From 494f381f213b3ca863e74048d193e246916b767b Mon Sep 17 00:00:00 2001
From: shangliang Xu <ghostxsl@users.noreply.github.com>
Date: Mon, 8 Aug 2022 16:24:26 +0800
Subject: [PATCH] [TIPC] fix random seed in train benchmark (#6603)

---
 ...gpu_normal_amp_infer_python_linux_gpu_cpu.txt |  2 +-
 ...gpu_normal_amp_infer_python_linux_gpu_cpu.txt |  2 +-
 ...gpu_normal_amp_infer_python_linux_gpu_cpu.txt |  2 +-
 ...gpu_normal_amp_infer_python_linux_gpu_cpu.txt |  2 +-
 ...gpu_normal_amp_infer_python_linux_gpu_cpu.txt |  2 +-
 ...gpu_normal_amp_infer_python_linux_gpu_cpu.txt |  2 +-
 ...gpu_normal_amp_infer_python_linux_gpu_cpu.txt |  2 +-
 ...gpu_normal_amp_infer_python_linux_gpu_cpu.txt |  2 +-
 ...gpu_normal_amp_infer_python_linux_gpu_cpu.txt |  2 +-
 ...gpu_normal_amp_infer_python_linux_gpu_cpu.txt |  2 +-
 ...gpu_normal_amp_infer_python_linux_gpu_cpu.txt |  2 +-
 test_tipc/test_train_inference_python.sh         | 16 ++++++++++++----
 12 files changed, 23 insertions(+), 15 deletions(-)

diff --git a/test_tipc/configs/keypoint/tinypose_128x96_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt b/test_tipc/configs/keypoint/tinypose_128x96_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
index b91057703..7be6cfed1 100644
--- a/test_tipc/configs/keypoint/tinypose_128x96_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
+++ b/test_tipc/configs/keypoint/tinypose_128x96_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
@@ -10,7 +10,7 @@ TrainReader.batch_size:lite_train_lite_infer=2|lite_train_whole_infer=2|whole_tr
 pretrain_weights:https://paddledet.bj.bcebos.com/models/keypoint/tinypose_128x96.pdparams
 trained_model_name:model_final.pdparams
 train_infer_img_dir:./dataset/coco/test2017/
-amp_level:O2
+null:null
 ##
 trainer:norm_train
 norm_train:tools/train.py -c test_tipc/configs/keypoint/tinypose_128x96.yml -o
diff --git a/test_tipc/configs/mask_rcnn/mask_rcnn_r50_fpn_1x_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt b/test_tipc/configs/mask_rcnn/mask_rcnn_r50_fpn_1x_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
index af061e055..796fd008c 100644
--- a/test_tipc/configs/mask_rcnn/mask_rcnn_r50_fpn_1x_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
+++ b/test_tipc/configs/mask_rcnn/mask_rcnn_r50_fpn_1x_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
@@ -10,7 +10,7 @@ TrainReader.batch_size:lite_train_lite_infer=2|lite_train_whole_infer=2|whole_tr
 pretrain_weights:https://paddledet.bj.bcebos.com/models/mask_rcnn_r50_fpn_1x_coco.pdparams
 trained_model_name:model_final.pdparams
 train_infer_img_dir:./dataset/coco/test2017/
-amp_level:O2
+null:null
 ##
 trainer:norm_train
 norm_train:tools/train.py -c configs/mask_rcnn/mask_rcnn_r50_fpn_1x_coco.yml -o
diff --git a/test_tipc/configs/picodet/picodet_lcnet_1_5x_416_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt b/test_tipc/configs/picodet/picodet_lcnet_1_5x_416_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
index 28272ad58..19ff2c7b8 100644
--- a/test_tipc/configs/picodet/picodet_lcnet_1_5x_416_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
+++ b/test_tipc/configs/picodet/picodet_lcnet_1_5x_416_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
@@ -10,7 +10,7 @@ TrainReader.batch_size:lite_train_lite_infer=2|lite_train_whole_infer=2|whole_tr
 pretrain_weights:https://paddledet.bj.bcebos.com/models/picodet_lcnet_1_5x_416_coco.pdparams
 trained_model_name:model_final.pdparams
 train_infer_img_dir:./dataset/coco/test2017/
