profile
viewpoint

retonym/ai-matrix 0

To make it easy to benchmark AI accelerators

retonym/lp-opt-tool 0

Intel Low Precision Optimization Tool, targeting to provide a unified low precision inference interface cross different deep learning frameworks, and support auto-tune with specified accuracy criterion to find out best quantized model.

retonym/oneDNN 0

oneAPI Deep Neural Network Library (oneDNN)

pull request commenttensorflow/tensorflow

[INTEL_MKL] Fix TF_DISABLE_MKL not work in remapper.cc

@ezhulenev That's good. Thanks for your reply.

retonym

comment created time in 8 days

PR opened tensorflow/tensorflow

[INTEL_MKL] Fix TF_DISABLE_MKL not work in remapper.cc

This PR fix bug that TF_DISABLE_MKL=1 not work in remapper.cc.

Users can use such env variable to disable MKL-implement op, even when they are using tensorflow with MKL support.

This PR only affects tensorflow built with --config=mkl.

+8 -0

0 comment

1 changed file

pr created time in 16 days

create barnchIntel-tensorflow/tensorflow

branch : yunfeimao/fix_tf_disable_mkl

created branch time in 16 days

startedtensorflow/mlir-hlo

started time in a month

fork retonym/lp-opt-tool

Intel Low Precision Optimization Tool, targeting to provide a unified low precision inference interface cross different deep learning frameworks, and support auto-tune with specified accuracy criterion to find out best quantized model.

fork in 2 months

pull request commenttensorflow/tensorflow

[INTEL MKL] Add MKL Conv + Bias + LeakyRelu Fusion

The conflicts are fixed.

retonym

comment created time in 2 months

push eventIntel-tensorflow/tensorflow

Basit Ayantunde

commit sha 196054b54c9a5f9dd7b5e4044de6c6b1a1dc9e79

Merge pull request #5 from tensorflow/master pulled commits from Tensorflow:master

view details

Basit Ayantunde

commit sha b15c71d3fe8963ca7a9615646efc668302feafd3

Merge pull request #7 from tensorflow/master Merge PRs from Tensorflow:master

view details

Basit Ayantunde

commit sha 285319f41b5a0169ca7e82f188186bb21e097cf6

Merge pull request #9 from tensorflow/master pull commits from tensorflow/tensorflow:master

view details

Dmitry Zakharov

commit sha a729898a0904611f970c990b81f8d36b7f79fae9

Fix of problem with copying of just downloaded TCF

view details

Dmitry Zakharov

commit sha a3db6fe20edfe76e6cc4425261608eea4489fc81

- default heap size was increased to cover all tests - minor fix in examples readmes

view details

Dmitry Zakharov

commit sha a1045ae332fb5b16d0b15aa28b842cc67dc568c2

ARC EMSDP: Return uboot.env (was strangely deleted during merge)

view details

Dmitry Zakharov

commit sha dd37e351117357bdd9688183285d574a710e2400

Fix height slicing in fit branch for odd input

view details

Dmitry Zakharov

commit sha 0d5787ca06a06032a80d8ae99caeca14e37a054c

replace MLI switchers with specializations in person_detection_generation

view details

Dmitry Zakharov

commit sha 6971de110a15f1a8ed6b328c227f7c9924b87c2f

ARC EMSDP - MicroSpeech int8 integration - Patch sources to use specialization during project generation

view details

Dmitry Zakharov

commit sha 0df881cdf814d72716250375dd082636e18eed38

ARC EMSDP: update common LCF to be not limit codesize. Update MLI link to more recent one

