trevor-m commented on a change in pull request #6905:
URL: https://github.com/apache/incubator-tvm/pull/6905#discussion_r522257464
##
File path: python/tvm/relay/op/contrib/tensorrt.py
##
@@ -598,14 +646,26 @@ def strided_slice_annotate_fn(expr): # pylint:
rohanmukh commented on pull request #6905:
URL: https://github.com/apache/incubator-tvm/pull/6905#issuecomment-726223278
Thanks for the comments @trevor-m.
This is an automated message from the Apache Git Service.
To
comaniac commented on a change in pull request #6882:
URL: https://github.com/apache/incubator-tvm/pull/6882#discussion_r522365307
##
File path: python/tvm/auto_scheduler/measure_record.py
##
@@ -173,3 +180,70 @@ def load_best(filename, workload_key=None, target=None):
rohanmukh commented on a change in pull request #6905:
URL: https://github.com/apache/incubator-tvm/pull/6905#discussion_r522295438
##
File path: python/tvm/relay/op/contrib/tensorrt.py
##
@@ -715,6 +771,34 @@ def conv3d_transpose_annotate_fn(expr): # pylint:
giuseros opened a new pull request #6907:
URL: https://github.com/apache/incubator-tvm/pull/6907
- I refactored the assembly in arm_cpu/tensor_intrin.py to use LLVM+TIR
- Removed the `interleave` boolean parameter in the intrinsic to switch
among two different interleaving modes. LLVM
anijain2305 commented on a change in pull request #6905:
URL: https://github.com/apache/incubator-tvm/pull/6905#discussion_r522282005
##
File path: python/tvm/relay/op/contrib/tensorrt.py
##
@@ -134,15 +135,18 @@ def partition_for_tensorrt(
if params:
comaniac commented on pull request #6903:
URL: https://github.com/apache/incubator-tvm/pull/6903#issuecomment-726264707
> At a high level, it seems to me this may benefit from the refactor I
proposed in #6888. In particular, rather than switching the behaviour of
ScheduleGetter with a
kazum commented on pull request #6886:
URL: https://github.com/apache/incubator-tvm/pull/6886#issuecomment-726236977
@tqchen Thanks, it makes sense to me :) I'll send the change in #6871 later.
Installing libc6-dev-i386 is necessary to compile wasm32 anyway. Could you
merge the
rohanmukh commented on a change in pull request #6905:
URL: https://github.com/apache/incubator-tvm/pull/6905#discussion_r522306819
##
File path: python/tvm/relay/op/contrib/tensorrt.py
##
@@ -134,15 +135,18 @@ def partition_for_tensorrt(
if params:
mod["main"]
giuseros commented on pull request #6907:
URL: https://github.com/apache/incubator-tvm/pull/6907#issuecomment-726287978
cc @mbaret @u99127 @anijain2305
This is an automated message from the Apache Git Service.
To respond to
jwfromm commented on a change in pull request #6839:
URL: https://github.com/apache/incubator-tvm/pull/6839#discussion_r522315510
##
File path: python/tvm/relay/frontend/onnx.py
##
@@ -2199,6 +2199,241 @@ def body_fn(*loop_inputs):
return outputs
+class
jwfromm commented on pull request #6839:
URL: https://github.com/apache/incubator-tvm/pull/6839#issuecomment-726256112
Overall LGTM. I think the documentation on the NMS onnx conversion could be
more detailed since it's a little difficult to understand right now.
alexgl-github opened a new pull request #6908:
URL: https://github.com/apache/incubator-tvm/pull/6908
Thanks for contributing to TVM! Please refer to guideline
https://tvm.apache.org/docs/contribute/ for useful information and tips. After
the pull request is submitted, please request
mbrookhart opened a new pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906
This adds regression tests for a number of bugs uncovered while testing ONNX
Object Detection models, and adds regression tests for the fixes
@tqchen
mbrookhart commented on a change in pull request #6839:
URL: https://github.com/apache/incubator-tvm/pull/6839#discussion_r522504040
##
File path: tests/python/frontend/onnx/test_forward.py
##
@@ -3889,78 +3986,7 @@ def test_if():
if __name__ == "__main__":
-
masahi commented on a change in pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906#discussion_r522513051
##
File path: src/relay/transforms/fold_scale_axis.cc
##
@@ -243,7 +243,18 @@ class ForwardPrep : private ExprVisitor {
}
}
// Visitor
This is an automated email from the ASF dual-hosted git repository.
tqchen pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/incubator-tvm.git.
from 1fbe174 [AutoScheduler] Add winograd support in tuning networks
(#6877)
add 6e048cf Bump up tophup cuda
This is an automated email from the ASF dual-hosted git repository.
tqchen pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/incubator-tvm.git.
from 1fbe174 [AutoScheduler] Add winograd support in tuning networks
(#6877)
add 6e048cf Bump up tophup cuda
tqchen commented on pull request #6909:
URL: https://github.com/apache/incubator-tvm/pull/6909#issuecomment-726411462
cc @jcf94 @merrymercy
This is an automated message from the Apache Git Service.
