comaniac commented on pull request #7145:
URL: https://github.com/apache/tvm/pull/7145#issuecomment-749989310
Per offline discussion, now we only support (de)serialization of ComputeDAG
constructed by compute, because this limitation can largely simplify the design.
@merrymercy @jcf9
masahi opened a new pull request #7157:
URL: https://github.com/apache/tvm/pull/7157
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
jcf94 commented on pull request #7156:
URL: https://github.com/apache/tvm/pull/7156#issuecomment-749974856
Will update the log version after #7144 since this PR has modified the log
structure of SearchTask.
This is an automa
jcf94 opened a new pull request #7156:
URL: https://github.com/apache/tvm/pull/7156
AutoScheduler uses a cost model to guide the search search.
We now have NO_REWRITE, INSERT_TRANSFORM_STAGE, REWRITE_FOR_PRE_TRANSFORMED
three options when applying schedule from AutoScheduler.
In m
comaniac commented on pull request #7143:
URL: https://github.com/apache/tvm/pull/7143#issuecomment-749962826
Per offline discussion, we cast the abstract SearchPolicy to the actual
instance so that we can pass it to the packed function.
@merrymercy @jcf94 PTAL.
---
comaniac merged pull request #7151:
URL: https://github.com/apache/tvm/pull/7151
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
This is an automated email from the ASF dual-hosted git repository.
comaniac 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 4a7503d Add a FunctionPattern, remove unused attribu
comaniac commented on pull request #7151:
URL: https://github.com/apache/tvm/pull/7151#issuecomment-749952354
Thanks @mbrookhart @jwfromm
This is an automated message from the Apache Git Service.
To respond to the message, p
jwfromm commented on a change in pull request #7146:
URL: https://github.com/apache/tvm/pull/7146#discussion_r547662242
##
File path: tests/python/topi/python/test_topi_batch_matmul_tensorcore.py
##
@@ -0,0 +1,75 @@
+# Licensed to the Apache Software Foundation (ASF) under one
jwfromm commented on pull request #7146:
URL: https://github.com/apache/tvm/pull/7146#issuecomment-749941299
@Meteorix out of curiosity can you share some of your benchmarking results?
I'd love to know how much faster this performs than cublas.
jwfromm commented on a change in pull request #7147:
URL: https://github.com/apache/tvm/pull/7147#discussion_r547655745
##
File path: python/tvm/topi/cuda/conv2d_alter_op.py
##
@@ -345,4 +347,49 @@ def _conv2d_legalize(attrs, inputs, arg_types):
else:
junrushao1994 commented on a change in pull request #7153:
URL: https://github.com/apache/tvm/pull/7153#discussion_r547579013
##
File path: include/tvm/target/tag.h
##
@@ -139,7 +139,7 @@ inline TargetTagRegEntry& TargetTagRegEntry::set_name() {
}
#define TVM_TARGET_TAG_REG
masahi commented on a change in pull request #7154:
URL: https://github.com/apache/tvm/pull/7154#discussion_r547615181
##
File path: python/tvm/relay/frontend/pytorch_utils.py
##
@@ -25,3 +35,98 @@ def is_version_greater_than(ver):
return "".join(re.findall(r"(\d+\.)(\d+\.
masahi commented on a change in pull request #7154:
URL: https://github.com/apache/tvm/pull/7154#discussion_r547566314
##
File path: tests/python/frontend/pytorch/test_object_detection.py
##
@@ -102,38 +105,55 @@ def test_detection_models():
scripted_model = generate_jit_m
roger-zhao opened a new issue #7155:
URL: https://github.com/apache/tvm/issues/7155
https://github.com/apache/tvm/blob/08a69d4f92742c8c526d6a7c2a5805d00f5dc725/python/tvm/autotvm/task/space.py#L839
here, when _length is None, then initiate it to current search space length,
it's alm
codeislife99 commented on a change in pull request #7126:
URL: https://github.com/apache/tvm/pull/7126#discussion_r547597810
##
File path: include/tvm/topi/transform.h
##
@@ -1386,6 +1386,96 @@ inline Array meshgrid(const Array&
inputs, const std::string& in
return result;
codeislife99 commented on a change in pull request #7126:
URL: https://github.com/apache/tvm/pull/7126#discussion_r547596393
##
File path: include/tvm/topi/transform.h
##
@@ -1386,6 +1386,96 @@ inline Array meshgrid(const Array&
inputs, const std::string& in
return result;
codeislife99 commented on a change in pull request #7126:
URL: https://github.com/apache/tvm/pull/7126#discussion_r547596393
##
File path: include/tvm/topi/transform.h
##
@@ -1386,6 +1386,96 @@ inline Array meshgrid(const Array&
inputs, const std::string& in
return result;
kevinthesun commented on a change in pull request #7154:
URL: https://github.com/apache/tvm/pull/7154#discussion_r547594012
##
File path: python/tvm/relay/frontend/pytorch_utils.py
##
@@ -25,3 +35,98 @@ def is_version_greater_than(ver):
return "".join(re.findall(r"(\d+\.)(
jcf94 commented on a change in pull request #7145:
URL: https://github.com/apache/tvm/pull/7145#discussion_r547593358
##
File path: python/tvm/auto_scheduler/search_task.py
##
@@ -221,10 +221,6 @@ def __init__(
target_host = Target(target_host)
self.dag
jcf94 commented on a change in pull request #7145:
URL: https://github.com/apache/tvm/pull/7145#discussion_r547593358
##
File path: python/tvm/auto_scheduler/search_task.py
##
@@ -221,10 +221,6 @@ def __init__(
target_host = Target(target_host)
self.dag
masahi commented on pull request #7154:
URL: https://github.com/apache/tvm/pull/7154#issuecomment-749860080
I should mention that this rewrite is not run by default, so there is no
perf risk.
