jinhongyii opened a new pull request #9112:
URL: https://github.com/apache/tvm/pull/9112
Add a requirement for fuse primitive:
The domain of a loop to be fused cannot depend on another loop to be fused.
cc @junrushao1994
--
This is an automated message from the Apache Git
jinhongyii commented on pull request #9100:
URL: https://github.com/apache/tvm/pull/9100#issuecomment-927004958
sure
--
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
junrushao1994 commented on pull request #9100:
URL: https://github.com/apache/tvm/pull/9100#issuecomment-927004840
@jinhongyii would you like to coordinate with @spectrometerHBH and get
DetectIterMap fixed foundamentally? Here is what we can do:
1. Make a minimal reproducible
jinhongyii commented on pull request #9100:
URL: https://github.com/apache/tvm/pull/9100#issuecomment-927004390
The new unittest will fail until the DetectIterMap is fixed because
DebugVerify will call DetectIterMap.
--
This is an automated message from the Apache Git Service.
To
junrushao1994 commented on pull request #9111:
URL: https://github.com/apache/tvm/pull/9111#issuecomment-926978402
The PR is ready for review! CC @comaniac @Laurawly @jcf94 @mbrookhart
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on
huajsj commented on pull request #9108:
URL: https://github.com/apache/tvm/pull/9108#issuecomment-926973143
@comaniac @masahi , please take a look.
--
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
junrushao1994 opened a new pull request #9111:
URL: https://github.com/apache/tvm/pull/9111
This PR is part of the meta schedule project (#8473) that adds the
asynchronous program runner interface, as well as a reference
implementation of RPCRunner. LocalRunner will be implemented
junrushao1994 commented on a change in pull request #9061:
URL: https://github.com/apache/tvm/pull/9061#discussion_r715943914
##
File path: include/tvm/meta_schedule/database.h
##
@@ -0,0 +1,275 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more
junrushao1994 commented on a change in pull request #9061:
URL: https://github.com/apache/tvm/pull/9061#discussion_r715943662
##
File path: python/tvm/meta_schedule/database/database.py
##
@@ -0,0 +1,240 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or
idoudali opened a new pull request #9110:
URL: https://github.com/apache/tvm/pull/9110
…name
This commit adds option "--name" to bash.sh and build.sh to enable the user
specify the name of the container and set the hostname inside the
container as well.
This helps the
comaniac commented on a change in pull request #9061:
URL: https://github.com/apache/tvm/pull/9061#discussion_r715935050
##
File path: python/tvm/meta_schedule/database/json_file.py
##
@@ -0,0 +1,61 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more
junrushao1994 commented on pull request #9061:
URL: https://github.com/apache/tvm/pull/9061#issuecomment-926952259
BTW, here is the convention we are using in TensorIR and meta schedule
project:
- An abstract base class that describe how the interface looks like, e.g.
`Schedule`,
SWu opened a new issue #9109:
URL: https://github.com/apache/tvm/issues/9109
The implementation of `cross_entropy_with_logits` seems to be incorrect:
https://github.com/apache/tvm/blob/main/python/tvm/relay/op/nn/_nn.py#L912
It should be something like:
```
huajsj opened a new pull request #9108:
URL: https://github.com/apache/tvm/pull/9108
This patch is one of serial patch for PR 7892 splitting. this is second part
the pipeline executor, include the configuration JSON file load and pipeline
executor export and import.
--
This is an
This is an automated email from the ASF dual-hosted git repository.
masahi pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git.
from 774ff12 [Frontend][TFLite] fix #9078 (#9099)
add 70f2297 [BYOC] Fix incorrect conv2d padding handling of `dnnl
masahi merged pull request #9097:
URL: https://github.com/apache/tvm/pull/9097
--
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:
AndrewZhaoLuo edited a comment on pull request #9045:
URL: https://github.com/apache/tvm/pull/9045#issuecomment-926823851
Going to wait after #9095 to merge
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL
This is an automated email from the ASF dual-hosted git repository.
masahi pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git.
