tvm-bot commented on PR #13519:
URL: https://github.com/apache/tvm/pull/13519#issuecomment-1331759744
Thanks for contributing to TVM! Please refer to the contributing guidelines
https://tvm.apache.org/docs/contribute/ for useful information and tips. Please
request code reviews
shengxinhu opened a new pull request, #13519:
URL: https://github.com/apache/tvm/pull/13519
This commit aims to solve https://github.com/apache/tvm/issues/13480
--
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 commented on code in PR #13349:
URL: https://github.com/apache/tvm/pull/13349#discussion_r103551
##
include/tvm/driver/driver_api.h:
##
@@ -78,20 +78,22 @@ TVM_DLL transform::Sequential
HostModulePassManager(IRModule mixed_mod, Target t
* \brief Lower an
masahi merged PR #13511:
URL: https://github.com/apache/tvm/pull/13511
--
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 PR #13515:
URL: https://github.com/apache/tvm/pull/13515
--
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 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 8dc8d248a1 [LLVM] Switch to using New Pass Manager
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 e47eed13d9 Add methods to get the size of VTCM on
mehrdadh commented on PR #13518:
URL: https://github.com/apache/tvm/pull/13518#issuecomment-1331542884
cc @guberti for review
--
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
mehrdadh opened a new pull request, #13518:
URL: https://github.com/apache/tvm/pull/13518
This PR adds serial_number as a project option to Arduino project API. In
addition, it is added to micro pytest_plugin to enable testing with assigned
serial number.
--
This is an automated message
tvm-bot commented on PR #13518:
URL: https://github.com/apache/tvm/pull/13518#issuecomment-1331542564
Thanks for contributing to TVM! Please refer to the contributing guidelines
https://tvm.apache.org/docs/contribute/ for useful information and tips. Please
request code reviews
shingjan opened a new issue, #13517:
URL: https://github.com/apache/tvm/issues/13517
Seems like this subgraph failed at `rewrite_tensorize` during tuning.
### Actual behavior
```
1: operator()
at /home/yj/tvm/src/meta_schedule/postproc/rewrite_tensorize.cc:37
This is an automated email from the ASF dual-hosted git repository.
github-bot pushed a change to branch nightly-docker-update
in repository https://gitbox.apache.org/repos/asf/tvm.git
omit c7c67e3532 [ci][docker] Nightly Docker image update
add 435df5081a [microTVM] Use
tvm-bot commented on PR #13516:
URL: https://github.com/apache/tvm/pull/13516#issuecomment-1331465063
Thanks for contributing to TVM! Please refer to the contributing guidelines
https://tvm.apache.org/docs/contribute/ for useful information and tips. Please
request code reviews
Wanger-SJTU opened a new pull request, #13516:
URL: https://github.com/apache/tvm/pull/13516
While `nop` func connected after input, the input `DLTensor` will be
released after `CreateTVMOp`, which may
make program crash when call `SetInputZeroCopy` for `heap use after free`
--
This
kparzysz-quic commented on PR #13515:
URL: https://github.com/apache/tvm/pull/13515#issuecomment-1331458132
cc: @areusch @masahi
--
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
This is an automated email from the ASF dual-hosted git repository.
kparzysz 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 694d4bf5ea [tir] Add copy on write to all nodes
kparzysz-quic merged PR #13512:
URL: https://github.com/apache/tvm/pull/13512
--
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.
mehrdadh pushed a change to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git
from b9f89a2eb9 [Docs][Bug] Fix broken link to tvmc python (#13499)
add b587e33027 [microTVM][Tutorial] Fix micro_aot
mehrdadh merged PR #13513:
URL: https://github.com/apache/tvm/pull/13513
--
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:
tvm-bot commented on PR #13515:
URL: https://github.com/apache/tvm/pull/13515#issuecomment-1331367950
Thanks for contributing to TVM! Please refer to the contributing guidelines
https://tvm.apache.org/docs/contribute/ for useful information and tips. Please
request code reviews
kparzysz-quic opened a new pull request, #13515:
URL: https://github.com/apache/tvm/pull/13515
LLVM 16 has removed support for the legacy pass manager, and the LLVM
codegen in TVM will no longer even compile.
--
This is an automated message from the Apache Git Service.
