[GitHub] [tvm] Hzfengsy commented on pull request #14732: [MetaSchedule] Skip custom_rule in case of None

Hzfengsy commented on PR #14732: URL: https://github.com/apache/tvm/pull/14732#issuecomment-1527026549 cc @junrushao -- 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

[GitHub] [tvm] tvm-bot commented on pull request #14737: [Uniy][Op] Expand support of attention bias layout

tvm-bot commented on PR #14737: URL: https://github.com/apache/tvm/pull/14737#issuecomment-1527023230 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

[GitHub] [tvm] cyx-6 opened a new pull request, #14737: [Uniy][Op] Expand support of attention bias layout

cyx-6 opened a new pull request, #14737: URL: https://github.com/apache/tvm/pull/14737 This PR expands the support of attention bias layout. In the past, we only support 3 types of layout, with `BNSS'`, `BSS'`, `BS'`, e.g. `BNSS'` for (batch_size, num_head, seq_len, seq_len_kv). However,

[GitHub] [tvm] echuraev commented on pull request #14707: [Keras][Bugfix] fix a bug about alpha attribute in LeakyReLU which lead to passes conflict

echuraev commented on PR #14707: URL: https://github.com/apache/tvm/pull/14707#issuecomment-1526970118 @tvm-bot rerun -- 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

[GitHub] [tvm] yelite commented on pull request #14665: [Unity] Add MultiSpan and attach it to CallNode produced by FuseOps

yelite commented on PR #14665: URL: https://github.com/apache/tvm/pull/14665#issuecomment-1526860017 @areusch Just want to clarify, I don't think the approach I shared is strictly better than the one implemented in this PR. I just wanted to share some thought on this, because I faced the

[tvm] branch unity updated: [Unity] Process all Relax functions in CompositeFunctionAnnotator (#14736)

This is an automated email from the ASF dual-hosted git repository. masahi 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 9914a77b0f [Unity] Process all Relax functions in

[GitHub] [tvm] masahi merged pull request #14736: [Unity] Process all Relax functions in CompositeFunctionAnnotator

masahi merged PR #14736: URL: https://github.com/apache/tvm/pull/14736 -- 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] branch v0.12.0 updated: [cherry-pick][ARITH] Enhance CanonicalSimplify to Simplify ProdDiv and [ci] disable merge (#14715)

This is an automated email from the ASF dual-hosted git repository. driazati pushed a commit to branch v0.12.0 in repository https://gitbox.apache.org/repos/asf/tvm.git The following commit(s) were added to refs/heads/v0.12.0 by this push: new dd25968edc [cherry-pick][ARITH] Enhance

[GitHub] [tvm] driazati merged pull request #14715: [cherry-pick][ARITH] Enhance CanonicalSimplify to Simplify ProdDiv and [ci] disable merge

driazati merged PR #14715: URL: https://github.com/apache/tvm/pull/14715 -- 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:

[GitHub] [tvm] areusch closed pull request #14665: [Unity] Add MultiSpan and attach it to CallNode produced by FuseOps

areusch closed pull request #14665: [Unity] Add MultiSpan and attach it to CallNode produced by FuseOps URL: https://github.com/apache/tvm/pull/14665 -- 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] areusch commented on pull request #14665: [Unity] Add MultiSpan and attach it to CallNode produced by FuseOps

areusch commented on PR #14665: URL: https://github.com/apache/tvm/pull/14665#issuecomment-1526797435 i understand your point and that you're proposing a workable solution, but you haven't addressed the drawback of that solution which is that there's no forcing function to ensure that

[GitHub] [tvm] yongwww commented on pull request #14635: [CI] Upgrade CI

yongwww commented on PR #14635: URL: https://github.com/apache/tvm/pull/14635#issuecomment-1526702995 @tvm-bot rerun -- 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

[GitHub] [tvm] masahi commented on a diff in pull request #14730: [Unity] Add options to MS tuning pass to enable more fine-grained tuning

masahi commented on code in PR #14730: URL: https://github.com/apache/tvm/pull/14730#discussion_r1179771263 ## python/tvm/meta_schedule/relax_integration.py: ## @@ -193,6 +194,8 @@ def tune_relax( The maximum number of trials to run max_trials_per_task :

