提交 140d328d 编写于 作者: 刘琦

Merge branch 'add-0.9.0-release-note' into 'master'

Add release note of v0.9.0.

See merge request !675
Release Notes
=====
v0.6.0 (2018-04-04)
# v0.9.0 (2018-07-20)
------
1. Change mace header interfaces, only including necessary methods.
## Improvements
1. New work flow and documents.
2. Separate the model library from MACE library.
3. Reduce the size of static and dynamic library.
4. Support `ArgMax` Operations.
5. Support `Deconvolution` of Caffe.
6. Support NDK-17b.
v0.6.2 (2018-05-17)
------
* Return status instead of abort when allocate failed
## Incompatible Changes
1. Use file to store OpenCL tuned parameters and Add `SetOpenCLParameterPath` API.
## New APIs
1. Add a new `MaceEngine::Init` API with model data file.
## Bug Fixed
1. Not unmap the model data file when load model from files with CPU runtime.
2. 2D LWS tuning does not work.
3. Winograd convolution of GPU failed when open tuning.
4. Incorrect dynamic library of host.
v0.6.3 (2018-05-21)
## Acknowledgement
Appreciate for the following guys contribute code to make MACE better.
Zero King(@l2dy), James Bie(@JamesBie), Sun Aries(@SunAriesCN), Allen(@allen0125),
conansherry(@conansherry), 黎明灰烬(@jackwish)
# v0.8.0 (2018-05-31)
------
1. support `float` `data_type` when running in GPU
1. Change build and run tools
2. Handle runtime failure
v0.7.0 (2018-05-18)
# v0.7.0 (2018-05-18)
------
1. Change interface that report error type
2. Improve CPU performance
3. Merge CPU/GPU engine to on
v0.8.0 (2018-05-31)
# v0.6.3 (2018-05-21)
------
1. Change build and run tools
2. Handle runtime failure
1. support `float` `data_type` when running in GPU
# v0.6.2 (2018-05-17)
------
* Return status instead of abort when allocate failed
# v0.6.0 (2018-04-04)
------
1. Change mace header interfaces, only including necessary methods.
......@@ -65,4 +65,4 @@ Optional dependencies
.. note::
- For Android build, `ANDROID_NDK_HOME` must be confifigured by using ``export ANDROID_NDK_HOME=/path/to/ndk``
- It will link ``libc++`` instead of ``libgnustl`` if ``NDK version >= r17b`` and ``bazel version >= 0.13.0``
- It will link ``libc++`` instead of ``gnustl`` if ``NDK version >= r17b`` and ``bazel version >= 0.13.0``, please refer to `NDK cpp-support <https://developer.android.com/ndk/guides/cpp-support>`__.
......@@ -204,8 +204,9 @@ The generated model files will be stored in ``build/${library_name}/model`` fold
=============================
4. Build MACE into a library
=============================
You could Download the prebuilt MACE Library from `Github MACE release page <https://github.com/XiaoMi/mace/releases>`__.
Use bazel to build MACE source code into a library.
Or use bazel to build MACE source code into a library.
.. code:: sh
......@@ -259,9 +260,26 @@ to run and validate your model.
6. Deploy your model into applications
=======================================
You could run model on CPU, GPU and DSP (based on the `runtime` in your model deployment file).
However, there are some differences in different devices.
* **CPU**
Almost all of mobile SoCs use ARM-based CPU architecture, so your model could run on different SoCs in theory.
* **GPU**
Although most GPUs use OpenCL standard, but there are some SoCs not fully complying with the standard,
or the GPU is too low-level to use. So you should have some fallback strategies when the GPU run failed.
* **DSP**
MACE only support Qualcomm DSP.
In the converting and building steps, you've got the static/shared library, model files and
header files.
``${library_name}`` is the name you defined in the first line of your deployment YAML file.
.. note::
......@@ -313,12 +331,7 @@ Please refer to \ ``mace/examples/example.cc``\ for full usage. The following li
#include "mace/public/mace.h"
#include "mace/public/mace_runtime.h"
// 0. Set pre-compiled OpenCL binary program file paths when available
if (device_type == DeviceType::GPU) {
mace::SetOpenCLBinaryPaths(opencl_binary_paths);
}
// 1. Set compiled OpenCL kernel cache, this is used to reduce the
// 0. Set compiled OpenCL kernel cache, this is used to reduce the
// initialization time since the compiling is too slow. It's suggested
// to set this even when pre-compiled OpenCL program file is provided
// because the OpenCL version upgrade may also leads to kernel
......@@ -328,14 +341,14 @@ Please refer to \ ``mace/examples/example.cc``\ for full usage. The following li
new FileStorageFactory(file_path));
ConfigKVStorageFactory(storage_factory);
// 2. Declare the device type (must be same with ``runtime`` in configuration file)
// 1. Declare the device type (must be same with ``runtime`` in configuration file)
DeviceType device_type = DeviceType::GPU;
// 3. Define the input and output tensor names.
// 2. Define the input and output tensor names.
std::vector<std::string> input_names = {...};
std::vector<std::string> output_names = {...};
// 4. Create MaceEngine instance
// 3. Create MaceEngine instance
std::shared_ptr<mace::MaceEngine> engine;
MaceStatus create_engine_status;
......@@ -348,10 +361,10 @@ Please refer to \ ``mace/examples/example.cc``\ for full usage. The following li
device_type,
&engine);
if (create_engine_status != MaceStatus::MACE_SUCCESS) {
// Report error
// fall back to other strategy.
}
// 5. Create Input and Output tensor buffers
// 4. Create Input and Output tensor buffers
std::map<std::string, mace::MaceTensor> inputs;
std::map<std::string, mace::MaceTensor> outputs;
for (size_t i = 0; i < input_count; ++i) {
......@@ -376,7 +389,7 @@ Please refer to \ ``mace/examples/example.cc``\ for full usage. The following li
outputs[output_names[i]] = mace::MaceTensor(output_shapes[i], buffer_out);
}
// 6. Run the model
// 5. Run the model
MaceStatus status = engine.Run(inputs, &outputs);
More details are in :doc:`advanced_usage`.
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册