msb-public
/
PaddleSpeech
mirror of https://github.com/PaddlePaddle/PaddleSpeech
1
0
Code Issues Projects Releases Wiki Activity

format code

Browse Source
pull/2524/head
Hui Zhang 3 years ago
parent 616fc4594b
commit 56a0a02452
16 changed files with 53 additions and 55 deletions
Show Stats Download Patch File Download Diff File

4
speechx/speechx/base/basic_types.h
Unescape View File

@ -14,10 +14,10 @@
#pragma once
#include "kaldi/base/kaldi-types.h"
#include <limits>
#include "kaldi/base/kaldi-types.h"
typedef float BaseFloat;
typedef double double64;

1
speechx/speechx/base/common.h
Unescape View File

@ -47,6 +47,5 @@
#include "base/flags.h"
#include "base/log.h"
#include "base/macros.h"
#include "utils/file_utils.h"
#include "utils/math.h"

1
speechx/speechx/decoder/ctc_prefix_beam_search_decoder.h
Unescape View File

@ -17,7 +17,6 @@
#include "decoder/ctc_beam_search_opt.h"
#include "decoder/ctc_prefix_beam_search_score.h"
#include "decoder/decoder_itf.h"
#include "fst/symbol-table.h"
namespace ppspeech {

11
speechx/speechx/decoder/ctc_tlg_decoder.h
Unescape View File

@ -16,7 +16,6 @@
#include "base/common.h"
#include "decoder/decoder_itf.h"
#include "kaldi/decoder/lattice-faster-online-decoder.h"
#include "util/parse-options.h"
@ -35,7 +34,7 @@ struct TLGDecoderOptions {
std::string word_symbol_table{};
std::string fst_path{};
static TLGDecoderOptions InitFromFlags(){
static TLGDecoderOptions InitFromFlags() {
TLGDecoderOptions decoder_opts;
decoder_opts.word_symbol_table = FLAGS_word_symbol_table;
decoder_opts.fst_path = FLAGS_graph_path;
@ -45,9 +44,11 @@ struct TLGDecoderOptions {
decoder_opts.opts.max_active = FLAGS_max_active;
decoder_opts.opts.beam = FLAGS_beam;
decoder_opts.opts.lattice_beam = FLAGS_lattice_beam;
LOG(INFO) << "LatticeFasterDecoder max active: " << decoder_opts.opts.max_active ;
LOG(INFO) << "LatticeFasterDecoder beam: " << decoder_opts.opts.beam ;
LOG(INFO) << "LatticeFasterDecoder lattice_beam: " << decoder_opts.opts.lattice_beam ;
LOG(INFO) << "LatticeFasterDecoder max active: "
<< decoder_opts.opts.max_active;
LOG(INFO) << "LatticeFasterDecoder beam: " << decoder_opts.opts.beam;
LOG(INFO) << "LatticeFasterDecoder lattice_beam: "
<< decoder_opts.opts.lattice_beam;
return decoder_opts;
}

3
speechx/speechx/decoder/param.h
Unescape View File

@ -30,7 +30,7 @@ DEFINE_int32(receptive_field_length,
7,
"receptive field of two CNN(kernel=3) downsampling module.");
DEFINE_int32(subsampling_rate,
4,
4,
"two CNN(kernel=3) module downsampling rate.");
DEFINE_int32(nnet_decoder_chunk, 1, "paddle nnet forward chunk");
@ -62,7 +62,6 @@ DEFINE_double(beam, 15.0, "decoder beam");
DEFINE_double(lattice_beam, 7.5, "decoder beam");
// DecodeOptions flags
// DEFINE_int32(chunk_size, -1, "decoding chunk size");
DEFINE_int32(num_left_chunks, -1, "left chunks in decoding");

12
speechx/speechx/decoder/recognizer.h
Unescape View File

@ -32,15 +32,15 @@ struct RecognizerResource {
ModelOptions model_opts{};
TLGDecoderOptions tlg_opts{};
// CTCBeamSearchOptions beam_search_opts;
static RecognizerResource InitFromFlags(){
static RecognizerResource InitFromFlags() {
RecognizerResource resource;
resource.acoustic_scale = FLAGS_acoustic_scale;
resource.feature_pipeline_opts = FeaturePipelineOptions::InitFromFlags();
resource.feature_pipeline_opts =
FeaturePipelineOptions::InitFromFlags();
resource.model_opts = ModelOptions::InitFromFlags();
resource.tlg_opts = TLGDecoderOptions::InitFromFlags();
return resource;
resource.tlg_opts = TLGDecoderOptions::InitFromFlags();
return resource;
}
};

