├── README.md ├── TIMIT_preparation.py ├── cfg ├── SincNet_MIT.cfg ├── SincNet_TIMIT.cfg ├── SincNet_TIMIT_m005.cfg ├── SincNet_TIMIT_m010.cfg ├── SincNet_TIMIT_m015.cfg ├── SincNet_TIMIT_m020.cfg ├── SincNet_TIMIT_m025.cfg ├── SincNet_TIMIT_m030.cfg ├── SincNet_TIMIT_m035.cfg ├── SincNet_TIMIT_m040.cfg ├── SincNet_TIMIT_m045.cfg ├── SincNet_TIMIT_m050.cfg ├── SincNet_TIMIT_m055.cfg ├── SincNet_TIMIT_m060.cfg ├── SincNet_TIMIT_m065.cfg ├── SincNet_TIMIT_m070.cfg ├── SincNet_TIMIT_m075.cfg ├── SincNet_TIMIT_m080.cfg ├── SincNet_TIMIT_m085.cfg ├── SincNet_TIMIT_m090.cfg └── SincNet_TIMIT_m095.cfg ├── data_io.py ├── data_lists ├── TIMIT_all.scp ├── TIMIT_labels.npy ├── TIMIT_test.scp └── TIMIT_train.scp ├── dnn_models.py ├── exp ├── SincNet_TIMIT │ └── res.res ├── SincNet_TIMIT_m005 │ └── res.res ├── SincNet_TIMIT_m010 │ └── res.res ├── SincNet_TIMIT_m015 │ └── res.res ├── SincNet_TIMIT_m020 │ └── res.res ├── SincNet_TIMIT_m025 │ └── res.res ├── SincNet_TIMIT_m030 │ └── res.res ├── SincNet_TIMIT_m035 │ └── res.res ├── SincNet_TIMIT_m040 │ └── res.res ├── SincNet_TIMIT_m045 │ └── res.res ├── SincNet_TIMIT_m050 │ └── res.res ├── SincNet_TIMIT_m055 │ └── res.res ├── SincNet_TIMIT_m060 │ └── res.res ├── SincNet_TIMIT_m065 │ └── res.res ├── SincNet_TIMIT_m070 │ └── res.res ├── SincNet_TIMIT_m075 │ └── res.res ├── SincNet_TIMIT_m080 │ └── res.res ├── SincNet_TIMIT_m085 │ └── res.res ├── SincNet_TIMIT_m090 │ └── res.res └── SincNet_TIMIT_m095 │ └── res.res ├── requirements.txt └── speaker_id.py /README.md: -------------------------------------------------------------------------------- 1 | 2 | # Additive Margin SincNet (AM-SincNet) 3 | AM-SincNet is a new approach for speaker recognition problems which is based in the neural network architecture SincNet and the additive margin softmax (AM-Softmax) loss function. It uses the architecture of the SincNet, but with an improved AM-Softmax layer. 4 | 5 | This repository releases an example of code to perform a speaker recognition experiment on the TIMIT dataset. To run it with other datasets you can have a look at the instructions on the original SincNet repository (https://github.com/mravanelli/SincNet). 6 | 7 | We should thank [@mravanelli](https://github.com/mravanelli/) for the [SincNet implementation](https://github.com/mravanelli/SincNet). 8 | 9 | ## Requirements 10 | For running this experiment we used a Linux environment with Python 3.6. 11 | 12 | You can see a list of python dependencies at [requirements.txt](requirements.txt). 13 | 14 | To install it on conda virtual environment (`conda install --file requirements.txt`). 15 | 16 | To install it on pip virtual environment (`pip install -r requirements.txt`). 17 | 18 | ## How to Run 19 | To run it on TIMIT dataset we have first to pre-process the data, removing the start and ending silences moments and also normalizing the audio sentences. 20 | 21 | `` 22 | python TIMIT_preparation.py $TIMIT_FOLDER $OUTPUT_FOLDER data_lists/TIMIT_all.scp 23 | `` 24 | 25 | where: 26 | - *$TIMIT_FOLDER* is the folder of the original TIMIT corpus 27 | - *$OUTPUT_FOLDER* is the folder in which the normalized TIMIT will be stored 28 | - *data_lists/TIMIT_all.scp* is the list of the TIMIT files used for training/test the speaker id system. 29 | 30 | then, we can run the experiment itself by typing. 31 | 32 | `` 33 | python speaker_id.py --cfg=cfg/$CFG_FILE 34 | `` 35 | 36 | where: 37 | - *$CFG_FILE* is the name of the cfg configuration file which is located at cfg folder. 38 | 39 | We have made avaliable several cfg configuration files for the experiments, if you want to run the experiment with the traditional SincNet (with no use of the improved AM-Softmax layer) you must use the [*SincNet_TIMIT.cfg*](cfg/SincNet_TIMIT.cfg) file, otherwise you can use the [*SincNet_TIMIT_m0XX.cfg*](cfg/) file where the *XX* denotes the size of the margin parameter that will be used for the AM-Softmax layer. 40 | 41 | 42 | ## Results 43 | When training have a look at the *cfg* configuration file, the output paths for the model and the result (*res.res*) files are placed there. 44 | 45 | We have also made available some results from our experiments, you can check them at [*exp*](exp/) folder. The resume of the results are saved in the *res.res* files. 46 | 47 | 48 | ## How to use SincNet with a different dataset? 49 | In this repository, we used the TIMIT dataset as a tutorial to show how SincNet works. 50 | With the current version of the code, you can easily use a different corpus. To do it you should provide in input the corpora-specific input files (in wav format) and your own labels. You should thus modify the paths into the *.scp files you find in the data_lists folder. 51 | 52 | To assign to each sentence the right label, you also have to modify the dictionary "*TIMIT_labels.npy*". 