view details

Dmitry Zakharov

commit sha 5a3f58ef75da826fc3bc78af810510e8b7b5f27a

ARC MLI: scales to be present in 32 bits now

view details

Dmitry Zakharov

commit sha 5b8cb124e3957bcd139912c00116c250a155a130

Update arc_mli_iss_fix branch with the latest state from upstream master

view details

Thibaut Goetghebuer-Planchon

commit sha 860898de6f60d22b043be96fc37d565180dce35b

Add int16x8 support for RESIZE_NEAREST_NEIGHBOR operator

view details

Thibaut Goetghebuer-Planchon

commit sha f6c0221749543fd16a59daa0445574525f3a4f53

Add int16 as possible output type to the error message

view details

Dmitry Zakharov

commit sha 8f91b0b504fc578159a1796b37ac208294c7e785

Use embarc MLI RC3 for arc targets

view details

Dmitry Zakharov

commit sha 0577515fc3f60049aafc0cdf3b63f67f6d24044c

Update branch with the latest state from upstream master

view details

daria

commit sha 1d7577e984f0d5ec0c8262600136daea3514581a

Update according to tflm changes

view details

daria

commit sha 620b38ab988e03d77ec764e8a8a7a6c6594236df

updates for XY functions to align with tflm changes

view details

Elena Zhelezina

commit sha 9ca8d0cdc15267e68c3d18e81bb4f6edbf8c39aa

SLICE, TRANSPOSE operators are fixed for the given network. Change-Id: Ia33e6d0bb55273a52e8b7ac6b9226b7175cb8010

view details

daria

commit sha cb69b88d5a2206b293f5fd226a1ad3918c24e55f

added uint32 to int cast, mli_is_applicable check

view details

push time in 2 months

Pull request review commenttensorflow/tensorflow

[INTEL MKL] Add MKL Conv + Bias + LeakyRelu Fusion

 TEST_F(RemapperTest, FuseConv2DWithSqueezeAndBias) {   ASSERT_EQ(tensors.size(), 1);   test::ExpectTensorNear<float>(tensors[0], tensors_expected[0], 1e-6); }-#endif+#endif  // !INTEL_MKL

Yes, these fusions are not supported by MKL currently.

retonym

comment created time in 2 months

PullRequestReviewEvent

Pull request review commenttensorflow/tensorflow

[INTEL MKL] Add MKL Conv + Bias + LeakyRelu Fusion

 bool IsDeviceCompatible(const RemapperContext& ctx, Pattern& matched) { }  bool IsSupportedActivation(const NodeDef& node) {-// Disable LeakyRelu temporarily before MKL PR is merged.-#ifndef INTEL_MKL   return IsRelu(node) || IsRelu6(node) || IsElu(node) || IsLeakyRelu(node);

Yeah, I will fix the conflicts after https://github.com/tensorflow/tensorflow/pull/42173 is merged

retonym

comment created time in 2 months

PullRequestReviewEvent

Pull request review commenttensorflow/tensorflow

[INTEL MKL] Add MKL Conv + Bias + LeakyRelu Fusion

 class MklRemapperTest : public GrapplerTest {     if (add_op == kAddNOp) {       auto addn = ops::AddN(s.WithOpName(add_op),                             std::initializer_list<Input>{input_addn, bias_add});-      if (has_relu) {-        auto relu = ops::Relu(s.WithOpName("relu"), addn);-        ops::Identity(s.WithOpName("fetch"), relu);+      auto activate = s.WithOpName("activation");+      auto fetch = s.WithOpName("fetch");+      if (activation == "Relu") {+        ops::Identity(fetch, ops::Relu(activate, addn));+      } else if (activation == "Relu6") {+        ops::Identity(fetch, ops::Relu6(activate, addn));+      } else if (activation == "Elu") {+        ops::Identity(fetch, ops::Elu(activate, addn));+      } else if (activation == "LeakyRelu") {+        ops::Identity(fetch, ops::internal::LeakyRelu(activate, addn));       } else {-        ops::Identity(s.WithOpName("fetch"), addn);+        DCHECK(activation == "None");+        ops::Identity(fetch, addn);       }     } else if (add_op == kAddV2Op) {       auto add = ops::AddV2(s.WithOpName(add_op), input_addn, bias_add);-      if (has_relu) {-        auto relu = ops::Relu(s.WithOpName("relu"), add);-        ops::Identity(s.WithOpName("fetch"), relu);+      auto activate = s.WithOpName("activation");+      auto fetch = s.WithOpName("fetch");+      if (activation == "Relu") {+        ops::Identity(fetch, ops::Relu(activate, add));+      } else if (activation == "Relu6") {+        ops::Identity(fetch, ops::Relu6(activate, add));+      } else if (activation == "Elu") {+        ops::Identity(fetch, ops::Elu(activate, add));+      } else if (activation == "LeakyRelu") {+        ops::Identity(fetch, ops::internal::LeakyRelu(activate, add));       } else {-        ops::Identity(s.WithOpName("fetch"), add);+        DCHECK(activation == "None");+        ops::Identity(fetch, add);       }

