add benchmark for maxout/put_along_axis/take_along_axis

[m3ngyang]

add benchmark for maxout/put_along_axis/take_along_axis
测试结果已添加。

[paddle-bot-old]

✅ This PR's description meets the template requirements!
Please wait for other CI results.

[paddle-bot-old]

Thanks for your contribution! Please wait for the result of CI firstly.

[CLAassistant]

All committers have signed the CLA.

[m3ngyang]

本地测试，cpu 环境

(paddle) root@94866e7b8d4a:/home/benchmark/api/dynamic_tests_v2# bash run.sh maxout 0 speed run.sh: line 6: [: !=: unary operator expected -- Current directory: /home/benchmark/api/dynamic_tests_v2 -- Entering /home/benchmark/api/common -- Current directory: /home/benchmark/api/common run command: git rev-parse HEAD run command: git show -s --format=%ad -- Current directory: /home/benchmark/api/dynamic_tests_v2 =========================================================================== -- paddle version : 0.0.0 -- paddle commit : eaacf8bfee5c9583f7ebf0deff20b90db9d73478 -- benchmark commit : 425703cc536122f96fb375c150699c85c47d1f9f -- benchmark last update time : Tue Mar 1 17:20:00 2022 +0800 =========================================================================== run command: /usr/local/share/miniconda3/envs/paddle/bin/python /home/benchmark/api/dynamic_tests_v2/maxout.py --task speed --framework paddle --testing_mode dynamic --json_file /home/benchmark/api/tests_v2/configs/maxout.json --config_id 0 --profiler none --backward True --use_gpu False --repeat 1000 --allow_adaptive_repeat False --log_level 0 Cannot import pytorch, maybe paddle is not installed. Cannot import pytorch, maybe pytorch is not installed. Namespace(allow_adaptive_repeat=False, api_name=None, backward=True, config_id=0, convert_to_fp16=False, filename=None, framework='paddle', get_status_without_running=False, gpu_time=0, json_file='/home/benchmark/api/tests_v2/configs/maxout.json', log_level=0, profiler='none', repeat=1000, scheduling_times='{}', sync_interval=80, task='speed', testing_mode='dynamic', unknown_dim=16, use_gpu=False) ---- Initialize APIConfig from /home/benchmark/api/tests_v2/configs/maxout.json, config_id = 0. Namespace(allow_adaptive_repeat=False, api_name=None, backward=True, config_id=0, convert_to_fp16=False, filename=None, framework='paddle', get_status_without_running=False, gpu_time=0, json_file='/home/benchmark/api/tests_v2/configs/maxout.json', log_level=0, profiler='none', repeat=1000, scheduling_times='{}', sync_interval=80, task='speed', testing_mode='dynamic', unknown_dim=16, use_gpu=False) [paddle][maxout] maxout { run_tf: True run_torch: False repeat: 5000 groups: 2 axis: -1 x_shape: [32, 12, 128, 128] x_dtype: float32 atol: 1e-06 } {"framework": "paddle", "version": "0.0.0", "name": "maxout", "device": "CPU", "backward": true, "speed": {"repeat": 1000, "begin": 10, "end": 990, "total": 67.3420314886132, "wall_time": 0, "total_include_wall_time": 67.3420314886132, "gpu_time": 0.0}, "parameters": "x (Variable) - dtype: float32, shape: [32, 12, 128, 128]\naxis (int): -1\ngroups (int): 2\n"} (paddle) root@94866e7b8d4a:/home/benchmark/api/dynamic_tests_v2# bash run.sh take_along_axis 0 speed run.sh: line 6: [: !=: unary operator expected -- Current directory: /home/benchmark/api/dynamic_tests_v2 -- Entering /home/benchmark/api/common -- Current directory: /home/benchmark/api/common run command: git rev-parse HEAD run command: git show -s --format=%ad -- Current directory: /home/benchmark/api/dynamic_tests_v2 =========================================================================== -- paddle version : 0.0.0 -- paddle commit : eaacf8bfee5c9583f7ebf0deff20b90db9d73478 -- benchmark commit : 6085eff3f1646448e431d037dbee1a1e0a231f9e -- benchmark last update time : Tue Mar 1 17:20:00 2022 +0800 =========================================================================== run command: /usr/local/share/miniconda3/envs/paddle/bin/python /home/benchmark/api/dynamic_tests_v2/take_along_axis.py --task speed --framework paddle --testing_mode dynamic --json_file /home/benchmark/api/tests_v2/configs/take_along_axis.json --config_id 0 --profiler none --backward True --use_gpu False --repeat 1000 --allow_adaptive_repeat False --log_level 0 Cannot import pytorch, maybe paddle is not installed. Cannot import pytorch, maybe pytorch is not