[GitHub] [tvm] sunggg commented on a diff in pull request #14730: [Unity] Add options to MS tuning pass to enable more fine-grained tuning

sunggg commented on code in PR #14730: URL: https://github.com/apache/tvm/pull/14730#discussion_r1179730946 ## python/tvm/meta_schedule/relax_integration.py: ## @@ -193,6 +194,8 @@ def tune_relax( The maximum number of trials to run max_trials_per_task :

[GitHub] [tvm] tvm-bot commented on pull request #14736: [Unity] Process all Relax functions in CompositeFunctionAnnotator

tvm-bot commented on PR #14736: URL: https://github.com/apache/tvm/pull/14736#issuecomment-1526458367 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

[GitHub] [tvm] yelite opened a new pull request, #14736: [Unity] Process all Relax functions in CompositeFunctionAnnotator

yelite opened a new pull request, #14736: URL: https://github.com/apache/tvm/pull/14736 This PR makes CompositeFunctionAnnotator process all Relax functions rather than just the 'main' function. This enables running cutlass partitioning on LLaMA from web-llm. cc @masahi @sunggg

[GitHub] [tvm] echuraev commented on pull request #14707: [Keras][Bugfix] fix a bug about alpha attribute in LeakyReLU which lead to passes conflict

echuraev commented on PR #14707: URL: https://github.com/apache/tvm/pull/14707#issuecomment-1526393628 @tvm-bot rerun -- 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

[GitHub] [tvm] yongwww commented on pull request #14635: [CI] Upgrade CI

yongwww commented on PR #14635: URL: https://github.com/apache/tvm/pull/14635#issuecomment-1526363584 @tvm-bot rerun -- 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] Add system lib build option to relax (#14734)

This is an automated email from the ASF dual-hosted git repository. ruihangl 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 b40391961a [Unity] Add system lib build option to

[GitHub] [tvm] MasterJH5574 merged pull request #14734: [Unity] Add system lib build option to relax

MasterJH5574 merged PR #14734: URL: https://github.com/apache/tvm/pull/14734 -- 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:

[GitHub] [tvm] tvm-bot commented on pull request #14735: [Relay] Handle pad value coming from Tensor instead of scalar

tvm-bot commented on PR #14735: URL: https://github.com/apache/tvm/pull/14735#issuecomment-1526295268 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

[GitHub] [tvm] tvm-bot commented on pull request #14735: [Relay] Handle pad value coming from Tensor instead of scalar

tvm-bot commented on PR #14735: URL: https://github.com/apache/tvm/pull/14735#issuecomment-1526295232 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

[GitHub] [tvm] kparzysz-quic opened a new pull request, #14735: [Relay] Handle pad value coming from Tensor instead of scalar

kparzysz-quic opened a new pull request, #14735: URL: https://github.com/apache/tvm/pull/14735 The PadCompute function would pass empty index to obtain the pad value. This caused a crash when the pad value was given in a tensor with the following message: ``` Check failed:

[tvm] branch driazati/release_merge updated (d3e168b2db -> 0e06514796)

This is an automated email from the ASF dual-hosted git repository. driazati pushed a change to branch driazati/release_merge in repository https://gitbox.apache.org/repos/asf/tvm.git discard d3e168b2db [ci] Disable docker build discard 9ffd008d75 [cherry-pick][ARITH] Enhance

[tvm] branch driazati/release_merge updated (b431770942 -> d3e168b2db)

This is an automated email from the ASF dual-hosted git repository. driazati pushed a change to branch driazati/release_merge in repository https://gitbox.apache.org/repos/asf/tvm.git discard b431770942 [ci] Disable docker build add d3e168b2db [ci] Disable docker build This update added

[tvm] branch unity updated: [Unity] Fix ForceNarrowIndexToI32 so it ignores i16 (#14733)

This is an automated email from the ASF dual-hosted git repository. ruihangl 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 2232207afd [Unity] Fix ForceNarrowIndexToI32 so it

[GitHub] [tvm] MasterJH5574 merged pull request #14733: [Unity] Fix ForceNarrowIndexToI32 so it ignores i16