Thanks for the advice. I rewrite this part of codes.

retonym

comment created time in 2 months

PullRequestReviewEvent

Pull request review commenttensorflow/tensorflow

[INTEL MKL] Add MKL Conv + Bias + LeakyRelu Fusion

 REGISTER_TEST_ALL_TYPES(NodeRewrite_FusedConv2D_Positive1);               "i:1, i:1} } }"                                                 \               " attr { key: 'fused_ops'        value { list: {s: 'Relu'} } }" \               " attr { key: 'epsilon'          value { f: 0.001 }}"           \+              " attr { key: 'leakyrelu_alpha'  value { f: 0.2 }}"                    \

The \ are aligned.

retonym

comment created time in 2 months

PullRequestReviewEvent

push eventIntel-tensorflow/tensorflow

yunfeima

commit sha 8a0a2be8241920e14f7403a5410b467bb747bf89

Polish code in mkl remapper test

view details

push time in 2 months

pull request commenttensorflow/tensorflow

[INTEL MKL] Matmul + Tanh fusion

Thanks for the review.

retonym

comment created time in 2 months

PR opened tensorflow/tensorflow

[INTEL MKL] Add MKL Conv + Bias + LeakyRelu Fusion

Add Conv + Bias + LeakyRelu fusion MKL implementation

+218 -63

0 comment

10 changed files

pr created time in 2 months

create barnchIntel-tensorflow/tensorflow

branch : yunfeimao/leaky_relu_fusion_mkl

created branch time in 2 months

push eventIntel-tensorflow/tensorflow

Patrik Laurell

commit sha d076366a38ebce54c573cabbf6b4772a0ff8f37d

Support quantized int8 and uint8 in TFLu mean operator

view details

Patrik Laurell

commit sha f6b587511a196d7c24c241b7da3932ccc2349d02

Addressed review comments

view details

Patrik Laurell

commit sha 20733ea481dfb8f914a4e75ccaaaa489b92f35fd

Addressed hajuho's comment

view details

Patrik Laurell

commit sha 96804c6ffb7d1ff11d88ec2f4c6529f96f1006e1

Updated Makefile

view details

Patrik Laurell

commit sha 29669a79305cfa3099606e0330eee2e70ef27aae

Re-added changes to internal/reference/reduce.h

view details

Patrik Laurell

commit sha 48dff25b99fef01e33537b705c73b66894c4f92f

Address review comments

view details

Daniel Nguyen

commit sha 998ba46b21127726f1d9b107498b3f0000f98571

summary_op_bm

view details

Daniel Nguyen

commit sha 718a587fce8ad91cf4ed5ad88e854cd5868ada44

small syntax

view details

Elena Zhelezina

commit sha 293c7f5c2b40b506d4f0a02eede0dceff57e26a3

Added op tests for leaky_relu. Change-Id: I5b418e2d7e1978359e3876f0f42ffd0226930991