26
speechx/speechx/decoder/u2_recognizer.h
Unescape View File

@ -21,10 +21,9 @@
#include "decoder/ctc_prefix_beam_search_decoder.h"
#include "decoder/decoder_itf.h"
#include "frontend/audio/feature_pipeline.h"
#include "nnet/decodable.h"
#include "fst/fstlib.h"
#include "fst/symbol-table.h"
#include "nnet/decodable.h"
DECLARE_int32(nnet_decoder_chunk);
DECLARE_int32(num_left_chunks);
@ -63,9 +62,9 @@ struct DecodeOptions {
// CtcEndpointConfig ctc_endpoint_opts;
CTCBeamSearchOptions ctc_prefix_search_opts{};
static DecodeOptions InitFromFlags(){
static DecodeOptions InitFromFlags() {
DecodeOptions decoder_opts;
decoder_opts.chunk_size=FLAGS_nnet_decoder_chunk;
decoder_opts.chunk_size = FLAGS_nnet_decoder_chunk;
decoder_opts.num_left_chunks = FLAGS_num_left_chunks;
decoder_opts.ctc_weight = FLAGS_ctc_weight;
decoder_opts.rescoring_weight = FLAGS_rescoring_weight;
@ -86,15 +85,16 @@ struct U2RecognizerResource {
DecodeOptions decoder_opts{};
static U2RecognizerResource InitFromFlags() {
U2RecognizerResource resource;
resource.vocab_path = FLAGS_vocab_path;
resource.acoustic_scale = FLAGS_acoustic_scale;
resource.feature_pipeline_opts = ppspeech::FeaturePipelineOptions::InitFromFlags();
resource.model_opts = ppspeech::ModelOptions::InitFromFlags();
resource.decoder_opts = ppspeech::DecodeOptions::InitFromFlags();
return resource;
}
U2RecognizerResource resource;
resource.vocab_path = FLAGS_vocab_path;
resource.acoustic_scale = FLAGS_acoustic_scale;
resource.feature_pipeline_opts =
ppspeech::FeaturePipelineOptions::InitFromFlags();
resource.model_opts = ppspeech::ModelOptions::InitFromFlags();
resource.decoder_opts = ppspeech::DecodeOptions::InitFromFlags();
return resource;
}
};

2
speechx/speechx/frontend/audio/data_cache.h
Unescape View File

@ -56,4 +56,4 @@ class DataCache : public FrontendInterface {
DISALLOW_COPY_AND_ASSIGN(DataCache);
};
}
} // namespace ppspeech

22
speechx/speechx/frontend/audio/feature_pipeline.h
Unescape View File

@ -46,17 +46,17 @@ struct FeaturePipelineOptions {
FeatureCacheOptions feature_cache_opts{};
AssemblerOptions assembler_opts{};
static FeaturePipelineOptions InitFromFlags(){
static FeaturePipelineOptions InitFromFlags() {
FeaturePipelineOptions opts;
opts.cmvn_file = FLAGS_cmvn_file;
LOG(INFO) << "cmvn file: " << opts.cmvn_file;
LOG(INFO) << "cmvn file: " << opts.cmvn_file;
// frame options
kaldi::FrameExtractionOptions frame_opts;
frame_opts.dither = 0.0;
LOG(INFO) << "dither: " << frame_opts.dither;
LOG(INFO) << "dither: " << frame_opts.dither;
frame_opts.frame_shift_ms = 10;
LOG(INFO) << "frame shift ms: " << frame_opts.frame_shift_ms;
LOG(INFO) << "frame shift ms: " << frame_opts.frame_shift_ms;
opts.use_fbank = FLAGS_use_fbank;
LOG(INFO) << "feature type: " << (opts.use_fbank ? "fbank" : "linear");
if (opts.use_fbank) {
@ -76,15 +76,19 @@ struct FeaturePipelineOptions {
opts.linear_spectrogram_opts.frame_opts = frame_opts;
}
LOG(INFO) << "frame length ms: " << frame_opts.frame_length_ms;
LOG(INFO) << "frame length ms: " << frame_opts.frame_length_ms;
// assembler opts
opts.assembler_opts.subsampling_rate = FLAGS_subsampling_rate;
LOG(INFO) << "subsampling rate: " << opts.assembler_opts.subsampling_rate;
opts.assembler_opts.receptive_filed_length = FLAGS_receptive_field_length;
LOG(INFO) << "nnet receptive filed length: " << opts.assembler_opts.receptive_filed_length;
LOG(INFO) << "subsampling rate: "
<< opts.assembler_opts.subsampling_rate;
opts.assembler_opts.receptive_filed_length =
FLAGS_receptive_field_length;
LOG(INFO) << "nnet receptive filed length: "
<< opts.assembler_opts.receptive_filed_length;
opts.assembler_opts.nnet_decoder_chunk = FLAGS_nnet_decoder_chunk;
LOG(INFO) << "nnet chunk size: " << opts.assembler_opts.nnet_decoder_chunk;
LOG(INFO) << "nnet chunk size: "
<< opts.assembler_opts.nnet_decoder_chunk;
return opts;
}
};

