experiments/STT
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Merge pull request #2074 from mozilla/bindings-idiomatic

Browse Source 
Some bindings clean-up
This commit is contained in:
Reuben Morais 2019-04-25 13:20:50 +00:00 committed by GitHub
commit 656ab5734a
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
8 changed files with 108 additions and 107 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
native_client/README.md
View File

@ -100,7 +100,7 @@ After following the above build and installation instructions, the Node.JS bindi
``` ```
cd native_client/javascript cd native_client/javascript
make package make build
make npm-pack make npm-pack
``` ```

3
native_client/javascript/client.js
View File

@ -1,5 +1,4 @@
#!/usr/bin/env node #!/usr/bin/env node
'use strict'; 'use strict';
const Fs = require('fs'); const Fs = require('fs');
@ -69,6 +68,7 @@ function metadataToString(metadata) {
for (var i = 0; i < metadata.num_items; ++i) { for (var i = 0; i < metadata.num_items; ++i) {
retval += metadata.items[i].character; retval += metadata.items[i].character;
} }
Ds.FreeMetadata(metadata);
return retval; return retval;
} }
@ -135,5 +135,6 @@ audioStream.on('finish', () => {
} }
const inference_stop = process.hrtime(inference_start); const inference_stop = process.hrtime(inference_start);
console.error('Inference took %ds for %ds audio file.', totalTime(inference_stop), audioLength.toPrecision(4)); console.error('Inference took %ds for %ds audio file.', totalTime(inference_stop), audioLength.toPrecision(4));
Ds.DestroyModel(model);
process.exit(0); process.exit(0);
}); });

15
native_client/javascript/deepspeech.i
View File

@ -28,7 +28,6 @@ using namespace node;
// make sure the string returned by SpeechToText is freed // make sure the string returned by SpeechToText is freed
%typemap(newfree) char* "DS_FreeString($1);"; %typemap(newfree) char* "DS_FreeString($1);";
%typemap(newfree) Metadata* "DS_FreeMetadata($1);";
%newobject DS_SpeechToText; %newobject DS_SpeechToText;
%newobject DS_IntermediateDecode; %newobject DS_IntermediateDecode;
@ -61,19 +60,8 @@ using namespace node;
%append_output(SWIG_NewPointerObj(%as_voidptr(*$1), $*1_descriptor, 0)); %append_output(SWIG_NewPointerObj(%as_voidptr(*$1), $*1_descriptor, 0));
} }
// extend ModelState with a destructor so that DestroyModel will be called
// when the JavaScript object gets finalized.
%nodefaultctor ModelState; %nodefaultctor ModelState;
%nodefaultdtor ModelState; %nodefaultdtor ModelState;
struct ModelState {};
%extend ModelState {
~ModelState() {
DS_DestroyModel($self);
}
}
%nodefaultdtor Metadata; %nodefaultdtor Metadata;
%nodefaultctor Metadata; %nodefaultctor Metadata;
%nodefaultctor MetadataItem; %nodefaultctor MetadataItem;
@ -94,9 +82,6 @@ struct ModelState {};
v8::Handle<v8::Value> result = SWIGV8_ARRAY_NEW(); v8::Handle<v8::Value> result = SWIGV8_ARRAY_NEW();
return result; return result;
} }
~Metadata() {
DS_FreeMetadata($self);
}
} }
%rename ("%(strip:[DS_])s") ""; %rename ("%(strip:[DS_])s") "";

8
native_client/javascript/index.js
View File

@ -75,7 +75,13 @@ Model.prototype.finishStreamWithMetadata = function() {
return binding.FinishStreamWithMetadata.apply(null, arguments); return binding.FinishStreamWithMetadata.apply(null, arguments);
} }
function DestroyModel(model) {
return binding.DestroyModel(model._impl);
}
module.exports = { module.exports = {
Model: Model, Model: Model,
printVersions: binding.PrintVersions printVersions: binding.PrintVersions,
DestroyModel: DestroyModel,
FreeMetadata: binding.FreeMetadata
}; };

