Paddle

Commit Graph

Author	SHA1	Message	Date
lidanqing	04b6c29ee0	Improve mobilenetv2 INT8 performance by using INT8 relu as post-op (#17570 ) * add INT8 conv+relu6 fuse and enbale mobilentv2 INT8 test test=develop * change fasle and 0.0 to fuse_brelu and brelu_threshold test=develop change the "fuse_relu\|\|fuse_brelu" to "unsigned_output" test=develop * Use relu instead of brelu as INT8 post-op because INT8 brelu is not enabled in mkldnn v0.18 test=develop * continuous-integration fix test=develop	6 years ago
Michał Gallus	0c39b97b4e	[MKL-DNN] Add Fully Connected Op for inference only(#15226 ) * fuse mul and elementwise add to fc * Reimplement the FC forward operator * Fix FC MKLDNN integration by transposing weights * Add FC MKLDNN Pass test=develop * FC MKLDNN Pass: change memcpy to std::copy * Fix MKLDNN FC handling of mismatch input and weights dims * Lower tolerance for MKL-DNN in resnet50 test test=develop * Adjust FC to support MKLDNN Op placement test=develop * Adjust Placement Op to set use_mkldnn attribute for graph test=develop * MKLDNN FC: fix weights format so that gemm version is called test=develop * FC MKLDNN: Remove tolerance decrease from tester_helper * FC MKL-DNN: Refactor the code, change input reorder to weight reorder * MKL-DNN FC: Introduce operator caching test=develop * FC MKL-DNN: Fix the tensor type in ExpectedKernelType test=develop * FC MKL-DNN: fix style changes test=develop * FC MKL-DNN: fallback to native on non-supported dim sizes test=develop * FC MKLDNN: fix CMake paths test=develop * FC MKLDNN: Refine placement pass graph mkldnn attribute test=develop * Fix Transpiler error for fuse_conv_eltwise test=develop * Fix missing STL includes in files test=develop * FC MKL-DNN: Enable new output size computation Also, refine pass to comply with newest interface. test=develop * FC MKL-DNN: enable only when fc_mkldnn_pass is enabled * FC MKL-DNN: Allow Weights to use oi or io format * FC MKL-DNN: Adjust UT to work with correct dims test=develop * Enable MKL DEBUG for resnet50 analyzer test=develop * FC MKL-DNN: Improve Hashing function test=develop * FC MKL-DNN: Fix shape for fc weights in transpiler * FC MKL-DNN: Update input pointer in re-used fc primitive * Add log for not handling fc fuse for unsupported dims test=develop * FC MKL-DNN: Move transpose from pass to Op Kernel test=develop * FC MKL-DNN: Disable transpose in unit test test=develop * FC MKL-DNN: Remove fc_mkldnn_pass from default list * Correct Flag for fake data analyzer tests test=develop * FC MKL-DNN: Add comment about fc mkldnn pass disablement test=develop * FC MKL-DNN: Disable fc in int8 tests test=develop	6 years ago
Sylwester Fraczek	bccb0ba49a	fix quantize_squash_pass segfault when no tensor linked to Bias (#17292 ) * fix quantize_squash_pass segfault when there is no tensor linked do Bias input test=develop * add googlenet test test=develop * fix concat CreateKey not using input format test=develop	6 years ago
guomingz	2281ebf0f3	Enable the convolution/relu6(bounded_relu) fusion for FP32 on Intel platform. (#17130 ) * Relu6 is the bottleneck op for Mobilenet-v2. As the mkldnn supports the conv/relu6 fusion, we implement it fusion via cpass way. Due to the int8 enabling for this fusion will be supported in MKLDNN v0.20, so this PR is focused on the fp32 optimization. Below table shows the benchmark(FPS) which measured on skx-8180(28 cores) Batch size \| with fusion \| without fusion -- \| -- \| -- 1 \| 214.7 \| 53.4 50 \| 1219.727 \| 137.280 test=develop * Fix the format issue test=develop * Add the missing nolint comments. test=develop * Fix the typos. test=develop * Register the conv_brelu_mkldnn_fuse_pass for the MKLDNN engine. test=develop * Adjust the indentation. test=develop * Add the test_conv_brelu_mkldnn_fuse_pass case. test=develop * Slightly update the code per Baidu comments. Let the parameter definition embedded into the code. That's will make the code easy to understand. test=develop	6 years ago