1
speechx/speechx/frontend/audio/mfcc.h
Unescape View File

@ -14,7 +14,6 @@
#pragma once
#include "kaldi/feat/feature-mfcc.h"
#include "kaldi/feat/feature-mfcc.h"
#include "kaldi/matrix/kaldi-vector.h"

1
speechx/speechx/nnet/ds2_nnet.h
Unescape View File

@ -13,6 +13,7 @@
// limitations under the License.
#pragma once
#include <numeric>
#include "base/common.h"
#include "kaldi/matrix/kaldi-matrix.h"
#include "nnet/nnet_itf.h"

16
speechx/speechx/nnet/nnet_itf.h
Unescape View File

@ -48,25 +48,25 @@ struct ModelOptions {
bool enable_fc_padding{false};
bool enable_profile{false};
static ModelOptions InitFromFlags(){
static ModelOptions InitFromFlags() {
ModelOptions opts;
opts.subsample_rate = FLAGS_subsampling_rate;
LOG(INFO) << "subsampling rate: " << opts.subsample_rate;
LOG(INFO) << "subsampling rate: " << opts.subsample_rate;
opts.model_path = FLAGS_model_path;
LOG(INFO) << "model path: " << opts.model_path ;
LOG(INFO) << "model path: " << opts.model_path;
opts.param_path = FLAGS_param_path;
LOG(INFO) << "param path: " << opts.param_path ;
LOG(INFO) << "param path: " << opts.param_path;
LOG(INFO) << "DS2 param: ";
opts.cache_names = FLAGS_model_cache_names;
LOG(INFO) << " cache names: " << opts.cache_names;
LOG(INFO) << " cache names: " << opts.cache_names;
opts.cache_shape = FLAGS_model_cache_shapes;
LOG(INFO) << " cache shape: " << opts.cache_shape;
LOG(INFO) << " cache shape: " << opts.cache_shape;
opts.input_names = FLAGS_model_input_names;
LOG(INFO) << " input names: " << opts.input_names;
LOG(INFO) << " input names: " << opts.input_names;
opts.output_names = FLAGS_model_output_names;
LOG(INFO) << " output names: " << opts.output_names;
LOG(INFO) << " output names: " << opts.output_names;
return opts;
}
};

1
speechx/speechx/nnet/u2_nnet.h
Unescape View File

@ -16,7 +16,6 @@
#include "base/common.h"
#include "kaldi/matrix/kaldi-matrix.h"
#include "nnet/nnet_itf.h"
#include "paddle/extension.h"
#include "paddle/jit/all.h"

3
speechx/speechx/protocol/websocket/websocket_client.h
Unescape View File

@ -13,7 +13,6 @@
// limitations under the License.
#include "base/common.h"
#include "boost/asio/connect.hpp"
#include "boost/asio/ip/tcp.hpp"
#include "boost/beast/core.hpp"
@ -54,4 +53,4 @@ class WebSocketClient {
websocket::stream<tcp::socket> ws_{ioc_};
std::unique_ptr<std::thread> t_{nullptr};
};
}
} // namespace ppspeech

2
speechx/speechx/protocol/websocket/websocket_server.h
Unescape View File

@ -15,12 +15,10 @@
#pragma once
#include "base/common.h"
#include "boost/asio/connect.hpp"
#include "boost/asio/ip/tcp.hpp"
#include "boost/beast/core.hpp"
#include "boost/beast/websocket.hpp"
#include "decoder/recognizer.h"
#include "frontend/audio/feature_pipeline.h"

2
speechx/speechx/utils/file_utils.h
Unescape View File

@ -20,4 +20,4 @@ bool ReadFileToVector(const std::string& filename,
std::vector<std::string>* data);
std::string ReadFile2String(const std::string& path);
}
} // namespace ppspeech

Write Preview
Loading…
Cancel
Save