From 53a79f5830d9c0da7ff115005b74052997bad50d Mon Sep 17 00:00:00 2001
From: Liangliang He <lliang.he@gmail.com>
Date: Fri, 27 Apr 2018 16:36:38 +0800
Subject: [PATCH] Reduce model tuning time

---
 docs/user/introduction.md |  2 --
 mace/utils/tuner.h        | 11 ++++++++---
 tools/example.yaml        |  1 -
 tools/mace_tools.py       |  7 ++++++-
 4 files changed, 14 insertions(+), 7 deletions(-)

diff --git a/docs/user/introduction.md b/docs/user/introduction.md
index 599cee07..3ed1512f 100644
--- a/docs/user/introduction.md
+++ b/docs/user/introduction.md
@@ -168,7 +168,6 @@ target_abis: [armeabi-v7a, arm64-v8a]
 # 具体机型的soc编号，可以使用`adb shell getprop | grep ro.board.platform | cut -d [ -f3 | cut -d ] -f1`获取
 target_socs: [msm8998]
 embed_model_data: 1
-vlog_level: 0
 models: # 一个配置文件可以包含多个模型的配置信息，最终生成的库中包含多个模型
   first_net: # 模型的标签，在调度模型的时候，会用这个变量
     platform: tensorflow
@@ -218,7 +217,6 @@ models: # 一个配置文件可以包含多个模型的配置信息，最终生
 | ---------- |:--------------:|
 | target_abis | 运行的ABI，可选包括安卓设备的armeabi-v7a，arm64-v8a等，以及开发人员的电脑终端（电脑终端使用‘host’表示）。可以同时指定多个ABI |
 | embed_model_data | 是否将模型里的数据嵌入到代码中，默认为1 |
-| vlog_level | 设置log打印的级别 |
 | platform | 模型对应的框架名称 [tensorflow | caffe] |
 | model_file_path | 模型的路径，可以是一个http或https的下载链接 |
 | weight_file_path | 权重文件的路径，可以是一个http或https的下载链接(caffe model)|
diff --git a/mace/utils/tuner.h b/mace/utils/tuner.h
index db4f25fa..719d7699 100644
--- a/mace/utils/tuner.h
+++ b/mace/utils/tuner.h
@@ -141,13 +141,18 @@ class Tuner {
       double *time_us,
       std::vector<param_type> *tuning_result) {
     RetType res;
+    int iter = 0;
     int64_t total_time_us = 0;
-    for (int i = 0; i < num_runs; ++i) {
+    for (iter = 0; iter < num_runs; ++iter) {
       res = func(params, timer, tuning_result);
       total_time_us += timer->AccumulatedMicros();
+      if (iter >= 1 && total_time_us > 100000 || total_time_us > 200000) {
+        ++iter;
+        break;
+      }
     }
 
-    *time_us = total_time_us * 1.0 / num_runs;
+    *time_us = total_time_us * 1.0 / iter;
     return res;
   }
 
@@ -167,7 +172,7 @@ class Tuner {
     for (auto param : params) {
       double tmp_time = 0.0;
       // warm up
-      Run<RetType>(func, param, timer, 2, &tmp_time, &tuning_result);
+      Run<RetType>(func, param, timer, 1, &tmp_time, &tuning_result);
 
       // run
       RetType tmp_res =
diff --git a/tools/example.yaml b/tools/example.yaml
index 6fe860d0..b0fda8b3 100644
--- a/tools/example.yaml
+++ b/tools/example.yaml
@@ -5,7 +5,6 @@
 target_abis: [armeabi-v7a, arm64-v8a]
 target_socs: [MSM8953]
 embed_model_data: 1
-vlog_level: 0
 models:
   preview_net:
     platform: tensorflow
diff --git a/tools/mace_tools.py b/tools/mace_tools.py
index 879cd72c..c8fbcdf4 100644
--- a/tools/mace_tools.py
+++ b/tools/mace_tools.py
@@ -349,6 +349,11 @@ def parse_args():
         type="bool",
         default="false",
         help="Collect report.")
+    parser.add_argument(
+        "--vlog_level",
+        type=int,
+        default=0,
+        help="VLOG level.")
     return parser.parse_known_args()
 
 
@@ -567,7 +572,7 @@ def main(unused_args):
     target_socs = get_target_socs(configs)
 
     embed_model_data = configs.get("embed_model_data", 1)
-    vlog_level = configs.get("vlog_level", 0)
+    vlog_level = FLAGS.vlog_level
     phone_data_dir = "/data/local/tmp/mace_run/"
     for target_abi in configs["target_abis"]:
         for target_soc in target_socs:
-- 
GitLab