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

[asr] remove useless conf of librispeech (#1227)

Browse Source 
* remve useless conf

* format code

* update conf

* update conf

* update conf
pull/1236/head
Hui Zhang 4 years ago committed by GitHub
parent 425b085f94
commit bb2a370b23
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
8 changed files with 76 additions and 88 deletions
Show Stats Download Patch File Download Diff File

4
examples/csmsc/voc5/README.md
Unescape View File

@ -127,8 +127,8 @@ HiFiGAN checkpoint contains files listed below.
```text
hifigan_csmsc_ckpt_0.1.1
├── default.yaml # default config used to train hifigan
├── feats_stats.npy # generator parameters of hifigan
├── default.yaml # default config used to train hifigan
├── feats_stats.npy # generator parameters of hifigan
└── snapshot_iter_2500000.pdz # statistics used to normalize spectrogram when training hifigan
```

42
examples/librispeech/asr1/conf/chunk_conformer.yaml
Unescape View File

@ -47,36 +47,32 @@ data:
dev_manifest: data/manifest.dev
test_manifest: data/manifest.test
collator:
vocab_filepath: data/lang_char/vocab.txt
unit_type: 'spm'
spm_model_prefix: 'data/lang_char/bpe_unigram_5000'
mean_std_filepath: ""
augmentation_config: conf/preprocess.yaml
batch_size: 16
raw_wav: True # use raw_wav or kaldi feature
spectrum_type: fbank #linear, mfcc, fbank
feat_dim: 80
delta_delta: False
dither: 1.0
target_sample_rate: 16000
max_freq: None
n_fft: None
stride_ms: 10.0
window_ms: 25.0
use_dB_normalization: True
target_dB: -20
random_seed: 0
keep_transcription_text: False
sortagrad: True
shuffle_method: batch_shuffle
num_workers: 2
sortagrad: 0 # Feed samples from shortest to longest ; -1: enabled for all epochs, 0: disabled, other: enabled for 'other' epochs
batch_size: 16
maxlen_in: 512 # if input length > maxlen-in, batchsize is automatically reduced
maxlen_out: 150 # if output length > maxlen-out, batchsize is automatically reduced
minibatches: 0 # for debug
batch_count: auto
batch_bins: 0
batch_frames_in: 0
batch_frames_out: 0
batch_frames_inout: 0
augmentation_config: conf/preprocess.yaml
num_workers: 0
subsampling_factor: 1
num_encs: 1
training:
n_epoch: 240
n_epoch: 120
accum_grad: 8
global_grad_clip: 5.0
optim: adam
@ -86,24 +82,16 @@ training:
scheduler: warmuplr
scheduler_conf:
warmup_steps: 25000
lr_decay: 1.0
log_interval: 100
checkpoint:
kbest_n: 50
latest_n: 5
decoding:
batch_size: 128
error_rate_type: wer
decoding_method: attention # 'attention', 'ctc_greedy_search', 'ctc_prefix_beam_search', 'attention_rescoring'
lang_model_path: data/lm/common_crawl_00.prune01111.trie.klm
alpha: 2.5
beta: 0.3
beam_size: 10
cutoff_prob: 1.0
cutoff_top_n: 0
num_proc_bsearch: 8
ctc_weight: 0.5 # ctc weight for attention rescoring decode mode.
decoding_chunk_size: -1 # decoding chunk size. Defaults to -1.
# <0: for decoding, use full chunk.

33
examples/librispeech/asr1/conf/chunk_transformer.yaml
Unescape View File

@ -34,36 +34,35 @@ model:
lsm_weight: 0.1 # label smoothing option
length_normalized_loss: false
data:
train_manifest: data/manifest.train
dev_manifest: data/manifest.dev
test_manifest: data/manifest.test
collator:
vocab_filepath: data/lang_char/vocab.txt
unit_type: 'spm'
spm_model_prefix: 'data/lang_char/bpe_unigram_5000'
mean_std_filepath: ""
augmentation_config: conf/preprocess.yaml
batch_size: 64
raw_wav: True # use raw_wav or kaldi feature
spectrum_type: fbank #linear, mfcc, fbank
feat_dim: 80
delta_delta: False
dither: 1.0
target_sample_rate: 16000
max_freq: None
n_fft: None
stride_ms: 10.0
window_ms: 25.0
use_dB_normalization: True
target_dB: -20
random_seed: 0
keep_transcription_text: False
sortagrad: True
shuffle_method: batch_shuffle
num_workers: 2
sortagrad: 0 # Feed samples from shortest to longest ; -1: enabled for all epochs, 0: disabled, other: enabled for 'other' epochs
batch_size: 64
maxlen_in: 512 # if input length > maxlen-in, batchsize is automatically reduced
maxlen_out: 150 # if output length > maxlen-out, batchsize is automatically reduced
minibatches: 0 # for debug
batch_count: auto
batch_bins: 0
batch_frames_in: 0
batch_frames_out: 0
batch_frames_inout: 0
augmentation_config: conf/preprocess.yaml
num_workers: 0
subsampling_factor: 1
num_encs: 1
training:
@ -102,5 +101,3 @@ decoding:
# 0: used for training, it's prohibited here.
num_decoding_left_chunks: -1 # number of left chunks for decoding. Defaults to -1.
simulate_streaming: true # simulate streaming inference. Defaults to False.