This is an automated message fro
zhiics commented on pull request #7154:
URL: https://github.com/apache/tvm/pull/7154#issuecomment-749858927
@masahi I think this is an plausible as well particularly it is only in the
parser. @kevinthesun please help take a look as well. Thanks.
masahi edited a comment on pull request #7154:
URL: https://github.com/apache/tvm/pull/7154#issuecomment-749858279
@zhiics Sure updated the description. Unfortunately I cannot claim that this
is perf improvement. The regression is only 200 us on CPU, so it may be just a
measurement noise,
masahi commented on pull request #7154:
URL: https://github.com/apache/tvm/pull/7154#issuecomment-749858279
@zhiics Sure updated the description. The regression is only 200 us on CPU,
so it may be just a measurement noise.
I have no idea why I'm not getting good speed up. IOU tests,
zhiics commented on pull request #7154:
URL: https://github.com/apache/tvm/pull/7154#issuecomment-749852977
@masahi Thanks for the perf improvement. Could you provide the CPU numbers
as well?
This is an automated message fro
This is an automated email from the ASF dual-hosted git repository.
jroesch 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 08a69d4 [Rust] Impl IsObjectRef for Array (#7138)
08a
jroesch merged pull request #7138:
URL: https://github.com/apache/tvm/pull/7138
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
jroesch commented on pull request #7138:
URL: https://github.com/apache/tvm/pull/7138#issuecomment-749847986
LGTM
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub an
masahi commented on a change in pull request #7154:
URL: https://github.com/apache/tvm/pull/7154#discussion_r547566314
##
File path: tests/python/frontend/pytorch/test_object_detection.py
##
@@ -102,38 +105,55 @@ def test_detection_models():
scripted_model = generate_jit_m
junrushao1994 edited a comment on pull request #7153:
URL: https://github.com/apache/tvm/pull/7153#issuecomment-749842135
Thank you Tristan for the hard work! Would love to ask some questions just
for clarification on the forum first :-)
Would you like to also copy your bullet points
junrushao1994 commented on pull request #7153:
URL: https://github.com/apache/tvm/pull/7153#issuecomment-749842135
Thank you Tristan for the hard work! Would love to ask some questions just
for clarification on the forum first :-)
--
masahi opened a new pull request #7154:
URL: https://github.com/apache/tvm/pull/7154
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
mbrookhart commented on a change in pull request #7126:
URL: https://github.com/apache/tvm/pull/7126#discussion_r547556384
##
File path: include/tvm/topi/transform.h
##
@@ -1386,6 +1386,96 @@ inline Array meshgrid(const Array&
inputs, const std::string& in
return result;
}
tkonolige opened a new pull request #7153:
URL: https://github.com/apache/tvm/pull/7153
- Added libbacktrace to 3rdparty
- Changed build settings to give absolute paths in debug symbols
- Move CHECK and LOG to tvm/support/logging.h
- Rename tvm::Error to tvm::CompileError
- Creat
tkonolige opened a new pull request #7152:
URL: https://github.com/apache/tvm/pull/7152
This PR is an alternate to #7108, that captures the name of a
TypedPackedFunc into a lambda instead of adding it as a field to the class.
Right now, naming a TypedPackedFunc is optional, but I suggest w
comaniac commented on pull request #7142:
URL: https://github.com/apache/tvm/pull/7142#issuecomment-749830449
I see what you meant. How about we just simply add a test in
`test_topi_conv2d_int8.py` that directly calls
`fallback_schedule_cpu_common_int8` that takes a workload generated by
Wheest commented on pull request #7142:
URL: https://github.com/apache/tvm/pull/7142#issuecomment-749822386
Thanks, I understand better what a good test for this PR would be: one that
fails on the current `main` branch but not this PR.
