Skip to content

Commit c62d4d2

Browse files
committed
update docs
1 parent 571decb commit c62d4d2

File tree

1 file changed

+8
-16
lines changed

1 file changed

+8
-16
lines changed

README.md

Lines changed: 8 additions & 16 deletions
Original file line numberDiff line numberDiff line change
@@ -6,7 +6,7 @@
66

77
## Description
88

9-
**PURIFY** is an open-source collection of routines written in `C++` available under the [license](#license) below. It implements different tools and high-level to perform radio interferometric imaging, _i.e._ to recover images from the Fourier measurements taken by radio interferometric telescopes.
9+
**PURIFY** is an open-source collection of routines written in `C++` available under the [license](#license) below. It implements different tools and high-level to perform radio interferometric imaging, _i.e._ to recover images from the Fourier measurements taken by radio interferometric telescopes.
1010

1111
**PURIFY** leverages recent developments in the field of compressive sensing and convex optimization. Low-level functionality to solve the resulting convex optimisation is factored into the open-source companion code, [**SOPT**](https://github.com/astro-informatics/sopt), also written by the authors of **PURIFY**. For further background please see the [reference](#references-and-citation) section.
1212

@@ -86,7 +86,7 @@ C++ dependencies as well as the **PURIFY** installation:
8686
You can turn the various options on and off by adding flags to the `conan install` command, e.g.
8787

8888
```bash
89-
conan install .. -of . --build missing -o cppflow=on -o openmp=on -o mpi=off
89+
conan install .. -of . --build missing -o onnxrt=on -o openmp=on -o mpi=off
9090
conan build .. -of .
9191
```
9292

@@ -118,25 +118,17 @@ On MacOS, you can also install most of the dependencies with Homebrew e.g.
118118
```
119119

120120

121-
### TensorFlow support
121+
### Machine-learning models support
122122

123-
The **SOPT** library includes an interface to TensorFlow for using trained models
123+
The **SOPT** library includes an interface to ONNXrt for using trained models
124124
as priors in the Forward-Backward optimization algorithm. To build **PURIFY** with
125-
TensorFlow capability, some extra steps are currently required.
126-
We aim to simplify the build process in a future release.
125+
ONNXrt capability, you need to enable `ONNXrt` support also in **SOPT**:
127126

128-
1. Install the [TensorFlow C API](https://www.tensorflow.org/install/lang_c)
129-
1. Clone the UCL fork of `cppflow` and create a `conan` package using
130-
131-
``` bash
132-
git clone git@github.com:UCL/cppflow.git
133-
conan create /path/to/cppflow/
134-
```
135-
1. Follow the nominal build instructions, making sure you enable the `cppflow`
127+
1. Follow the nominal build instructions, making sure you enable the `onnxrt`
136128
option when building **SOPT** ...
137129

138130
```bash
139-
conan create /path/to/purify/sopt/ --build missing -o cppflow=on
131+
conan create /path/to/purify/sopt/ --build missing -o onnxrt=on
140132
```
141133

142134
1. ... and **PURIFY**:
@@ -145,7 +137,7 @@ We aim to simplify the build process in a future release.
145137
cd /path/to/purify
146138
mkdir build
147139
cd build
148-
conan install .. --build missing -o cppflow=on
140+
conan install .. --build missing -o onnrt=on
149141
conan build ..
150142
```
151143

0 commit comments

Comments
 (0)