[GitHub] [tvm] masahi commented on pull request #15772: [Unity][QNN] Add QNN quantize/dequantize ops

masahi commented on PR #15772: URL: https://github.com/apache/tvm/pull/15772#issuecomment-1733063297 > I actually did not understand how we might be able to work with just "QDQ representation" with legalization. Would we be doing the work of both FQ2I and Canonicalize pass together in one s

[GitHub] [tvm] rutkoor commented on a diff in pull request #15679: [Unity] Support Padding Reversal in Alter-Op pass

rutkoor commented on code in PR #15679: URL: https://github.com/apache/tvm/pull/15679#discussion_r1335492704 ## python/tvm/relax/transform/legalize_ops/manipulate.py: ## @@ -205,3 +213,16 @@ def te_layout_transform(data, name): output_dtype = call_args[0].struct_info.dtype

[GitHub] [tvm] ibsidorenko commented on pull request #15772: [Unity][QNN] Add QNN quantize/dequantize ops

ibsidorenko commented on PR #15772: URL: https://github.com/apache/tvm/pull/15772#issuecomment-1733099520 Thank you all, @masahi and @quic-sanirudh for opinion and discussion! Since there is still no consensus on the need of QNN dialect, I suggest to add only R.quantize/R.dequantize ops

[GitHub] [tvm] HongHongHongL opened a new pull request, #15814: [Unity]Add fast_softmax op and FastMathTransform pass

HongHongHongL opened a new pull request, #15814: URL: https://github.com/apache/tvm/pull/15814 This PR adds fast_softmax op to Relax, using topi.nn.fast_softmax. We also add FastMathTransform pass to transform relax.nn.softmax to relax.nn.fast_softmax. -- This is an automated message f

[GitHub] [tvm] masahi commented on a diff in pull request #15679: [Unity] Support Padding Reversal in Alter-Op pass

masahi commented on code in PR #15679: URL: https://github.com/apache/tvm/pull/15679#discussion_r1335797265 ## python/tvm/relax/transform/legalize_ops/manipulate.py: ## @@ -205,3 +213,16 @@ def te_layout_transform(data, name): output_dtype = call_args[0].struct_info.dtype

[GitHub] [tvm] dmilosevic252 commented on pull request #15804: [Fix][Unity] Fix TVMError when loading ONNX model with CumSum operator

dmilosevic252 commented on PR #15804: URL: https://github.com/apache/tvm/pull/15804#issuecomment-1733617541 @jikechao Thank you for the feedback! I added a regression test. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and u

[GitHub] [tvm] Lunderberg commented on pull request #15756: [Draft][Unittest][Metal] Add minimal metal functionality test to CI

Lunderberg commented on PR #15756: URL: https://github.com/apache/tvm/pull/15756#issuecomment-1733940538 And the results look good, with the intentionally-incorrect codegen resulting in a failure found at compile-time of the `*.metallib`, during execution of the `xcrun` command. The

[GitHub] [tvm] Lunderberg commented on issue #15148: [Tracking Issue] TFLite operator support

Lunderberg commented on issue #15148: URL: https://github.com/apache/tvm/issues/15148#issuecomment-1733981183 The error message is below (full test output [here](https://ci.tlcpack.ai/blue/organizations/jenkins/tvm-arm/detail/PR-15762/4/tests/)). Unfortunately, this doesn't indicate a clea

[GitHub] [tvm] lhutton1 merged pull request #15779: [TVMC] enable dumping imported modules too

lhutton1 merged PR #15779: URL: https://github.com/apache/tvm/pull/15779 -- 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.

[tvm] branch main updated: [TVMC] enable dumping imported modules too (#15779)

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 cde83e1088 [TVMC] enable dumping imported modules too (

[GitHub] [tvm] lhutton1 commented on pull request #15779: [TVMC] enable dumping imported modules too

lhutton1 commented on PR #15779: URL: https://github.com/apache/tvm/pull/15779#issuecomment-1734034456 Thanks @srkreddy1238! -- 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.