To respond to the
tqchen commented on pull request #6904:
URL: https://github.com/apache/incubator-tvm/pull/6904#issuecomment-726412047
Thanks @solin319 can you send a regression testcase that prevents this from
happening later?
This is an
mbrookhart commented on pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906#issuecomment-726416707
@tqchen @masahi That sounds good, I'll split the other parts off into a new
PR, and we can wait on `take` until we have a more robust solution.
TaylorZowtuk opened a new pull request #6909:
URL: https://github.com/apache/incubator-tvm/pull/6909
While running scripts using both AutoScheduler and AutoTvm to consecutively
search for schedules for a number of operators/shapes, I observed different
behaviors during measurement
TaylorZowtuk commented on pull request #6909:
URL: https://github.com/apache/incubator-tvm/pull/6909#issuecomment-726338624
@tqchen @merrymercy Thoughts and review please?
This is an automated message from the Apache Git
merrymercy commented on a change in pull request #6882:
URL: https://github.com/apache/incubator-tvm/pull/6882#discussion_r522021963
##
File path: python/tvm/auto_scheduler/task_scheduler.py
##
@@ -337,10 +338,40 @@ def tune(self, tune_option, search_policy="default"):
tkonolige opened a new pull request #6910:
URL: https://github.com/apache/incubator-tvm/pull/6910
This PR flows span information from tvmscript through to tir nodes. It also
modifies many of the tir construction functions to accept a span.
@spectrometerHBH @junrushao1994 @tqchen
trevor-m commented on a change in pull request #6905:
URL: https://github.com/apache/incubator-tvm/pull/6905#discussion_r522477235
##
File path: python/tvm/relay/op/contrib/tensorrt.py
##
@@ -715,6 +771,34 @@ def conv3d_transpose_annotate_fn(expr): # pylint:
tqchen merged pull request #6908:
URL: https://github.com/apache/incubator-tvm/pull/6908
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
masahi commented on a change in pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906#discussion_r522514891
##
File path: src/relay/op/tensor/transform.cc
##
@@ -1076,7 +1076,7 @@ Examples::
.set_support_level(3)
.add_type_rel("Take", TakeRel)
mbrookhart commented on a change in pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906#discussion_r522500934
##
File path: src/relay/op/tensor/transform.cc
##
@@ -1076,7 +1076,7 @@ Examples::
.set_support_level(3)
.add_type_rel("Take",
trevor-m commented on a change in pull request #6905:
URL: https://github.com/apache/incubator-tvm/pull/6905#discussion_r522510919
##
File path: tests/python/contrib/test_tensorrt.py
##
@@ -874,13 +960,82 @@ def test_densenet121():
run_and_verify_model("densenet121")
mbrookhart commented on a change in pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906#discussion_r522516014
##
File path: src/relay/op/tensor/transform.cc
##
@@ -1076,7 +1076,7 @@ Examples::
.set_support_level(3)
.add_type_rel("Take",
mbrookhart opened a new pull request #6911:
URL: https://github.com/apache/incubator-tvm/pull/6911
@tqchen follow up while we discuss #6906
This is an automated message from the Apache Git Service.
To respond to the
masahi commented on a change in pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906#discussion_r522419567
##
File path: src/relay/op/tensor/transform.cc
##
@@ -1076,7 +1076,7 @@ Examples::
.set_support_level(3)
.add_type_rel("Take", TakeRel)
masahi commented on a change in pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906#discussion_r522420882
##
File path: src/relay/op/tensor/transform.cc
##
@@ -1076,7 +1076,7 @@ Examples::
.set_support_level(3)
.add_type_rel("Take", TakeRel)
masahi commented on a change in pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906#discussion_r522517058
##
File path: src/relay/op/tensor/transform.cc
##
@@ -1076,7 +1076,7 @@ Examples::
.set_support_level(3)
.add_type_rel("Take", TakeRel)
tqchen commented on pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906#issuecomment-726416125
@mbrookhart perhaps we can separate out the other things (fold scale axis
and bound) into separate PRs so we can merge them in first. It would also makes
the PR itself
mbrookhart commented on a change in pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906#discussion_r522517222
##
File path: src/relay/op/tensor/transform.cc
##
@@ -1076,7 +1076,7 @@ Examples::
.set_support_level(3)
.add_type_rel("Take",
tqchen commented on pull request #6890:
URL: https://github.com/apache/incubator-tvm/pull/6890#issuecomment-726415566
cc @mbrookhart @jroesch
This is an automated message from the Apache Git Service.
