[GitHub] [tvm] masahi opened a new issue #7312: [FLAKY] test_dlpack.py

masahi opened a new issue #7312: URL: https://github.com/apache/tvm/issues/7312 https://ci.tlcpack.ai/blue/organizations/jenkins/tvm/detail/PR-7303/12/pipeline This is an automated message from the Apache Git Service. To

[GitHub] [tvm] FrozenGene commented on a change in pull request #6998: [TFLite] Strided slice handling of shrink_axis_mask improved

FrozenGene commented on a change in pull request #6998: URL: https://github.com/apache/tvm/pull/6998#discussion_r560684155 ## File path: tests/python/integration/test_reduce.py ## @@ -191,11 +191,11 @@ def test_rfactor_factor_axis(): n = tvm.runtime.convert(1027) A =

[GitHub] [tvm] anijain2305 commented on pull request #7303: [TOPI] Make cumsum IR reusable, add thrust scan

anijain2305 commented on pull request #7303: URL: https://github.com/apache/tvm/pull/7303#issuecomment-763297967 Thanks :) LGTM :) This is an automated message from the Apache Git Service. To respond to the message,

[GitHub] [tvm] masahi edited a comment on pull request #7303: [TOPI] Make cumsum IR reusable, add thrust scan

masahi edited a comment on pull request #7303: URL: https://github.com/apache/tvm/pull/7303#issuecomment-763291062 @anijain2305 I added an empty tensor test in https://github.com/apache/tvm/pull/7303/commits/a6c740348282c2d13f22883e62c7c910b73ad8c2 OpenCL seems to have a problem

[GitHub] [tvm] masahi commented on pull request #7303: [TOPI] Make cumsum IR reusable, add thrust scan

masahi commented on pull request #7303: URL: https://github.com/apache/tvm/pull/7303#issuecomment-763291062 @anijain2305 I added an empty tensor test in https://github.com/apache/tvm/pull/7303/commits/20afc3243a17f48084204855f498c7f9af1cad7a OpenCL seems to have a problem with 0

[GitHub] [tvm] icemelon9 opened a new pull request #7311: [COMMUNITY] @tkonolige -> Reviewer

icemelon9 opened a new pull request #7311: URL: https://github.com/apache/tvm/pull/7311 Please join us to welcome Tristan Konolige (@tkonolige) as a new reviewer. Tristan has been actively contributing to TVM core compiler and TVM script. - [Commits

[GitHub] [tvm] masahi commented on pull request #7303: [TOPI] Make cumsum IR reusable, add thrust scan

masahi commented on pull request #7303: URL: https://github.com/apache/tvm/pull/7303#issuecomment-763264832 > Once it is merged, I can try on my end with TF models as well. Perf improvement is not expected, since it only improves `get_valid_count` slightly if you use thrust scan

[GitHub] [tvm] anijain2305 commented on pull request #7303: [TOPI] Make cumsum IR reusable, add thrust scan

anijain2305 commented on pull request #7303: URL: https://github.com/apache/tvm/pull/7303#issuecomment-763263064 Yes, I eyeballed the changes wrt to empty tensor and they looked good. So, I am happy to approve this PR. Once it is merged, I can try on my end with TF models as well.

[GitHub] [tvm] masahi commented on pull request #7303: [TOPI] Make cumsum IR reusable, add thrust scan

masahi commented on pull request #7303: URL: https://github.com/apache/tvm/pull/7303#issuecomment-763257824 hmm interesting, I've never created a test case with empty tensor, is that possible? Note that the IR is copied straight from https://github.com/apache/tvm/pull/7303, so the

[GitHub] [tvm] comaniac commented on a change in pull request #7297: [FIX,AUTOTVM] Add flop counts to cublas

comaniac commented on a change in pull request #7297: URL: https://github.com/apache/tvm/pull/7297#discussion_r560599941 ## File path: python/tvm/topi/cuda/dense.py ## @@ -44,6 +44,8 @@ def dense_cublas(cfg, data, weight, bias=None, out_dtype=None): matmul =