[GitHub] [tvm] Lunderberg opened a new pull request, #15815: [Unity] Extend EliminateCommonSubexpr to operate on relax::Expr

Lunderberg opened a new pull request, #15815: URL: https://github.com/apache/tvm/pull/15815 Prior to this commit, the `EliminateCommonSubexpr` utility could only apply to `relax::Function` instances. This commit extends the allowed usage to apply to any `relax::Expr` that contains variable

[GitHub] [tvm] farshidsp commented on a diff in pull request #15764: [Disco] Add AllGather

farshidsp commented on code in PR #15764: URL: https://github.com/apache/tvm/pull/15764#discussion_r1336337450 ## python/tvm/relax/transform/legalize_ops/ccl.py: ## @@ -46,6 +46,30 @@ def _allreduce(_bb: BlockBuilder, call: Call) -> Expr: ) +@register_legalize("relax.cc

[GitHub] [tvm] farshidsp commented on pull request #15764: [Disco] Add AllGather

farshidsp commented on PR #15764: URL: https://github.com/apache/tvm/pull/15764#issuecomment-1734393526 Thanks @Lunderberg for reviewing. Your changes have been addressed. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and u

[GitHub] [tvm] LeshengJin commented on a diff in pull request #15764: [Disco] Add AllGather

LeshengJin commented on code in PR #15764: URL: https://github.com/apache/tvm/pull/15764#discussion_r1336456333 ## python/tvm/relax/op/ccl/ccl.py: ## @@ -44,6 +47,27 @@ def allreduce(x, op_type: str = "sum"): # pylint: disable=invalid-name return _ffi_api.allreduce(x, op_

[GitHub] [tvm] jinhongyii commented on pull request #15781: [Unity][Dlight] Fix inline consumer in matmul tensorize rule

jinhongyii commented on PR #15781: URL: https://github.com/apache/tvm/pull/15781#issuecomment-1734650372 @Hzfengsy Please review again -- 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

[GitHub] [tvm] Hzfengsy merged pull request #15781: [Unity][Dlight] Fix inline consumer in matmul tensorize rule

Hzfengsy merged PR #15781: URL: https://github.com/apache/tvm/pull/15781 -- 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.

[tvm] branch unity updated: [Unity][Dlight] Fix inline consumer in matmul tensorize rule (#15781)

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 e809d64658 [Unity][Dlight] Fix inline consumer in mat

[GitHub] [tvm] yongwww commented on issue #15101: [Unity] [Tracking Issue] Heterogeneous execution for Relax

yongwww commented on issue #15101: URL: https://github.com/apache/tvm/issues/15101#issuecomment-1734693653 @qelk123 Thank you for sharing your progress and challenges on the end-to-end multi-device support. It's promising to hear that it's operational for simpler graphs. The VDevice propaga

[GitHub] [tvm] Archermmt commented on pull request #15813: [Unity][MSC][pre M1.2] Reconstruct codegen

Archermmt commented on PR #15813: URL: https://github.com/apache/tvm/pull/15813#issuecomment-1734757371 @Lunderberg could you please check why the error occurs in CI test: tests/python/relax/test_analysis_suggest_layout_transforms.py::test_non_affine_access [01:22:12] /workspace/src/

[GitHub] [tvm] vinx13 opened a new pull request, #15816: Revert "[TensorIR][Visitor] Visit buffer members in `match_buffer`'s in block visitor functions (#15153)

vinx13 opened a new pull request, #15816: URL: https://github.com/apache/tvm/pull/15816 This reverts commit 34637d7ee38f2636b1948548a39c15838d7a8db6. It is better not visiting fields of buffer when visiting `MatchBufferRegion` since it's more common that the buffer is not mutated in `

[tvm] branch nightly updated (ae89c1e56d -> cde83e1088)

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 ae89c1e56d [OpenCL] Don't initialize OpenCL runtime on host (#15745) add cde83e1088 [TVMC] enable dumping impor

[GitHub] [tvm] LeshengJin opened a new pull request, #15817: [TE] Support using tir::Var as CreatePrimFunc args

LeshengJin opened a new pull request, #15817: URL: https://github.com/apache/tvm/pull/15817 - Allow `CreatePrimFunc` args to be a mixture of `te::Tensor` and `tir::Var`. - Integrate `CreatePrimFunc` and `CreateRelaxPrimFunc` into one function. -- This is an automated message from the Ap

[GitHub] [tvm] junrushao commented on pull request #15817: [TE] Support using tir::Var as CreatePrimFunc args

junrushao commented on PR #15817: URL: https://github.com/apache/tvm/pull/15817#issuecomment-1734926227 I'm a big fan of this change! CC: @yongwww @masahi @Hzfengsy -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the