I've been working on devising a test like this,
tkonolige commented on a change in pull request #7107:
URL: https://github.com/apache/tvm/pull/7107#discussion_r547540336
##
File path: tests/scripts/task_ci_python_setup.sh
##
@@ -31,3 +31,4 @@ set -o pipefail
echo "Addtiional setup in" ${CI_IMAGE_NAME}
python3 -m pip inst
comaniac commented on pull request #7142:
URL: https://github.com/apache/tvm/pull/7142#issuecomment-749775593
As you pointed out, the workload doesn't handle asymmetric padding as the
compute implementation, which looks like a bug to me. However, it never
triggers CI errors before, meaning
comaniac commented on pull request #7126:
URL: https://github.com/apache/tvm/pull/7126#issuecomment-749772392
> @comaniac Every time I select request changes on GitHub, it switches it
"suggested changes". Maybe because I am not a committer?
Yes that's the case, but we'll do our best
Wheest commented on pull request #7142:
URL: https://github.com/apache/tvm/pull/7142#issuecomment-749763887
Happy to add a test case if necessary, though I'm still to get familiar with
the testing infrastructure for TVM.
Existing specific TOPI conv2d implementations are tested with a
mbrookhart opened a new pull request #7151:
URL: https://github.com/apache/tvm/pull/7151
Thanks!
cc @comaniac
This is an automated message from the Apache Git Service.
To respond to the message, please log on to G
tkonolige commented on a change in pull request #7149:
URL: https://github.com/apache/tvm/pull/7149#discussion_r547482617
##
File path: src/relay/op/tensor/transform.cc
##
@@ -1553,6 +1553,59 @@ RELAY_REGISTER_OP("meshgrid")
.set_attr("FTVMCompute", MeshgridCompute)
.
tkonolige commented on a change in pull request #7126:
URL: https://github.com/apache/tvm/pull/7126#discussion_r547480430
##
File path: python/tvm/relay/op/transform.py
##
@@ -1320,3 +1320,84 @@ def adv_index(inputs):
Output tensor.
"""
return _make.adv_index
tkonolige commented on pull request #7126:
URL: https://github.com/apache/tvm/pull/7126#issuecomment-749747412
@comaniac Every time I select request changes on GitHub, it switches it
"suggested changes". Maybe because I am not a committer?
-
tkonolige commented on a change in pull request #7126:
URL: https://github.com/apache/tvm/pull/7126#discussion_r547476827
##
File path: python/tvm/relay/op/transform.py
##
@@ -1320,3 +1320,83 @@ def adv_index(inputs):
Output tensor.
"""
return _make.adv_index
jwfromm commented on pull request #7086:
URL: https://github.com/apache/tvm/pull/7086#issuecomment-749726587
I think @tqchen and @icemelon9 need to take another look to confirm that
these changes look good before we can merge. I believe this solution addresses
their concerns while removing
comaniac commented on pull request #7149:
URL: https://github.com/apache/tvm/pull/7149#issuecomment-749725438
The Relay part LGTM. However, since I'm not familiar with the implementation
of those operators, I would ask @tkonolige and @mbrookhart to review this PR.
comaniac commented on pull request #7126:
URL: https://github.com/apache/tvm/pull/7126#issuecomment-749724064
@tkonolige @mbrookhart PTAL and approve or request changes explicitly.
Thanks.
This is an automated message from t
comaniac commented on a change in pull request #7145:
URL: https://github.com/apache/tvm/pull/7145#discussion_r547452564
##
File path: python/tvm/auto_scheduler/search_task.py
##
@@ -221,10 +221,6 @@ def __init__(
target_host = Target(target_host)
self.d
codeislife99 opened a new pull request #7150:
URL: https://github.com/apache/tvm/pull/7150
This PR builds the TF Frontend for sparse_reshape op( #7125 ) [TF:
https://www.tensorflow.org/api_docs/python/tf/sparse/reshape]
Thi
tkonolige commented on a change in pull request #7148:
URL: https://github.com/apache/tvm/pull/7148#discussion_r547395050
##
File path: python/tvm/topi/cuda/sparse.py
##
@@ -311,6 +339,8 @@ def sparse_dense_padded(data, weight_data, weight_indices,
weight_indptr):
output
tkonolige commented on a change in pull request #7125:
URL: https://github.com/apache/tvm/pull/7125#discussion_r547393612
##
File path: python/tvm/relay/op/transform.py
##
@@ -1320,3 +1320,52 @@ def adv_index(inputs):
Output tensor.