53 | The labels are specified within a python dictionary that contains sentence ids as keys (e.g., "*si1027*") and speaker_ids as values. Each speaker_id is an integer, ranging from 0 to N_spks-1. In the TIMIT dataset, you can easily retrieve the speaker id from the path (e.g., *train/dr1/fcjf0/si1027.wav* is the sentence_id "*si1027*" uttered by the speaker "*fcjf0*"). For other datasets, you should be able to retrieve in such a way this dictionary containing pairs of speakers and sentence ids. 54 | 55 | You should then modify the config file (*cfg/SincNet_TIMIT.cfg*) according to your new paths. Remember also to change the field "*class_lay=462*" according to the number of speakers N_spks you have in your dataset. 56 | 57 | 58 | ## Cite us 59 | 60 | If you use this code or part of it, please cite us! 61 | 62 | ``` 63 | @INPROCEEDINGS{8852112, 64 | author={J. A. {Chagas Nunes} and D. {Macêdo} and C. {Zanchettin}}, 65 | booktitle={2019 International Joint Conference on Neural Networks (IJCNN)}, 66 | title={Additive Margin SincNet for Speaker Recognition}, 67 | year={2019}, 68 | volume={}, 69 | number={}, 70 | pages={1-5}, 71 | keywords={}, 72 | doi={10.1109/IJCNN.2019.8852112}, 73 | ISSN={}, 74 | month={July},} 75 | ``` 76 | 77 | You can also find the paper at [IEEE](https://ieeexplore.ieee.org/document/8852112) or the preprint at [arXiv](https://arxiv.org/abs/1901.10826). 78 | -------------------------------------------------------------------------------- /TIMIT_preparation.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python3 2 | 3 | # TIMIT_preparation 4 | # Mirco Ravanelli 5 | # Mila - University of Montreal 6 | 7 | # July 2018 8 | 9 | # Description: 10 | # This code prepares TIMIT for the following speaker identification experiments. 11 | # It removes start and end silences according to the information reported in the *.wrd files and normalizes the amplitude of each sentence. 12 | 13 | # How to run it: 14 | # python TIMIT_preparation.py $TIMIT_FOLDER $OUTPUT_FOLDER data_lists/TIMIT_all.scp 15 | 16 | import shutil 17 | import os 18 | import soundfile as sf 19 | import numpy as np 20 | import sys 21 | from os import listdir, makedirs, remove, rmdir 22 | from os.path import isfile, exists 23 | 24 | def ReadList(list_file): 25 | f=open(list_file,"r") 26 | lines=f.readlines() 27 | list_sig=[] 28 | for x in lines: 29 | list_sig.append(x.rstrip()) 30 | f.close() 31 | return list_sig 32 | 33 | def copy_folder(in_folder,out_folder): 34 | if not(os.path.isdir(out_folder)): 35 | shutil.copytree(in_folder, out_folder, ignore=ig_f) 36 | 37 | def ig_f(dir, files): 38 | return [f for f in files if os.path.isfile(os.path.join(dir, f))] 39 | 40 | def maybe_make_directory(dir_path): 41 | if not exists(dir_path): 42 | makedirs(dir_path) 43 | 44 | in_folder=sys.argv[1] 45 | out_folder=sys.argv[2] 46 | list_file=sys.argv[3] 47 | 48 | # Read List file 49 | list_sig=ReadList(list_file) 50 | 51 | # Replicate input folder structure to output folder 52 | copy_folder(in_folder,out_folder) 53 | 54 | 55 | # Speech Data Reverberation Loop 56 | for i in range(len(list_sig)): 57 | 58 | # Open the wav file 59 | wav_file=in_folder+'/'+list_sig[i] 60 | wav_file = wav_file.upper() 61 | 62 | [signal, fs] = sf.read(wav_file) 63 | signal=signal.astype(np.float64) 64 | 65 | # Signal normalization 66 | signal=signal/np.abs(np.max(signal)) 67 | 68 | # Read wrd file 69 | wrd_file=wav_file.replace(".WAV",".WRD") 70 | wrd_sig=ReadList(wrd_file) 71 | beg_sig=int(wrd_sig[0].split(' ')[0]) 72 | end_sig=int(wrd_sig[-1].split(' ')[1]) 73 | 74 | # Remove silences 75 | signal=signal[beg_sig:end_sig] 76 | 77 | 78 | # Save normalized speech 79 | file_out=out_folder+'/'+list_sig[i] 80 | 81 | file_out = file_out.lower() 82 | final_folder = file_out.split('/') 83 | final_folder = '/'.join(final_folder[:len(final_folder)-1]) + '/' 84 | maybe_make_directory(final_folder) 85 | sf.write(file_out, signal, fs) 86 | 87 | print("Done %s" % (file_out)) 88 | -------------------------------------------------------------------------------- /cfg/SincNet_MIT.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/MIT_train.scp 3 | te_lst=data_lists/MIT_test.scp 4 | lab_dict=data_lists/MIT_labels.npy 5 | data_folder=/home/joao/datasets/mit_processed 6 | output_folder=exp/SincNet_MIT2/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=48 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | begin_epochs=0 47 | N_epochs=361 48 | N_batches=800 49 | N_eval_epoch=8 50 | seed=1234 51 | AMSoftmax=False 52 | AMSoftmax_m = 0 -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=/home/joao/datasets/timit_processed 6 | output_folder=exp/SincNet_TIMIT/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=361 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=False 51 | AMSoftmax_m = 0.5 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m005.