To respond to the
tvm-bot commented on PR #13514:
URL: https://github.com/apache/tvm/pull/13514#issuecomment-1331222051
Thanks for contributing to TVM! Please refer to the contributing guidelines
https://tvm.apache.org/docs/contribute/ for useful information and tips. Please
request code reviews
mkatanbaf opened a new pull request, #13514:
URL: https://github.com/apache/tvm/pull/13514
adds support for tuning microTVm models using meta-schedule
--
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 PR #13511:
URL: https://github.com/apache/tvm/pull/13511#issuecomment-1331180156
Can we expose the API to python too?
--
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
mehrdadh commented on PR #13513:
URL: https://github.com/apache/tvm/pull/13513#issuecomment-1331171007
@gromero thanks for the review! I added more info in the PR description.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub
tqchen commented on issue #13508:
URL: https://github.com/apache/tvm/issues/13508#issuecomment-1331136601
Ideally we should keep simplifier light weight. In this case, disabling
`kApplyConstraintsToBooleanBranches` makes sense
--
This is an automated message from the Apache Git
tqchen commented on code in PR #13012:
URL: https://github.com/apache/tvm/pull/13012#discussion_r1035150091
##
src/tir/transforms/install_debug_spans.h:
##
@@ -0,0 +1,141 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license
tqchen commented on PR #13012:
URL: https://github.com/apache/tvm/pull/13012#issuecomment-1331127918
Given that we are moving more towards TVMScript, it would be great if we can
trace the span in the TVMScript printer instead?
--
This is an automated message from the Apache Git Service.
AndrewZhaoLuo commented on code in PR #13469:
URL: https://github.com/apache/tvm/pull/13469#discussion_r1035126503
##
src/relay/qnn/op/batch_matmul.cc:
##
@@ -175,27 +200,32 @@ Expr QnnBatchMatmulCanonicalize(const Attrs& attrs, const
Array& new_args,
const auto*
mehrdadh opened a new pull request, #13513:
URL: https://github.com/apache/tvm/pull/13513
Added config_main_stack_size to autotune since running graph debugger
requires more workspace. Also added to micro_aot because of KWS model.
--
This is an automated message from the Apache Git
tvm-bot commented on PR #13512:
URL: https://github.com/apache/tvm/pull/13512#issuecomment-1331079358
Thanks for contributing to TVM! Please refer to the contributing guidelines
https://tvm.apache.org/docs/contribute/ for useful information and tips. Please
request code reviews
driazati opened a new pull request, #13512:
URL: https://github.com/apache/tvm/pull/13512
This enables copy on write methods for all nodes since some were missing
it before (see #13012 for more context)
--
This is an automated message from the Apache Git Service.
To respond to the
driazati commented on code in PR #13012:
URL: https://github.com/apache/tvm/pull/13012#discussion_r1035084604
##
src/tir/transforms/install_debug_spans.cc:
##
@@ -0,0 +1,150 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license
comaniac commented on issue #13508:
URL: https://github.com/apache/tvm/issues/13508#issuecomment-1331032516
@Lunderberg your analysis makes a lot of sense. After removing
`kApplyConstraintsToBooleanBranches`, the lowering time of the example became
13 seconds, which looks much more
kparzysz-quic commented on code in PR #13012:
URL: https://github.com/apache/tvm/pull/13012#discussion_r1034898851
##
include/tvm/ir/expr.h:
##
@@ -526,6 +526,7 @@ class IntImm : public PrimExpr {
TVM_DLL IntImm(DataType dtype, int64_t value, Span span = Span());
vvchernov commented on PR #13472:
URL: https://github.com/apache/tvm/pull/13472#issuecomment-1331017399
Hello @AndrewZhaoLuo! I've done this fix to support quantized
distilbert-like model with mixed-precision in QLinearMatMut. The model is
supported by ONNX RT. After this fix and fix from
csullivan commented on PR #13325:
URL: https://github.com/apache/tvm/pull/13325#issuecomment-1331015785
cc @echuraev @elvin-n does this change look good to you and pass internal
texture tests?
--
This is an automated message from the Apache Git Service.
