blob: add1159e1eb3e7a127fc233e67c3f9a23220fdf8 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
|
# moco-tf
_moco-tf_ translates a TensorFlow model into _loco_
## Purpose
_moco-tf_ is to convert TensorFlow generated model file to in-memory _loco_ IR Graph.
## How to use
```cxx
#include <moco/tf/Frontend.h>
...
::moco::tf::Frontend moco;
std::string pb_path = "path_to_pb_file_to_load";
auto loco_graph = moco.load(sig, pb_path, ::moco::tf::Frontend::FileType::Binary);
```
## Dependency
Please refer [requires.cmake](./requires.cmake) for dependant modules.
## Naming rules
### TensorFlow node names
Use `REGISTER_OP` argument used in TensorFlow source `core` folder.
```
cd tensorflow/core
grep -Rn "REGISTER_OP"
```
To see single Op, `Conv2D` for example
```
cd tensorflow/core
grep -Rn "REGISTER_OP" | grep "Conv2D"
```
### Names related with TensorFlow nodes
Like `GraphBuilder` and `Canonicalization`, TensorFlow node names can be used as
prefix or suffix.
- `Conv2DGraphBuilder`
- `Conv2DCanonicalizier`
### TensorFlow Dialect IR
Use `TF` prefix with TensorFlow Dialect node names
- `TFAvgPool`
- `TFConv2D`
|
