh-arm closed pull request #15641: [DO_NOT_MERGE][Flaky][CI] Disable
flaky autotvm test test_multi_filter
URL: https://github.com/apache/tvm/pull/15641
--
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
ashutosh-arm closed issue #15611: [Flaky Test]
`tests/python/unittest/test_autotvm_droplet_tuner.py::test_multi_filter`
URL: https://github.com/apache/tvm/issues/15611
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
ibsidorenko opened a new pull request, #15686:
URL: https://github.com/apache/tvm/pull/15686
This commit adds 2 new Relax ops:
1. "relax.op.smooth" (R.smooth): Multiply elements from a tensor by a scale
(if it operates like "multiply") or pass input as is (if it operates like
Lunderberg commented on PR #15646:
URL: https://github.com/apache/tvm/pull/15646#issuecomment-1708824478
@ci-bot rerun
--
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
junrushao commented on code in PR #15685:
URL: https://github.com/apache/tvm/pull/15685#discussion_r1317632043
##
src/target/llvm/llvm_module.cc:
##
@@ -45,6 +45,7 @@
#include
#include
#include
+#include "llvm/TargetParser/X86TargetParser.h"
Review Comment:
nit: style
cbalint13 commented on code in PR #15685:
URL: https://github.com/apache/tvm/pull/15685#discussion_r1317642603
##
src/target/llvm/llvm_module.cc:
##
@@ -45,6 +45,7 @@
#include
#include
#include
+#include "llvm/TargetParser/X86TargetParser.h"
Review Comment:
done.
MasterJH5574 opened a new pull request, #15687:
URL: https://github.com/apache/tvm/pull/15687
This PR enhances the fallback dlight GPU rule so that it can support more
non-trivial spatial workloads.
Particularly, to this end, this PR makes the following changes:
* for function
lhutton1 commented on issue #12567:
URL: https://github.com/apache/tvm/issues/12567#issuecomment-1708592870
Hi @gessha, I suspect you need to build tvm with LLVM support. Can you try
adding something like the following to config.cmake:
```
set(USE_LLVM llvm-config-16)
```
and
junrushao commented on code in PR #15673:
URL: https://github.com/apache/tvm/pull/15673#discussion_r1317585783
##
src/runtime/disco/loader.cc:
##
@@ -187,5 +231,16 @@ TVM_REGISTER_GLOBAL("runtime.disco.ShardLoaderLoad")
return
cbalint13 opened a new pull request, #15685:
URL: https://github.com/apache/tvm/pull/15685
Hi folks,
This PR leverage **LLVM** itself for CPU features lookup, **replacing
hard-coded** lists.
In order to keep maintainability with X86 families & features we can rely on
LLVM itself.
masahi opened a new pull request, #15707:
URL: https://github.com/apache/tvm/pull/15707
https://github.com/apache/tvm/pull/15657 modified the signature of a mod
generated by `LiftTransformParams` pass to take unpacked params as input rather
than tuple params. Some cutlass tests needs
rutkoor commented on PR #15679:
URL: https://github.com/apache/tvm/pull/15679#issuecomment-1711288315
@tvm-bot rerun
--
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
masahi commented on code in PR #15678:
URL: https://github.com/apache/tvm/pull/15678#discussion_r1319438923
##
python/tvm/relax/transform/optimize_layout_transform.py:
##
@@ -0,0 +1,147 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor
junrushao merged PR #15673:
URL: https://github.com/apache/tvm/pull/15673
--
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:
echuraev merged PR #15683:
URL: https://github.com/apache/tvm/pull/15683
--
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:
sjain58 opened a new pull request, #15708:
URL: https://github.com/apache/tvm/pull/15708
Simplify Conv->bias_add->mul->add to Conv->bias_add sequence if one of the
inputs to Conv, bias_add, mul and add are constant scalars.
def @main(%q1: Tensor[(1, 3, 224, 224), float32]) {
%0 =
rutkoor commented on PR #15679:
URL: https://github.com/apache/tvm/pull/15679#issuecomment-1711281848
@tvm-bot rerun
--
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
rutkoor commented on PR #15679:
URL: https://github.com/apache/tvm/pull/15679#issuecomment-1711285581
@tvm-bot re-build
--
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
rutkoor commented on PR #15679:
URL: https://github.com/apache/tvm/pull/15679#issuecomment-1711284797
@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
Anndrey24 opened a new pull request, #15711:
URL: https://github.com/apache/tvm/pull/15711
Whenever both dotprod and i8mm were available together on a target (e.g.