view details

Lukas Geiger

commit sha c7eba425b96cf5f71b3f6511c171cebc63a36ab9

Support TF logs in TerminateOnNaN callback

view details

Elena Zhelezina

commit sha bf02b27dfc45d4387ecb3191f41e7eb2801bd34c

Addressed reviewer's comments. Change-Id: Ic1761eea5f7164906d5434137a077b304772e26f

view details

Vignesh Kothapalli

commit sha dc4ce6f3bf89260b880f80056965237c5d23754c

extended test cases and added eager mode tests

view details

Vignesh Kothapalli

commit sha 5e6ea264de2e7bcfc180768e3bdcae95d9afe7e0

restructured test combinations

view details

Vignesh Kothapalli

commit sha 75df7906f10c86be191425fd4e93a5b109360464

fixed pylint issues

view details

danielyou0230

commit sha ae9cc4a84d24fdb97f07235c8dfc248a0d593b56

TFLite: reduced redundant calculation in integer_ops conv.h

view details

Kamil Rakoczy

commit sha 6e339264157a48cae09754fda4b7b41bed3586d6

Move person model and ruy to tools/make/Makefile. This commit fixes a `make` complaining about overriding recipe target for person_model_grayscale and ruy third party downloads. Signed-off-by: Kamil Rakoczy <krakoczy@antmicro.com>

view details

Kamil Rakoczy

commit sha ebec29b480fca6c8c63a32d5110af38f4e8205d9

Limit max depth of find to 2. This commit fixes `make` overriding recipe warnings for target specific Makefiles. Without limiting max depth of directories to search, `find` is also finding target specific Makefiles that are included directly by the main example’s Makefile. Signed-off-by: Kamil Rakoczy <krakoczy@antmicro.com>

view details

SiCong Li

commit sha 8fd0cd30b70181f2246b687c0db3e88a85a4dc0a

Add symmetric int16 support to tflu softmax reference kernel

view details

Daniel Nguyen

commit sha 75c1f5d4728a7630fac3a2c19ac45df4cbd764f0

clean up only

view details

Vignesh Kothapalli

commit sha 17b5eb57b577dc4ebf04494f43f25e5c131bd0e4

removed combinations for ragged based tests

view details

push time in 2 months

push eventIntel-tensorflow/tensorflow

retonym

commit sha cd21cdef42545fc6b733bfffce2eeaf81b09d8a5

Change the BUILD deps order for beauty

view details

push time in 2 months

push eventIntel-tensorflow/tensorflow

Katherine Tian

commit sha 61d63b37643568c47f0fc293924d89837cd60c0a

binary add and zeros like update

view details

A. Unique TensorFlower

commit sha 6861b3f69d8db26045316d8b4db10d4d5ce20c65

Add device_type to TfStatsDatabase. PiperOrigin-RevId: 326113527 Change-Id: I794cd6fa3724c8189971990abf65e3180c0b82dc

view details

Jared Duke

commit sha bd62ee5bfa8e855ecfda67e3a28095ecb31fb531

Fix several narrowing issues on 32-bit devices PiperOrigin-RevId: 326117351 Change-Id: Iaa094b1e7664ec0bc9f0c91e6cd972fcae19b5b9

view details

Yixing Fu

commit sha c09eabcd28667c81e86a2cbafb5a68c03186e038

changes according to comment

view details

Allen Lavoie

commit sha c604d85364963b8d80eb009f01ed4e70177dc785

Parallel device: hide device.name, move scoping to the ParallelDevice object Naming is an implementation detail that is expected to change, and having enter/exit on the object itself is a better API. PiperOrigin-RevId: 326118080 Change-Id: Ica7cf901960631c9896a4efcad435bf3aaddde67

view details

Yunxing Dai

commit sha 29784fade22771cbe97f6d2bf236fd11a9e8116f

Support dynamic leading dimension for tensorlist. PiperOrigin-RevId: 326118256 Change-Id: Icea13ee03e23aaac56f17dc5f58bd9182a4ba02a

