tipc: support benchmark with fp32/fp16 training

test_tipc/README.md

@@ -31,7 +31,7 @@
 ```
 ./test_tipc/
 ├── common_func.sh                      #test_*.sh会调用到的公共函数
-├── config     # 配置文件目录
+├── configs     # 配置文件目录
 │   ├── MobileNetV3         # MobileNetV3系列模型测试配置文件目录
 │   │   ├── MobileNetV3_large_x1_0_train_infer_python.txt                                    #基础训练预测配置文件
 │   │   ├── MobileNetV3_large_x1_0_train_linux_gpu_fleet_amp_infer_python_linux_gpu_cpu.txt  #多机多卡训练预测配置文件

test_tipc/configs/HRNet/HRNet_W18_C_train_infer_python.txt

@@ -50,5 +50,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 -o Global.benchmark:False
 null:null
 null:null
+===========================train_benchmark_params==========================
+batch_size:64
+fp_items:fp32|fp16
+epoch:1
+--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
+flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
 random_infer_input:[{float32,[3,224,224]}]
\ No newline at end of file

test_tipc/configs/MobileNetV1/MobileNetV1_train_infer_python.txt

@@ -52,7 +52,7 @@ null:null
 null:null
 ===========================train_benchmark_params==========================
 batch_size:64|128
-fp_items:fp32
+fp_items:fp32|fp16
 epoch:1
 --profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
 flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096

test_tipc/configs/MobileNetV2/MobileNetV2_train_infer_python.txt

@@ -52,7 +52,7 @@ null:null
 null:null
 ===========================train_benchmark_params==========================
 batch_size:64|128
-fp_items:fp32
+fp_items:fp32|fp16
 epoch:1
 --profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
 flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096

test_tipc/configs/MobileNetV3/MobileNetV3_large_x1_0_train_infer_python.txt

@@ -52,7 +52,7 @@ null:null
 null:null
 ===========================train_benchmark_params==========================
 batch_size:256|640
-fp_items:fp32
+fp_items:fp32|fp16
 epoch:1
 --profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
 flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096

test_tipc/configs/PPHGNet/PPHGNet_small_train_infer_python.txt

@@ -49,5 +49,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran
 -o Global.save_log_path:null
 -o Global.benchmark:False
 null:null
+===========================train_benchmark_params==========================
+batch_size:128
+fp_items:fp32|fp16
+epoch:1
+--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
+flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
 random_infer_input:[{float32,[3,224,224]}]

test_tipc/configs/PPHGNet/PPHGNet_tiny_train_infer_python.txt

@@ -49,5 +49,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml -o PreProcess.tran
 -o Global.save_log_path:null
 -o Global.benchmark:False
 null:null
+===========================train_benchmark_params==========================
+batch_size:128
+fp_items:fp32|fp16
+epoch:1
+--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
+flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
 random_infer_input:[{float32,[3,224,224]}]

test_tipc/configs/PPLCNet/PPLCNet_x0_25_train_infer_python.txt

@@ -49,5 +49,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 -o Global.save_log_path:null
 -o Global.benchmark:False
 null:null
+===========================train_benchmark_params==========================
+batch_size:512
+fp_items:fp32|fp16
+epoch:1
+--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
+flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
 random_infer_input:[{float32,[3,224,224]}]
\ No newline at end of file

test_tipc/configs/PPLCNet/PPLCNet_x0_35_train_infer_python.txt

@@ -49,5 +49,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 -o Global.save_log_path:null
 -o Global.benchmark:False
 null:null
+===========================train_benchmark_params==========================
+batch_size:512
+fp_items:fp32|fp16
+epoch:1
+--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
+flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
 random_infer_input:[{float32,[3,224,224]}]
\ No newline at end of file

test_tipc/configs/PPLCNet/PPLCNet_x0_5_train_infer_python.txt

@@ -49,5 +49,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 -o Global.save_log_path:null
 -o Global.benchmark:False
 null:null
+===========================train_benchmark_params==========================
+batch_size:512
+fp_items:fp32|fp16
+epoch:1
+--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
+flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
 random_infer_input:[{float32,[3,224,224]}]
\ No newline at end of file

test_tipc/configs/PPLCNet/PPLCNet_x0_75_train_infer_python.txt

@@ -49,5 +49,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 -o Global.save_log_path:null
 -o Global.benchmark:False
 null:null
+===========================train_benchmark_params==========================
+batch_size:512
+fp_items:fp32|fp16
+epoch:1
+--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
+flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
 random_infer_input:[{float32,[3,224,224]}]
\ No newline at end of file

test_tipc/configs/PPLCNet/PPLCNet_x1_0_train_infer_python.txt

@@ -49,5 +49,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 -o Global.save_log_path:null