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=/home/joao/datasets/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m005/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=361 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.05 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m010.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=/home/joao/datasets/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m010/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=361 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.10 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m015.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=/home/joao/datasets/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m015/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=361 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.15 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m020.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=/home/joao/datasets/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m020/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=361 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.20 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m025.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=/home/joao/datasets/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m025/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=361 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.25 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m030.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=/home/joao/datasets/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m030/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=361 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.30 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m035.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=dataset/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m035/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=360 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.35 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m040.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=dataset/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m040/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=360 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.40 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m045.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=dataset/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m045/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=360 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.45 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m050.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=dataset/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m050/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=360 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.50 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m055.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=dataset/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m055/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=360 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.55 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m060.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=dataset/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m060/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=360 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.60 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m065.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=dataset/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m065/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=360 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.65 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m070.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=dataset/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m070/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=360 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.70 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m075.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=dataset/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m075/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=360 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.75 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m080.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=dataset/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m080/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=360 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.80 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m085.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=/home/joao/datasets/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m085/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=361 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.85 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m090.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=/home/joao/datasets/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m090/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=361 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.90 52 | -------------------------------------------------------------------------------- /cfg/SincNet_TIMIT_m095.cfg: -------------------------------------------------------------------------------- 1 | [data] 2 | tr_lst=data_lists/TIMIT_train.scp 3 | te_lst=data_lists/TIMIT_test.scp 4 | lab_dict=data_lists/TIMIT_labels.npy 5 | data_folder=/home/joao/datasets/timit_processed 6 | output_folder=exp/SincNet_TIMIT_m095/ 7 | pt_file=none 8 | 9 | [windowing] 10 | fs=16000 11 | cw_len=200 12 | cw_shift=10 13 | 14 | [cnn] 15 | cnn_N_filt=80,60,60 16 | cnn_len_filt=251,5,5 17 | cnn_max_pool_len=3,3,3 18 | cnn_use_laynorm_inp=True 19 | cnn_use_batchnorm_inp=False 20 | cnn_use_laynorm=True,True,True 21 | cnn_use_batchnorm=False,False,False 22 | cnn_act=leaky_relu,leaky_relu,leaky_relu 23 | cnn_drop=0.0,0.0,0.0 24 | 25 | [dnn] 26 | fc_lay=2048,2048,2048 27 | fc_drop=0.0,0.0,0.0 28 | fc_use_laynorm_inp=True 29 | fc_use_batchnorm_inp=False 30 | fc_use_batchnorm=True,True,True 31 | fc_use_laynorm=False,False,False 32 | fc_act=leaky_relu,leaky_relu,leaky_relu 33 | 34 | [class] 35 | class_lay=462 36 | class_drop=0.0 37 | class_use_laynorm_inp=False 38 | class_use_batchnorm_inp=False 39 | class_use_batchnorm=False 40 | class_use_laynorm=False 41 | class_act=softmax 42 | 43 | [optimization] 44 | lr=0.001 45 | batch_size=128 46 | N_epochs=361 47 | N_batches=800 48 | N_eval_epoch=8 49 | seed=1234 50 | AMSoftmax=True 51 | AMSoftmax_m = 0.95 52 | -------------------------------------------------------------------------------- /data_io.py: -------------------------------------------------------------------------------- 1 | import configparser as ConfigParser 2 | from optparse import OptionParser 3 | import numpy as np 4 | #import scipy.io.wavfile 5 | import torch 6 | 7 | def ReadList(list_file): 8 | f=open(list_file,"r") 9 | lines=f.readlines() 10 | list_sig=[] 11 | for x in lines: 12 | list_sig.append(x.rstrip()) 13 | f.close() 14 | return list_sig 15 | 16 | 17 | def read_conf(): 18 | 19 | parser=OptionParser() 20 | parser.add_option("--cfg") # Mandatory 21 | (options,args)=parser.parse_args() 22 | 23 | cfg_file=options.cfg 24 | Config = ConfigParser.ConfigParser() 25 | Config.read(cfg_file) 26 | 27 | #[data] 28 | options.tr_lst=Config.get('data', 'tr_lst') 29 | options.te_lst=Config.get('data', 'te_lst') 30 | options.lab_dict=Config.get('data', 'lab_dict') 31 | options.data_folder=Config.get('data', 'data_folder') 32 | options.output_folder=Config.get('data', 'output_folder') 33 | options.pt_file=Config.get('data', 'pt_file') 34 | 35 | #[windowing] 36 | options.fs=Config.get('windowing', 'fs') 37 | options.cw_len=Config.get('windowing', 'cw_len') 38 | options.cw_shift=Config.get('windowing', 'cw_shift') 39 | 40 | #[cnn] 41 | options.cnn_N_filt=Config.get('cnn', 'cnn_N_filt') 42 | options.cnn_len_filt=Config.get('cnn', 'cnn_len_filt') 43 | options.cnn_max_pool_len=Config.get('cnn', 'cnn_max_pool_len') 44 | options.cnn_use_laynorm_inp=Config.get('cnn', 'cnn_use_laynorm_inp') 45 | options.cnn_use_batchnorm_inp=Config.get('cnn', 'cnn_use_batchnorm_inp') 46 | options.cnn_use_laynorm=Config.get('cnn', 'cnn_use_laynorm') 47 | options.cnn_use_batchnorm=Config.get('cnn', 'cnn_use_batchnorm') 48 | options.cnn_act=Config.get('cnn', 'cnn_act') 49 | options.cnn_drop=Config.get('cnn', 'cnn_drop') 50 | 51 | 52 | #[dnn] 53 | options.fc_lay=Config.get('dnn', 'fc_lay') 54 | options.fc_drop=Config.get('dnn', 'fc_drop') 55 | options.fc_use_laynorm_inp=Config.get('dnn', 'fc_use_laynorm_inp') 56 | options.fc_use_batchnorm_inp=Config.get('dnn', 'fc_use_batchnorm_inp') 57 | options.fc_use_batchnorm=Config.get('dnn', 'fc_use_batchnorm') 58 | options.fc_use_laynorm=Config.get('dnn', 'fc_use_laynorm') 59 | options.fc_act=Config.get('dnn', 'fc_act') 60 | 61 | #[class] 62 | options.class_lay=Config.get('class', 'class_lay') 63 | options.class_drop=Config.get('class', 'class_drop') 64 | options.class_use_laynorm_inp=Config.get('class', 'class_use_laynorm_inp') 65 | options.class_use_batchnorm_inp=Config.get('class', 'class_use_batchnorm_inp') 66 | options.class_use_batchnorm=Config.get('class', 'class_use_batchnorm') 67 | options.class_use_laynorm=Config.get('class', 'class_use_laynorm') 68 | options.class_act=Config.get('class', 'class_act') 69 | 70 | 71 | #[optimization] 72 | options.lr=Config.get('optimization', 'lr') 73 | options.batch_size=Config.get('optimization', 'batch_size') 74 | options.N_epochs=Config.get('optimization', 'N_epochs') 75 | options.N_batches=Config.get('optimization', 'N_batches') 76 | options.N_eval_epoch=Config.get('optimization', 'N_eval_epoch') 77 | options.seed=Config.get('optimization', 