installed. Namespace(allow_adaptive_repeat=False, api_name=None, backward=True, config_id=0, convert_to_fp16=False, filename=None, framework='paddle', get_status_without_running=False, gpu_time=0, json_file='/home/benchmark/api/tests_v2/configs/take_along_axis.json', log_level=0, profiler='none', repeat=1000, scheduling_times='{}', sync_interval=80, task='speed', testing_mode='dynamic', unknown_dim=16, use_gpu=False) ---- Initialize APIConfig from /home/benchmark/api/tests_v2/configs/take_along_axis.json, config_id = 0. Namespace(allow_adaptive_repeat=False, api_name=None, backward=True, config_id=0, convert_to_fp16=False, filename=None, framework='paddle', get_status_without_running=False, gpu_time=0, json_file='/home/benchmark/api/tests_v2/configs/take_along_axis.json', log_level=0, profiler='none', repeat=1000, scheduling_times='{}', sync_interval=80, task='speed', testing_mode='dynamic', unknown_dim=16, use_gpu=False) [paddle][take_along_axis] take_along_axis { run_tf: True run_torch: False axis: 0 arr_shape: [200, 300] arr_dtype: float32 indices_shape: [1, 1] indices_dtype: int atol: 1e-06 } {"framework": "paddle", "version": "0.0.0", "name": "take_along_axis", "device": "CPU", "backward": true, "speed": {"repeat": 1000, "begin": 10, "end": 990, "total": 0.23803297354250538, "wall_time": 0, "total_include_wall_time": 0.23803297354250538, "gpu_time": 0.0}, "parameters": "arr (Variable) - dtype: float32, shape: [200, 300]\nindices (Variable) - dtype: int, shape: [1, 1]\naxis (int): 0\n"} (paddle) root@94866e7b8d4a:/home/benchmark/api/dynamic_tests_v2# bash run.sh put_along_axis 0 speed run.sh: line 6: [: !=: unary operator expected -- Current directory: /home/benchmark/api/dynamic_tests_v2 -- Entering /home/benchmark/api/common -- Current directory: /home/benchmark/api/common run command: git rev-parse HEAD run command: git show -s --format=%ad -- Current directory: /home/benchmark/api/dynamic_tests_v2 =========================================================================== -- paddle version : 0.0.0 -- paddle commit : eaacf8bfee5c9583f7ebf0deff20b90db9d73478 -- benchmark commit : 6085eff3f1646448e431d037dbee1a1e0a231f9e -- benchmark last update time : Tue Mar 1 17:20:00 2022 +0800 =========================================================================== run command: /usr/local/share/miniconda3/envs/paddle/bin/python /home/benchmark/api/dynamic_tests_v2/put_along_axis.py --task speed --framework paddle --testing_mode dynamic --json_file /home/benchmark/api/tests_v2/configs/put_along_axis.json --config_id 0 --profiler none --backward True --use_gpu False --repeat 1000 --allow_adaptive_repeat False --log_level 0 Cannot import pytorch, maybe paddle is not installed. Cannot import pytorch, maybe pytorch is not installed. Namespace(allow_adaptive_repeat=False, api_name=None, backward=True, config_id=0, convert_to_fp16=False, filename=None, framework='paddle', get_status_without_running=False, gpu_time=0, json_file='/home/benchmark/api/tests_v2/configs/put_along_axis.json', log_level=0, profiler='none', repeat=1000, scheduling_times='{}', sync_interval=80, task='speed', testing_mode='dynamic', unknown_dim=16, use_gpu=False) ---- Initialize APIConfig from /home/benchmark/api/tests_v2/configs/put_along_axis.json, config_id = 0. Namespace(allow_adaptive_repeat=False, api_name=None, backward=True, config_id=0, convert_to_fp16=False, filename=None, framework='paddle', get_status_without_running=False, gpu_time=0, json_file='/home/benchmark/api/tests_v2/configs/put_along_axis.json', log_level=0, profiler='none', repeat=1000, scheduling_times='{}', sync_interval=80, task='speed', testing_mode='dynamic', unknown_dim=16, use_gpu=False) [paddle][put_along_axis] put_along_axis { run_tf: True run_torch: False values: 99.0 axis: 0 arr_shape: [200, 300] arr_dtype: float32 indices_shape: [1, 1] indices_dtype: int atol: 1e-06 } {"framework": "paddle", "version": "0.0.0", "name": "put_along_axis", "device": "CPU", "backward": true, "speed": {"repeat": 1000, "begin": 10, "end": 990, "total": 0.3393506517215651, "wall_time": 0, "total_include_wall_time": 0.3393506517215651, "gpu_time": 0.0}, "parameters": "arr (Variable) - dtype: float32, shape: [200, 300]\nindices (Variable) - dtype: int, shape: [1, 1]\naxis (int): 0\nvalues (float32): 99.0\n"}