`"llvm --device=arm_cpu --mtriple=aarch64-linux-gnu
-mattr=+v8.2a,+dotprod,+i8mm"`), the native int8 conv2d implementation
Lunderberg commented on PR #15702:
URL: https://github.com/apache/tvm/pull/15702#issuecomment-1711689431
Whoops! Thank you, and I'm glad we have the functionality!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
Lunderberg closed pull request #15702: [Unity][Utility] Implement operator<<
for tvm::runtime::ShapeTuple
URL: https://github.com/apache/tvm/pull/15702
--
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
Lunderberg commented on PR #15700:
URL: https://github.com/apache/tvm/pull/15700#issuecomment-1711718670
@tqchen For the background PR, the slice index is only included in the saved
parameters if two conditions are met.
1. The symbolic variable is required for a computation that
junrushao merged PR #15486:
URL: https://github.com/apache/tvm/pull/15486
--
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:
junrushao merged PR #15687:
URL: https://github.com/apache/tvm/pull/15687
--
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:
Lunderberg opened a new pull request, #15688:
URL: https://github.com/apache/tvm/pull/15688
Prior to this commit, writing a subclass of `relax::ExprVisitor` or
`relax::ExprMutator` required separate overrides for visiting a
`relax::DataflowVar` and a `relax::Var`. In the majority of
Lunderberg commented on PR #15672:
URL: https://github.com/apache/tvm/pull/15672#issuecomment-1709137698
@ci-bot rerun
--
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
slyubomirsky opened a new pull request, #15689:
URL: https://github.com/apache/tvm/pull/15689
As part of #15319, this PR implements liveness analysis, which is
implemented using a dataflow analysis framework similar to that described by
Adrian Sampson in these lecture notes:
cbalint13 commented on PR #15685:
URL: https://github.com/apache/tvm/pull/15685#issuecomment-1711439202
> The CI fails because the LLVM version on CI is pretty low (==10). I'm
curious if there's any variant of this API on LLVM 10? If not, we should bump
LLVM to 15 or 16
Folks,
Lunderberg commented on PR #15577:
URL: https://github.com/apache/tvm/pull/15577#issuecomment-1708887108
@Hzfengsy Thank you, and I do think the value should be a `tir.PrimExpr` and
not just a `tir.Var` for a few reasons.
* API consistency when applying `BindSymbolicVars`. When a
masahi merged PR #15671:
URL: https://github.com/apache/tvm/pull/15671
--
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:
Lunderberg commented on PR #15627:
URL: https://github.com/apache/tvm/pull/15627#issuecomment-1708956357
Thank you for the detailed response, @tqchen.
I think my main concern is that there are several types of predictability,
depending on the specific audience.
1. As a
csullivan merged PR #15657:
URL: https://github.com/apache/tvm/pull/15657
--
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:
csullivan merged PR #15626:
URL: https://github.com/apache/tvm/pull/15626
--
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:
csullivan commented on PR #15626:
URL: https://github.com/apache/tvm/pull/15626#issuecomment-1708960891
Thanks @Lunderberg and @sunggg! This is merged.
--
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
tqchen commented on PR #15700:
URL: https://github.com/apache/tvm/pull/15700#issuecomment-1711550038
looking at the background PR, I am not too sure if we want to hardcode slice
index in the parameters. This is because in many cases they can be dynamically
given by the runtime.