To respond to the message, please
janetsc commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1035036297
##
src/runtime/hexagon/ops/conv2d.h:
##
@@ -75,15 +77,31 @@ inline void* to_ptr(uintptr_t v) { return
reinterpret_cast(v); }
inline uintptr_t to_uint(void* ptr) {
lhutton1 commented on PR #13212:
URL: https://github.com/apache/tvm/pull/13212#issuecomment-1330982489
Thanks for the updates @sergey-grovety @arina-grovety, looks great! I
started reviewing this evening but didn't fully get through it, will pick up
where I left off tomorrow
--
This is
quic-sanirudh commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1035023792
##
src/runtime/hexagon/ops/conv2d.h:
##
@@ -75,15 +77,31 @@ inline void* to_ptr(uintptr_t v) { return
reinterpret_cast(v); }
inline uintptr_t to_uint(void*
janetsc commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1034920993
##
src/runtime/hexagon/ops/conv2d.h:
##
@@ -133,7 +155,48 @@ inline uintptr_t hwio_at(const DLTensor& f, int y, int x,
int i, int o) {
* @param width
* @param depth
tvm-bot commented on PR #13511:
URL: https://github.com/apache/tvm/pull/13511#issuecomment-1330959206
Thanks for contributing to TVM! Please refer to the contributing guidelines
https://tvm.apache.org/docs/contribute/ for useful information and tips. Please
request code reviews
janetsc opened a new pull request, #13511:
URL: https://github.com/apache/tvm/pull/13511
Expose methods so that callers can get both the size of VTCM on device as
well as the allocated size of the pool for the session.
cc: @masahi
--
This is an automated message from the Apache
tvm-bot commented on PR #13510:
URL: https://github.com/apache/tvm/pull/13510#issuecomment-1330914154
Thanks for contributing to TVM! Please refer to the contributing guidelines
https://tvm.apache.org/docs/contribute/ for useful information and tips. Please
request code reviews
wrongtest-intellif opened a new pull request, #13510:
URL: https://github.com/apache/tvm/pull/13510
Hi, this change try add optional `consumer_blocks` param to `cache_write`
symmetrically, following change on `cache_read` of
https://github.com/apache/tvm/pull/12505.
This is helpful
janetsc commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1034920993
##
src/runtime/hexagon/ops/conv2d.h:
##
@@ -133,7 +155,48 @@ inline uintptr_t hwio_at(const DLTensor& f, int y, int x,
int i, int o) {
* @param width
* @param depth
This is an automated email from the ASF dual-hosted git repository.
gromero 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 b9f89a2eb9 [Docs][Bug] Fix broken link to tvmc python
gromero merged PR #13499:
URL: https://github.com/apache/tvm/pull/13499
--
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:
janetsc commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1034919255
##
src/runtime/hexagon/ops/conv2d.h:
##
@@ -75,15 +77,31 @@ inline void* to_ptr(uintptr_t v) { return
reinterpret_cast(v); }
inline uintptr_t to_uint(void* ptr) {
janetsc commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1034919255
##
src/runtime/hexagon/ops/conv2d.h:
##
@@ -75,15 +77,31 @@ inline void* to_ptr(uintptr_t v) { return
reinterpret_cast(v); }
inline uintptr_t to_uint(void* ptr) {
quic-sanirudh commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1034912923
##
src/runtime/hexagon/ops/conv2d.h:
##
@@ -75,15 +77,31 @@ inline void* to_ptr(uintptr_t v) { return
reinterpret_cast(v); }
inline uintptr_t to_uint(void*
Lunderberg commented on issue #13508:
URL: https://github.com/apache/tvm/issues/13508#issuecomment-1330823429
Looks like the performance degredation is from `RemoveNoOp`. Even though
the data-flow is disabled by default, the analyzer of `IRMutatorWithAnalyzer`
still collects scoped
srkreddy1238 commented on PR #13325:
URL: https://github.com/apache/tvm/pull/13325#issuecomment-1330818742
@junrushao can you take a look 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
cconvey commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1034881781
##
cmake/modules/Hexagon.cmake:
##
@@ -178,6 +178,15 @@ if(BUILD_FOR_HEXAGON)
"${TVMRT_SOURCE_DIR}/hexagon/ops/*.cc"
)
+ include_directories(
+
This is an automated email from the ASF dual-hosted git repository.