[m3ngyang]

gpu 测试：

• maxout

(base) λ 48d12ee09b84 /home/paddle/benchmark/api/dynamic_tests_v2 {add_ops} bash run.sh maxout 1 speed grep: warning: GREP_OPTIONS is deprecated; please use an alias or script grep: warning: GREP_OPTIONS is deprecated; please use an alias or script -- Current directory: /home/paddle/benchmark/api/dynamic_tests_v2 -- Entering /home/paddle/benchmark/api/common -- Current directory: /home/paddle/benchmark/api/common run command: git rev-parse HEAD run command: git show -s --format=%ad -- Current directory: /home/paddle/benchmark/api/dynamic_tests_v2 grep: warning: GREP_OPTIONS is deprecated; please use an alias or script =========================================================================== -- paddle version : 0.0.0 -- paddle commit : eaacf8bfee5c9583f7ebf0deff20b90db9d73478 -- benchmark commit : 69b086c4ed78b3425955e9f3d683453e94a65598 -- benchmark last update time : Tue Mar 1 17:20:00 2022 +0800 =========================================================================== run command: nvidia-smi -L run command: nvprof --profile-from-start off /root/miniconda3/bin/python /home/paddle/benchmark/api/dynamic_tests_v2/maxout.py --task speed --framework paddle --testing_mode dynamic --json_file /home/paddle/benchmark/api/tests_v2/configs/maxout.json --config_id 1 --profiler nvprof --backward True --use_gpu True --repeat 1000 --allow_adaptive_repeat False --log_level 0 Type Time(%) Time Calls Avg Min Max Name GPU activities: 60.71% 15.7347s 1000 15.735ms 15.561ms 15.750ms void paddle::operators::math::KernelMaxoutGrad<float>(int, float const *, float const , float const , paddle::operators::math::KernelMaxoutGrad<float>*, int, int, int, int, int) 22.63% 5.86613s 1000 5.8661ms 5.8611ms 5.8945ms void paddle::operators::math::KernelMaxOut<float>(int, float const *, int, int, int, int, int, paddle::operators::math::KernelMaxOut<float>*) 16.66% 4.31877s 1000 4.3188ms 4.2655ms 4.5649ms void Eigen::internal::EigenMetaKernel<Eigen::TensorEvaluator<Eigen::TensorAssignOp<Eigen::TensorMap<Eigen::Tensor<float, int=1, int=1, long>, int=0, Eigen::MakePointer>, Eigen::TensorCwiseNullaryOp<Eigen::internal::scalar_constant_op<float>, Eigen::TensorMap<Eigen::Tensor<float, int=1, int=1, long>, int=0, Eigen::MakePointer> const > const > const , Eigen::GpuDevice>, long>(float, int=1) total gpu_time: 25917.8060 ms run command: /root/miniconda3/bin/python /home/paddle/benchmark/api/dynamic_tests_v2/maxout.py --task speed --framework paddle --testing_mode dynamic --json_file /home/paddle/benchmark/api/tests_v2/configs/maxout.json --config_id 1 --profiler none --backward True --use_gpu True --repeat 1000 --allow_adaptive_repeat False --log_level 0 --gpu_time 25917.805962773844 grep: warning: GREP_OPTIONS is deprecated; please use an alias or script W0304 08:43:11.716053 619 gpu_context.cc:240] Please NOTE: device: 0, GPU Compute Capability: 7.0, Driver API Version: 10.1, Runtime API Version: 10.1 W0304 08:43:11.721065 619 gpu_context.cc:268] device: 0, cuDNN Version: 7.6. Cannot import pytorch, maybe paddle is not installed. Cannot import pytorch, maybe pytorch is not installed. Namespace(allow_adaptive_repeat=False, api_name=None, backward=True, config_id=1, convert_to_fp16=False, filename=None, framework='paddle', get_status_without_running=False, gpu_time=25917.805962773844, json_file='/home/paddle/benchmark/api/tests_v2/configs/maxout.json', log_level=0, profiler='none', repeat=1000, scheduling_times='{}', sync_interval=80, task='speed', testing_mode='dynamic', unknown_dim=16, use_gpu=True) ---- Initialize APIConfig from /home/paddle/benchmark/api/tests_v2/configs/maxout.json, config_id = 1. Namespace(allow_adaptive_repeat=False, api_name=None, backward=True, config_id=1, convert_to_fp16=False, filename=None, framework='paddle', get_status_without_running=False, gpu_time=25917.805962773844, json_file='/home/paddle/benchmark/api/tests_v2/configs/maxout.json', log_level=0, profiler='none', repeat=1000, scheduling_times='{}', sync_interval=80, task='speed', testing_mode='dynamic', unknown_dim=16, use_gpu=True) [paddle][maxout] maxout { run_tf: True run_torch: False repeat: 5000 groups: 2 axis: -1 x_shape: [32, 8, 1024, 3072] x_dtype: float32 atol: 1e-06 } {"framework": "paddle", "version": "0.0.0", "name": "maxout", "device": "GPU", "backward": true, "speed": {"repeat": 1000, "begin": 10, "end": 990, "total": 25.951521980519196, "wall_time": 0, "total_include_wall_time": 25.951521980519196, "gpu_time": 25.917805962773844}, "parameters": "x (Variable) - dtype: float32, shape: [32, 8, 1024, 3072]\naxis (int): -1\ngroups (int): 2\n"}

