mingfeima / convnet-benchmark-py Goto Github PK

I run the script for deploying resnext101, but I can't get any gains with mkldnn and cache-weight.
Here is my runtime information:
./run.sh --inference --single
inference only
using single batch size
using OMP_NUM_THREADS=32
using KMP_AFFINITY=granularity=fine,compact,1,0
using KMP_BLOCKTIME=1

Running on device: Intel(R) Xeon(R) Gold 6142 CPU @ 2.60GHz
Running on torch: 1.4.0
Running on torchvision: 0.5.0

ModelType: resnext101, Kernels: nn Input shape: 1x3x224x224
nn :forward: 236.60 (ms) 4.23 (imgs/s)
nn :backward: 0.00 (ms)
nn :update: 0.00 (ms)
nn :total: 236.60 (ms) 4.23 (imgs/s)

---

./run.sh --inference --single --mkldnn
ModelType: resnext101, Kernels: nn Input shape: 1x3x224x224
nn :forward: 242.39 (ms) 4.13 (imgs/s)
nn :backward: 0.00 (ms)
nn :update: 0.00 (ms)
nn :total: 242.39 (ms) 4.13 (imgs/s)

---

./run.sh --inference --single --mkldnn --cache-weight
ModelType: resnext101, Kernels: nn Input shape: 1x3x224x224
nn :forward: 236.43 (ms) 4.23 (imgs/s)
nn :backward: 0.00 (ms)
nn :update: 0.00 (ms)
nn :total: 236.43 (ms) 4.23 (imgs/s)

There is no benifits with mkldnn or cache-weight. I install pytorch by pip, is it the case that build pytorch without icc?
Here is my pytorch built informations:
PyTorch built with:

• GCC 7.3
• Intel(R) Math Kernel Library Version 2019.0.4 Product Build 20190411 for Intel(R) 64 architecture applications
• Intel(R) MKL-DNN v0.21.1 (Git Hash 7d2fd500bc78936d1d648ca713b901012f470dbc)
• OpenMP 201511 (a.k.a. OpenMP 4.5)
• NNPACK is enabled
• CUDA Runtime 10.1
• NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_61,code=sm_61;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_37,code=compute_37
• CuDNN 7.6.3
• Magma 2.5.1
• Build settings: BLAS=MKL, BUILD_NAMEDTENSOR=OFF, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -fopenmp -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -O2 -fPIC -Wno-narrowing -Wall -Wextra -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Wno-stringop-overflow, DISABLE_NUMA=1, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, USE_STATIC_DISPATCH=OFF,

---

@mingfeima Can u help me to find why I can't get the benifits that you mentioned？