ashutoshp pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/main by this push:
new f6f7feafb2 [CMSIS-NN] Support int16 handling for
ashutosh-arm merged PR #13498:
URL: https://github.com/apache/tvm/pull/13498
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail:
cconvey commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1034851072
##
src/runtime/hexagon/ops/conv2d.h:
##
@@ -75,15 +77,31 @@ inline void* to_ptr(uintptr_t v) { return
reinterpret_cast(v); }
inline uintptr_t to_uint(void* ptr) {
quic-sanirudh commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1034828798
##
src/runtime/hexagon/ops/conv2d.h:
##
@@ -144,7 +207,42 @@ void blockize_hwc_16b(void* out, void* inp_flat, int
height, int width, int dept
* @param width
*
quic-sanirudh commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1034828402
##
src/runtime/hexagon/ops/conv2d.h:
##
@@ -133,7 +155,48 @@ inline uintptr_t hwio_at(const DLTensor& f, int y, int x,
int i, int o) {
* @param width
* @param
Lunderberg commented on issue #13508:
URL: https://github.com/apache/tvm/issues/13508#issuecomment-1330735356
Thank you, and I'm seeing the same behavior with this example. Using
f5a102c83, the lowering step runs in 0.18s, while using 101e3a4ad (just after
PR#13217) the lowering step runs
quic-sanirudh commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1034826123
##
cmake/modules/Hexagon.cmake:
##
@@ -178,6 +178,15 @@ if(BUILD_FOR_HEXAGON)
"${TVMRT_SOURCE_DIR}/hexagon/ops/*.cc"
)
+ include_directories(
+
cconvey commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1034756650
##
cmake/modules/Hexagon.cmake:
##
@@ -178,6 +178,15 @@ if(BUILD_FOR_HEXAGON)
"${TVMRT_SOURCE_DIR}/hexagon/ops/*.cc"
)
+ include_directories(
+
bkmgit commented on issue #13492:
URL: https://github.com/apache/tvm/issues/13492#issuecomment-1330669303
CPU build fails with the current changes.
--
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
gromero commented on PR #13499:
URL: https://github.com/apache/tvm/pull/13499#issuecomment-1330618555
@leandron Since there are changes since you're last review, would you like
to take another look? Thanks!
--
This is an automated message from the Apache Git Service.
To respond to the
bkmgit commented on PR #13499:
URL: https://github.com/apache/tvm/pull/13499#issuecomment-1330544029
@gromero Thanks for the feedback. Done.
--
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
masahi commented on issue #13508:
URL: https://github.com/apache/tvm/issues/13508#issuecomment-1330543584
hmm strange, the new flag `use_dataflow_analysis` in `RemoveNoOp` is set to
false by default, so I thought it shouldn't affect the default lowering in any
way.
--
This is an
quic-sanirudh commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1034657546
##
src/runtime/hexagon/ops/conv2d_quant_hvx.cc:
##
@@ -0,0 +1,317 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor
quic-sanirudh commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1034655362
##
src/runtime/hexagon/ops/conv2d.h:
##
@@ -75,15 +77,31 @@ inline void* to_ptr(uintptr_t v) { return
reinterpret_cast(v); }
inline uintptr_t to_uint(void*
quic-sanirudh commented on code in PR #13256:
URL: https://github.com/apache/tvm/pull/13256#discussion_r1034650897
##
src/runtime/hexagon/ops/conv2d.h:
##
@@ -75,15 +77,31 @@ inline void* to_ptr(uintptr_t v) { return
reinterpret_cast(v); }
inline uintptr_t to_uint(void*
neildhickey commented on code in PR #13498:
URL: https://github.com/apache/tvm/pull/13498#discussion_r1034608038
##
tests/python/contrib/test_cmsisnn/test_pooling.py:
##
@@ -103,18 +105,17 @@ def test_op_int8(
compiler_cpu,
cpu_flags,
):
-"""Tests QNN pooling op
rebel-jhlee commented on PR #12254:
URL: https://github.com/apache/tvm/pull/12254#issuecomment-1330368319
@leandron could you tell me it can be merged or not?
Our group always patching this branch to the state-of-the-art TVM code, to
parse the relay with custom type
--
This is an
Mousius commented on issue #13501:
URL: https://github.com/apache/tvm/issues/13501#issuecomment-1330280391
The demo intentionally didn't use the Project API so as to demonstrate how
to integrate TVM into an existing project rather than generating it in TVM. I
think it's valuable to keep it
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 c0ba8a1951 [docs][Adreno] Remove unnecessary
masahi merged PR #13509:
URL: https://github.com/apache/tvm/pull/13509
--
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:
74 matches
Mail list logo