I
wrongtest-intellif opened a new pull request, #15709:
URL: https://github.com/apache/tvm/pull/15709
(no comment)
--
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
Anndrey24 opened a new pull request, #15710:
URL: https://github.com/apache/tvm/pull/15710
When padding the input data, the int8 conv2d interleaved schedule tries to
split the `data_im2col` cols axis by a factor of 16 in order to then vectorize
over those splits. However, the size of the
tqchen commented on PR #15700:
URL: https://github.com/apache/tvm/pull/15700#issuecomment-1711774724
Definitely agree that the parameters like `lora_scaling`, `temperature`
needs to be passed in.
These are usually parameters that needs to be decoupled from the weights
themselves.
tqchen commented on PR #15700:
URL: https://github.com/apache/tvm/pull/15700#issuecomment-1711783558
To expand a bit, since I think the examples are great illustrating the
overall case, there are usually two category of parameter inputs to a function
- C0: weights, usually fixed for
Hzfengsy merged PR #15934:
URL: https://github.com/apache/tvm/pull/15934
--
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:
ekalda commented on PR #15918:
URL: https://github.com/apache/tvm/pull/15918#issuecomment-1764716961
> * So, I had to look into adding zextend, sextend, truncate plus the
vectorpermute, vectorshuffle instead.
> The good point is that these are lowered to exactly what is needed
Lunderberg commented on PR #15916:
URL: https://github.com/apache/tvm/pull/15916#issuecomment-1764836058
Rebased onto `unity` head to ensure the CI tests haven't broken in the
meantime.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on
leandron commented on PR #15930:
URL: https://github.com/apache/tvm/pull/15930#issuecomment-1764457271
Cc @Liam-Sturge, I will close this so that we can submit a change coming
from the tlcpack repo.
--
This is an automated message from the Apache Git Service.
To respond to the message,
leandron closed pull request #15930: [Do not merge] Update ci-arm image to use
tag 20231013-060139-71caa19f9
URL: https://github.com/apache/tvm/pull/15930
--
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
ysh329 commented on issue #15812:
URL: https://github.com/apache/tvm/issues/15812#issuecomment-1764657318
Hi all, with help of @Hzfengsy, branch v0.14.0 was created. However, because
PR about version modification merged without sqaush, related commit is
unverified, more concretely
Hzfengsy merged PR #15933:
URL: https://github.com/apache/tvm/pull/15933
--
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:
guoyaol commented on PR #15884:
URL: https://github.com/apache/tvm/pull/15884#issuecomment-1757864541
@junrushao @MasterJH5574
--
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
tqchen commented on PR #15912:
URL: https://github.com/apache/tvm/pull/15912#issuecomment-1757926404
Would be great to cross check kv to see if we can work with the offset
parameter in DLTensor. The main reason is that the data ptr may not points to
an addressable location (it is not an
tlopex opened a new pull request, #15915:
URL: https://github.com/apache/tvm/pull/15915
Support Reverse sequence quantization operation as part of #15148
--
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
kparzysz-quic commented on code in PR #15901:
URL: https://github.com/apache/tvm/pull/15901#discussion_r1355671093
##
src/tir/schedule/primitive/loop_transformation.cc:
##
@@ -454,6 +454,56 @@ Array Split(ScheduleState self, const StmtSRef&
loop_sref, const Array
return
Thrsu closed issue #15895: [Bug] [Unity] TypeError encountered when converting
ONNX model with MaxPool operator
URL: https://github.com/apache/tvm/issues/15895
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL
Hzfengsy merged PR #15884:
URL: https://github.com/apache/tvm/pull/15884
--
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:
Hzfengsy merged PR #15893:
URL: https://github.com/apache/tvm/pull/15893
--
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:
junrushao commented on issue #15716:
URL: https://github.com/apache/tvm/issues/15716#issuecomment-1758184178
Yeah it doesn't do anything concrete to the object system...
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use
leandron merged PR #15821:
URL: https://github.com/apache/tvm/pull/15821
--
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:
501 - 600 of 16502 matches
Mail list logo