change all recipes

examples/aishell/asr0/conf/deepspeech2.yaml

@@ -1,68 +1,64 @@
 # https://yaml.org/type/float.html
-data:
- train_manifest: data/manifest.train
- dev_manifest: data/manifest.dev
- test_manifest: data/manifest.test
- min_input_len: 0.0
- max_input_len: 27.0 # second
- min_output_len: 0.0
- max_output_len: .inf
- min_output_input_ratio: 0.00
- max_output_input_ratio: .inf
+###########################################
+# Data #
+###########################################
+train_manifest: data/manifest.train
+dev_manifest: data/manifest.dev
+test_manifest: data/manifest.test
+min_input_len: 0.0
+max_input_len: 27.0 # second
+min_output_len: 0.0
+max_output_len: .inf
+min_output_input_ratio: 0.00
+max_output_input_ratio: .inf
 
-collator:
- batch_size: 64 # one gpu
- mean_std_filepath: data/mean_std.json
- unit_type: char
- vocab_filepath: data/lang_char/vocab.txt 
- augmentation_config: conf/augmentation.json
- random_seed: 0
- spm_model_prefix: 
- spectrum_type: linear
- feat_dim: 
- delta_delta: False
- stride_ms: 10.0
- window_ms: 20.0
- n_fft: None
- max_freq: None
- target_sample_rate: 16000
- use_dB_normalization: True
- target_dB: -20
- dither: 1.0
- keep_transcription_text: False
- sortagrad: True
- shuffle_method: batch_shuffle
- num_workers: 2
+###########################################
+# Dataloader #
+###########################################
+batch_size: 64 # one gpu
+mean_std_filepath: data/mean_std.json
+unit_type: char
+vocab_filepath: data/lang_char/vocab.txt 
+augmentation_config: conf/augmentation.json
+random_seed: 0
+spm_model_prefix: 
+spectrum_type: linear
+feat_dim: 
+delta_delta: False
+stride_ms: 10.0
+window_ms: 20.0
+n_fft: None
+max_freq: None
+target_sample_rate: 16000
+use_dB_normalization: True
+target_dB: -20
+dither: 1.0
+keep_transcription_text: False
+sortagrad: True
+shuffle_method: batch_shuffle
+num_workers: 2
 
-model:
- num_conv_layers: 2
- num_rnn_layers: 3
- rnn_layer_size: 1024
- use_gru: True 
- share_rnn_weights: False
- blank_id: 0
- ctc_grad_norm_type: instance 
+############################################
+# Network Architecture #
+############################################
+num_conv_layers: 2
+num_rnn_layers: 3
+rnn_layer_size: 1024
+use_gru: True 
+share_rnn_weights: False
+blank_id: 0
+ctc_grad_norm_type: instance 
 
-training:
- n_epoch: 80
- accum_grad: 1
- lr: 2e-3
- lr_decay: 0.83
- weight_decay: 1e-06
- global_grad_clip: 3.0
- log_interval: 100
- checkpoint:
- kbest_n: 50
- latest_n: 5
-
-decoding:
- batch_size: 128
- error_rate_type: cer 
- decoding_method: ctc_beam_search
- lang_model_path: data/lm/zh_giga.no_cna_cmn.prune01244.klm
- alpha: 1.9
- beta: 5.0
- beam_size: 300
- cutoff_prob: 0.99
- cutoff_top_n: 40
- num_proc_bsearch: 10
+###########################################
+# Training #
+###########################################
+n_epoch: 80
+accum_grad: 1
+lr: 2e-3
+lr_decay: 0.83
+weight_decay: 1e-06
+global_grad_clip: 3.0
+log_interval: 100
+checkpoint:
+ kbest_n: 50
+ latest_n: 5

examples/aishell/asr0/conf/deepspeech2_online.yaml

@@ -1,70 +1,68 @@
 # https://yaml.org/type/float.html
-data:
- train_manifest: data/manifest.train
- dev_manifest: data/manifest.dev
- test_manifest: data/manifest.test
- min_input_len: 0.0
- max_input_len: 27.0 # second
- min_output_len: 0.0
- max_output_len: .inf
- min_output_input_ratio: 0.00
- max_output_input_ratio: .inf
+###########################################
+# Data #
+###########################################
+train_manifest: data/manifest.train
+dev_manifest: data/manifest.dev
+test_manifest: data/manifest.test
+min_input_len: 0.0
+max_input_len: 27.0 # second
+min_output_len: 0.0
+max_output_len: .inf
+min_output_input_ratio: 0.00
+max_output_input_ratio: .inf
 