---

• take_along_axis

(base) λ 48d12ee09b84 /home/paddle/benchmark/api/dynamic_tests_v2 {add_ops} bash run.sh take_along_axis 1 speed grep: warning: GREP_OPTIONS is deprecated; please use an alias or script grep: warning: GREP_OPTIONS is deprecated; please use an alias or script -- Current directory: /home/paddle/benchmark/api/dynamic_tests_v2 -- Entering /home/paddle/benchmark/api/common -- Current directory: /home/paddle/benchmark/api/common run command: git rev-parse HEAD run command: git show -s --format=%ad -- Current directory: /home/paddle/benchmark/api/dynamic_tests_v2 grep: warning: GREP_OPTIONS is deprecated; please use an alias or script =========================================================================== -- paddle version : 0.0.0 -- paddle commit : eaacf8bfee5c9583f7ebf0deff20b90db9d73478 -- benchmark commit : 69b086c4ed78b3425955e9f3d683453e94a65598 -- benchmark last update time : Tue Mar 1 17:20:00 2022 +0800 =========================================================================== run command: nvidia-smi -L run command: nvprof --profile-from-start off /root/miniconda3/bin/python /home/paddle/benchmark/api/dynamic_tests_v2/take_along_axis.py --task speed --framework paddle --testing_mode dynamic --json_file /home/paddle/benchmark/api/tests_v2/configs/take_along_axis.json --config_id 1 --profiler nvprof --backward True --use_gpu True --repeat 1000 --allow_adaptive_repeat False --log_level 0 Type Time(%) Time Calls Avg Min Max Name GPU activities: 23.40% 34.194ms 1000 34.193us 33.472us 35.712us [CUDA memcpy DtoD] 22.49% 32.875ms 1000 32.874us 32.160us 35.104us void Eigen::internal::EigenMetaKernel<Eigen::TensorEvaluator<Eigen::TensorAssignOp<Eigen::TensorMap<Eigen::Tensor<float, int=2, int=1, int>, int=16, Eigen::MakePointer>, Eigen::TensorBroadcastingOp<Eigen::DSizes<long, int=2> const , Eigen::TensorMap<Eigen::Tensor<float const , int=2, int=1, int>, int=16, Eigen::MakePointer> const > const > const , Eigen::GpuDevice>, int>(float, int=2) 21.11% 30.856ms 1000 30.856us 29.120us 32.320us void paddle::operators::GatherScatterGPUKernel<float, long, paddle::operators::TensorAssign, bool=0>(float*, int, long const *, float, long, int, int, long, long, paddle::operators::TensorAssign const &) 21.10% 30.843ms 1000 30.843us 29.120us 31.712us void paddle::operators::GatherScatterGPUKernel<float, long, paddle::operators::ReduceAdd, bool=1>(float*, int, long const *, float, long, int, int, long, long, paddle::operators::ReduceAdd const &) 10.28% 15.029ms 1000 15.028us 14.688us 15.552us void Eigen::internal::EigenMetaKernel<Eigen::TensorEvaluator<Eigen::TensorAssignOp<Eigen::TensorMap<Eigen::Tensor<float, int=1, int=1, long>, int=0, Eigen::MakePointer>, Eigen::TensorCwiseNullaryOp<Eigen::internal::scalar_constant_op<float>, Eigen::TensorMap<Eigen::Tensor<float, int=1, int=1, long>, int=0, Eigen::MakePointer> const > const > const , Eigen::GpuDevice>, long>(float, int=1) 1.61% 2.3474ms 1000 2.3470us 2.1760us 2.8800us void Eigen::internal::EigenMetaKernel<Eigen::TensorEvaluator<Eigen::TensorAssignOp<Eigen::TensorMap<Eigen::Tensor<long, int=2, int=1, int>, int=16, Eigen::MakePointer>, Eigen::TensorBroadcastingOp<Eigen::DSizes<long, int=2> const , Eigen::TensorMap<Eigen::Tensor<long const , int=2, int=1, int>, int=16, Eigen::MakePointer> const > const > const , Eigen::GpuDevice>, int>(long, int=2) total gpu_time: 146.1282 ms run command: /root/miniconda3/bin/python /home/paddle/benchmark/api/dynamic_tests_v2/take_along_axis.py --task speed --framework paddle --testing_mode dynamic --json_file /home/paddle/benchmark/api/tests_v2/configs/take_along_axis.json --config_id 1 --profiler none --backward True --use_gpu True --repeat 1000 --allow_adaptive_repeat False --log_level 0 --gpu_time 146.12820512820514 grep: warning: GREP_OPTIONS is deprecated; please use an alias or script W0304 08:45:14.841684 706 gpu_context.cc:240] Please NOTE: device: 0, GPU Compute Capability: 7.0, Driver API Version: 10.1, Runtime API Version: 10.1 W0304 08:45:14.846143 706 gpu_context.cc:268] device: 0, cuDNN Version: 7.6. Cannot import pytorch, maybe paddle is not installed. Cannot import pytorch, maybe pytorch is not installed. Namespace(allow_adaptive_repeat=False, api_name=None, backward=True, config_id=1, convert_to_fp16=False, filename=None, framework='paddle', get_status_without_running=False, gpu_time=146.12820512820514, json_file='/home/paddle/benchmark/api/tests_v2/configs/take_along_axis.json', log_level=0, profiler='none', repeat=1000, scheduling_times='{}', sync_interval=80, task='speed', testing_mode='dynamic', unknown_dim=16, use_gpu=True) ---- Initialize APIConfig from /home/paddle/benchmark/api/tests_v2/configs/take_along_axis.json, config_id = 1. Namespace(allow_adaptive_repeat=False, api_name=None, backward=True, config_id=1, convert_to_fp16=False, filename=None, framework='paddle', get_status_without_running=False, gpu_time=146.12820512820514, json_file='/home/paddle/benchmark/api/tests_v2/configs/take_along_axis.json', log_level=0, profiler='none', repeat=1000, scheduling_times='{}', sync_interval=80, task='speed', testing_mode='dynamic', unknown_dim=16, use_gpu=True) [paddle][take_along_axis] take_along_axis { run_tf: True run_torch: False axis: 0 arr_shape: [1024, 3072] arr_dtype: float32 indices_shape: [1, 1] indices_dtype: int atol: 1e-06 } {"framework": "paddle", "version": "0.0.0", "name": "take_along_axis", "device": "GPU", "backward": true, "speed": {"repeat": 1000, "begin": 10, "end": 990, "total": 0.2995177191130969, "wall_time": 0, "total_include_wall_time": 0.2995177191130969, "gpu_time": 0.14612820512820515}, "parameters": "arr (Variable) - dtype: float32, shape: [1024, 3072]\nindices (Variable) - dtype: int, shape: [1, 1]\naxis (int): 0\n"}