"""
return _make.adv_index
comaniac commented on a change in pull request #7144:
URL: https://github.com/apache/tvm/pull/7144#discussion_r547386833
##
File path: python/tvm/auto_scheduler/measure_record.py
##
@@ -98,6 +98,46 @@ def __iter__(self):
yield ret[0], ret[1] # (input, result)
ANSHUMAN87 commented on a change in pull request #7107:
URL: https://github.com/apache/tvm/pull/7107#discussion_r547302078
##
File path: tests/scripts/task_ci_python_setup.sh
##
@@ -31,3 +31,4 @@ set -o pipefail
echo "Addtiional setup in" ${CI_IMAGE_NAME}
python3 -m pip ins
This is an automated email from the ASF dual-hosted git repository.
mbaret pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git.
from 66744d9 [TFLite] pack operation extedned with const args (#6984)
add bc43ed4 [BYOC] [ACL] include_non_call_ops = Fa
mbaret merged pull request #7121:
URL: https://github.com/apache/tvm/pull/7121
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 s
mbaret commented on pull request #7121:
URL: https://github.com/apache/tvm/pull/7121#issuecomment-749529776
Thanks @d-smirnov
This is an automated message from the Apache Git Service.
To respond to the message, please log on
This is an automated email from the ASF dual-hosted git repository.
sijusamuel pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git.
from 968b6f6 Add `is_floating_point()` test and better type support in
`verify_model_vm()` (#7134)
add 66744d9 [TFL
siju-samuel merged pull request #6984:
URL: https://github.com/apache/tvm/pull/6984
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
codeislife99 commented on pull request #7149:
URL: https://github.com/apache/tvm/pull/7149#issuecomment-749495823
cc: @trevor-m @zhiics @comaniac @anijain2305 PTAL !
This is an automated message from the Apache Git Service.
codeislife99 opened a new pull request #7149:
URL: https://github.com/apache/tvm/pull/7149
This PR is for adding support for sparse segment sum OP
(https://www.tensorflow.org/api_docs/python/tf/sparse/segment_sum?hl=bn) as a
part of a larger effort to add sparse operator support. (#7125, #
codeislife99 commented on a change in pull request #7126:
URL: https://github.com/apache/tvm/pull/7126#discussion_r547197106
##
File path: src/relay/op/tensor/transform.cc
##
@@ -1553,6 +1553,63 @@ RELAY_REGISTER_OP("meshgrid")
.set_attr("FTVMCompute", MeshgridCompute)
codeislife99 commented on a change in pull request #7126:
URL: https://github.com/apache/tvm/pull/7126#discussion_r547196976
##
File path: src/relay/op/tensor/transform.cc
##
@@ -1553,6 +1553,63 @@ RELAY_REGISTER_OP("meshgrid")
.set_attr("FTVMCompute", MeshgridCompute)
masahi commented on pull request #7134:
URL: https://github.com/apache/tvm/pull/7134#issuecomment-749440133
Thanks @TylerADavis
This is an automated message from the Apache Git Service.
To respond to the message, please log
This is an automated email from the ASF dual-hosted git repository.
masahi 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 968b6f6 Add `is_floating_point()` test and better type
masahi merged pull request #7134:
URL: https://github.com/apache/tvm/pull/7134
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 s
junrushao1994 closed issue #6943:
URL: https://github.com/apache/tvm/issues/6943
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
junrushao1994 commented on issue #6943:
URL: https://github.com/apache/tvm/issues/6943#issuecomment-749415022
Thanks for reporting. We use the discussion forum
(https://discuss.tvm.apache.org/) for general usage issues. Please open a
thread in the forum. Thanks!
-
junrushao1994 closed issue #6931:
URL: https://github.com/apache/tvm/issues/6931
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
junrushao1994 commented on issue #6931:
URL: https://github.com/apache/tvm/issues/6931#issuecomment-749414663
Thanks for reporting. We use the discussion forum
(https://discuss.tvm.apache.org/) for general usage issues. Please open a
thread in the forum. Thanks!
-
junrushao1994 commented on issue #7057:
URL: https://github.com/apache/tvm/issues/7057#issuecomment-749413853
AFAIK it happens when an error is thrown in the constructor. Thanks for
reporting. We use the discussion forum (https://discuss.tvm.apache.org/) for
general usage issues. Please op
junrushao1994 closed issue #7057:
URL: https://github.com/apache/tvm/issues/7057
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
junrushao1994 commented on issue #7077:
URL: https://github.com/apache/tvm/issues/7077#issuecomment-749413539
Thanks for reporting. We use the discussion forum
(https://discuss.tvm.apache.org/) for general usage issues. Please open a
thread in the forum. Thanks!
-
junrushao1994 closed issue #7077:
URL: https://github.com/apache/tvm/issues/7077
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
76 matches
Mail list logo