from 5871608 [Relay] Prepare for merging context_analysis.cc and
device_annotation.cc (#9077)
add 774ff12
masahi closed issue #9078:
URL: https://github.com/apache/tvm/issues/9078
--
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:
masahi merged pull request #9099:
URL: https://github.com/apache/tvm/pull/9099
--
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:
This is an automated email from the ASF dual-hosted git repository.
masahi pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git.
from 8a62f10 [TFLite] Fix padding calculation in Transpose Conv (#9089)
add 5871608 [Relay] Prepare for merging
masahi merged pull request #9077:
URL: https://github.com/apache/tvm/pull/9077
--
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:
This is an automated email from the ASF dual-hosted git repository.
masahi pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git.
from d3d7e8e [PYTHON][FFI] Skip numpy.ascontiguousarray if C_CONTIGUOUS ==
True (#9073)
add 8a62f10 [TFLite] Fix
masahi merged pull request #9089:
URL: https://github.com/apache/tvm/pull/9089
--
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:
tkonolige commented on a change in pull request #36:
URL: https://github.com/apache/tvm-rfcs/pull/36#discussion_r715865170
##
File path: rfcs/0036-tvm-script-namespace.md
##
@@ -0,0 +1,89 @@
+- Feature Name: TVM Script Namespace
+- Start Date: 2021-09-23
+- RFC PR:
Lunderberg commented on a change in pull request #9091:
URL: https://github.com/apache/tvm/pull/9091#discussion_r715844177
##
File path: src/tir/transforms/storage_flatten.cc
##
@@ -50,6 +50,913 @@ using runtime::StorageRank;
using runtime::StorageScope;
using
AndrewZhaoLuo commented on a change in pull request #9102:
URL: https://github.com/apache/tvm/pull/9102#discussion_r715840346
##
File path: python/tvm/relay/frontend/paddlepaddle.py
##
@@ -64,26 +129,193 @@ def _get_pad_size(in_size, dilated_kernel_size,
stride_size):
Lunderberg commented on a change in pull request #9091:
URL: https://github.com/apache/tvm/pull/9091#discussion_r715837497
##
File path: src/tir/transforms/storage_flatten.cc
##
@@ -50,6 +50,913 @@ using runtime::StorageRank;
using runtime::StorageScope;
using
anwang2009 commented on a change in pull request #8952:
URL: https://github.com/apache/tvm/pull/8952#discussion_r715835129
##
File path: python/tvm/relay/frontend/onnx.py
##
@@ -3462,6 +3462,66 @@ def _impl_v10(cls, inputs, attr, params):
return _qnn.op.quantize(out,
AndrewZhaoLuo commented on pull request #9102:
URL: https://github.com/apache/tvm/pull/9102#issuecomment-926849019
Thanks for the contribution! This is great, but can you split this up into
like ~10 PRs so we can effectively review this?
--
This is an automated message from the Apache
Lunderberg commented on a change in pull request #9091:
URL: https://github.com/apache/tvm/pull/9091#discussion_r715831607
##
File path: src/tir/transforms/storage_flatten.cc
##
@@ -50,6 +50,913 @@ using runtime::StorageRank;
using runtime::StorageScope;
using
Mousius opened a new pull request #9107:
URL: https://github.com/apache/tvm/pull/9107
Google Mock is the mocking/helper framework that gets bundled with Google
Test, it used to be separate but now isn't. I ran into the issue of Google
Mock not being configured fully in the i386 build of
mehrdadh commented on issue #9084:
URL: https://github.com/apache/tvm/issues/9084#issuecomment-926826684
I'll take a look at this. thanks for reporting it
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL
AndrewZhaoLuo commented on pull request #9045:
URL: https://github.com/apache/tvm/pull/9045#issuecomment-926823851
Going to wait after #9039 to merge
--
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
