mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-08-09 01:52:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update doc for exporting to ncnn (#1072)

Browse Source
This commit is contained in:
Fangjun Kuang 2023-05-19 16:22:08 +08:00 committed by GitHub
parent ae1949ddcc
commit a5bbfc6f7e
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 6 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
docs/source/model-export/export-ncnn-zipformer.rst
View File

@ -276,7 +276,7 @@ The result looks like below:
7767517 7767517
2029 2547 2029 2547
SherpaMetaData sherpa_meta_data1 0 0 0=2 1=32 2=4 3=7 -23316=5,2,4,3,2,4 -23317=5,384,384,384,384,384 -23318=5,192,192,192,192,192 -23319=5,1,2,4,8,2 -23320=5,31,31,31,31,31 SherpaMetaData sherpa_meta_data1 0 0 0=2 1=32 2=4 3=7 15=1 -23316=5,2,4,3,2,4 -23317=5,384,384,384,384,384 -23318=5,192,192,192,192,192 -23319=5,1,2,4,8,2 -23320=5,31,31,31,31,31
Input in0 0 1 in0 Input in0 0 1 in0
**Explanation** **Explanation**
@ -300,6 +300,9 @@ The result looks like below:
- ``3=7``, 3 is the key and 7 is the value of for the amount of padding - ``3=7``, 3 is the key and 7 is the value of for the amount of padding
used in the Conv2DSubsampling layer. It should be 7 for zipformer used in the Conv2DSubsampling layer. It should be 7 for zipformer
if you don't change zipformer.py. if you don't change zipformer.py.
- ``15=1``, attribute 15, this is the model version. Starting from
`sherpa-ncnn`_ v2.0, we require that the model version has to
be >= 1.
- ``-23316=5,2,4,3,2,4``, attribute 16, this is an array attribute. - ``-23316=5,2,4,3,2,4``, attribute 16, this is an array attribute.
It is attribute 16 since -23300 - (-23316) = 16. It is attribute 16 since -23300 - (-23316) = 16.
The first element of the array is the length of the array, which is 5 in our case. The first element of the array is the length of the array, which is 5 in our case.
@ -338,6 +341,8 @@ The result looks like below:
+----------+--------------------------------------------+ +----------+--------------------------------------------+
| 3 | 7 (if you don't change code) | | 3 | 7 (if you don't change code) |
+----------+--------------------------------------------+ +----------+--------------------------------------------+
| 15 | 1 (The model version) |
+----------+--------------------------------------------+
|-23316 | ``--num-encoder-layer`` | |-23316 | ``--num-encoder-layer`` |
+----------+--------------------------------------------+ +----------+--------------------------------------------+
|-23317 | ``--encoder-dims`` | |-23317 | ``--encoder-dims`` |