[GitHub] [tvm] masahi commented on pull request #15429: [Unity][DLight] Update GEMV rules

2023-07-28 Thread via GitHub
masahi commented on PR #15429: URL: https://github.com/apache/tvm/pull/15429#issuecomment-1656572692 On RTX 4080 this PR improved over my last week's result as follows: 7B ``` | mode| seqlen | genlen | p50: tok/s | | FT |128 |128

[GitHub] [tvm] tvm-bot commented on pull request #15437: [MetaSchedule] Fix mma default rule and disable tuning abort

2023-07-28 Thread via GitHub
tvm-bot commented on PR #15437: URL: https://github.com/apache/tvm/pull/15437#issuecomment-1656568847 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 from

[GitHub] [tvm] cblmemo commented on issue #15426: [Bug] InternalError: Check failed: (loops.size() == 3 || !state->is_mma) is false: The MMA tensor core only supports SSR loops now

2023-07-28 Thread via GitHub
cblmemo commented on issue #15426: URL: https://github.com/apache/tvm/issues/15426#issuecomment-1656561740 Yes, I believe it is a bug introduced in that PR... Basically, `SSR loops` means a triple nested loop, in which the first two layers are `Spacial` loop, and the third layers are

[GitHub] [tvm] supersat opened a new issue, #15436: [Bug] Unable to build TVM with LLVM 16

2023-07-28 Thread via GitHub
supersat opened a new issue, #15436: URL: https://github.com/apache/tvm/issues/15436 I am attempting to rebuild TVM from scratch with LLVM 16, which was also built from scratch. This used to work, but now I'm getting a cmake error, as shown below. I believe this might be related to #15164.

[tvm] branch nightly updated (4a558204fd -> 5b431f5dd0)

2023-07-28 Thread github-bot
This is an automated email from the ASF dual-hosted git repository. github-bot pushed a change to branch nightly in repository https://gitbox.apache.org/repos/asf/tvm.git from 4a558204fd [Hexagon] Add default vtcm capacity for targets (#15414) add 64ac43a243 [CI] Bump Flax and Jaxlib ve

[GitHub] [tvm] sunggg commented on a diff in pull request #15435: [Unity] cuda graph support for cublas

2023-07-28 Thread via GitHub
sunggg commented on code in PR #15435: URL: https://github.com/apache/tvm/pull/15435#discussion_r1278245037 ## src/runtime/contrib/cublas/cublas.cc: ## @@ -239,8 +239,12 @@ void CallCublasLt(cublasLtHandle_t hdl, const DLTensor* A, const DLTensor* B, co auto B_data = static_

[GitHub] [tvm] Hzfengsy commented on pull request #15427: [Unity][Dlight] Improve Dlight Tensorization Rule

2023-07-28 Thread via GitHub
Hzfengsy commented on PR #15427: URL: https://github.com/apache/tvm/pull/15427#issuecomment-1656536868 Thanks @adfwer233 -- 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

[tvm] branch unity updated: [Unity][Dlight] Improve Dlight Tensorization Rule (#15427)

2023-07-28 Thread syfeng
This is an automated email from the ASF dual-hosted git repository. syfeng pushed a commit to branch unity in repository https://gitbox.apache.org/repos/asf/tvm.git The following commit(s) were added to refs/heads/unity by this push: new 7bb37a54b4 [Unity][Dlight] Improve Dlight Tensorizati

[GitHub] [tvm] Hzfengsy merged pull request #15427: [Unity][Dlight] Improve Dlight Tensorization Rule

2023-07-28 Thread via GitHub
Hzfengsy merged PR #15427: URL: https://github.com/apache/tvm/pull/15427 -- 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: commits-unsubscr...@tvm.apache.