MasterJH5574 merged PR #14733: URL: https://github.com/apache/tvm/pull/14733 -- 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:

[GitHub] [tvm] zxybazh commented on a diff in pull request #14730: [Unity] Add options to MS tuning pass to enable more fine-grained tuning

zxybazh commented on code in PR #14730: URL: https://github.com/apache/tvm/pull/14730#discussion_r1179512748 ## python/tvm/meta_schedule/relax_integration.py: ## @@ -193,6 +194,8 @@ def tune_relax( The maximum number of trials to run max_trials_per_task :

[GitHub] [tvm] yongwww commented on pull request #14651: [CI] Modify test cases to accommodate the CI upgrades

yongwww commented on PR #14651: URL: https://github.com/apache/tvm/pull/14651#issuecomment-1526091903 It is ready for review now, CI is supposed to be green soon. cc: @tqchen @masahi -- This is an automated message from the Apache Git Service. To respond to the message, please log on to

[GitHub] [tvm] driazati closed pull request #14672: [ci] Add sccache logging

driazati closed pull request #14672: [ci] Add sccache logging URL: https://github.com/apache/tvm/pull/14672 -- 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,

[GitHub] [tvm] driazati closed pull request #14253: [ci] Add -LARGE suffix to instance labels

driazati closed pull request #14253: [ci] Add -LARGE suffix to instance labels URL: https://github.com/apache/tvm/pull/14253 -- 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

[GitHub] [tvm] driazati closed pull request #13903: [ci] Disable GPU unit tests

driazati closed pull request #13903: [ci] Disable GPU unit tests URL: https://github.com/apache/tvm/pull/13903 -- 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,

[GitHub] [tvm] tqchen commented on pull request #14734: [Unity] Add system lib build option to relax

tqchen commented on PR #14734: URL: https://github.com/apache/tvm/pull/14734#issuecomment-1526036486 Good suggestion, that is a great topic for discussion. -- 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 #14734: [Unity] Add system lib build option to relax

tvm-bot commented on PR #14734: URL: https://github.com/apache/tvm/pull/14734#issuecomment-1525985728 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

[GitHub] [tvm] tqchen opened a new pull request, #14734: [Unity] Add system lib build option to relax

tqchen opened a new pull request, #14734: URL: https://github.com/apache/tvm/pull/14734 This PR add system lib build option to relax also adds an util that creates an archived static library in compilation -- This is an automated message from the Apache Git Service. To respond to

[GitHub] [tvm] AndrewZhaoLuo commented on issue #14661: [Bug][ONNX] cann't parse strides for LpPool

AndrewZhaoLuo commented on issue #14661: URL: https://github.com/apache/tvm/issues/14661#issuecomment-1525949224 Ah the answer is as of opset 11, strides is an optional attribute. If it is not present, you just treat it as 1 along every spatial axis. Should be simple to fix, but

[tvm] branch unity updated: [Unity] Allow modifying function signature by AMP to accept fp16 inputs (#14719)

This is an automated email from the ASF dual-hosted git repository. jwfromm 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 9c36056a12 [Unity] Allow modifying function

[GitHub] [tvm] jwfromm merged pull request #14719: [Unity] Allow modifying function signature by AMP to accept fp16 inputs

jwfromm merged PR #14719: URL: https://github.com/apache/tvm/pull/14719 -- 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:

[GitHub] [tvm] tvm-bot commented on pull request #14733: [Unity] Fix ForceNarrowIndexToI32 so it ignores i16

tvm-bot commented on PR #14733: URL: https://github.com/apache/tvm/pull/14733#issuecomment-1525739819 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

[GitHub] [tvm] tqchen opened a new pull request, #14733: [Unity] Fix ForceNarrowIndexToI32 so it ignores i16

tqchen opened a new pull request, #14733: URL: https://github.com/apache/tvm/pull/14733 This PR fixes the forece narrow to i32 pass so it ignores integer with smaller bitwidth -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub

[GitHub] [tvm] ysh329 commented on pull request #14715: [cherry-pick][ARITH] Enhance CanonicalSimplify to Simplify ProdDiv and [ci] disable merge

ysh329 commented on PR #14715: URL: https://github.com/apache/tvm/pull/14715#issuecomment-1525723215 > > > @ysh329 I believe this PR is exposing a latent bug that was being hidden by the merge before, I think we can fix it by cherry picking #14538, that works locally for me at least > >

[GitHub] [tvm] echuraev commented on pull request #14707: [Keras][Bugfix] fix a bug about alpha attribute in LeakyReLU which lead to passes conflict

echuraev commented on PR #14707: URL: https://github.com/apache/tvm/pull/14707#issuecomment-1525717999 @tvm-bot rerun -- 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

[GitHub] [tvm] tvm-bot commented on pull request #14732: [MetaSchedule] Skip custom_rule in case of None

tvm-bot commented on PR #14732: URL: https://github.com/apache/tvm/pull/14732#issuecomment-1525645712 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

[GitHub] [tvm] apeskov opened a new pull request, #14732: [MetaSchedule] Skip custom_rule in case of None

apeskov opened a new pull request, #14732: URL: https://github.com/apache/tvm/pull/14732 Custom rule mechanics allows to have "schedule_rule" attribute equals "None". This value will be ignored by `ApplyCustomRuleNode`. But also this block will be ignore by all other rules. It happens

[GitHub] [tvm] echuraev commented on pull request #14707: [Keras][Bugfix] fix a bug about alpha attribute in LeakyReLU which lead to passes conflict

echuraev commented on PR #14707: URL: https://github.com/apache/tvm/pull/14707#issuecomment-1525590894 @tvm-bot rerun -- 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

[GitHub] [tvm] tvm-bot commented on pull request #14731: [Unity][CUTLASS] Add layer norm support

tvm-bot commented on PR #14731: URL: https://github.com/apache/tvm/pull/14731#issuecomment-1525559501 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

[GitHub] [tvm] blufissure commented on issue #14729: [Bug] different results when conducting the attribute and ifthenelse statement

blufissure commented on issue #14729: URL: https://github.com/apache/tvm/issues/14729#issuecomment-1525509623 > Hi @blufissure. I cannot reproduce this crash in the latest TVM (i.e., 0.13.dev0). I guess this bug has been fixed. In the latest version, TVM will not check the negative

[GitHub] [tvm] jikechao commented on issue #14729: [Bug] different results when conducting the attribute and ifthenelse statement

jikechao commented on issue #14729: URL: https://github.com/apache/tvm/issues/14729#issuecomment-1525496481 Hi @blufissure. I cannot reproduce this crash in the latest TVM (i.e., 0.13.dev0). I guess this bug has been fixed. -- This is an automated message from the Apache Git Service. To

[GitHub] [tvm] tvm-bot commented on pull request #14730: [Unity] Add options to MS tuning pass to enable more fine-grained tuning

tvm-bot commented on PR #14730: URL: https://github.com/apache/tvm/pull/14730#issuecomment-1525105552 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

[GitHub] [tvm] blufissure opened a new issue, #14729: [Bug] different results when conducting the attribute and ifthenelse statement

blufissure opened a new issue, #14729: URL: https://github.com/apache/tvm/issues/14729 Hello, I’m studying TVM and its low-level operators recently. I tried to construct some test files but I was puzzled by the results of some of the files. When I built different statements, different

[GitHub] [tvm] jikechao commented on pull request #14724: [bugfix][keras] fix wrong documentation and add layout checking

jikechao commented on PR #14724: URL: https://github.com/apache/tvm/pull/14724#issuecomment-1524861301 This PR led to many test cases crashing. Thus, I'm confused if this fix is valid. Wish for your comments. -- This is an automated message from the Apache Git Service. To respond to the

[GitHub] [tvm] lightzhan-intellif commented on pull request #14693: [TIR] Enhance TVMScript Buffer Slice Access

lightzhan-intellif commented on PR #14693: URL: https://github.com/apache/tvm/pull/14693#issuecomment-1524799005 > Thanks @lightzhan-intellif for the fix! I think it would be good to also test that the scripted output of `VectorizeLoop` can be parsed as this is something that is currently