---

• put_along_axis

(base) λ 48d12ee09b84 /home/paddle/benchmark/api/dynamic_tests_v2 {add_ops} bash run.sh put_along_axis 1 speed grep: warning: GREP_OPTIONS is deprecated; please use an alias or script grep: warning: GREP_OPTIONS is deprecated; please use an alias or script -- Current directory: /home/paddle/benchmark/api/dynamic_tests_v2 -- Entering /home/paddle/benchmark/api/common -- Current directory: /home/paddle/benchmark/api/common run command: git rev-parse HEAD run command: git show -s --format=%ad -- Current directory: /home/paddle/benchmark/api/dynamic_tests_v2 grep: warning: GREP_OPTIONS is deprecated; please use an alias or script =========================================================================== -- paddle version : 0.0.0 -- paddle commit : eaacf8bfee5c9583f7ebf0deff20b90db9d73478 -- benchmark commit : 69b086c4ed78b3425955e9f3d683453e94a65598 -- benchmark last update time : Tue Mar 1 17:20:00 2022 +0800 =========================================================================== run command: nvidia-smi -L run command: nvprof --profile-from-start off /root/miniconda3/bin/python /home/paddle/benchmark/api/dynamic_tests_v2/put_along_axis.py --task speed --framework paddle --testing_mode dynamic --json_file /home/paddle/benchmark/api/tests_v2/configs/put_along_axis.json --config_id 1 --profiler nvprof --backward True --use_gpu True --repeat 1000 --allow_adaptive_repeat False --log_level 0 Type Time(%) Time Calls Avg Min Max Name GPU activities: 49.11% 68.423ms 2000 34.211us 33.504us 35.680us [CUDA memcpy DtoD] 22.89% 31.888ms 1000 31.888us 31.615us 32.127us void paddle::operators::ScatterInputGradGPUKernel<float, long>(float*, int, long const *, long, int, int, long, long) 22.85% 31.835ms 1000 31.834us 31.616us 32.064us void paddle::operators::GatherScatterGPUKernel<float, long, paddle::operators::TensorAssign, bool=1>(float*, int, long const *, float, long, int, int, long, long, paddle::operators::TensorAssign const &) 2.15% 2.9977ms 1000 2.9970us 2.9440us 5.3440us void Eigen::internal::EigenMetaKernel<Eigen::TensorEvaluator<Eigen::TensorAssignOp<Eigen::TensorMap<Eigen::Tensor<float, int=2, int=1, int>, int=16, Eigen::MakePointer>, Eigen::TensorBroadcastingOp<Eigen::DSizes<long, int=2> const , Eigen::TensorMap<Eigen::Tensor<float const , int=2, int=1, int>, int=16, Eigen::MakePointer> const > const > const , Eigen::GpuDevice>, int>(float, int=2) 1.70% 2.3701ms 1000 2.3700us 2.2720us 2.8480us void Eigen::internal::EigenMetaKernel<Eigen::TensorEvaluator<Eigen::TensorAssignOp<Eigen::TensorMap<Eigen::Tensor<long, int=2, int=1, int>, int=16, Eigen::MakePointer>, Eigen::TensorBroadcastingOp<Eigen::DSizes<long, int=2> const , Eigen::TensorMap<Eigen::Tensor<long