-collator:
- batch_size: 64 # one gpu
- mean_std_filepath: data/mean_std.json
- unit_type: char
- vocab_filepath: data/lang_char/vocab.txt 
- augmentation_config: conf/augmentation.json
- random_seed: 0
- spm_model_prefix: 
- spectrum_type: linear #linear, mfcc, fbank
- feat_dim: 
- delta_delta: False
- stride_ms: 10.0
- window_ms: 20.0
- n_fft: None
- max_freq: None
- target_sample_rate: 16000
- use_dB_normalization: True
- target_dB: -20
- dither: 1.0
- keep_transcription_text: False
- sortagrad: True
- shuffle_method: batch_shuffle
- num_workers: 0
+###########################################
+# Dataloader #
+###########################################
+batch_size: 64 # one gpu
+mean_std_filepath: data/mean_std.json
+unit_type: char
+vocab_filepath: data/lang_char/vocab.txt 
+augmentation_config: conf/augmentation.json
+random_seed: 0
+spm_model_prefix: 
+spectrum_type: linear #linear, mfcc, fbank
+feat_dim: 
+delta_delta: False
+stride_ms: 10.0
+window_ms: 20.0
+n_fft: None
+max_freq: None
+target_sample_rate: 16000
+use_dB_normalization: True
+target_dB: -20
+dither: 1.0
+keep_transcription_text: False
+sortagrad: True
+shuffle_method: batch_shuffle
+num_workers: 0
 
-model:
- num_conv_layers: 2
- num_rnn_layers: 5
- rnn_layer_size: 1024
- rnn_direction: forward # [forward, bidirect]
- num_fc_layers: 0
- fc_layers_size_list: -1,
- use_gru: False 
- blank_id: 0
+############################################
+# Network Architecture #
+############################################
+num_conv_layers: 2
+num_rnn_layers: 5
+rnn_layer_size: 1024
+rnn_direction: forward # [forward, bidirect]
+num_fc_layers: 0
+fc_layers_size_list: -1,
+use_gru: False 
+blank_id: 0
 
 
-training:
- n_epoch: 65
- accum_grad: 1
- lr: 5e-4
- lr_decay: 0.93
- weight_decay: 1e-06
- global_grad_clip: 3.0
- log_interval: 100
- checkpoint:
- kbest_n: 50
- latest_n: 5
+###########################################
+# Training #
+###########################################
+n_epoch: 65
+accum_grad: 1
+lr: 5e-4
+lr_decay: 0.93
+weight_decay: 1e-06
+global_grad_clip: 3.0
+log_interval: 100
+checkpoint:
+ kbest_n: 50
+ latest_n: 5
+
 
-decoding:
- batch_size: 32
- error_rate_type: cer 
- decoding_method: ctc_beam_search
- lang_model_path: data/lm/zh_giga.no_cna_cmn.prune01244.klm
- alpha: 2.2 #1.9
- beta: 4.3
- beam_size: 300
- cutoff_prob: 0.99
- cutoff_top_n: 40
- num_proc_bsearch: 10

examples/aishell/asr0/conf/tuning/chunk_decode.yaml

@@ -0,0 +1,10 @@
+chunk_batch_size: 32
+error_rate_type: cer 
+decoding_method: ctc_beam_search
+lang_model_path: data/lm/zh_giga.no_cna_cmn.prune01244.klm
+alpha: 2.2 #1.9
+beta: 4.3
+beam_size: 300
+cutoff_prob: 0.99
+cutoff_top_n: 40
+num_proc_bsearch: 10