8
native_client/python/client.py
View File

@ -39,6 +39,7 @@ N_FEATURES = 26
# Size of the context window used for producing timesteps in the input vector # Size of the context window used for producing timesteps in the input vector
N_CONTEXT = 9 N_CONTEXT = 9
def convert_samplerate(audio_path): def convert_samplerate(audio_path):
sox_cmd = 'sox {} --type raw --bits 16 --channels 1 --rate 16000 --encoding signed-integer --endian little --compression 0.0 --no-dither - '.format(quote(audio_path)) sox_cmd = 'sox {} --type raw --bits 16 --channels 1 --rate 16000 --encoding signed-integer --endian little --compression 0.0 --no-dither - '.format(quote(audio_path))
try: try:
@ -50,11 +51,9 @@ def convert_samplerate(audio_path):
return 16000, np.frombuffer(output, np.int16) return 16000, np.frombuffer(output, np.int16)
def metadata_to_string(metadata): def metadata_to_string(metadata):
retval = '' return ''.join(item.character for item in metadata.items)
for item in range(metadata.num_items):
retval += metadata.items[item].character
return retval
class VersionAction(argparse.Action): class VersionAction(argparse.Action):
@ -65,6 +64,7 @@ class VersionAction(argparse.Action):
printVersions() printVersions()
exit(0) exit(0)
def main(): def main():
parser = argparse.ArgumentParser(description='Running DeepSpeech inference.') parser = argparse.ArgumentParser(description='Running DeepSpeech inference.')
parser.add_argument('--model', required=True, parser.add_argument('--model', required=True,

14
native_client/python/impl.i
View File

@ -33,17 +33,19 @@ import_array();
%append_output(SWIG_NewPointerObj(%as_voidptr(*$1), $*1_descriptor, 0)); %append_output(SWIG_NewPointerObj(%as_voidptr(*$1), $*1_descriptor, 0));
} }
%extend struct MetadataItem {
MetadataItem* __getitem__(size_t i) {
return &$self[i];
}
}
%typemap(out) Metadata* { %typemap(out) Metadata* {
// owned, extended destructor needs to be called by SWIG // owned, extended destructor needs to be called by SWIG
%append_output(SWIG_NewPointerObj(%as_voidptr($1), $1_descriptor, SWIG_POINTER_OWN)); %append_output(SWIG_NewPointerObj(%as_voidptr($1), $1_descriptor, SWIG_POINTER_OWN));
} }
%typemap(out) MetadataItem* %{
$result = PyList_New(arg1->num_items);
for (int i = 0; i < arg1->num_items; ++i) {
PyObject* o = SWIG_NewPointerObj(SWIG_as_voidptr(&arg1->items[i]), SWIGTYPE_p_MetadataItem, 0);
PyList_SetItem($result, i, o);
}
%}
%extend struct Metadata { %extend struct Metadata {
~Metadata() { ~Metadata() {
DS_FreeMetadata($self); DS_FreeMetadata($self);

1
native_client/python/setup.cfg
View File

@ -1,6 +1,5 @@
[build_ext] [build_ext]
include-dirs=./ include-dirs=./
swig-opts=-c++ -keyword
build-lib=temp_build build-lib=temp_build
build-temp=temp_build build-temp=temp_build