[GitHub] [tvm] junrushao commented on pull request #15433: [Unity] nn.Module Module and Effect

2023-07-28 Thread via GitHub
junrushao commented on PR #15433: URL: https://github.com/apache/tvm/pull/15433#issuecomment-1656511303 There's some issue with the CI. I squashed and rebased this PR trying to make this work -- This is an automated message from the Apache Git Service. To respond to the message, please lo

[GitHub] [tvm] junrushao commented on pull request #15429: [Unity][DLight] Update GEMV rules

2023-07-28 Thread via GitHub
junrushao commented on PR #15429: URL: https://github.com/apache/tvm/pull/15429#issuecomment-1656507395 CC @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 comment. To unsub

[GitHub] [tvm] masahi commented on a diff in pull request #15435: [Unity] cuda graph support for cublas

2023-07-28 Thread via GitHub
masahi commented on code in PR #15435: URL: https://github.com/apache/tvm/pull/15435#discussion_r1278192503 ## src/runtime/contrib/cublas/cublas.cc: ## @@ -239,8 +239,12 @@ void CallCublasLt(cublasLtHandle_t hdl, const DLTensor* A, const DLTensor* B, co auto B_data = static_

[GitHub] [tvm] tvm-bot commented on pull request #15435: [Unity] cuda graph support for cublas

2023-07-28 Thread via GitHub
tvm-bot commented on PR #15435: URL: https://github.com/apache/tvm/pull/15435#issuecomment-1656398146 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 from

[GitHub] [tvm] sunggg opened a new pull request, #15435: [Unity] cuda graph support for cublas

2023-07-28 Thread via GitHub
sunggg opened a new pull request, #15435: URL: https://github.com/apache/tvm/pull/15435 This PR enables cuda graph with cublas. -- 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 comme

[tvm] branch unity updated: [Unity] nn.Module Torch Integration (#15424)

2023-07-28 Thread yaxingcai
This is an automated email from the ASF dual-hosted git repository. yaxingcai pushed a commit to branch unity in repository https://gitbox.apache.org/repos/asf/tvm.git The following commit(s) were added to refs/heads/unity by this push: new 7a4c85e5fb [Unity] nn.Module Torch Integration (#1

[GitHub] [tvm] cyx-6 merged pull request #15424: [Unity] nn.Module Torch Integration

2023-07-28 Thread via GitHub
cyx-6 merged PR #15424: URL: https://github.com/apache/tvm/pull/15424 -- 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: commits-unsubscr...@tvm.apache.org

[GitHub] [tvm] tqchen merged pull request #15423: [Submodule] Add Flash attention v2

2023-07-28 Thread via GitHub
tqchen merged PR #15423: URL: https://github.com/apache/tvm/pull/15423 -- 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: commits-unsubscr...@tvm.apache.or

[tvm] branch main updated: [Submodule] Add Flash attention v2 (#15423)

2023-07-28 Thread tqchen
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 5b431f5dd0 [Submodule] Add Flash attention v2 (#15423)

[GitHub] [tvm] tvm-bot commented on pull request #15434: [Unity][Frontends][Onnx] Improve ConstantOfShape behavior

2023-07-28 Thread via GitHub
tvm-bot commented on PR #15434: URL: https://github.com/apache/tvm/pull/15434#issuecomment-1656234149 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 from

[GitHub] [tvm] tvm-bot commented on pull request #15433: [Unity] nn.Module Module and Effect

2023-07-28 Thread via GitHub
tvm-bot commented on PR #15433: URL: https://github.com/apache/tvm/pull/15433#issuecomment-1656034018 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 from

[GitHub] [tvm] cyx-6 opened a new pull request, #15433: [Unity] nn.Module Module and Effect

2023-07-28 Thread via GitHub
cyx-6 opened a new pull request, #15433: URL: https://github.com/apache/tvm/pull/15433 This PR introduces the Module and Effect in the nn frontend. It adds the built-in `Linear`, `RMSNorm` and `Embedding` modules. And it adds a `KVCache` effect. Also, this PR fix the interface of `R.