[GitHub] [tvm] mbrookhart commented on pull request #7303: [TOPI] Make cumsum IR reusable, add thrust scan

mbrookhart commented on pull request #7303: URL: https://github.com/apache/tvm/pull/7303#issuecomment-763245797 Yeah, scanning on the non-inner axis will have a cache locality performance hit, but I'm honestly not sure if that would be better or worse than the overhead from doing a pair

[GitHub] [tvm] anijain2305 commented on pull request #7293: [TFLite] Added ability to infer shapes for arguments

anijain2305 commented on pull request #7293: URL: https://github.com/apache/tvm/pull/7293#issuecomment-763244935 Thanks @d-smirnov @jwfromm This is an automated message from the Apache Git Service. To respond to the

[tvm] branch main updated (2290cc0 -> f8c55db)

This is an automated email from the ASF dual-hosted git repository. anijain2305 pushed a change to branch main in repository https://gitbox.apache.org/repos/asf/tvm.git. from 2290cc0 [TOPI] Minor perf improvement for GPU scatter (#7233) add f8c55db [TFLite] Added ability to infer

[GitHub] [tvm] anijain2305 merged pull request #7293: [TFLite] Added ability to infer shapes for arguments

anijain2305 merged pull request #7293: URL: https://github.com/apache/tvm/pull/7293 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

[GitHub] [tvm] ZihengJiang commented on pull request #6370: [TOPI] Add einsum operator

ZihengJiang commented on pull request #6370: URL: https://github.com/apache/tvm/pull/6370#issuecomment-763230104 @hanke580 Nice work! I did not realize that we already has an implementation for `einsum`. I will take a closer look at it!

[GitHub] [tvm] tqchen commented on pull request #6370: [TOPI] Add einsum operator

tqchen commented on pull request #6370: URL: https://github.com/apache/tvm/pull/6370#issuecomment-763226229 also cc @ZihengJiang This is an automated message from the Apache Git Service. To respond to the message, please

[GitHub] [tvm] gromero commented on a change in pull request #7266: [µTVM] Add TVMPlatformGenerateRandom, a non-cryptographic random number generator.

gromero commented on a change in pull request #7266: URL: https://github.com/apache/tvm/pull/7266#discussion_r560563123 ## File path: tests/micro/qemu/zephyr-runtime/src/main.c ## @@ -161,6 +162,27 @@ tvm_crt_error_t TVMPlatformTimerStop(double* elapsed_time_seconds) {

[GitHub] [tvm] gromero commented on a change in pull request #7266: [µTVM] Add TVMPlatformGenerateRandom, a non-cryptographic random number generator.

gromero commented on a change in pull request #7266: URL: https://github.com/apache/tvm/pull/7266#discussion_r560563123 ## File path: tests/micro/qemu/zephyr-runtime/src/main.c ## @@ -161,6 +162,27 @@ tvm_crt_error_t TVMPlatformTimerStop(double* elapsed_time_seconds) {

[GitHub] [tvm] gromero commented on a change in pull request #7266: [µTVM] Add TVMPlatformGenerateRandom, a non-cryptographic random number generator.

gromero commented on a change in pull request #7266: URL: https://github.com/apache/tvm/pull/7266#discussion_r560564611 ## File path: tests/micro/qemu/zephyr-runtime/src/main.c ## @@ -161,6 +162,27 @@ tvm_crt_error_t TVMPlatformTimerStop(double* elapsed_time_seconds) {

[GitHub] [tvm] dlexplorer opened a new issue #7310: Error of compilation ONNX BERT with tune information

dlexplorer opened a new issue #7310: URL: https://github.com/apache/tvm/issues/7310 commit [[AutoScheduler] Add layout rewrite support for dense and batch matmul…](https://github.com/apache/tvm/commit/7dcafb017a05ac0d5ecd7cfe8d8741d33a24bbad) breaks compilation of ONNX BERT with autotune