30
examples/librispeech/asr1/conf/conformer.yaml
Unescape View File

@ -34,6 +34,7 @@ model:
# hybrid CTC/attention
model_conf:
ctc_weight: 0.3
ctc_grad_norm_type: null
lsm_weight: 0.1 # label smoothing option
length_normalized_loss: false
@ -50,24 +51,23 @@ collator:
spm_model_prefix: 'data/lang_char/bpe_unigram_5000'
mean_std_filepath: ""
augmentation_config: conf/preprocess.yaml
batch_size: 16
raw_wav: True # use raw_wav or kaldi feature
spectrum_type: fbank #linear, mfcc, fbank
feat_dim: 80
delta_delta: False
dither: 1.0
target_sample_rate: 16000
max_freq: None
n_fft: None
stride_ms: 10.0
window_ms: 25.0
use_dB_normalization: True
target_dB: -20
random_seed: 0
keep_transcription_text: False
sortagrad: True
shuffle_method: batch_shuffle
num_workers: 2
sortagrad: 0 # Feed samples from shortest to longest ; -1: enabled for all epochs, 0: disabled, other: enabled for 'other' epochs
batch_size: 16
maxlen_in: 512 # if input length > maxlen-in, batchsize is automatically reduced
maxlen_out: 150 # if output length > maxlen-out, batchsize is automatically reduced
minibatches: 0 # for debug
batch_count: auto
batch_bins: 0
batch_frames_in: 0
batch_frames_out: 0
batch_frames_inout: 0
augmentation_config: conf/preprocess.yaml
num_workers: 0
subsampling_factor: 1
num_encs: 1
training:

29
examples/librispeech/asr1/conf/transformer.yaml
Unescape View File

@ -51,24 +51,23 @@ collator:
spm_model_prefix: 'data/lang_char/bpe_unigram_5000'
mean_std_filepath: ""
augmentation_config: conf/preprocess.yaml
batch_size: 32
raw_wav: True # use raw_wav or kaldi feature
spectrum_type: fbank #linear, mfcc, fbank
feat_dim: 80
delta_delta: False
dither: 1.0
target_sample_rate: 16000
max_freq: None
n_fft: None
stride_ms: 10.0
window_ms: 25.0
use_dB_normalization: True
target_dB: -20
random_seed: 0
keep_transcription_text: False
sortagrad: True
shuffle_method: batch_shuffle
num_workers: 2
sortagrad: 0 # Feed samples from shortest to longest ; -1: enabled for all epochs, 0: disabled, other: enabled for 'other' epochs
batch_size: 32
maxlen_in: 512 # if input length > maxlen-in, batchsize is automatically reduced
maxlen_out: 150 # if output length > maxlen-out, batchsize is automatically reduced
minibatches: 0 # for debug
batch_count: auto
batch_bins: 0
batch_frames_in: 0
batch_frames_out: 0
batch_frames_inout: 0
augmentation_config: conf/preprocess.yaml
num_workers: 0
subsampling_factor: 1
num_encs: 1
training:

2
paddlespeech/t2s/frontend/zh_frontend.py
Unescape View File

@ -106,7 +106,7 @@ class Frontend():
for seg in segments:
phones = []
# Replace all English words in the sentence
seg = re.sub('[a-zA-Z]+','',seg)
seg = re.sub('[a-zA-Z]+', '', seg)
seg_cut = psg.lcut(seg)
initials = []
finals = []

10
paddlespeech/t2s/models/fastspeech2/fastspeech2.py
Unescape View File

@ -942,7 +942,12 @@ class StyleFastSpeech2Inference(FastSpeech2Inference):
"""
spk_id = paddle.to_tensor(spk_id)
normalized_mel, d_outs, p_outs, e_outs = self.acoustic_model.inference(
text, durations=None, pitch=None, energy=None, spk_emb=spk_emb, spk_id=spk_id)
text,
durations=None,
pitch=None,
energy=None,
spk_emb=spk_emb,
spk_id=spk_id)
# priority: groundtruth > scale/bias > previous output
# set durations
if isinstance(durations, np.ndarray):
@ -996,8 +1001,7 @@ class StyleFastSpeech2Inference(FastSpeech2Inference):
energy=energy,
use_teacher_forcing=True,
spk_emb=spk_emb,
spk_id=spk_id
)
spk_id=spk_id)
logmel = self.normalizer.inverse(normalized_mel)
return logmel

Write Preview
Loading…
Cancel
Save