view details

TensorFlower Gardener

commit sha 6974852f964d9d12e3a96edf017faa892076ced2

Merge pull request #42195 from WindQAQ:canonicalize-shape-n-with-partial-static-input-shape PiperOrigin-RevId: 326120497 Change-Id: I0d70c8557a5d0e73bf75e08aeed149e267fc082c

view details

Daniel Nguyen

commit sha a7e4df924652a151677b3e0b95609db27cb04631

switched from unsigned char to TF_Bool. Added input checking for TF_AllocatorAttributes

view details

Akshay Modi

commit sha b297140e1a1f0e916d0f60e743e7763745608b45

Allow NdarraySpec to be written in saved model. PiperOrigin-RevId: 326121293 Change-Id: I7a4351a9ab3e0381ff5616f67d0e61880f3bb649

view details

A. Unique TensorFlower

commit sha 672041b9524a30f6c567b98b2ebf12f032b8a7fd

[XLA:SPMD] Reshard tile to partial replicate. PiperOrigin-RevId: 326121818 Change-Id: I5e26a7c93d5bb54841df0b7da31708e754f8e4a2

view details

Jose Baiocchi

commit sha 518d3339772433c790bff0bb70fb8b57d0fdc082

Add XStats to XEventMetadata PiperOrigin-RevId: 326121854 Change-Id: I0c00f7ea095dab412afeaa9a1213c2ee4c44235f

view details

Yanhua Sun

commit sha 24d9188b68a82f602205e5116b8a54725e023377

avoid one extra function call to optimize functions. Also, this is more consistent with everywhere else on how the checks are done PiperOrigin-RevId: 326125083 Change-Id: I5636d8160a83f5fc810526c9f5993080bbd7f697

view details

TensorFlower Gardener

commit sha 7d744e47968535709ab7054953199be93ae884a5

Merge pull request #42225 from freckie:patch-1 PiperOrigin-RevId: 326128275 Change-Id: I19e08c893d73a606507ae730849882b142effae7

view details

Rahul Joshi

commit sha db04ee6238f95c6ac23664eada064b15636b69d8

[MLIR] Extend ResourceDeviceInference to handle calls. - Handle calls by propagating device attribute to callee arguments, similar to functional If and While. - Also extend the unit test to exercise calls and IfRegion (which works without any code changes in this pass). PiperOrigin-RevId: 326128697 Change-Id: I65188acc5ee453fdb3d4151279869f0669ffe71d

view details

Tim Shen

commit sha 85cfe48f62bb433bd996297ee41b052cd644d006

[PJRT] Disable MSan for two tests. PiperOrigin-RevId: 326129785 Change-Id: Ib3912bc4fa6847f4ac4d2ccd40ec9f925ca96e88

view details

Yujing Zhang

commit sha 497d94e9acc1c062791a62e9d8d339a16f223c47

Separate function tests from c_api_remote_test PiperOrigin-RevId: 326133427 Change-Id: Ia611cd513ef84d393f0f32ba9410c037f339fe19

view details

Rachel Lim

commit sha cee5f17479966a40125a842fba3bfdd050faf079

[tf.data] Split RangeBenchmark into two different benchmark methods PiperOrigin-RevId: 326135090 Change-Id: I5e7735468f953e5155fcbbb2c61a6084332096c0

view details

Jonathan Chu

commit sha 3710d824e457b656f66ae56b3df082899de16cf8

Loop through additional arguments once, rather than looping through kwargs

view details

Sanjoy Das

commit sha 68984541641975e9e80a80eb739c25e3fab9c92f

Use the correct path for common_win_cuda11.bat PiperOrigin-RevId: 326135121 Change-Id: I51aeebde1bbbfb0c5be0575d49f4e3d940988fcf

view details

A. Unique TensorFlower

commit sha ebfe3af239f8ee3bd45148ad2da9ff9e9db4edfa

Raise an error in tf.linalg.expm if the L1 norm of the matrix is infinite or NaN instead of looping infinitely. PiperOrigin-RevId: 326135199 Change-Id: I740d0a5502f14aa3b45fff9496b3f7c307304909

view details

push time in 2 months

push eventIntel-tensorflow/tensorflow

retonym

commit sha be4cde01a4b00483dfeaf53becb0e2e6e70c0168

Fix mkl_fused_ops_test fail caused by dependency

view details

push time in 2 months

pull request commenttensorflow/tensorflow

[INTEL MKL] Enable Conv + (Bias+BN) + LeakyRelu Fusion with Eigen implementation in CPU (Resubmit)