areusch commented on pull request #9090:
URL: https://github.com/apache/tvm/pull/9090#issuecomment-926813062
@gromero thanks, please fix the CI issue
--
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
areusch commented on a change in pull request #7742:
URL: https://github.com/apache/tvm/pull/7742#discussion_r715206463
##
File path: tests/micro/stm32/test_code_emitter.py
##
@@ -27,7 +28,8 @@
# 1 = INFO messages are not printed
# 2 = INFO and WARNING messages are not
shashwat14 removed a comment on issue #9078:
URL: https://github.com/apache/tvm/issues/9078#issuecomment-926806805
I think the axis is hard coded to 1 assuming the task is a simple
classification task. However, the softmax can occur over any axis. In the above
example it occurs for every
shashwat14 commented on issue #9078:
URL: https://github.com/apache/tvm/issues/9078#issuecomment-926806805
I think the axis is hard coded to 1 assuming the task is a simple
classification task. However, the softmax can occur over any axis. In the above
example it occurs for every "pixel"
csullivan commented on a change in pull request #9091:
URL: https://github.com/apache/tvm/pull/9091#discussion_r715763492
##
File path: src/tir/transforms/storage_flatten.cc
##
@@ -50,6 +50,913 @@ using runtime::StorageRank;
using runtime::StorageScope;
using
gayatripk1 commented on a change in pull request #9063:
URL: https://github.com/apache/tvm/pull/9063#discussion_r715749082
##
File path: python/tvm/relay/frontend/onnx.py
##
@@ -3430,6 +3430,25 @@ def _impl_v10(cls, inputs, attr, params):
return _qnn.op.quantize(out,
tmoreau89 removed a comment on pull request #9091:
URL: https://github.com/apache/tvm/pull/9091#issuecomment-926758244
Interestingly there is one CI unit test failure:
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use
tmoreau89 commented on pull request #9091:
URL: https://github.com/apache/tvm/pull/9091#issuecomment-926758244
Interestingly there is one CI unit test failure:
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL
Mousius opened a new pull request #9106:
URL: https://github.com/apache/tvm/pull/9106
Using the new name transformations added in #9088, the C interface API is
now generated in C++ rather than in Python.
Follow up PRs will clean up any remaining name transformation
inconsistencies.
mbs-octoml commented on a change in pull request #9103:
URL: https://github.com/apache/tvm/pull/9103#discussion_r715708496
##
File path: include/tvm/driver/driver_api.h
##
@@ -44,6 +44,31 @@
namespace tvm {
+/*!
+ * \brief Returns the optimized IRModule for original fused
Mousius commented on a change in pull request #9103:
URL: https://github.com/apache/tvm/pull/9103#discussion_r715590473
##
File path: python/tvm/relay/build_module.py
##
@@ -282,6 +282,8 @@ def get_executor_from_target(target, target_host):
return executor
+# Which
tmoreau89 opened a new pull request #9105:
URL: https://github.com/apache/tvm/pull/9105
Please join us to welcome @jtuyls as a new reviewer to TVM. Jorn has been
maintaining the Xilinx Vitis AI integration with TVM.
- [Commits
tqchen closed issue #8879:
URL: https://github.com/apache/tvm/issues/8879
--
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:
leandron commented on pull request #8815:
URL: https://github.com/apache/tvm/pull/8815#issuecomment-926678457
> @leandron I checked the ci-cpu docker image and the PyXIR version being
used is still v0.1.6 so the CI needs to be updated. Could you update the CI?
To update the images
jtuyls commented on pull request #8815:
URL: https://github.com/apache/tvm/pull/8815#issuecomment-926652944
@leandron I checked the ci-cpu docker image and the PyXIR version being used
is still v0.1.6 so the CI needs to be updated. Could you update the CI?