To respond to the
This is an automated email from the ASF dual-hosted git repository.
tqchen pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/incubator-tvm.git.
from 6e048cf Bump up tophup cuda version (#6908)
add f952fa7 [TFLite runtime] Allow to set number of threads
tqchen merged pull request #6901:
URL: https://github.com/apache/incubator-tvm/pull/6901
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
zhiics commented on a change in pull request #6905:
URL: https://github.com/apache/incubator-tvm/pull/6905#discussion_r522473247
##
File path: python/tvm/relay/op/contrib/tensorrt.py
##
@@ -152,13 +153,50 @@ def partition_for_tensorrt(
with
mbrookhart commented on a change in pull request #6839:
URL: https://github.com/apache/incubator-tvm/pull/6839#discussion_r522503532
##
File path: python/tvm/topi/cuda/nms.py
##
@@ -519,14 +557,90 @@ def non_max_suppression(
coord_start,
id_index,
masahi commented on a change in pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906#discussion_r522515694
##
File path: src/relay/op/tensor/transform.cc
##
@@ -1076,7 +1076,7 @@ Examples::
.set_support_level(3)
.add_type_rel("Take", TakeRel)
masahi commented on a change in pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906#discussion_r522514891
##
File path: src/relay/op/tensor/transform.cc
##
@@ -1076,7 +1076,7 @@ Examples::
.set_support_level(3)
.add_type_rel("Take", TakeRel)
rohanmukh commented on a change in pull request #6905:
URL: https://github.com/apache/incubator-tvm/pull/6905#discussion_r522597861
##
File path: python/tvm/relay/op/contrib/tensorrt.py
##
@@ -715,6 +771,34 @@ def conv3d_transpose_annotate_fn(expr): # pylint:
rohanmukh commented on a change in pull request #6905:
URL: https://github.com/apache/incubator-tvm/pull/6905#discussion_r522653603
##
File path: python/tvm/relay/op/contrib/tensorrt.py
##
@@ -152,13 +153,50 @@ def partition_for_tensorrt(
with
masahi commented on pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906#issuecomment-726427732
@mbrookhart maybe you can summarize what the issue is with `take` + dynamic
(for me + others to comment)? I haven't looked into `take` topi compute in
detail, so I
masahi commented on issue #6913:
URL: https://github.com/apache/incubator-tvm/issues/6913#issuecomment-726567977
maybe related https://github.com/apache/incubator-tvm/pull/6911
This is an automated message from the Apache
masahi edited a comment on pull request #6906:
URL: https://github.com/apache/incubator-tvm/pull/6906#issuecomment-726427732
@mbrookhart maybe you can summarize what the issue is with `take` + dynamic
(for me + others to comment)? I haven't looked into `take` topi compute in
detail, so I
lixiaoquan opened a new issue #6913:
URL: https://github.com/apache/incubator-tvm/issues/6913
Patch to reproduce:
```
diff --git a/tests/python/unittest/test_arith_const_int_bound.py
b/tests/python/unittest/test_arith_const_int_bound.py
index 84fc7fd64..0bcdff487 100644
---
solin319 commented on pull request #6904:
URL: https://github.com/apache/incubator-tvm/pull/6904#issuecomment-726550528
add cudnn test in test_topi_conv3d_ncdhw.py
This is an automated message from the Apache Git Service.
