Comments (7)
openppl-public
commented on August 16, 2024
comand:
./pplnn-build/tools/pplnn --onnx-model tests/testdata/conv.onnx --enable-profiling --warmuptimes 3
I tried this command and it worked correctly in my machine. Cound you delete the pplnn-build directory and re-run the build command ./build.sh and try the command above again?
from ppl.nn.
ltj2013
commented on August 16, 2024
You can enable the profiling flag in CMakeLists.
https://github.com/openppl-public/ppl.nn/blob/513e6122dd90f619da0e550a8a3dc5e3184e5c58/CMakeLists.txt#L11
from ppl.nn.
codgeek
commented on August 16, 2024
You can enable the profiling flag in CMakeLists.
https://github.com/openppl-public/ppl.nn/blob/513e6122dd90f619da0e550a8a3dc5e3184e5c58/CMakeLists.txt#L11
seems not work by this flag. this test model conv.onnx still got stuck.
-options="-DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=${pplnn_build_dir}/install $*"
+options="-DCMAKE_BUILD_TYPE=Release -DPPLNN_ENABLE_KERNEL_PROFILING=ON -DCMAKE_INSTALL_PREFIX=${pplnn_build_dir}/install $*"
tests/testdata/conv.onnx --enable-profiling --warmuptimes 3
[INFO][2021-07-05 22:19:38.555][pplnn.cc:700] ppl.nn version: 513e6122dd90f619da0e550a8a3dc5e3184e5c58-dirty
[INFO][2021-07-05 22:19:38.555][pplnn.cc:127] ***** register X86Engine *****
[INFO][2021-07-05 22:19:38.555][simple_graph_partitioner.cc:107] total partition(s) of graph[torch-jit-export]: 1.
[INFO][2021-07-05 22:19:38.555][pplnn.cc:540] ----- input info -----
[INFO][2021-07-05 22:19:38.555][pplnn.cc:543] input[0]:
[INFO][2021-07-05 22:19:38.555][pplnn.cc:544] name: input
[INFO][2021-07-05 22:19:38.555][pplnn.cc:551] dim(s): 1 3 4 4
[INFO][2021-07-05 22:19:38.555][pplnn.cc:553] DataType: FLOAT32
[INFO][2021-07-05 22:19:38.555][pplnn.cc:554] DataFormat: NDARRAY
[INFO][2021-07-05 22:19:38.555][pplnn.cc:555] NumBytesIncludePadding: 192
[INFO][2021-07-05 22:19:38.555][pplnn.cc:556] NumBytesExcludePadding: 192
[INFO][2021-07-05 22:19:38.555][pplnn.cc:559] ----- output info -----
[INFO][2021-07-05 22:19:38.555][pplnn.cc:562] output[0]:
[INFO][2021-07-05 22:19:38.556][pplnn.cc:563] name: 5
[INFO][2021-07-05 22:19:38.556][pplnn.cc:570] dim(s): 1 3 5 5
[INFO][2021-07-05 22:19:38.556][pplnn.cc:572] DataType: FLOAT32
[INFO][2021-07-05 22:19:38.556][pplnn.cc:573] DataFormat: N16CX
[INFO][2021-07-05 22:19:38.556][pplnn.cc:574] NumBytesIncludePadding: 1600
[INFO][2021-07-05 22:19:38.556][pplnn.cc:575] NumBytesExcludePadding: 300
[INFO][2021-07-05 22:19:38.556][pplnn.cc:578] ----------------------
[INFO][2021-07-05 22:19:38.556][pplnn.cc:808] Run() costs: 0.010000 ms.
[INFO][2021-07-05 22:19:38.556][pplnn.cc:816] Run ok
[INFO][2021-07-05 22:19:38.556][pplnn.cc:820] Warm up start for 3 times.
[INFO][2021-07-05 22:19:38.556][pplnn.cc:827] Warm up end.
[INFO][2021-07-05 22:19:38.556][pplnn.cc:835] Profiling start
from ppl.nn.
codgeek
commented on August 16, 2024
comand:
./pplnn-build/tools/pplnn --onnx-model tests/testdata/conv.onnx --enable-profiling --warmuptimes 3I tried this command and it worked correctly in my machine. Cound you delete the
pplnn-builddirectory and re-run the build command./build.shand try the command above again?
I've tried, tests/testdata/conv.onnx still got stuck.
