TFLite Model Benchmark Toolは、TensorFlow Lite形式のモデル(.tflite
モデル)のベンチマークを取ることができるツールです。各オペレーション毎にかかる時間を測定することもできます。on desktopでの実行方法をメモしておきます。.tflite
モデルがすでに手元にある前提です。
Bazelのインストール
TFLite Model Benchmark Toolを使用するにはBazelというビルドツールが必要です。公式のインストール手順に従って以下のコードを実行するとすぐにインストールできました。
brew tap bazelbuild/tap
brew tap-pin bazelbuild/tap
brew install bazelbuild/tap/bazel
Benchmark Toolをビルドして動かす
まずTensorFlowのリポジトリをクローンしてきます。
git clone https://github.com/tensorflow/tensorflow.git
cd tensorflow
Benchmark Toolのチュートリアル通りに、バイナリファイルをビルドします。この時--copt=-DTFLITE_PROFILING_ENABLED
というオプションをつけてビルドすることで、オペレーション毎の実行速度を計測することができます。
bazel build -c opt \
--copt=-DTFLITE_PROFILING_ENABLED \
tensorflow/lite/tools/benchmark:benchmark_model
あとは自分で用意した.tflite
モデルを指定してBenchmark Toolを動かすだけです。
bazel-bin/tensorflow/lite/tools/benchmark/benchmark_model \
--graph=mobilenet_quant_v1_224.tflite \
--num_threads=4
そうすると、以下のように各オペレーション毎の実行速度が表示されるかと思います。
============================== Run Order ==============================
[node type] [start] [first] [avg ms] [%] [cdf%] [mem KB] [times called] [Name]
CONV_2D 0.000 4.269 4.269 0.107% 0.107% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_0/Relu6]
DEPTHWISE_CONV_2D 4.270 2.150 2.150 0.054% 0.161% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_1_depthwise/Relu6]
CONV_2D 6.421 6.107 6.107 0.153% 0.314% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_1_pointwise/Relu6]
DEPTHWISE_CONV_2D 12.528 1.366 1.366 0.034% 0.348% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_2_depthwise/Relu6]
CONV_2D 13.895 4.195 4.195 0.105% 0.454% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_2_pointwise/Relu6]
DEPTHWISE_CONV_2D 18.091 1.260 1.260 0.032% 0.485% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_3_depthwise/Relu6]
CONV_2D 19.352 6.652 6.652 0.167% 0.652% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_3_pointwise/Relu6]
DEPTHWISE_CONV_2D 26.005 0.698 0.698 0.018% 0.670% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_4_depthwise/Relu6]
CONV_2D 26.703 3.344 3.344 0.084% 0.754% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_4_pointwise/Relu6]
DEPTHWISE_CONV_2D 30.047 0.646 0.646 0.016% 0.770% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_5_depthwise/Relu6]
CONV_2D 30.694 5.800 5.800 0.145% 0.915% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_5_pointwise/Relu6]
DEPTHWISE_CONV_2D 36.495 0.331 0.331 0.008% 0.924% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_6_depthwise/Relu6]
CONV_2D 36.826 2.838 2.838 0.071% 0.995% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_6_pointwise/Relu6]
DEPTHWISE_CONV_2D 39.665 0.439 0.439 0.011% 1.006% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_7_depthwise/Relu6]
CONV_2D 40.105 5.293 5.293 0.133% 1.139% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_7_pointwise/Relu6]
DEPTHWISE_CONV_2D 45.399 0.352 0.352 0.009% 1.147% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_8_depthwise/Relu6]
CONV_2D 45.752 5.322 5.322 0.133% 1.281% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_8_pointwise/Relu6]
DEPTHWISE_CONV_2D 51.075 0.357 0.357 0.009% 1.290% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_9_depthwise/Relu6]
CONV_2D 51.432 5.693 5.693 0.143% 1.433% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_9_pointwise/Relu6]
DEPTHWISE_CONV_2D 57.126 0.366 0.366 0.009% 1.442% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_10_depthwise/Relu6]
CONV_2D 57.493 5.472 5.472 0.137% 1.579% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_10_pointwise/Relu6]
DEPTHWISE_CONV_2D 62.966 0.364 0.364 0.009% 1.588% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_11_depthwise/Relu6]
CONV_2D 63.330 5.404 5.404 0.136% 1.724% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_11_pointwise/Relu6]
DEPTHWISE_CONV_2D 68.735 0.155 0.155 0.004% 1.728% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_12_depthwise/Relu6]
CONV_2D 68.891 2.970 2.970 0.074% 1.802% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_12_pointwise/Relu6]
DEPTHWISE_CONV_2D 71.862 0.206 0.206 0.005% 1.807% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_13_depthwise/Relu6]
CONV_2D 72.069 5.888 5.888 0.148% 1.955% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_13_pointwise/Relu6]
AVERAGE_POOL_2D 77.958 0.036 0.036 0.001% 1.956% 0.000 0 [MobilenetV1/Logits/AvgPool_1a/AvgPool]
CONV_2D 77.994 1.445 1.445 0.036% 1.992% 0.000 0 [MobilenetV1/Logits/Conv2d_1c_1x1/BiasAdd]
RESHAPE 79.440 0.002 0.002 0.000% 1.992% 0.000 0 [MobilenetV1/Predictions/Reshape]
SOFTMAX 79.443 0.029 0.029 0.001% 1.993% 0.000 0 [MobilenetV1/Predictions/Softmax]
============================== Top by Computation Time ==============================
[node type] [start] [first] [avg ms] [%] [cdf%] [mem KB] [times called] [Name]
CONV_2D 19.352 6.652 6.652 0.167% 0.167% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_3_pointwise/Relu6]
CONV_2D 6.421 6.107 6.107 0.153% 0.320% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_1_pointwise/Relu6]
CONV_2D 72.069 5.888 5.888 0.148% 0.468% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_13_pointwise/Relu6]
CONV_2D 30.694 5.800 5.800 0.145% 0.613% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_5_pointwise/Relu6]
CONV_2D 51.432 5.693 5.693 0.143% 0.756% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_9_pointwise/Relu6]
CONV_2D 57.493 5.472 5.472 0.137% 0.893% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_10_pointwise/Relu6]
CONV_2D 63.330 5.404 5.404 0.136% 1.029% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_11_pointwise/Relu6]
CONV_2D 45.752 5.322 5.322 0.133% 1.162% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_8_pointwise/Relu6]
CONV_2D 40.105 5.293 5.293 0.133% 1.295% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_7_pointwise/Relu6]
CONV_2D 0.000 4.269 4.269 0.107% 1.402% 0.000 0 [MobilenetV1/MobilenetV1/Conv2d_0/Relu6]
Number of nodes executed: 31
============================== Summary by node type ==============================
[Node type] [count] [avg ms] [avg %] [cdf %] [mem KB] [times called]
CONV_2D 15 1.406 89.270% 89.270% 0.000 0
DEPTHWISE_CONV_2D 13 0.169 10.730% 100.000% 0.000 0
SOFTMAX 1 0.000 0.000% 100.000% 0.000 0
RESHAPE 1 0.000 0.000% 100.000% 0.000 0
AVERAGE_POOL_2D 1 0.000 0.000% 100.000% 0.000 0
Timings (microseconds): count=50 first=79449 curr=81350 min=77385 max=88213 avg=79732 std=1929
Memory (bytes): count=0
31 nodes observed
Average inference timings in us: Warmup: 83235, Init: 38467, no stats: 79760.9
各項目の意味も何となくですがわかりますね。
Enjoy TFLite!!