To
FrozenGene commented on a change in pull request #6882:
URL: https://github.com/apache/incubator-tvm/pull/6882#discussion_r522587153
##
File path: python/tvm/auto_scheduler/task_scheduler.py
##
@@ -337,10 +338,40 @@ def tune(self, tune_option, search_policy="default"):
merrymercy commented on a change in pull request #6882:
URL: https://github.com/apache/incubator-tvm/pull/6882#discussion_r522655783
##
File path: python/tvm/auto_scheduler/measure_record.py
##
@@ -14,8 +14,13 @@
# KIND, either express or implied. See the License for the
#
anijain2305 opened a new pull request #6912:
URL: https://github.com/apache/incubator-tvm/pull/6912
When compiling a dynamic model with TRT, the primitive function has
constants already bound in the graph. TVM codegen typically lifts those into
variables in ANF pass, so we never see this
merrymercy commented on a change in pull request #6877:
URL: https://github.com/apache/incubator-tvm/pull/6877#discussion_r521989104
##
File path: src/auto_scheduler/search_policy/sketch_policy.cc
##
@@ -116,8 +116,8 @@ SketchPolicy::SketchPolicy(SearchTask task, CostModel
merrymercy commented on a change in pull request #6877:
URL: https://github.com/apache/incubator-tvm/pull/6877#discussion_r521989104
##
File path: src/auto_scheduler/search_policy/sketch_policy.cc
##
@@ -116,8 +116,8 @@ SketchPolicy::SketchPolicy(SearchTask task, CostModel
mbaret commented on pull request #6903:
URL: https://github.com/apache/incubator-tvm/pull/6903#issuecomment-725988146
At a high level, it seems to me this may benefit from the refactor I
proposed in https://github.com/apache/incubator-tvm/pull/6888. In particular,
rather than switching
merrymercy commented on a change in pull request #6882:
URL: https://github.com/apache/incubator-tvm/pull/6882#discussion_r522021963
##
File path: python/tvm/auto_scheduler/task_scheduler.py
##
@@ -337,10 +338,40 @@ def tune(self, tune_option, search_policy="default"):
merrymercy commented on a change in pull request #6882:
URL: https://github.com/apache/incubator-tvm/pull/6882#discussion_r521995550
##
File path: src/auto_scheduler/feature.cc
##
@@ -1345,11 +1345,6 @@ void GetPerStoreFeaturesFromStates(const Array&
states, const SearchTask&
merrymercy commented on a change in pull request #6882:
URL: https://github.com/apache/incubator-tvm/pull/6882#discussion_r521995550
##
File path: src/auto_scheduler/feature.cc
##
@@ -1345,11 +1345,6 @@ void GetPerStoreFeaturesFromStates(const Array&
states, const SearchTask&
merrymercy commented on a change in pull request #6882:
URL: https://github.com/apache/incubator-tvm/pull/6882#discussion_r522021963
##
File path: python/tvm/auto_scheduler/task_scheduler.py
##
@@ -337,10 +338,40 @@ def tune(self, tune_option, search_policy="default"):
merrymercy commented on a change in pull request #6882:
URL: https://github.com/apache/incubator-tvm/pull/6882#discussion_r522021963
##
File path: python/tvm/auto_scheduler/task_scheduler.py
##
@@ -337,10 +338,40 @@ def tune(self, tune_option, search_policy="default"):
solin319 opened a new pull request #6904:
URL: https://github.com/apache/incubator-tvm/pull/6904
when build 3d_conv with cudnn, following error occurred
1. pad_d, pad_h, pad_w = (padding, padding, padding) if isinstance(padding,
int) else padding
ValueError: too many values to
tqchen commented on pull request #6886:
URL: https://github.com/apache/incubator-tvm/pull/6886#issuecomment-726090951
Thanks @kazum what you said makes sense. We could actually change the LLVM
codegen. Under the systemlib mode, it is not necessary to export these two
global variables and
merrymercy merged pull request #6877:
URL: https://github.com/apache/incubator-tvm/pull/6877
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
merrymercy commented on a change in pull request #6882:
URL: https://github.com/apache/incubator-tvm/pull/6882#discussion_r522067286
##
File path: tutorials/auto_scheduler/tune_network_cuda.py
##
@@ -0,0 +1,286 @@
+# Licensed to the Apache Software Foundation (ASF) under one
This is an automated email from the ASF dual-hosted git repository.
lmzheng pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/incubator-tvm.git.
from a7b22ab Dynamic gpu tests, add dynamic strided slice to topi (#6870)
add 1fbe174 [AutoScheduler] Add
merrymercy commented on a change in pull request #6882:
URL: https://github.com/apache/incubator-tvm/pull/6882#discussion_r522067286
##
File path: tutorials/auto_scheduler/tune_network_cuda.py
##
@@ -0,0 +1,286 @@
+# Licensed to the Apache Software Foundation (ASF) under one
rohanmukh opened a new pull request #6905:
URL: https://github.com/apache/incubator-tvm/pull/6905
refactoring test tensort code
added comments to dynamic check wrapper
log.warn changed to logger.info
TRT codegen taking slice_mode into account
TRT codegen to
rohanmukh commented on pull request #6905:
URL: https://github.com/apache/incubator-tvm/pull/6905#issuecomment-726167838
@trevor-m @anijain2305 @zhiics
This is an automated message from the Apache Git Service.
To respond
70 matches
Mail list logo