const , int=2, int=1, int>, int=16, Eigen::MakePointer> const > const > const , Eigen::GpuDevice>, int>(long, int=2) 1.30% 1.8160ms 1000 1.8160us 1.7600us 2.4640us [CUDA memcpy HtoD] total gpu_time: 139.3260 ms run command: /root/miniconda3/bin/python /home/paddle/benchmark/api/dynamic_tests_v2/put_along_axis.py --task speed --framework paddle --testing_mode dynamic --json_file /home/paddle/benchmark/api/tests_v2/configs/put_along_axis.json --config_id 1 --profiler none --backward True --use_gpu True --repeat 1000 --allow_adaptive_repeat False --log_level 0 --gpu_time 139.32600285074324 grep: warning: GREP_OPTIONS is deprecated; please use an alias or script W0304 08:46:06.412171 793 gpu_context.cc:240] Please NOTE: device: 0, GPU Compute Capability: 7.0, Driver API Version: 10.1, Runtime API Version: 10.1 W0304 08:46:06.416616 793 gpu_context.cc:268] device: 0, cuDNN Version: 7.6. Cannot import pytorch, maybe paddle is not installed. Cannot import pytorch, maybe pytorch is not installed. Namespace(allow_adaptive_repeat=False, api_name=None, backward=True, config_id=1, convert_to_fp16=False, filename=None, framework='paddle', get_status_without_running=False, gpu_time=139.32600285074324, json_file='/home/paddle/benchmark/api/tests_v2/configs/put_along_axis.json', log_level=0, profiler='none', repeat=1000, scheduling_times='{}', sync_interval=80, task='speed', testing_mode='dynamic', unknown_dim=16, use_gpu=True) ---- Initialize APIConfig from /home/paddle/benchmark/api/tests_v2/configs/put_along_axis.json, config_id = 1. Namespace(allow_adaptive_repeat=False, api_name=None, backward=True, config_id=1, convert_to_fp16=False, filename=None, framework='paddle', get_status_without_running=False, gpu_time=139.32600285074324, json_file='/home/paddle/benchmark/api/tests_v2/configs/put_along_axis.json', log_level=0, profiler='none', repeat=1000, scheduling_times='{}', sync_interval=80, task='speed', testing_mode='dynamic', unknown_dim=16, use_gpu=True) [paddle][put_along_axis] put_along_axis { run_tf: True run_torch: False values: 1000.0 axis: 0 arr_shape: [1024, 3072] arr_dtype: float32 indices_shape: [1, 1] indices_dtype: int atol: 1e-06 } {"framework": "paddle", "version": "0.0.0", "name": "put_along_axis", "device": "GPU", "backward": true, "speed": {"repeat": 1000, "begin": 10, "end": 990, "total": 0.4084915530924894, "wall_time": 0, "total_include_wall_time": 0.4084915530924894, "gpu_time": 0.13932600285074323}, "parameters": "arr (Variable) - dtype: float32, shape: [1024, 3072]\nindices (Variable) - dtype: int, shape: [1, 1]\naxis (int): 0\nvalues (float32): 1000.0\n"}

[ZzSean]

copyright年份

[ZzSean]

同上

[ZzSean]

同上

[ZzSean]

再加个一个config，稍微大一点的

[ZzSean]

LGTM