examples/aishell/asr0/conf/tuning/decode.yaml

@@ -0,0 +1,10 @@
+decode_batch_size: 128
+error_rate_type: cer 
+decoding_method: ctc_beam_search
+lang_model_path: data/lm/zh_giga.no_cna_cmn.prune01244.klm
+alpha: 1.9
+beta: 5.0
+beam_size: 300
+cutoff_prob: 0.99
+cutoff_top_n: 40
+num_proc_bsearch: 10

examples/aishell/asr0/local/test.sh

@@ -1,16 +1,17 @@
 #!/bin/bash
 
-if [ $# != 3 ];then
- echo "usage: ${0} config_path ckpt_path_prefix model_type"
+if [ $# != 4 ];then
+ echo "usage: ${0} config_path decode_config_path ckpt_path_prefix model_type"
  exit -1
 fi
 
 ngpu=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
 echo "using $ngpu gpus..."
 
 config_path=$1
-ckpt_prefix=$2
-model_type=$3
+decode_config_path=$2
+ckpt_prefix=$3
+model_type=$4
 
 # download language model
 bash local/download_lm_ch.sh
@@ -21,6 +22,7 @@ fi
 python3 -u ${BIN_DIR}/test.py \
 --ngpu ${ngpu} \
 --config ${config_path} \
+--decode_cfg ${decode_config_path} \
 --result_file ${ckpt_prefix}.rsl \
 --checkpoint_path ${ckpt_prefix} \
 --model_type ${model_type}

examples/aishell/asr0/local/test_export.sh

@@ -1,16 +1,17 @@
 #!/bin/bash
 
-if [ $# != 3 ];then
- echo "usage: ${0} config_path ckpt_path_prefix model_type"
+if [ $# != 4 ];then
+ echo "usage: ${0} config_path decode_config_path ckpt_path_prefix model_type"
  exit -1
 fi
 
 ngpu=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
 echo "using $ngpu gpus..."
 
 config_path=$1
-jit_model_export_path=$2
-model_type=$3
+decode_config_path=$2
+jit_model_export_path=$3
+model_type=$4
 
 # download language model
 bash local/download_lm_ch.sh > /dev/null 2>&1
@@ -21,6 +22,7 @@ fi
 python3 -u ${BIN_DIR}/test_export.py \
 --ngpu ${ngpu} \
 --config ${config_path} \
+--decode_cfg ${decode_config_path} \
 --result_file ${jit_model_export_path}.rsl \
 --export_path ${jit_model_export_path} \
 --model_type ${model_type}

examples/aishell/asr0/local/test_hub_ori

@@ -0,0 +1,47 @@
+#!/bin/bash
+
+if [ $# != 4 ];then
+ echo "usage: ${0} config_path ckpt_path_prefix model_type audio_file"
+ exit -1
+fi
+
+ngpu=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
+echo "using $ngpu gpus..."
+
+config_path=$1
+ckpt_prefix=$2
+model_type=$3
+audio_file=$4
+
+mkdir -p data
+wget -nc https://paddlespeech.bj.bcebos.com/datasets/single_wav/zh/demo_01_03.wav -P data/
+if [ $? -ne 0 ]; then
+ exit 1
+fi
+
+if [ ! -f ${audio_file} ]; then
+ echo "Plase input the right audio_file path"
+ exit 1
+fi
+
+# download language model
+bash local/download_lm_ch.sh
+if [ $? -ne 0 ]; then
+ exit 1
+fi
+
+python3 -u ${BIN_DIR}/test_hub.py \
+--nproc ${ngpu} \
+--config ${config_path} \
+--result_file ${ckpt_prefix}.rsl \
+--checkpoint_path ${ckpt_prefix} \
+--model_type ${model_type} \
+--audio_file ${audio_file}
+
+if [ $? -ne 0 ]; then
+ echo "Failed in evaluation!"
+ exit 1
+fi
+
+
+exit 0