[GitHub] [tvm] Liam-Sturge closed issue #15420: [Bug] [CI] Error installing Jax for Ubuntu ci_cpu Docker build

2023-07-28 Thread via GitHub
Liam-Sturge closed issue #15420: [Bug] [CI] Error installing Jax for Ubuntu ci_cpu Docker build URL: https://github.com/apache/tvm/issues/15420 -- 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 sp

[GitHub] [tvm] Liam-Sturge commented on issue #15420: [Bug] [CI] Error installing Jax for Ubuntu ci_cpu Docker build

2023-07-28 Thread via GitHub
Liam-Sturge commented on issue #15420: URL: https://github.com/apache/tvm/issues/15420#issuecomment-1655939601 No longer seeing the Installation error after bumping the version of Jaxlib and Flax. -- This is an automated message from the Apache Git Service. To respond to the message, plea

[GitHub] [tvm] kparzysz-quic commented on pull request #15369: [Relax] Macro recfactor, R.macro prototype

2023-07-28 Thread via GitHub
kparzysz-quic commented on PR #15369: URL: https://github.com/apache/tvm/pull/15369#issuecomment-1655889197 Thanks! Created a PR for the non-Relax part here: https://github.com/apache/tvm/pull/15432 -- This is an automated message from the Apache Git Service. To respond to the message, p

[GitHub] [tvm] kparzysz-quic commented on pull request #15432: [TIR] Generalize implementation of T.macro to work with other dialects

2023-07-28 Thread via GitHub
kparzysz-quic commented on PR #15432: URL: https://github.com/apache/tvm/pull/15432#issuecomment-1655887857 cc: @Lunderberg @junrushao @yzh119 @cyx @slyubomirsky -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL

[GitHub] [tvm] tvm-bot commented on pull request #15432: [TIR] Generalize implementation of T.macro to work with other dialects

2023-07-28 Thread via GitHub
tvm-bot commented on PR #15432: URL: https://github.com/apache/tvm/pull/15432#issuecomment-1655886541 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 from

[GitHub] [tvm] tvm-bot commented on pull request #15432: [TIR] Generalize implementation of T.macro to work with other dialects

2023-07-28 Thread via GitHub
tvm-bot commented on PR #15432: URL: https://github.com/apache/tvm/pull/15432#issuecomment-1655886557 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 from

[GitHub] [tvm] kparzysz-quic opened a new pull request, #15432: [TIR] Generalize implementation of T.macro to work with other dialects

2023-07-28 Thread via GitHub
kparzysz-quic opened a new pull request, #15432: URL: https://github.com/apache/tvm/pull/15432 As a background info---the script parser works by visiting a "statement" (or top-level expression) at a time. The expression parts of the statement are evaluated, and then the IR corresponding to

[GitHub] [tvm] tvm-bot commented on pull request #15431: [BugFix][AutoScheduler][TE] Fix InferRootBound wrong checking for sta…

2023-07-28 Thread via GitHub
tvm-bot commented on PR #15431: URL: https://github.com/apache/tvm/pull/15431#issuecomment-1655698421 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 from

[GitHub] [tvm] tsu-bin opened a new pull request, #15431: [BugFix][AutoScheduler][TE] Fix InferRootBound wrong checking for sta…

2023-07-28 Thread via GitHub
tsu-bin opened a new pull request, #15431: URL: https://github.com/apache/tvm/pull/15431 To reproduce the issue, just construct a simple softmax op and run auto-scheduler, below is the code snippet ``` @auto_scheduler.register_workload def create_softmax_op(*shape): logits =

[tvm] branch main updated (d1f7ef4986 -> 6e33de61af)

2023-07-28 Thread leandron
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 d1f7ef4986 [XGBoost,MetaSchedule] Support xgb set tree method (#15133) add 6e33de61af [Frontend][Keras] Add support

[GitHub] [tvm] leandron merged pull request #15422: [Frontend][Keras] Add support for swish actiivation

2023-07-28 Thread via GitHub
leandron merged PR #15422: URL: https://github.com/apache/tvm/pull/15422 -- 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: commits-unsubscr...@tvm.apache.

[GitHub] [tvm] skystoler closed pull request #14880: [AutoScheduler] Update crossover operator in GA for Ansor

2023-07-28 Thread via GitHub
skystoler closed pull request #14880: [AutoScheduler] Update crossover operator in GA for Ansor URL: https://github.com/apache/tvm/pull/14880 -- 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 spec

[tvm] branch main updated: [XGBoost,MetaSchedule] Support xgb set tree method (#15133)

2023-07-28 Thread xiyou
This is an automated email from the ASF dual-hosted git repository. xiyou 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 d1f7ef4986 [XGBoost,MetaSchedule] Support xgb set tree m

