m53297601
/
Paddle
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from 'd96ee24f0b'
${ noResults }
Paddle/paddle/fluid/inference/api
History
Tao Luo 1e1974c998
Merge pull request #12563 from luotao1/anakin_test 		7 years ago
..
demo_ci Modify style (#12465) 7 years ago
CMakeLists.txt modify the anakin_model download dir 7 years ago
README.md move contrib/inference to paddle/fluid/inference/api 7 years ago
api.cc fix inference double free bug (#12613) 7 years ago
api_anakin_engine.cc Improve anakin feature (#11961) 7 years ago
api_anakin_engine.h Improve anakin feature (#11961) 7 years ago
api_anakin_engine_tester.cc Improve anakin feature (#11961) 7 years ago
api_impl.cc fea/lightly support lod (#12451) 7 years ago
api_impl.h bugfix/tensorrt analysis fix subgraph trigger (#12266) 7 years ago
api_impl_tester.cc inference-api code clean (#12274) 7 years ago
api_tensorrt_subgraph_engine.cc inference analyzer as bin (#12450) 7 years ago
api_tensorrt_subgraph_engine_tester.cc fea/lightly support lod (#12451) 7 years ago
api_tester.cc inference-api code clean (#12274) 7 years ago
high_level_api.md fix dead link in high_level_api.md 7 years ago
high_level_api_cn.md fix dead link in high_level_api.md 7 years ago
paddle_inference_api.h hide misc APIs (#12540) 7 years ago

README.md

Embed Paddle Inference in Your Application

Paddle inference offers the APIs in C and C++ languages.

One can easily deploy a model trained by Paddle following the steps as below:

  1. Optimize the native model;
  2. Write some codes for deployment.

Let's explain the steps in detail.

Optimize the native Fluid Model

The native model that get from the training phase needs to be optimized for that.

  • Clean the noise such as the cost operators that do not need inference;
  • Prune unnecessary computation fork that has nothing to do with the output;
  • Remove extraneous variables;
  • Memory reuse for native Fluid executor;
  • Translate the model storage format to some third-party engine's, so that the inference API can utilize the engine for acceleration;

We have an official tool to do the optimization, call paddle_inference_optimize --help for more information.

Write some codes

Read paddle_inference_api.h for more information.