Hi Rajeshwar and Eugene, the UT failures are not related to this PR. Please help tp approve this PR.

retonym

comment created time in 2 months

startedgrailbio/bazel-compilation-database

started time in 2 months

starteddowndemo/Cpp-Templates-2ed

started time in 2 months

pull request commenttensorflow/tensorflow

Enable Conv + Biasadd + LeakyRelu Fusion with Eigen implementation in CPU

I resubmit the PR https://github.com/tensorflow/tensorflow/pull/42551 to fix the errors. Thanks.

retonym

comment created time in 2 months

PR opened tensorflow/tensorflow

Enable Conv + (Biaa+BN) + LeakyRelu Fusion with Eigen implementation in CPU (Resubmit)

This PR fixes the error in https://github.com/tensorflow/tensorflow/pull/42489

This error is caused by not sending activation parameter with conv + bn + leakyrelu pattern in remapper.cc

https://github.com/tensorflow/tensorflow/commit/4d022d6e2cbc924fbff1ffa8c1d98383c4ecaeae is the squeeze commit for all commits in https://github.com/tensorflow/tensorflow/pull/42489

https://github.com/tensorflow/tensorflow/commit/6aabcb1923fce33dd3cd17ab6875f5d2ec9d49b1 updates the unit test to capture the error found in Google internal test.

https://github.com/tensorflow/tensorflow/commit/62db9a72f1d5d980ee6c8f8f9c69510b8220c5d5 fixes the error.

+254 -29

0 comment

11 changed files

pr created time in 2 months

create barnchIntel-tensorflow/tensorflow

branch : yunfeimao/leaky_relu_fusion_eigen2

created branch time in 2 months

push eventIntel-tensorflow/tensorflow

yunfeima

commit sha f8800e423c589973a7704881979788476a8e383a

Update remapper tests for copy leakyrelu alpha

view details

yunfeima

commit sha 086954fe52db0cb8a2d631622f685fb3a60d6d5f

Fix missing activation in conv bn leakyrelu copyattr

view details

push time in 2 months

startedWalton1128/CPP-Templates-2nd--

started time in 2 months

push eventIntel-tensorflow/tensorflow

retonym

commit sha e2f4cdae2a1f7268b322e7f8155b43b49a25d112

Disable LeakyRelu before MKL PR is merged

view details

push time in 2 months

PR opened tensorflow/tensorflow

Enable Conv + Biasadd + LeakyRelu Fusion with Eigen implementation in CPU

This PR enables Conv + Biasadd + LeakyRelu fusion in CPU. The fusion pattern is very common in YOLOv3 model.

+229 -29

0 comment

11 changed files

pr created time in 2 months

push eventIntel-tensorflow/tensorflow

retonym

commit sha 5a4d636dd1620a679f37e3319c17c5a95717054d

add some comments

view details

push time in 2 months

create barnchIntel-tensorflow/tensorflow

branch : yunfeimao/leaky_relu_fusion_eigen

created branch time in 2 months

startedgoogle/cpu_features

started time in 3 months

PR opened tensorflow/tensorflow

[INTEL MKL] Matmul + Tanh fusion

This PR enables Matmul + Tanh fusion in MKL

+48 -3

0 comment

7 changed files

pr created time in 3 months

create barnchIntel-tensorflow/tensorflow

branch : yunfeimao/matmul_tanh_fusion

created branch time in 3 months

more