-amp_level:O2
+null:null
 ##
 trainer:norm_train
 norm_train:tools/train.py -c configs/picodet/legacy_model/more_config/picodet_lcnet_1_5x_416_coco.yml -o
diff --git a/test_tipc/configs/picodet/picodet_s_320_coco_lcnet_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt b/test_tipc/configs/picodet/picodet_s_320_coco_lcnet_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
index cac5a70bb..cf61a5ab1 100644
--- a/test_tipc/configs/picodet/picodet_s_320_coco_lcnet_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
+++ b/test_tipc/configs/picodet/picodet_s_320_coco_lcnet_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
@@ -10,7 +10,7 @@ TrainReader.batch_size:lite_train_lite_infer=2|lite_train_whole_infer=2|whole_tr
 pretrain_weights:https://paddledet.bj.bcebos.com/models/picodet_s_320_coco_lcnet.pdparams
 trained_model_name:model_final.pdparams
 train_infer_img_dir:./dataset/coco/test2017/
-amp_level:O2
+null:null
 ##
 trainer:norm_train
 norm_train:tools/train.py -c configs/picodet/picodet_s_320_coco_lcnet.yml -o
diff --git a/test_tipc/configs/picodet/picodet_s_320_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt b/test_tipc/configs/picodet/picodet_s_320_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
index b1a87cdd7..4c3cf24d8 100644
--- a/test_tipc/configs/picodet/picodet_s_320_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
+++ b/test_tipc/configs/picodet/picodet_s_320_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
@@ -10,7 +10,7 @@ TrainReader.batch_size:lite_train_lite_infer=2|lite_train_whole_infer=2|whole_tr
 pretrain_weights:https://paddledet.bj.bcebos.com/models/picodet_s_320_coco.pdparams
 trained_model_name:model_final.pdparams
 train_infer_img_dir:./dataset/coco/test2017/
-amp_level:O2
+null:null
 ##
 trainer:norm_train
 norm_train:tools/train.py -c configs/picodet/legacy_model/picodet_s_320_coco.yml -o
diff --git a/test_tipc/configs/ppyolo/ppyolo_mbv3_large_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt b/test_tipc/configs/ppyolo/ppyolo_mbv3_large_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
index 6ee39190d..e89918d47 100644
--- a/test_tipc/configs/ppyolo/ppyolo_mbv3_large_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
+++ b/test_tipc/configs/ppyolo/ppyolo_mbv3_large_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
@@ -10,7 +10,7 @@ TrainReader.batch_size:lite_train_lite_infer=2|lite_train_whole_infer=2|whole_tr
 pretrain_weights:https://paddledet.bj.bcebos.com/models/ppyolo_mbv3_large_coco.pdparams
 trained_model_name:model_final.pdparams
 train_infer_img_dir:./dataset/coco/test2017/
-amp_level:O2
+null:null
 ##
 trainer:norm_train
 norm_train:tools/train.py -c configs/ppyolo/ppyolo_mbv3_large_coco.yml -o
diff --git a/test_tipc/configs/ppyolo/ppyolo_r50vd_dcn_1x_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt b/test_tipc/configs/ppyolo/ppyolo_r50vd_dcn_1x_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
index 6a4235a0b..2aac9ba9f 100644
--- a/test_tipc/configs/ppyolo/ppyolo_r50vd_dcn_1x_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
+++ b/test_tipc/configs/ppyolo/ppyolo_r50vd_dcn_1x_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
@@ -10,7 +10,7 @@ TrainReader.batch_size:lite_train_lite_infer=2|lite_train_whole_infer=2|whole_tr
 pretrain_weights:https://paddledet.bj.bcebos.com/models/ppyolo_r50vd_dcn_1x_coco.pdparams
 trained_model_name:model_final.pdparams
 train_infer_img_dir:./dataset/coco/test2017/