164
native_client/python/setup.py
View File

@ -7,93 +7,101 @@ import os
import subprocess import subprocess
import sys import sys
try: def main():
import numpy
try: try:
numpy_include = numpy.get_include() import numpy
except AttributeError: try:
numpy_include = numpy.get_numpy_include() numpy_include = numpy.get_include()
except ImportError: except AttributeError:
numpy_include = '' numpy_include = numpy.get_numpy_include()
assert 'NUMPY_INCLUDE' in os.environ except ImportError:
numpy_include = ''
assert 'NUMPY_INCLUDE' in os.environ
def read(fname): def read(fname):
return open(os.path.join(os.path.dirname(__file__), fname)).read() return open(os.path.join(os.path.dirname(__file__), fname)).read()
numpy_include = os.getenv('NUMPY_INCLUDE', numpy_include) numpy_include = os.getenv('NUMPY_INCLUDE', numpy_include)
numpy_min_ver = os.getenv('NUMPY_DEP_VERSION', '') numpy_min_ver = os.getenv('NUMPY_DEP_VERSION', '')
project_name = 'deepspeech' project_name = 'deepspeech'
if '--project_name' in sys.argv: if '--project_name' in sys.argv:
project_name_idx = sys.argv.index('--project_name') project_name_idx = sys.argv.index('--project_name')
project_name = sys.argv[project_name_idx + 1] project_name = sys.argv[project_name_idx + 1]
sys.argv.remove('--project_name') sys.argv.remove('--project_name')
sys.argv.pop(project_name_idx) sys.argv.pop(project_name_idx)
with open('../../VERSION', 'r') as ver: with open('../../VERSION', 'r') as ver:
project_version = ver.read().strip() project_version = ver.read().strip()
class BuildExtFirst(build): class BuildExtFirst(build):
sub_commands = [('build_ext', build.has_ext_modules), sub_commands = [('build_ext', build.has_ext_modules),
('build_py', build.has_pure_modules), ('build_py', build.has_pure_modules),
('build_clib', build.has_c_libraries), ('build_clib', build.has_c_libraries),
('build_scripts', build.has_scripts)] ('build_scripts', build.has_scripts)]
# Properly pass arguments for linking, setuptools will perform some checks # Properly pass arguments for linking, setuptools will perform some checks
def lib_dirs_split(a): def lib_dirs_split(a):
if os.name == 'posix': if os.name == 'posix':
return a.split('-L')[1:] return a.split('-L')[1:]
if os.name == 'nt': if os.name == 'nt':
return [] return []
def libs_split(a): raise AssertionError('os.name == java not expected')
if os.name == 'posix':
return a.split('-l')[1:]
if os.name == 'nt': def libs_split(a):
return a.split('.lib')[0:1] if os.name == 'posix':
return a.split('-l')[1:]
ds_ext = Extension('deepspeech._impl', if os.name == 'nt':
['impl.i'], return a.split('.lib')[0:1]
include_dirs = [ numpy_include, '../' ],
library_dirs = list(map(lambda x: x.strip(), lib_dirs_split(os.getenv('MODEL_LDFLAGS', '')))),
libraries = list(map(lambda x: x.strip(), libs_split(os.getenv('MODEL_LIBS', ''))))
)
setup(name = project_name, raise AssertionError('os.name == java not expected')
description = 'A library for running inference on a DeepSpeech model',
long_description = read('../../README.md'), ds_ext = Extension(name='deepspeech._impl',
long_description_content_type = 'text/markdown; charset=UTF-8', sources=['impl.i'],
author = 'Mozilla', include_dirs=[numpy_include, '../'],
version = project_version, library_dirs=list(map(lambda x: x.strip(), lib_dirs_split(os.getenv('MODEL_LDFLAGS', '')))),
package_dir = {'deepspeech': '.'}, libraries=list(map(lambda x: x.strip(), libs_split(os.getenv('MODEL_LIBS', '')))),
cmdclass = {'build': BuildExtFirst}, swig_opts=['-c++', '-keyword', '-builtin'])
license = 'MPL-2.0',
url = 'https://github.com/mozilla/DeepSpeech', setup(name=project_name,
project_urls = { description='A library for running inference on a DeepSpeech model',
'Documentation': 'https://github.com/mozilla/DeepSpeech/tree/v{}#project-deepspeech'.format(project_version), long_description=read('../../README.md'),
'Tracker': 'https://github.com/mozilla/DeepSpeech/issues', long_description_content_type='text/markdown; charset=UTF-8',
'Repository': 'https://github.com/mozilla/DeepSpeech/tree/v{}'.format(project_version), author='Mozilla',
'Discussions': 'https://discourse.mozilla.org/c/deep-speech', version=project_version,
}, package_dir={'deepspeech': '.'},
ext_modules = [ds_ext], cmdclass={'build': BuildExtFirst},
py_modules = ['deepspeech', 'deepspeech.client', 'deepspeech.impl'], license='MPL-2.0',
entry_points={'console_scripts':['deepspeech = deepspeech.client:main']}, url='https://github.com/mozilla/DeepSpeech',
install_requires = ['numpy%s' % numpy_min_ver], project_urls={
include_package_data = True, 'Documentation': 'https://github.com/mozilla/DeepSpeech/tree/v{}#project-deepspeech'.format(project_version),
classifiers = [ 'Tracker': 'https://github.com/mozilla/DeepSpeech/issues',
'Development Status :: 3 - Alpha', 'Repository': 'https://github.com/mozilla/DeepSpeech/tree/v{}'.format(project_version),
'Environment :: Console', 'Discussions': 'https://discourse.mozilla.org/c/deep-speech',
'Intended Audience :: Developers', },
'Intended Audience :: Science/Research', ext_modules=[ds_ext],
'License :: OSI Approved :: Mozilla Public License 2.0 (MPL 2.0)', py_modules=['deepspeech', 'deepspeech.client', 'deepspeech.impl'],
'Programming Language :: Python :: 2.7', entry_points={'console_scripts':['deepspeech=deepspeech.client:main']},
'Programming Language :: Python :: 3.4', install_requires=['numpy%s' % numpy_min_ver],
'Programming Language :: Python :: 3.5', include_package_data=True,
'Programming Language :: Python :: 3.6', classifiers=[
'Topic :: Multimedia :: Sound/Audio :: Speech', 'Development Status :: 3 - Alpha',
'Topic :: Scientific/Engineering :: Human Machine Interfaces', 'Environment :: Console',
'Topic :: Scientific/Engineering', 'Intended Audience :: Developers',
'Topic :: Utilities', 'Intended Audience :: Science/Research',
]) 'License :: OSI Approved :: Mozilla Public License 2.0 (MPL 2.0)',
'Programming Language :: Python :: 2.7',
'Programming Language :: Python :: 3.4',
'Programming Language :: Python :: 3.5',
'Programming Language :: Python :: 3.6',
'Topic :: Multimedia :: Sound/Audio :: Speech',
'Topic :: Scientific/Engineering :: Human Machine Interfaces',
'Topic :: Scientific/Engineering',
'Topic :: Utilities',
])
if __name__ == '__main__':
main()