[GitHub] [tvm] masahi edited a comment on pull request #7303: [TOPI] Make cumsum IR reusable, add thrust scan

masahi edited a comment on pull request #7303: URL: https://github.com/apache/tvm/pull/7303#issuecomment-763106769 1. Right now, inclusive scan can be supported by `exclusive_scan(data) + data`. I think that is fine for now, given that our scan IR is far from stable and we don't want to

[GitHub] [tvm] masahi commented on pull request #7303: [TOPI] Make cumsum IR reusable, add thrust scan

masahi commented on pull request #7303: URL: https://github.com/apache/tvm/pull/7303#issuecomment-763106769 1. Right now, inclusive scan can be supported by `exclusive_scan(data) + data`. I think that is fine for now, given that our scan IR is far from stable and we don't want to maintain

[GitHub] [tvm] mbrookhart commented on pull request #7233: [TOPI] Minor perf improvement for GPU scatter

mbrookhart commented on pull request #7233: URL: https://github.com/apache/tvm/pull/7233#issuecomment-763065131 Thanks @masahi @tkonolige @FrozenGene This is an automated message from the Apache Git Service. To respond to

[tvm] branch main updated: [TOPI] Minor perf improvement for GPU scatter (#7233)

This is an automated email from the ASF dual-hosted git repository. mbrookhart 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 2290cc0 [TOPI] Minor perf improvement for GPU

[GitHub] [tvm] mbrookhart merged pull request #7233: [TOPI] Minor perf improvement for GPU scatter

mbrookhart merged pull request #7233: URL: https://github.com/apache/tvm/pull/7233 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

[GitHub] [tvm] tqchen edited a comment on pull request #7309: [FIX] Add GetFullType to print the type of a container and its contents

tqchen edited a comment on pull request #7309: URL: https://github.com/apache/tvm/pull/7309#issuecomment-763036932 The deletion on non-virtual destructor is fine as long as they are held via ObjectPtr (the deleter is stored inside the object itself). Unfortunately directly printing

[GitHub] [tvm] tqchen commented on pull request #7309: [FIX] Add GetFullType to print the type of a container and its contents

tqchen commented on pull request #7309: URL: https://github.com/apache/tvm/pull/7309#issuecomment-763036932 The deletion on non-virtual destructor is fine as long as they are held via ObjectPtr (the deleter is stored inside the object itself). Unfortunately directly printing out the

[GitHub] [tvm] altanh commented on a change in pull request #7287: [PRNG] Add check to PRNG to make sure that unsigned integer arithmetic is wrapping

altanh commented on a change in pull request #7287: URL: https://github.com/apache/tvm/pull/7287#discussion_r560391942 ## File path: python/tvm/topi/random/kernel.py ## @@ -234,6 +242,18 @@ def gen_ir(gen_ptr, out_gen_ptr, out_array_ptr): out_gen =

[GitHub] [tvm] tkonolige opened a new pull request #7309: [FIX] Add GetFullType to print the type of a container and its contents

tkonolige opened a new pull request #7309: URL: https://github.com/apache/tvm/pull/7309 I've added a new function `GetFullType` on all object that prints the type of the object along with the types of objects it contains. I've only updated on assert to use it so far, I'll update the rest

[GitHub] [tvm] areusch commented on a change in pull request #7266: [µTVM] Add TVMPlatformGenerateRandom, a non-cryptographic random number generator.

areusch commented on a change in pull request #7266: URL: https://github.com/apache/tvm/pull/7266#discussion_r560359820 ## File path: src/runtime/crt/host/main.cc ## @@ -93,6 +94,20 @@ tvm_crt_error_t TVMPlatformTimerStop(double* elapsed_time_seconds) {

[GitHub] [tvm] areusch commented on a change in pull request #7266: [µTVM] Add TVMPlatformGenerateRandom, a non-cryptographic random number generator.