Jacek Czaja	87a44b1149	[MKL-DNN] Added reusing of primitive descriptors (fp32) (#16667 ) * - Reuse of conv PD - conv transpose pd reused - Added PD reusing of softmax and Batch Norm - Refactoring and removal of not needed routines of mkl-dnn ops test=develop - Fix to reusing conv test=develop - Lint fixes test=develop - Further lint fixes test=develop - Lint fixes test=develop - lint fixes test=develop - Lint workaround test=develop * - Fix after review on including boost as third party header test=develop * - Fix after review. Name change to something more descriptive test=develop	6 years ago
Jacek Czaja	2632327429	[MKL-DNN] Tensor modifications revert (#16462 ) * Revert "[MKL-DNN] Fix to crash of Transformer when mkldnn is to be used (#16233)" This reverts commit `13816dd4ac`. Apart from enabling transformer for MKL-DNN * Revert "- MKL-DNN pooling updated to set_prim_desc" This reverts commit `c63f6b2039`. Conflicts: paddle/fluid/operators/mkldnn/concat_mkldnn_op.cc * Revert "[MKL-DNN] MKL-DNN specific Tensor modification (#15429)" test=develop This reverts commit `dec9cf53c8`. * - concat compilation fix - lint test=develop - Lint fixes test=develop - Lint fixes test=develop - Fix Transpose MKLDNN op test=develop	6 years ago
xiaolil1	e235882c18	Enable MKL-DNN INT8 Concat Kernel. (#16156 ) * Enable INT8 Concat Kernel to improve the performance of MobileNet-SSD. test=develop * Optimize UT format. test=develop * Fix UT file address issue. test=develop * Refine the license year. test=develop * Optimize code for new API. test=develop * Restructure INT8 Concat kernel. test=develop	6 years ago
xiaolil1	e818fa1004	Enable INT8 transpose kernel for MobileNet-SSD improvement. (#16159 ) * Enable INT8 transpose kernel for MobileNet-SSD improvement. test=develop * Refine the license year. test=develop * Delete redundant code. test=develop * Add axis check. test=develop	6 years ago
luotao1	d9f0e7252a	refine with comments test=develop	6 years ago
luotao1	6fa52f83ba	Merge branch 'develop' into fc_infershape	6 years ago
Wojciech Uss	2579ade45f	Add cpu_quantize_pass for C-API quantization (#16127 ) * Add cpu_quantize_pass for C-API quantization test=develop * add cpu_quantize_pass test * fix lint: add include memory unorderd_map and unordered_set test=develop * fuse_relu 1 test=develop * tuned 2 without squash * fixes test=develop * remove unused vars test=develop * refactored test=develop * fix lint c-style cast -> C++ style cast test=develop * remove QuantMax and c style casts test=develop * last usage of QuantMax removed test=develop * Fix Analysis Predictor UT Check if memory_optimize_pass has already been added to the analysis config before adding a new one, so that it is not added multiple times. test=develop * change map to unordered_map fix the forgotten part of cpu_quantize_pass_tester.cc test=develop * removed quantized attribute * fixed cpu_quantize_pass_tester and op attr comments test=develop * removed redundant line test=debug * removed gmock test=develop * fix after merge	6 years ago