Occasionally i tried another mobileNetV2 model, Profiling run to the end. It seems the conv.onnx sample has something wrong
[INFO][2021-07-05 22:29:08.104][pplnn.cc:700] ppl.nn version: 513e6122dd90f619da0e550a8a3dc5e3184e5c58-dirty
[INFO][2021-07-05 22:29:08.104][pplnn.cc:127] ***** register X86Engine *****
[INFO][2021-07-05 22:29:08.116][simple_graph_partitioner.cc:107] total partition(s) of graph[torch-jit-export]: 1.
[INFO][2021-07-05 22:29:08.139][pplnn.cc:540] ----- input info -----
[INFO][2021-07-05 22:29:08.139][pplnn.cc:543] input[0]:
[INFO][2021-07-05 22:29:08.139][pplnn.cc:544] name: input
[INFO][2021-07-05 22:29:08.139][pplnn.cc:551] dim(s): 1 3 224 224
[INFO][2021-07-05 22:29:08.139][pplnn.cc:553] DataType: FLOAT32
[INFO][2021-07-05 22:29:08.139][pplnn.cc:554] DataFormat: NDARRAY
[INFO][2021-07-05 22:29:08.139][pplnn.cc:555] NumBytesIncludePadding: 602112
[INFO][2021-07-05 22:29:08.139][pplnn.cc:556] NumBytesExcludePadding: 602112
[INFO][2021-07-05 22:29:08.139][pplnn.cc:559] ----- output info -----
[INFO][2021-07-05 22:29:08.139][pplnn.cc:562] output[0]:
[INFO][2021-07-05 22:29:08.139][pplnn.cc:563] name: output
[INFO][2021-07-05 22:29:08.139][pplnn.cc:570] dim(s): 1 1000
[INFO][2021-07-05 22:29:08.139][pplnn.cc:572] DataType: FLOAT32
[INFO][2021-07-05 22:29:08.139][pplnn.cc:573] DataFormat: NDARRAY
[INFO][2021-07-05 22:29:08.139][pplnn.cc:574] NumBytesIncludePadding: 4000
[INFO][2021-07-05 22:29:08.139][pplnn.cc:575] NumBytesExcludePadding: 4000
[INFO][2021-07-05 22:29:08.139][pplnn.cc:578] ----------------------
[INFO][2021-07-05 22:29:08.139][pplnn.cc:808] Run() costs: 9.195000 ms.
[INFO][2021-07-05 22:29:08.139][pplnn.cc:816] Run ok
[INFO][2021-07-05 22:29:08.139][pplnn.cc:820] Warm up start for 3 times.
[INFO][2021-07-05 22:29:08.161][pplnn.cc:827] Warm up end.
[INFO][2021-07-05 22:29:08.161][pplnn.cc:835] Profiling start
[INFO][2021-07-05 22:29:09.164][pplnn.cc:851] Duration: 1002.473000 ms
[INFO][2021-07-05 22:29:09.164][pplnn.cc:861] Average run cost: 7.317321 ms.
[INFO][2021-07-05 22:29:09.164][pplnn.cc:864] Profiling End
from ppl.nn.
codgeek
commented on August 16, 2024
You can enable the profiling flag in CMakeLists.
https://github.com/openppl-public/ppl.nn/blob/513e6122dd90f619da0e550a8a3dc5e3184e5c58/CMakeLists.txt#L11
Actually, I tried and find this PPLNN_ENABLE_KERNEL_PROFILING macro is designed for exporting detailed timing for each op, other than the profiling of just run multi loops, the --enable-profiling is corresponding to the second kind.
from ppl.nn.
ltj2013
commented on August 16, 2024
You can enable the profiling flag in CMakeLists.
https://github.com/openppl-public/ppl.nn/blob/513e6122dd90f619da0e550a8a3dc5e3184e5c58/CMakeLists.txt#L11Actually, I tried and find this
PPLNN_ENABLE_KERNEL_PROFILINGmacro is designed for exporting detailed timing for each op, other than the profiling of just run multi loops, the--enable-profilingis corresponding to the second kind.
Yes, PPLNN_ENABLE_KERNEL_PROFILING macro is designed for exporting detailed timing for each op. And --enable-profiling is also used to print the kernel profiling in detail.
from ppl.nn.
Alcanderian
commented on August 16, 2024
comand:
./pplnn-build/tools/pplnn --onnx-model tests/testdata/conv.onnx --enable-profiling --warmuptimes 3I tried this command and it worked correctly in my machine. Cound you delete the
pplnn-builddirectory and re-run the build command./build.shand try the command above again?I've tried, tests/testdata/conv.onnx still got stuck.