--
This is an automated
qrsforever opened a new issue #9104:
URL: https://github.com/apache/tvm/issues/9104
https://github.com/apache/tvm/blob/d3d7e8eb6c201506dc706a055e16eed189dcdb0b/python/tvm/relay/frontend/tensorflow_ops.py#L1959-L1970
Code Example:
```python
model3d =
spectrometerHBH commented on pull request #9041:
URL: https://github.com/apache/tvm/pull/9041#issuecomment-926639300
cc @junrushao1994 @MasterJH5574 @Hzfengsy
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL
tqchen commented on pull request #8777:
URL: https://github.com/apache/tvm/pull/8777#issuecomment-926625279
Coming late to the discussion, thanks for the great work. It would be nice
to discuss the naming a bit. Given we are moving towards a serious first class
PT support. Some namespace
leandron commented on pull request #8815:
URL: https://github.com/apache/tvm/pull/8815#issuecomment-926622771
> @leandron Has the CI been updated to include the changes in #8814? Then I
will fix the conflicts here and retrigger the build.
Sorry I lost track of recent updates.
This is an automated email from the ASF dual-hosted git repository.
tqchen 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 d3d7e8e [PYTHON][FFI] Skip numpy.ascontiguousarray if
tqchen merged pull request #9073:
URL: https://github.com/apache/tvm/pull/9073
--
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 commented on pull request #9073:
URL: https://github.com/apache/tvm/pull/9073#issuecomment-926621208
Thanks @wangxiang2713
--
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
jtuyls commented on pull request #8815:
URL: https://github.com/apache/tvm/pull/8815#issuecomment-926620710
@leandron Has the CI been updated to include the changes in
https://github.com/apache/tvm/pull/8814? Then I will fix the conflicts here and
retrigger the build.
--
This is an
masahi edited a comment on issue #8296:
URL: https://github.com/apache/tvm/issues/8296#issuecomment-926605871
@AndrewZhaoLuo I briefly looked at bfloat16. While fp16 vs bf16 makes no
difference for the conversion pass, it seems it is going to take a lot of
effort to compile and run a bf16
masahi commented on issue #8296:
URL: https://github.com/apache/tvm/issues/8296#issuecomment-926605871
@AndrewZhaoLuo I briefly looked at bfloat16. While fp16 vs bf16 makes no
difference for the conversion pass, it seems it is going to take a lot of
effort to compile and run a bf16 model
masahi closed issue #8294:
URL: https://github.com/apache/tvm/issues/8294
--
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:
masahi commented on issue #8294:
URL: https://github.com/apache/tvm/issues/8294#issuecomment-926600506
I think we can close this now
--
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
masahi edited a comment on issue #8294:
URL: https://github.com/apache/tvm/issues/8294#issuecomment-926599422
I finally finished collecting data on FP16 performance using Tensocore.
Since the NHWC conv2d tensorcore schedule requires a batch size multiple of at
least 8, all batch sizes are
masahi commented on issue #8294:
URL: https://github.com/apache/tvm/issues/8294#issuecomment-926599422
I finally finished collecting data on FP16 performance using Tensocore.
Since the NHWC conv2d tensorcore schedule requires a batch size multiple of at
least 8, all batch sizes are 8. The
This is an automated email from the ASF dual-hosted git repository.
leandron 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 7c6a334 Move the allocates of AoT codegen to be
leandron merged pull request #9065:
URL: https://github.com/apache/tvm/pull/9065
--
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:
liubowen520 commented on pull request #7831:
URL: https://github.com/apache/tvm/pull/7831#issuecomment-926574313
Hi @hypercubestart , Great work! I'm doing with 4bit in TVM now. But I found
there are two points can be improved in asnumpy.
1. Asnumpy don't support the conversion of
mikepapadim opened a new pull request #9103:
URL: https://github.com/apache/tvm/pull/9103
This PR builds on top of @electriclilies initial work to move and unify
build into C++ `driver_api.cc` from `build_module.py`.
It removes redundant optimizations passes and introduces helpers
jiangjiajun commented on pull request #9083:
URL: https://github.com/apache/tvm/pull/9083#issuecomment-926540453
> As mentioned in #9098, can you also change `gen_requirements.py` to add
paddlepaddle into the "tvmc" section, together with all other frontends?
No problem, I'll modify
Mousius commented on a change in pull request #9077:
URL: https://github.com/apache/tvm/pull/9077#discussion_r715445096
##
File path: include/tvm/relay/attrs/annotation.h
##
@@ -32,14 +32,55 @@ namespace tvm {
namespace relay {
/*!