 -o Global.benchmark:False
 null:null
+===========================train_benchmark_params==========================
+batch_size:512
+fp_items:fp32|fp16
+epoch:1
+--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
+flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
 random_infer_input:[{float32,[3,224,224]}]
\ No newline at end of file

test_tipc/configs/PPLCNet/PPLCNet_x1_5_train_infer_python.txt

@@ -49,5 +49,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 -o Global.save_log_path:null
 -o Global.benchmark:False
 null:null
+===========================train_benchmark_params==========================
+batch_size:512
+fp_items:fp32|fp16
+epoch:1
+--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
+flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
 random_infer_input:[{float32,[3,224,224]}]
\ No newline at end of file

test_tipc/configs/PPLCNet/PPLCNet_x2_0_train_infer_python.txt

@@ -49,5 +49,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 -o Global.save_log_path:null
 -o Global.benchmark:False
 null:null
+===========================train_benchmark_params==========================
+batch_size:512
+fp_items:fp32|fp16
+epoch:1
+--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
+flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
 random_infer_input:[{float32,[3,224,224]}]
\ No newline at end of file

test_tipc/configs/PPLCNet/PPLCNet_x2_5_train_infer_python.txt

@@ -49,5 +49,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 -o Global.save_log_path:null
 -o Global.benchmark:False
 null:null
+===========================train_benchmark_params==========================
+batch_size:512
+fp_items:fp32|fp16
+epoch:1
+--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
+flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
 random_infer_input:[{float32,[3,224,224]}]
\ No newline at end of file

test_tipc/configs/PPLCNetV2/PPLCNetV2_base_train_infer_python.txt

@@ -49,5 +49,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 -o Global.save_log_path:null
 -o Global.benchmark:False
 null:null
+===========================train_benchmark_params==========================
+batch_size:500
+fp_items:fp32|fp16
+epoch:1
+--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
+flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
-random_infer_input:[{float32,[3,224,224]}]
+random_infer_input:[{float32,[3,224,224]}]
\ No newline at end of file

test_tipc/configs/ResNet/ResNet152_train_infer_python.txt

@@ -51,8 +51,8 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 null:null
 null:null
 ===========================train_benchmark_params==========================
-batch_size:32
-fp_items:fp32
+batch_size:32|64
+fp_items:fp32|fp16
 epoch:1
 --profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
 flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096

test_tipc/configs/ResNet/ResNet50_train_infer_python.txt

@@ -51,8 +51,8 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 null:null
 null:null
 ===========================train_benchmark_params==========================
-batch_size:128
-fp_items:fp32
+batch_size:128|64
+fp_items:fp32|fp16
 epoch:1
 --profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
 flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096

test_tipc/configs/ResNet/ResNet50_vd_train_infer_python.txt

@@ -50,5 +50,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 -o Global.benchmark:False
 null:null
 null:null
+===========================train_benchmark_params==========================
+batch_size:64
+fp_items:fp32|fp16
+epoch:1
+--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
+flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
 random_infer_input:[{float32,[3,224,224]}]
\ No newline at end of file

test_tipc/configs/ShuffleNet/ShuffleNetV2_x1_0_train_infer_python.txt

@@ -52,9 +52,9 @@ null:null
 null:null
 ===========================train_benchmark_params==========================
 batch_size:256|1536
-fp_items:fp32
+fp_items:fp32|fp16
 epoch:2
 --profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
 flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
-random_infer_input:[{float32,[3,224,224]}]
+random_infer_input:[{float32,[3,224,224]}]
\ No newline at end of file

test_tipc/configs/SwinTransformer/SwinTransformer_tiny_patch4_window7_224_train_infer_python.txt

@@ -51,8 +51,8 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 null:null
 null:null
 ===========================train_benchmark_params==========================
-batch_size:64|104
-fp_items:fp32
+batch_size:64|104|128
+fp_items:fp32|fp16
 epoch:1
 --profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
 flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096

test_tipc/configs/Twins/alt_gvt_small_train_infer_python.txt

@@ -50,5 +50,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
 -o Global.benchmark:False
 null:null
 null:null
+===========================train_benchmark_params==========================
+batch_size:128
+fp_items:fp32|fp16
+epoch:1
+--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
+flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
 ===========================infer_benchmark_params==========================
 random_infer_input:[{float32,[3,224,224]}]
\ No newline at end of file