Occasionally i tried another mobileNetV2 model, Profiling run to the end. It seems the conv.onnx sample has something wrong
[INFO][2021-07-05 22:29:08.104][pplnn.cc:700] ppl.nn version: 513e6122dd90f619da0e550a8a3dc5e3184e5c58-dirty [INFO][2021-07-05 22:29:08.104][pplnn.cc:127] ***** register X86Engine ***** [INFO][2021-07-05 22:29:08.116][simple_graph_partitioner.cc:107] total partition(s) of graph[torch-jit-export]: 1. [INFO][2021-07-05 22:29:08.139][pplnn.cc:540] ----- input info ----- [INFO][2021-07-05 22:29:08.139][pplnn.cc:543] input[0]: [INFO][2021-07-05 22:29:08.139][pplnn.cc:544] name: input [INFO][2021-07-05 22:29:08.139][pplnn.cc:551] dim(s): 1 3 224 224 [INFO][2021-07-05 22:29:08.139][pplnn.cc:553] DataType: FLOAT32 [INFO][2021-07-05 22:29:08.139][pplnn.cc:554] DataFormat: NDARRAY [INFO][2021-07-05 22:29:08.139][pplnn.cc:555] NumBytesIncludePadding: 602112 [INFO][2021-07-05 22:29:08.139][pplnn.cc:556] NumBytesExcludePadding: 602112 [INFO][2021-07-05 22:29:08.139][pplnn.cc:559] ----- output info ----- [INFO][2021-07-05 22:29:08.139][pplnn.cc:562] output[0]: [INFO][2021-07-05 22:29:08.139][pplnn.cc:563] name: output [INFO][2021-07-05 22:29:08.139][pplnn.cc:570] dim(s): 1 1000 [INFO][2021-07-05 22:29:08.139][pplnn.cc:572] DataType: FLOAT32 [INFO][2021-07-05 22:29:08.139][pplnn.cc:573] DataFormat: NDARRAY [INFO][2021-07-05 22:29:08.139][pplnn.cc:574] NumBytesIncludePadding: 4000 [INFO][2021-07-05 22:29:08.139][pplnn.cc:575] NumBytesExcludePadding: 4000 [INFO][2021-07-05 22:29:08.139][pplnn.cc:578] ---------------------- [INFO][2021-07-05 22:29:08.139][pplnn.cc:808] Run() costs: 9.195000 ms. [INFO][2021-07-05 22:29:08.139][pplnn.cc:816] Run ok [INFO][2021-07-05 22:29:08.139][pplnn.cc:820] Warm up start for 3 times. [INFO][2021-07-05 22:29:08.161][pplnn.cc:827] Warm up end. [INFO][2021-07-05 22:29:08.161][pplnn.cc:835] Profiling start [INFO][2021-07-05 22:29:09.164][pplnn.cc:851] Duration: 1002.473000 ms [INFO][2021-07-05 22:29:09.164][pplnn.cc:861] Average run cost: 7.317321 ms. [INFO][2021-07-05 22:29:09.164][pplnn.cc:864] Profiling End
I've tried simple conv on my machine. I found it will take about 30 seconds to finish the profiling because it take tooooooooooooooo little time in each run, lmao.
You will find the time is slowly growing if you print the run_dur at pplnn.cc:847, just be patient:)
from ppl.nn.
Related Issues (20)
- linux compile error protobuf static assertion failed HOT 3
- malloc_consolidate(): invalid chunk size HOT 2
- pplnn save-input 得到的NDARRAY的 shape不正确 HOT 1
- 如何使用cmake的将ppl.nn和依赖ppl.nn的代码一同编译？ HOT 3
- Segmentation fault at ppl::nn::x86::X86Kernel::DumpOutputTensors HOT 5
- 获取模型推理结果（GetOutputs)耗时长 HOT 2
- Install Error HOT 1
- The compilation passed, but an error was reported in test phase HOT 2
- Floating point exception (core dumped) ？ HOT 4
- 使用x86 engine运行resnet50 fp16 onnx模型 core dump
- (Ask) why InferInheritedType handle int8 to fp16 out? HOT 3
- Got wrong output shape when run a Gemm op(transB=0) use cuda HOT 4
- Crash with ONNX Split operator
- 关于全局engine，其他线程引用导致的性能下降问题 HOT 4
- 推理误差排查
- 多模型pipeline的示例
- ARM平台是否可以跑int8的推理
- cuda build error HOT 1
- When I run build.sh to compile the project, a compilation error occured. HOT 1
- build pplnn error
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from ppl.nn.