luotao1	721c2c00ef	refine fc_infershape test=develop	6 years ago
xiaolil1	a177d48217	Add Requantize OP (#15318 ) * Enable INT8 ReQuantize OP test=develop * Clean code test=develop * Add comments test=develop * Revert "Clean code" test=develop This reverts commit a7a49b8aa214f9730cb84e11ea96da564fe4b4d9. * Modify requantize op test test=develop * fix requantize UT by moving public function to public test file. test=develop * Fix test fail due to file address change. test=develop * Change file address for requantize op. test=develop	6 years ago
xiaolil1	1abddd8d97	Optimize Quantize Op with primitive reuse. (#15929 ) test=develop	6 years ago
xiaoli.liu@intel.com	c4187dbd7c	Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into dequantize-reuse	6 years ago
xiaolil1	6724be2b0d	INT8 Pool kernel Key Creation Optimization. (#15883 ) * Optimize key creation of INT8 pool kernel to improve the peformance of ResNet-50 and MobileNet, especially for latency. test=develop * Optimize key creation of pool fp32 grad. test=develop	6 years ago
Tao Luo	ba90e05281	Merge pull request #15917 from jczaja/prv-tensor-mkldnn-ops [MKL-DNN] Adjusting ops to Tensor modifications	6 years ago
Tao Luo	effec86600	Merge pull request #15913 from liangan1/func_coverage Enable function coverage for U8/S8 ConvMKLDNNOpKernel	6 years ago
Jacek Czaja	c63f6b2039	- MKL-DNN pooling updated to set_prim_desc - MKLDNN ops revisited - disabled softmax modifications - disabled elementwise_add - reverted LRN modifications - reverted SUM primitive - Partial reviing of softmax - Enable softmax - Softmax changes - LRN is back - LRN partially disabled - LRN is back - LRN fix - compilation fixes - Sum fixed(hopefully) - Enabling (partially) elementwise_add - Fixes to elemenwise_add - Lint fixes quantize fix - compilation fix test=develop Disabling pooling - Disabled quantize op test=develop	6 years ago
xiaoli.liu@intel.com	70759d181b	Optimize INT8 DeQuantize Op with primitive reuse. test=develop	6 years ago
Michal Gallus	6ebe9877bb	Improve code reuse at MKL-DNN sum test=develop	6 years ago
liangan1	4acc522087	Enable function coverage for U8/S8 ConvMKLDNNOpKernel test=develop	6 years ago
Jacek Czaja	dec9cf53c8	[MKL-DNN] MKL-DNN specific Tensor modification (#15429 ) * - Implemented draft of primitive desc keeping in Tensor test=develop - TransposeMKLDNNHandler::AcquireSrcMemory was reimplemented - Added nchw and nc formats setting for sake of compatiblity Fixed unit tests - Worakaround to problem with 5D data in conv - Added 3D and 1D MKL-DNN formats for name handles for tensor test=develop - Fix to UTs test=develop - Conv fp32 op was updated Cosmetic fixes test=develop - tensor mkldnn cosmetics test=develop - Moved most of mkl-dnn specific code from Tensor to mkl-dnn utils * - Lint fixes test=develop * - setting prim dec in Tensor , sets also layout to kMKLDNN test=develop * - Moved creation of prim desc totally out of Tensor test=develop * - Cosmetic fixes adter review test=develop	6 years ago
Sylwester Fraczek	74672d1aff	Change (smart_ptr.get()) -> smart_ptr reason: dereferencing smart pointer is the same as the underlying pointer test=develop	6 years ago
Krzysztof Binias	1578c60bdd	Add new ut and remove unnecessary code test=develop	6 years ago
Michal Gallus	7a8eff36a6	Fix old FC backward weights descriptor creation test=develop	6 years ago
Krzysztof Binias	69b7c595d6	Small fix test=develop	6 years ago
Krzysztof Binias	b1bdcd4de8	Make separate folders for mkldnn codes test=develop	6 years ago

28 Commits (5eb81fe595a95758ee01450f600850273c97a197)