-amp_level:O2
+null:null
 ##
 trainer:norm_train
 norm_train:tools/train.py -c configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml -o
diff --git a/test_tipc/configs/ppyolo/ppyolo_tiny_650e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt b/test_tipc/configs/ppyolo/ppyolo_tiny_650e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
index cb0659a7b..dcd3252fc 100644
--- a/test_tipc/configs/ppyolo/ppyolo_tiny_650e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
+++ b/test_tipc/configs/ppyolo/ppyolo_tiny_650e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
@@ -10,7 +10,7 @@ TrainReader.batch_size:lite_train_lite_infer=2|lite_train_whole_infer=2|whole_tr
 pretrain_weights:https://paddledet.bj.bcebos.com/models/ppyolo_tiny_650e_coco.pdparams
 trained_model_name:model_final.pdparams
 train_infer_img_dir:./dataset/coco/test2017/
-amp_level:O2
+null:null
 ##
 trainer:norm_train
 norm_train:tools/train.py -c configs/ppyolo/ppyolo_tiny_650e_coco.yml -o
diff --git a/test_tipc/configs/ppyolo/ppyolov2_r50vd_dcn_365e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt b/test_tipc/configs/ppyolo/ppyolov2_r50vd_dcn_365e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
index 721db09e0..b6423b95a 100644
--- a/test_tipc/configs/ppyolo/ppyolov2_r50vd_dcn_365e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
+++ b/test_tipc/configs/ppyolo/ppyolov2_r50vd_dcn_365e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
@@ -10,7 +10,7 @@ TrainReader.batch_size:lite_train_lite_infer=2|lite_train_whole_infer=2|whole_tr
 pretrain_weights:https://paddledet.bj.bcebos.com/models/ppyolov2_r50vd_dcn_365e_coco.pdparams
 trained_model_name:model_final.pdparams
 train_infer_img_dir:./dataset/coco/test2017/
-amp_level:O2
+null:null
 ##
 trainer:norm_train
 norm_train:tools/train.py -c configs/ppyolo/ppyolov2_r50vd_dcn_365e_coco.yml -o
diff --git a/test_tipc/configs/ppyoloe/ppyoloe_crn_s_300e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt b/test_tipc/configs/ppyoloe/ppyoloe_crn_s_300e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
index 815a23937..f11713ba9 100644
--- a/test_tipc/configs/ppyoloe/ppyoloe_crn_s_300e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
+++ b/test_tipc/configs/ppyoloe/ppyoloe_crn_s_300e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
@@ -10,7 +10,7 @@ TrainReader.batch_size:lite_train_lite_infer=2|lite_train_whole_infer=2|whole_tr
 pretrain_weights:https://paddledet.bj.bcebos.com/models/ppyoloe_crn_s_300e_coco.pdparams
 trained_model_name:model_final.pdparams
 train_infer_img_dir:./dataset/coco/test2017/
-amp_level:O2
+null:null
 ##
 trainer:norm_train
 norm_train:tools/train.py -c configs/ppyoloe/ppyoloe_crn_s_300e_coco.yml -o
diff --git a/test_tipc/configs/yolov3/yolov3_darknet53_270e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt b/test_tipc/configs/yolov3/yolov3_darknet53_270e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
index b3142b6ea..7987def92 100644
--- a/test_tipc/configs/yolov3/yolov3_darknet53_270e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
+++ b/test_tipc/configs/yolov3/yolov3_darknet53_270e_coco_train_linux_gpu_normal_amp_infer_python_linux_gpu_cpu.txt
@@ -10,7 +10,7 @@ TrainReader.batch_size:lite_train_lite_infer=2|lite_train_whole_infer=2|whole_tr
 pretrain_weights:https://paddledet.bj.bcebos.com/models/yolov3_darknet53_270e_coco.pdparams
 trained_model_name:model_final.pdparams
 train_infer_img_dir:./dataset/coco/test2017/