[GitHub] [tvm] zxybazh merged pull request #15133: [XGBoost,MetaSchedule] Support xgb set tree method

2023-07-28 Thread via GitHub
zxybazh merged PR #15133: URL: https://github.com/apache/tvm/pull/15133 -- 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: commits-unsubscr...@tvm.apache.o

[GitHub] [tvm] jikechao opened a new issue, #15430: [Bug] [Relay][Frontend][Keras] Dot operator lead to a crash when crate executor

2023-07-28 Thread via GitHub
jikechao opened a new issue, #15430: URL: https://github.com/apache/tvm/issues/15430 A model with `Dot` operator can be loaded to TVM but crashes when creating the graph executor. ### model structure and RelayIR infor: ![image](https://github.com/apache/tvm/assets/29506758/9b2

[GitHub] [tvm] twmht commented on issue #15426: [Bug] InternalError: Check failed: (loops.size() == 3 || !state->is_mma) is false: The MMA tensor core only supports SSR loops now

2023-07-28 Thread via GitHub
twmht commented on issue #15426: URL: https://github.com/apache/tvm/issues/15426#issuecomment-1655248585 maybe this error was introduced from https://github.com/apache/tvm/pull/14673 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to Gi

[GitHub] [tvm] tvm-bot commented on pull request #15429: [Unity][DLight] Update GEMV rules

2023-07-28 Thread via GitHub
tvm-bot commented on PR #15429: URL: https://github.com/apache/tvm/pull/15429#issuecomment-1655241892 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 from

[GitHub] [tvm] lhutton1 commented on pull request #15421: [CI] Bump Flax and Jaxlib versions to fix Jaxlib install error

2023-07-28 Thread via GitHub
lhutton1 commented on PR #15421: URL: https://github.com/apache/tvm/pull/15421#issuecomment-1655227334 Thanks @Liam-Sturge @yongwww -- 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 c

[tvm] branch main updated: [CI] Bump Flax and Jaxlib versions to fix Jaxlib install error (#15421)

2023-07-28 Thread lukhut
This is an automated email from the ASF dual-hosted git repository. lukhut 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 64ac43a243 [CI] Bump Flax and Jaxlib versions to fix Ja

[GitHub] [tvm] lhutton1 merged pull request #15421: [CI] Bump Flax and Jaxlib versions to fix Jaxlib install error

2023-07-28 Thread via GitHub
lhutton1 merged PR #15421: URL: https://github.com/apache/tvm/pull/15421 -- 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: commits-unsubscr...@tvm.apache.

[GitHub] [tvm] Aleksei-grovety opened a new pull request, #15428: [microNPU][ETHOSU] Fix concatenation with reused buffers

2023-07-28 Thread via GitHub
Aleksei-grovety opened a new pull request, #15428: URL: https://github.com/apache/tvm/pull/15428 Add a pass to copy concatenation arguments which are used more than once in concatenation operation to prevent a situation where an argument used in multiple concatenations will be written to on

[GitHub] [tvm] tvm-bot commented on pull request #15428: [microNPU][ETHOSU] Fix concatenation with reused buffers

2023-07-28 Thread via GitHub
tvm-bot commented on PR #15428: URL: https://github.com/apache/tvm/pull/15428#issuecomment-1655224114 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 from

[GitHub] [tvm] Aarsh2001 commented on pull request #15415: [Frontend][ONNX] add onnx Mish operator

2023-07-28 Thread via GitHub
Aarsh2001 commented on PR #15415: URL: https://github.com/apache/tvm/pull/15415#issuecomment-1655211423 Hey @Hzfengsy, just added a test -- 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 speci

[GitHub] [tvm] tvm-bot commented on pull request #15427: [Unity][Dlight] Improve Dlight Tensorization Rule

2023-07-28 Thread via GitHub
tvm-bot commented on PR #15427: URL: https://github.com/apache/tvm/pull/15427#issuecomment-1655200594 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 from

[GitHub] [tvm] tvm-bot commented on pull request #15378: [Relay] Generate reduced testcase for compilation failure

2023-07-28 Thread via GitHub
tvm-bot commented on PR #15378: URL: https://github.com/apache/tvm/pull/15378#issuecomment-1655192181 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 from