areusch commented on a change in pull request #7266: URL: https://github.com/apache/tvm/pull/7266#discussion_r560359820 ## File path: src/runtime/crt/host/main.cc ## @@ -93,6 +94,20 @@ tvm_crt_error_t TVMPlatformTimerStop(double* elapsed_time_seconds) {

[GitHub] [tvm] tkonolige opened a new pull request #7308: [FIX] Infer input shape in sparse_dense_padded's alter_op if one does not exist

tkonolige opened a new pull request #7308: URL: https://github.com/apache/tvm/pull/7308 If there are multiple alter_ops in a model, the first alteration does not run type inference for the subsequent ones. In this case, we don't have the shape information, so we run the inferencer

[GitHub] [tvm] mbrookhart commented on pull request #7303: [TOPI] Make cumsum IR reusable, add thrust scan

mbrookhart commented on pull request #7303: URL: https://github.com/apache/tvm/pull/7303#issuecomment-762969945 Scan is probably the most hand-optimized kernel in thrust, I'm thrilled to be within 10x for a cross-GPU kernel. Overall I'm happy with this, but I have 2 thoughts. 1.

[tvm] branch main updated: [Relay][Frontend][Onnx] Compare against onnxruntime more consistently during testing (#7300)

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 f91b51d [Relay][Frontend][Onnx] Compare against

[GitHub] [tvm] tqchen merged pull request #7300: [Relay][Frontend][Onnx] Compare against onnxruntime more consistently during testing

tqchen merged pull request #7300: URL: https://github.com/apache/tvm/pull/7300 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

[GitHub] [tvm] tqchen commented on pull request #7108: [RUNTIME] Add function name to PackedFunc and TypedPackedFunc

tqchen commented on pull request #7108: URL: https://github.com/apache/tvm/pull/7108#issuecomment-762871262 closing in favor of the alternative capturing approach This is an automated message from the Apache Git Service. To

[GitHub] [tvm] tqchen closed pull request #7108: [RUNTIME] Add function name to PackedFunc and TypedPackedFunc

tqchen closed pull request #7108: URL: https://github.com/apache/tvm/pull/7108 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

[GitHub] [tvm] tqchen closed issue #7302: [TIR][REFACTOR][RFC] ForNode -- Introduce Annotations and ThreadBinding to for_type

tqchen closed issue #7302: URL: https://github.com/apache/tvm/issues/7302 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

[GitHub] [tvm] tqchen commented on issue #7302: [TIR][REFACTOR][RFC] ForNode -- Introduce Annotations and ThreadBinding to for_type

tqchen commented on issue #7302: URL: https://github.com/apache/tvm/issues/7302#issuecomment-762863140 https://github.com/apache/tvm/pull/7306 This is an automated message from the Apache Git Service. To respond to the

[GitHub] [tvm] juannzou removed a comment on issue #7182: VTA test fail while running the 2D convolution testbench on Pynq

juannzou removed a comment on issue #7182: URL: https://github.com/apache/tvm/issues/7182#issuecomment-753602648 Hi @junrushao1994 I thought it is a bug, since I tried with different settings and still get errors. But I just posted it on the discuss forum, according to your

[GitHub] [tvm] masahi commented on pull request #7233: [TOPI] Minor perf improvement for GPU scatter

masahi commented on pull request #7233: URL: https://github.com/apache/tvm/pull/7233#issuecomment-762707056 @tkonolige @mbrookhart I separated the two scatter implementations, things should look clean now. The sequential one is chosen by default, and I confirmed that by tuning the scatter

[GitHub] [tvm] masahi commented on a change in pull request #7233: [TOPI] Minor perf improvement for GPU scatter

masahi commented on a change in pull request #7233: URL: https://github.com/apache/tvm/pull/7233#discussion_r560015060 ## File path: python/tvm/topi/cuda/scatter.py ## @@ -312,19 +313,18 @@ def gen_ir_4d(data, indices, updates, axis, out, update_func): out_ptr =