-amp_level:O2
+null:null
 ##
 trainer:norm_train
 norm_train:tools/train.py -c configs/yolov3/yolov3_darknet53_270e_coco.yml -o
diff --git a/test_tipc/test_train_inference_python.sh b/test_tipc/test_train_inference_python.sh
index d92ef2c3e..6da0b030f 100644
--- a/test_tipc/test_train_inference_python.sh
+++ b/test_tipc/test_train_inference_python.sh
@@ -271,17 +271,25 @@ else
                 save_log="${LOG_PATH}/${trainer}_gpus_${gpu}_autocast_${autocast}"
                 if [ ${autocast} = "amp" ] || [ ${autocast} = "fp16" ]; then
                     set_autocast="--amp"
-                    set_train_params1="amp_level=O2"
+                    set_amp_level="amp_level=O2"
                 else
                     set_autocast=" "
+                    set_amp_level=" "
+                fi
+                if [ ${MODE} = "benchmark_train" ]; then
+                    set_shuffle="TrainReader.shuffle=False"
+                    set_enable_ce="--enable_ce=True"
+                else
+                    set_shuffle=" "
+                    set_enable_ce=" "
                 fi
 
                 set_save_model=$(func_set_params "${save_model_key}" "${save_log}")
                 nodes="1"
                 if [ ${#gpu} -le 2 ];then  # train with cpu or single gpu
-                    cmd="${python} ${run_train} LearningRate.base_lr=0.0001 log_iter=1 ${set_use_gpu} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_batchsize} ${set_filename} ${set_train_params1} ${set_autocast}"
+                    cmd="${python} ${run_train} LearningRate.base_lr=0.0001 log_iter=1 ${set_use_gpu} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_batchsize} ${set_filename} ${set_shuffle} ${set_amp_level} ${set_enable_ce} ${set_autocast} ${set_train_params1}"
                 elif [ ${#ips} -le 15 ];then  # train with multi-gpu
-                    cmd="${python} -m paddle.distributed.launch --gpus=${gpu} ${run_train} log_iter=1 ${set_use_gpu} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_batchsize} ${set_filename} ${set_train_params1} ${set_autocast}"
+                    cmd="${python} -m paddle.distributed.launch --gpus=${gpu} ${run_train} log_iter=1 ${set_use_gpu} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_batchsize} ${set_filename} ${set_shuffle} ${set_amp_level} ${set_enable_ce} ${set_autocast} ${set_train_params1}"
                 else     # train with multi-machine
                     IFS=","
                     ips_array=(${ips})
@@ -289,7 +297,7 @@ else
                     save_log="${LOG_PATH}/${trainer}_gpus_${gpu}_autocast_${autocast}_nodes_${nodes}"
                     IFS="|"
                     set_save_model=$(func_set_params "${save_model_key}" "${save_log}")
-                    cmd="${python} -m paddle.distributed.launch --ips=${ips} --gpus=${gpu} ${run_train} log_iter=1 ${set_use_gpu} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_batchsize} ${set_filename} ${set_train_params1} ${set_autocast}"
+                    cmd="${python} -m paddle.distributed.launch --ips=${ips} --gpus=${gpu} ${run_train} log_iter=1 ${set_use_gpu} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_batchsize} ${set_filename} ${set_shuffle} ${set_amp_level} ${set_enable_ce} ${set_autocast} ${set_train_params1}"
                 fi
                 # run train
                 train_log_path="${LOG_PATH}/${trainer}_gpus_${gpu}_autocast_${autocast}_nodes_${nodes}.log"
-- 
GitLab