File size: 2,671 Bytes
d596074 |
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 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 |
Export to ONNX
==============
In this section, we describe how to export models to `ONNX`_.
.. hint::
Before you continue, please run:
.. code-block:: bash
pip install onnx
In each recipe, there is a file called ``export-onnx.py``, which is used
to export trained models to `ONNX`_.
There is also a file named ``onnx_pretrained.py``, which you can use
the exported `ONNX`_ model in Python with `onnxruntime`_ to decode sound files.
sherpa-onnx
-----------
We have a separate repository `sherpa-onnx`_ for deploying your exported models
on various platforms such as:
- iOS
- Android
- Raspberry Pi
- Linux/macOS/Windows
Please see the documentation of `sherpa-onnx`_ for details:
`<https://k2-fsa.github.io/sherpa/onnx/index.html>`_
Example
-------
In the following, we demonstrate how to export a streaming Zipformer pre-trained
model from
`<https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless7-2022-11-11>`_
to `ONNX`_.
Download the pre-trained model
------------------------------
.. hint::
We assume you have installed `git-lfs`_.
.. code-block:: bash
cd egs/librispeech/ASR
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
pushd $repo
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "exp/pretrained.pt"
cd exp
ln -s pretrained.pt epoch-99.pt
popd
Export the model to ONNX
------------------------
.. code-block:: bash
./pruned_transducer_stateless7_streaming/export-onnx.py \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--use-averaged-model 0 \
--epoch 99 \
--avg 1 \
--decode-chunk-len 32 \
--exp-dir $repo/exp/
.. warning::
``export-onnx.py`` from different recipes has different options.
In the above example, ``--decode-chunk-len`` is specific for the
streaming Zipformer. Other models won't have such an option.
It will generate the following 3 files in ``$repo/exp``
- ``encoder-epoch-99-avg-1.onnx``
- ``decoder-epoch-99-avg-1.onnx``
- ``joiner-epoch-99-avg-1.onnx``
Decode sound files with exported ONNX models
--------------------------------------------
.. code-block:: bash
./pruned_transducer_stateless7_streaming/onnx_pretrained.py \
--encoder-model-filename $repo/exp/encoder-epoch-99-avg-1.onnx \
--decoder-model-filename $repo/exp/decoder-epoch-99-avg-1.onnx \
--joiner-model-filename $repo/exp/joiner-epoch-99-avg-1.onnx \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav
|