- * \brief Options for the device
Mousius commented on a change in pull request #9077:
URL: https://github.com/apache/tvm/pull/9077#discussion_r715444003
##
File path: src/relay/op/annotation/annotation.cc
##
@@ -36,15 +38,40 @@
namespace tvm {
namespace relay {
-// relay.annotation.on_device
Mousius commented on a change in pull request #9077:
URL: https://github.com/apache/tvm/pull/9077#discussion_r715442621
##
File path: include/tvm/runtime/ndarray.h
##
@@ -38,9 +38,19 @@
#include
namespace tvm {
-namespace runtime {
-typedef DLDevice Device;
+// alias
Mousius commented on a change in pull request #9077:
URL: https://github.com/apache/tvm/pull/9077#discussion_r715442276
##
File path: src/relay/op/memory/device_copy.h
##
@@ -0,0 +1,79 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more
Mousius commented on a change in pull request #9077:
URL: https://github.com/apache/tvm/pull/9077#discussion_r715441811
##
File path: include/tvm/relay/attrs/annotation.h
##
@@ -32,14 +32,55 @@ namespace tvm {
namespace relay {
/*!
- * \brief Options for the device
This is an automated email from the ASF dual-hosted git repository.
leandron pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git.
from af3c819 Support colons inside of TVMC input shape name arguments
(#9080)
add 83ba991 Add `extern "C"` to C
This is an automated email from the ASF dual-hosted git repository.
leandron pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git.
from 04ae028 add paddlepaddle to python/gen_requirements.py (#9098)
add af3c819 Support colons inside of TVMC input
leandron commented on pull request #9094:
URL: https://github.com/apache/tvm/pull/9094#issuecomment-926460575
Thanks @Mousius and @junrushao1994! Merged now!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL
leandron merged pull request #9094:
URL: https://github.com/apache/tvm/pull/9094
--
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:
leandron commented on pull request #9080:
URL: https://github.com/apache/tvm/pull/9080#issuecomment-926459861
Merged now, thanks @Mousius @binarybana @mbrookhart!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
leandron merged pull request #9080:
URL: https://github.com/apache/tvm/pull/9080
--
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:
leandron commented on pull request #9083:
URL: https://github.com/apache/tvm/pull/9083#issuecomment-926452420
As mentioned in #9098, can you also change `gen_requirements.py` to add
paddlepaddle into the "tvmc" section, together with all other frontends?
--
This is an automated message
This is an automated email from the ASF dual-hosted git repository.
leandron pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git.
from f6aa2f3 [Relay] Fix compiler warning in ExtractOperators (#9075)
add 04ae028 add paddlepaddle to
leandron merged pull request #9098:
URL: https://github.com/apache/tvm/pull/9098
--
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:
mikepapadim commented on pull request #9031:
URL: https://github.com/apache/tvm/pull/9031#issuecomment-926440784
PTAL
--
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
junrushao1994 commented on a change in pull request #9061:
URL: https://github.com/apache/tvm/pull/9061#discussion_r715398873
##
File path: python/tvm/meta_schedule/database/json_file.py
##
@@ -0,0 +1,37 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or
junrushao1994 commented on a change in pull request #9061:
URL: https://github.com/apache/tvm/pull/9061#discussion_r715396178
##
File path: include/tvm/meta_schedule/database.h
##
@@ -0,0 +1,210 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more
junrushao1994 commented on pull request #9061:
URL: https://github.com/apache/tvm/pull/9061#issuecomment-926419330
@mbrookhart @comaniac I updated the patch according to your comments. Major
changes:
- Remove `WorkloadRegistry`, `WorkloadToken` and the annoying `token_id_`
- Rename
junrushao1994 commented on a change in pull request #9041:
URL: https://github.com/apache/tvm/pull/9041#discussion_r715349916
##
File path: src/tir/schedule/state.cc
##
@@ -206,25 +188,11 @@ class StateCreator : private StmtVisitor {
* \param stmt A for-loop statement or a
87 matches
Mail list logo