'seed') 78 | options.AMSoftmax = Config.get('optimization', 'AMSoftmax') 79 | options.AMSoftmax_m = Config.get('optimization', 'AMSoftmax_m') 80 | 81 | return options 82 | 83 | 84 | def str_to_bool(s): 85 | if s == 'True': 86 | return True 87 | elif s == 'False': 88 | return False 89 | else: 90 | raise ValueError 91 | 92 | 93 | def create_batches_rnd(batch_size,data_folder,wav_lst,N_snt,wlen,lab_dict,fact_amp): 94 | 95 | # Initialization of the minibatch (batch_size,[0=>x_t,1=>x_t+N,1=>random_samp]) 96 | sig_batch=np.zeros([batch_size,wlen]) 97 | lab_batch=np.zeros(batch_size) 98 | 99 | snt_id_arr=np.random.randint(N_snt, size=batch_size) 100 | 101 | rand_amp_arr = np.random.uniform(1.0-fact_amp,1+fact_amp,batch_size) 102 | 103 | for i in range(batch_size): 104 | 105 | # select a random sentence from the list (joint distribution) 106 | [fs,signal]=scipy.io.wavfile.read(data_folder+wav_lst[snt_id_arr[i]]) 107 | signal=signal.astype(float)/32768 108 | 109 | # accesing to a random chunk 110 | snt_len=signal.shape[0] 111 | snt_beg=np.random.randint(snt_len-wlen-1) #randint(0, snt_len-2*wlen-1) 112 | snt_end=snt_beg+wlen 113 | 114 | sig_batch[i,:]=signal[snt_beg:snt_end]*rand_amp_arr[i] 115 | lab_batch[i]=lab_dict[wav_lst[snt_id_arr[i]]] 116 | 117 | inp=torch.from_numpy(sig_batch).float().cuda().contiguous() # Current Frame 118 | lab=torch.from_numpy(lab_batch).float().cuda().contiguous() 119 | 120 | return inp,lab 121 | -------------------------------------------------------------------------------- /data_lists/TIMIT_labels.npy: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/joaoantoniocn/AM-SincNet/5a6404288cad4cc5f5335a97cc1e02605f4c16ab/data_lists/TIMIT_labels.npy -------------------------------------------------------------------------------- /data_lists/TIMIT_test.scp: -------------------------------------------------------------------------------- 1 | train/dr1/fcjf0/si1027.wav 2 | train/dr1/fcjf0/si1657.wav 3 | train/dr1/fcjf0/si648.wav 4 | train/dr1/fdaw0/si1271.wav 5 | train/dr1/fdaw0/si1406.wav 6 | train/dr1/fdaw0/si2036.wav 7 | train/dr1/fdml0/si1149.wav 8 | train/dr1/fdml0/si1779.wav 9 | train/dr1/fdml0/si2075.wav 10 | train/dr1/fecd0/si1418.wav 11 | train/dr1/fecd0/si2048.wav 12 | train/dr1/fecd0/si788.wav 13 | train/dr1/fetb0/si1148.wav 14 | train/dr1/fetb0/si1778.wav 15 | train/dr1/fetb0/si518.wav 16 | train/dr1/fjsp0/si1434.wav 17 | train/dr1/fjsp0/si1763.wav 18 | train/dr1/fjsp0/si804.wav 19 | train/dr1/fkfb0/si1608.wav 20 | train/dr1/fkfb0/si2238.wav 21 | train/dr1/fkfb0/si978.wav 22 | train/dr1/fmem0/si1377.wav 23 | train/dr1/fmem0/si2007.wav 24 | train/dr1/fmem0/si747.wav 25 | train/dr1/fsah0/si1244.wav 26 | train/dr1/fsah0/si1874.wav 27 | train/dr1/fsah0/si614.wav 28 | train/dr1/fsjk1/si1025.wav 29 | train/dr1/fsjk1/si2285.wav 30 | train/dr1/fsjk1/si696.wav 31 | train/dr1/fsma0/si1621.wav 32 | train/dr1/fsma0/si2251.wav 33 | train/dr1/fsma0/si991.wav 34 | train/dr1/ftbr0/si1402.wav 35 | train/dr1/ftbr0/si2181.wav 36 | train/dr1/ftbr0/si921.wav 37 | train/dr1/fvfb0/si1032.wav 38 | train/dr1/fvfb0/si1510.wav 39 | train/dr1/fvfb0/si2292.wav 40 | train/dr1/fvmh0/si1466.wav 41 | train/dr1/fvmh0/si2096.wav 42 | train/dr1/fvmh0/si836.wav 43 | train/dr1/mcpm0/si1194.wav 44 | train/dr1/mcpm0/si1824.wav 45 | train/dr1/mcpm0/si564.wav 46 | train/dr1/mdac0/si1261.wav 47 | train/dr1/mdac0/si1837.wav 48 | train/dr1/mdac0/si631.wav 49 | train/dr1/mdpk0/si1053.wav 50 | train/dr1/mdpk0/si1683.wav 51 | train/dr1/mdpk0/si552.wav 52 | train/dr1/medr0/si1374.wav 53 | train/dr1/medr0/si2004.wav 54 | train/dr1/medr0/si744.wav 55 | train/dr1/mgrl0/si1497.wav 56 | train/dr1/mgrl0/si2127.wav 57 | train/dr1/mgrl0/si867.wav 58 | train/dr1/mjeb1/si1467.wav 59 | train/dr1/mjeb1/si2097.wav 60 | train/dr1/mjeb1/si837.wav 61 | train/dr1/mjwt0/si1291.wav 62 | train/dr1/mjwt0/si1381.wav 63 | train/dr1/mjwt0/si751.wav 64 | train/dr1/mkls0/si1437.wav 65 | train/dr1/mkls0/si1533.wav 66 | train/dr1/mkls0/si2067.wav 67 | train/dr1/mklw0/si1571.wav 68 | train/dr1/mklw0/si1844.wav 69 | train/dr1/mklw0/si2201.wav 70 | train/dr1/mmgg0/si1079.wav 71 | train/dr1/mmgg0/si1709.wav 72 | train/dr1/mmgg0/si2339.wav 73 | train/dr1/mmrp0/si2034.wav 74 | train/dr1/mmrp0/si717.wav 75 | train/dr1/mmrp0/si774.wav 76 | train/dr1/mpgh0/si1554.wav 77 | train/dr1/mpgh0/si675.wav 78 | train/dr1/mpgh0/si924.wav 79 | train/dr1/mpgr0/si1410.wav 80 | train/dr1/mpgr0/si2040.wav 81 | train/dr1/mpgr0/si780.wav 82 | train/dr1/mpsw0/si1067.wav 83 | train/dr1/mpsw0/si1697.wav 84 | train/dr1/mpsw0/si2327.wav 85 | train/dr1/mrai0/si1954.wav 86 | train/dr1/mrai0/si2052.wav 87 | train/dr1/mrai0/si792.wav 88 | train/dr1/mrcg0/si1428.wav 89 | train/dr1/mrcg0/si2058.wav 90 | train/dr1/mrcg0/si798.wav 91 | train/dr1/mrdd0/si1050.wav 92 | train/dr1/mrdd0/si1680.wav 93 | train/dr1/mrdd0/si2310.wav 94 | train/dr1/mrso0/si1206.wav 95 | train/dr1/mrso0/si1659.wav 96 | train/dr1/mrso0/si2289.wav 97 | train/dr1/mrws0/si1102.wav 98 | train/dr1/mrws0/si1732.wav 99 | train/dr1/mrws0/si472.wav 100 | train/dr1/mtjs0/si1192.wav 101 | train/dr1/mtjs0/si1822.wav 102 | train/dr1/mtjs0/si562.wav 103 | train/dr1/mtpf0/si1235.wav 104 | train/dr1/mtpf0/si1865.wav 105 | train/dr1/mtpf0/si605.wav 106 | train/dr1/mtrr0/si1548.wav 107 | train/dr1/mtrr0/si2178.wav 108 | train/dr1/mtrr0/si918.wav 109 | train/dr1/mwad0/si1062.wav 110 | train/dr1/mwad0/si1749.wav 111 | train/dr1/mwad0/si2322.wav 112 | train/dr1/mwar0/si1045.wav 113 | train/dr1/mwar0/si1675.wav 114 | train/dr1/mwar0/si2305.wav 115 | train/dr2/faem0/si1392.wav 116 | train/dr2/faem0/si2022.wav 117 | train/dr2/faem0/si762.wav 118 | train/dr2/fajw0/si1263.wav 119 | train/dr2/fajw0/si1893.wav 120 | train/dr2/fajw0/si633.wav 121 | train/dr2/fcaj0/si1479.wav 122 | train/dr2/fcaj0/si1804.wav 123 | train/dr2/fcaj0/si849.wav 124 | train/dr2/fcmm0/si1083.wav 125 | train/dr2/fcmm0/si1957.wav 126 | train/dr2/fcmm0/si453.wav 127 | train/dr2/fcyl0/si1297.wav 128 | train/dr2/fcyl0/si1927.wav 129 | train/dr2/fcyl0/si667.wav 130 | train/dr2/fdas1/si1461.wav 131 | train/dr2/fdas1/si2091.wav 132 | train/dr2/fdas1/si831.wav 133 | train/dr2/fdnc0/si1278.wav 134 | train/dr2/fdnc0/si1908.wav 135 | train/dr2/fdnc0/si2287.wav 136 | train/dr2/fdxw0/si1511.wav 137 | train/dr2/fdxw0/si2141.wav 138 | train/dr2/fdxw0/si881.wav 139 | train/dr2/feac0/si1245.wav 140 | train/dr2/feac0/si1875.wav 141 | train/dr2/feac0/si615.wav 142 | train/dr2/fhlm0/si1560.wav 143 | train/dr2/fhlm0/si2190.wav 144 | train/dr2/fhlm0/si930.wav 145 | train/dr2/fjkl0/si1562.wav 146 | train/dr2/fjkl0/si2192.wav 147 | train/dr2/fjkl0/si932.wav 148 | train/dr2/fkaa0/si1208.wav 149 | train/dr2/fkaa0/si1838.wav 150 | train/dr2/fkaa0/si578.wav 151 | train/dr2/flma0/si1243.wav 152 | train/dr2/flma0/si1873.wav 153 | train/dr2/flma0/si613.wav 154 | train/dr2/flmc0/si1372.wav 155 | train/dr2/flmc0/si2002.wav 156 | train/dr2/flmc0/si742.wav 157 | train/dr2/fmjb0/si1177.wav 158 | train/dr2/fmjb0/si1807.wav 159 | train/dr2/fmjb0/si547.wav 160 | train/dr2/fmkf0/si1018.wav 161 | train/dr2/fmkf0/si1536.wav 162 | train/dr2/fmkf0/si906.wav 163 | train/dr2/fmmh0/si1537.wav 164 | train/dr2/fmmh0/si2167.wav 165 | train/dr2/fmmh0/si907.wav 166 | train/dr2/fpjf0/si1046.wav 167 | train/dr2/fpjf0/si1259.wav 168 | train/dr2/fpjf0/si1676.wav 169 | train/dr2/frll0/si1514.wav 170 | train/dr2/frll0/si805.wav 171 | train/dr2/frll0/si884.wav 172 | train/dr2/fscn0/si1886.wav 173 | train/dr2/fscn0/si626.wav 174 | train/dr2/fscn0/si705.wav 175 | train/dr2/fskl0/si1529.wav 176 | train/dr2/fskl0/si2159.wav 177 | train/dr2/fskl0/si899.wav 178 | train/dr2/fsrh0/si1719.wav 179 | train/dr2/fsrh0/si1931.wav 180 | train/dr2/fsrh0/si671.wav 181 | train/dr2/ftmg0/si1532.wav 182 | train/dr2/ftmg0/si2162.wav 183 | train/dr2/ftmg0/si902.wav 184 | train/dr2/marc0/si1188.wav 185 | train/dr2/marc0/si1818.wav 186 | train/dr2/marc0/si558.wav 187 | train/dr2/mbjv0/si1247.wav 188 | train/dr2/mbjv0/si1877.wav 189 | train/dr2/mbjv0/si617.wav 190 | train/dr2/mcew0/si1442.wav 191 | train/dr2/mcew0/si2072.wav 192 | train/dr2/mcew0/si812.wav 193 | train/dr2/mctm0/si1350.wav 194 | train/dr2/mctm0/si1980.wav 195 | train/dr2/mctm0/si720.wav 196 | train/dr2/mdbp0/si1158.wav 197 | train/dr2/mdbp0/si1788.wav 198 | train/dr2/mdbp0/si528.wav 199 | train/dr2/mdem0/si1868.wav 200 | train/dr2/mdem0/si608.wav 201 | train/dr2/mdem0/si800.wav 202 | train/dr2/mdlb0/si1306.wav 203 | train/dr2/mdlb0/si1936.wav 204 | train/dr2/mdlb0/si676.wav 205 | train/dr2/mdlc2/si1614.wav 206 | train/dr2/mdlc2/si2244.wav 207 | train/dr2/mdlc2/si984.wav 208 | train/dr2/mdmt0/si1832.wav 209 | train/dr2/mdmt0/si2341.wav 210 | train/dr2/mdmt0/si572.wav 211 | train/dr2/mdps0/si1651.wav 212 | train/dr2/mdps0/si1979.wav 213 | train/dr2/mdps0/si719.wav 214 | train/dr2/mdss0/si1881.wav 215 | train/dr2/mdss0/si2087.wav 216 | train/dr2/mdss0/si621.wav 217 | train/dr2/mdwd0/si1260.wav 218 | train/dr2/mdwd0/si1890.wav 219 | train/dr2/mdwd0/si557.wav 220 | train/dr2/mefg0/si465.wav 221 | train/dr2/mefg0/si491.wav 222 | train/dr2/mefg0/si598.wav 223 | train/dr2/mhrm0/si1475.wav 224 | train/dr2/mhrm0/si2218.wav 225 | train/dr2/mhrm0/si958.wav 226 | train/dr2/mjae0/si1524.wav 227 | train/dr2/mjae0/si1999.wav 228 | train/dr2/mjae0/si2154.wav 229 | train/dr2/mjbg0/si1232.wav 230 | train/dr2/mjbg0/si1724.wav 231 | train/dr2/mjbg0/si1862.wav 232 | train/dr2/mjde0/si1120.wav 233 | train/dr2/mjde0/si463.wav 234 | train/dr2/mjde0/si490.wav 235 | train/dr2/mjeb0/si1286.wav 236 | train/dr2/mjeb0/si1916.wav 237 | train/dr2/mjeb0/si656.wav 238 | train/dr2/mjhi0/si1328.wav 239 | train/dr2/mjhi0/si555.wav 240 | train/dr2/mjhi0/si698.wav 241 | train/dr2/mjma0/si1495.wav 242 | train/dr2/mjma0/si2125.wav 243 | train/dr2/mjma0/si865.wav 244 | train/dr2/mjmd0/si1028.wav 245 | train/dr2/mjmd0/si1658.wav 246 | train/dr2/mjmd0/si2288.wav 247 | train/dr2/mjpm0/si1368.wav 248 | train/dr2/mjpm0/si1998.wav 249 | train/dr2/mjpm0/si738.wav 250 | train/dr2/mjrp0/si1835.wav 251 | train/dr2/mjrp0/si1845.wav 252 | train/dr2/mjrp0/si585.wav 253 | train/dr2/mkah0/si1528.wav 254 | train/dr2/mkah0/si2158.wav 255 | train/dr2/mkah0/si898.wav 256 | train/dr2/mkaj0/si1414.wav 257 | train/dr2/mkaj0/si2044.wav 258 | train/dr2/mkaj0/si784.wav 259 | train/dr2/mkdt0/si2153.wav 260 | train/dr2/mkdt0/si814.wav 261 | train/dr2/mkdt0/si893.wav 262 | train/dr2/mkjo0/si1517.wav 263 | train/dr2/mkjo0/si2147.wav 264 | train/dr2/mkjo0/si887.wav 265 | train/dr2/mmaa0/si1588.wav 266 | train/dr2/mmaa0/si2105.wav 267 | train/dr2/mmaa0/si845.wav 268 | train/dr2/mmag0/si1126.wav 269 | train/dr2/mmag0/si1756.wav 270 | train/dr2/mmag0/si496.wav 271 | train/dr2/mmds0/si1343.wav 272 | train/dr2/mmds0/si1973.wav 273 | train/dr2/mmds0/si713.wav 274 | train/dr2/mmgk0/si1322.wav 275 | train/dr2/mmgk0/si1952.wav 276 | train/dr2/mmgk0/si692.wav 277 | train/dr2/mmxs0/si2136.wav 278 | train/dr2/mmxs0/si629.wav 279 | train/dr2/mmxs0/si876.wav 280 | train/dr2/mppc0/si1412.wav 281 | train/dr2/mppc0/si2042.wav 282 | train/dr2/mppc0/si782.wav 283 | train/dr2/mprb0/si1205.wav 284 | train/dr2/mprb0/si1215.wav 285 | train/dr2/mprb0/si575.wav 286 | train/dr2/mrab0/si1224.wav 287 | train/dr2/mrab0/si1854.wav 288 | train/dr2/mrab0/si594.wav 289 | train/dr2/mrcw0/si1371.wav 290 | train/dr2/mrcw0/si2001.wav 291 | train/dr2/mrcw0/si741.wav 292 | train/dr2/mrfk0/si1076.wav 293 | train/dr2/mrfk0/si1706.wav 294 | train/dr2/mrfk0/si2336.wav 295 | train/dr2/mrgs0/si1356.wav 296 | train/dr2/mrgs0/si1986.wav 297 | train/dr2/mrgs0/si726.wav 298 | train/dr2/mrhl0/si1515.wav 299 | train/dr2/mrhl0/si2145.wav 300 | train/dr2/mrhl0/si885.wav 301 | train/dr2/mrjh0/si1519.wav 302 | train/dr2/mrjh0/si889.wav 303 | train/dr2/mrjh0/si914.wav 304 | train/dr2/mrjm0/si1095.wav 305 | train/dr2/mrjm0/si1228.wav 306 | train/dr2/mrjm0/si1858.wav 307 | train/dr2/mrjm1/si1298.wav 308 | train/dr2/mrjm1/si1928.wav 309 | train/dr2/mrjm1/si668.wav 310 | train/dr2/mrjt0/si1498.wav 311 | train/dr2/mrjt0/si1805.wav 312 | train/dr2/mrjt0/si868.wav 313 | train/dr2/mrlj0/si1420.wav 314 | train/dr2/mrlj0/si2050.wav 315 | train/dr2/mrlj0/si790.wav 316 | train/dr2/mrlr0/si1196.wav 317 | train/dr2/mrlr0/si1826.wav 318 | train/dr2/mrlr0/si566.wav 319 | train/dr2/mrms0/si1113.wav 320 | train/dr2/mrms0/si2057.wav 321 | train/dr2/mrms0/si2100.wav 322 | train/dr2/msat0/si1526.wav 323 | train/dr2/msat0/si2156.wav 324 | train/dr2/msat0/si896.wav 325 | train/dr2/mtat1/si1409.wav 326 | train/dr2/mtat1/si1627.wav 327 | train/dr2/mtat1/si779.wav 328 | train/dr2/mtbc0/si1173.wav 329 | train/dr2/mtbc0/si1803.wav 330 | train/dr2/mtbc0/si543.wav 331 | train/dr2/mtdb0/si1401.wav 332 | train/dr2/mtdb0/si2031.wav 333 | train/dr2/mtdb0/si771.wav 334 | train/dr2/mtjg0/si1520.wav 335 | train/dr2/mtjg0/si2157.wav 336 | train/dr2/mtjg0/si890.wav 337 | train/dr2/mwsb0/si1626.wav 338 | train/dr2/mwsb0/si2256.wav 339 | train/dr2/mwsb0/si996.wav 340 | train/dr2/mzmb0/si1166.wav 341 | train/dr2/mzmb0/si1796.wav 342 | train/dr2/mzmb0/si536.wav 343 | train/dr3/falk0/si1086.wav 344 | train/dr3/falk0/si456.wav 345 | train/dr3/falk0/si658.wav 346 | train/dr3/fcke0/si1111.wav 347 | train/dr3/fcke0/si1741.wav 348 | train/dr3/fcke0/si481.wav 349 | train/dr3/fcmg0/si1142.wav 350 | train/dr3/fcmg0/si1242.wav 351 | train/dr3/fcmg0/si1872.wav 352 | train/dr3/fdfb0/si1318.wav 353 | train/dr3/fdfb0/si1948.wav 354 | train/dr3/fdfb0/si2010.wav 355 | train/dr3/fdjh0/si1565.wav 356 | train/dr3/fdjh0/si2195.wav 357 | train/dr3/fdjh0/si935.wav 358 | train/dr3/feme0/si1505.wav 359 | train/dr3/feme0/si2135.wav 360 | train/dr3/feme0/si875.wav 361 | train/dr3/fgcs0/si1486.wav 362 | train/dr3/fgcs0/si2116.wav 363 | train/dr3/fgcs0/si856.wav 364 | train/dr3/fgrw0/si1152.wav 365 | train/dr3/fgrw0/si1782.wav 366 | train/dr3/fgrw0/si1990.wav 367 | train/dr3/fjlg0/si1506.wav 368 | train/dr3/fjlg0/si1889.wav 369 | train/dr3/fjlg0/si2306.wav 370 | train/dr3/fjlr0/si1231.wav 371 | train/dr3/fjlr0/si1861.wav 372 | train/dr3/fjlr0/si601.wav 373 | train/dr3/flac0/si1339.wav 374 | train/dr3/flac0/si2161.wav 375 | train/dr3/flac0/si901.wav 376 | train/dr3/fljd0/si1516.wav 377 | train/dr3/fljd0/si2146.wav 378 | train/dr3/fljd0/si886.wav 379 | train/dr3/fltm0/si1070.wav 380 | train/dr3/fltm0/si1700.wav 381 | train/dr3/fltm0/si2330.wav 382 | train/dr3/fmjf0/si1254.wav 383 | train/dr3/fmjf0/si1884.wav 384 | train/dr3/fmjf0/si624.wav 385 | train/dr3/fntb0/si1203.wav 386 | train/dr3/fntb0/si573.wav 387 | train/dr3/fntb0/si679.wav 388 | train/dr3/fpaz0/si1593.wav 389 | train/dr3/fpaz0/si2223.wav 390 | train/dr3/fpaz0/si963.wav 391 | train/dr3/fsjs0/si1171.wav 392 | train/dr3/fsjs0/si1801.wav 393 | train/dr3/fsjs0/si541.wav 394 | train/dr3/fsjw0/si1333.wav 395 | train/dr3/fsjw0/si1963.wav 396 | train/dr3/fsjw0/si703.wav 397 | train/dr3/fskc0/si1416.wav 398 | train/dr3/fskc0/si2046.wav 399 | train/dr3/fskc0/si786.wav 400 | train/dr3/fsls0/si1056.wav 401 | train/dr3/fsls0/si1686.wav 402 | train/dr3/fsls0/si2316.wav 403 | train/dr3/madc0/si1367.wav 404 | train/dr3/madc0/si1997.wav 405 | train/dr3/madc0/si737.wav 406 | train/dr3/makb0/si1016.wav 407 | train/dr3/makb0/si1646.wav 408 | train/dr3/makb0/si2276.wav 409 | train/dr3/makr0/si1352.wav 410 | train/dr3/makr0/si1982.wav 411 | train/dr3/makr0/si722.wav 412 | train/dr3/mapv0/si1293.wav 413 | train/dr3/mapv0/si1923.wav 414 | train/dr3/mapv0/si663.wav 415 | train/dr3/mbef0/si1281.wav 416 | train/dr3/mbef0/si1911.wav 417 | train/dr3/mbef0/si651.wav 418 | train/dr3/mcal0/si1138.wav 419 | train/dr3/mcal0/si1768.wav 420 | train/dr3/mcal0/si508.wav 421 | train/dr3/mcdc0/si1292.wav 422 | train/dr3/mcdc0/si1922.wav 423 | train/dr3/mcdc0/si662.wav 424 | train/dr3/mcdd0/si1513.wav 425 | train/dr3/mcdd0/si2143.wav 426 | train/dr3/mcdd0/si883.wav 427 | train/dr3/mcef0/si1135.wav 428 | train/dr3/mcef0/si1765.wav 429 | train/dr3/mcef0/si842.wav 430 | train/dr3/mdbb1/si1006.wav 431 | train/dr3/mdbb1/si1636.wav 432 | train/dr3/mdbb1/si2056.wav 433 | train/dr3/mddc0/si1419.wav 434 | train/dr3/mddc0/si2049.wav 435 | train/dr3/mddc0/si789.wav 436 | train/dr3/mdef0/si1123.wav 437 | train/dr3/mdef0/si1563.wav 438 | train/dr3/mdef0/si2193.wav 439 | train/dr3/mdhs0/si1530.wav 440 | train/dr3/mdhs0/si2160.wav 441 | train/dr3/mdhs0/si900.wav 442 | train/dr3/mdjm0/si1455.wav 443 | train/dr3/mdjm0/si2085.wav 444 | train/dr3/mdjm0/si825.wav 445 | train/dr3/mdlc0/si1395.wav 446 | train/dr3/mdlc0/si2025.wav 447 | train/dr3/mdlc0/si765.wav 448 | train/dr3/mdlh0/si1960.wav 449 | train/dr3/mdlh0/si574.wav 450 | train/dr3/mdlh0/si700.wav 451 | train/dr3/mdns0/si1011.wav 452 | train/dr3/mdns0/si2271.wav 453 | train/dr3/mdns0/si873.wav 454 | train/dr3/mdss1/si1327.wav 455 | train/dr3/mdss1/si1713.wav 456 | train/dr3/mdss1/si697.wav 457 | train/dr3/mdtb0/si1200.wav 458 | train/dr3/mdtb0/si1830.wav 459 | train/dr3/mdtb0/si570.wav 460 | train/dr3/mdwm0/si1546.wav 461 | train/dr3/mdwm0/si2176.wav 462 | train/dr3/mdwm0/si916.wav 463 | train/dr3/mfmc0/si1132.wav 464 | train/dr3/mfmc0/si1762.wav 465 | train/dr3/mfmc0/si502.wav 466 | train/dr3/mgaf0/si1282.wav 467 | train/dr3/mgaf0/si1912.wav 468 | train/dr3/mgaf0/si652.wav 469 | train/dr3/mhjb0/si1017.wav 470 | train/dr3/mhjb0/si1647.wav 471 | train/dr3/mhjb0/si2277.wav 472 | train/dr3/mhmr0/si1119.wav 473 | train/dr3/mhmr0/si1692.wav 474 | train/dr3/mhmr0/si489.wav 475 | train/dr3/milb0/si2163.wav 476 | train/dr3/milb0/si807.wav 477 | train/dr3/milb0/si903.wav 478 | train/dr3/mjda0/si1031.wav 479 | train/dr3/mjda0/si1661.wav 480 | train/dr3/mjda0/si2291.wav 481 | train/dr3/mjjb0/si1139.wav 482 | train/dr3/mjjb0/si1277.wav 483 | train/dr3/mjjb0/si1769.wav 484 | train/dr3/mjkr0/si1201.wav 485 | train/dr3/mjkr0/si1831.wav 486 | train/dr3/mjkr0/si571.wav 487 | train/dr3/mjlg1/si1012.wav 488 | train/dr3/mjlg1/si1642.wav 489 | train/dr3/mjlg1/si2272.wav 490 | train/dr3/mjrh1/si1558.wav 491 | train/dr3/mjrh1/si1774.wav 492 | train/dr3/mjrh1/si514.wav 493 | train/dr3/mkls1/si1545.wav 494 | train/dr3/mkls1/si2175.wav 495 | train/dr3/mkls1/si915.wav 496 | train/dr3/mkxl0/si1185.wav 497 | train/dr3/mkxl0/si1815.wav 498 | train/dr3/mkxl0/si1958.wav 499 | train/dr3/mlns0/si1407.wav 500 | train/dr3/mlns0/si2037.wav 501 | train/dr3/mlns0/si777.wav 502 | train/dr3/mmam0/si1597.wav 503 | train/dr3/mmam0/si1668.wav 504 | train/dr3/mmam0/si2227.wav 505 | train/dr3/mmar0/si1336.wav 506 | train/dr3/mmar0/si1966.wav 507 | train/dr3/mmar0/si706.wav 508 | train/dr3/mmeb0/si1357.wav 509 | train/dr3/mmeb0/si1987.wav 510 | train/dr3/mmeb0/si727.wav 511 | train/dr3/mmjb1/si1408.wav 512 | train/dr3/mmjb1/si2038.wav 513 | train/dr3/mmjb1/si778.wav 514 | train/dr3/mmsm0/si1106.wav 515 | train/dr3/mmsm0/si1736.wav 516 | train/dr3/mmsm0/si476.wav 517 | train/dr3/mprd0/si1431.wav 518 | train/dr3/mprd0/si2061.wav 519 | train/dr3/mprd0/si801.wav 520 | train/dr3/mrbc0/si1665.wav 521 | train/dr3/mrbc0/si1859.wav 522 | train/dr3/mrbc0/si599.wav 523 | train/dr3/mrds0/si1167.wav 524 | train/dr3/mrds0/si1797.wav 525 | train/dr3/mrds0/si537.wav 526 | train/dr3/mree0/si1104.wav 527 | train/dr3/mree0/si1734.wav 528 | train/dr3/mree0/si1959.wav 529 | train/dr3/mreh1/si1599.wav 530 | train/dr3/mreh1/si2229.wav 531 | train/dr3/mreh1/si969.wav 532 | train/dr3/mrjb1/si1020.wav 533 | train/dr3/mrjb1/si1413.wav 534 | train/dr3/mrjb1/si2021.wav 535 | train/dr3/mrtc0/si1458.wav 536 | train/dr3/mrtc0/si2088.wav 537 | train/dr3/mrtc0/si828.wav 538 | train/dr3/mrtj0/si1551.wav 539 | train/dr3/mrtj0/si2032.wav 540 | train/dr3/mrtj0/si772.wav 541 | train/dr3/mrwa0/si1603.wav 542 | train/dr3/mrwa0/si2233.wav 543 | train/dr3/mrwa0/si973.wav 544 | train/dr3/msfv0/si1262.wav 545 | train/dr3/msfv0/si1892.wav 546 | train/dr3/msfv0/si632.wav 547 | train/dr3/mtjm0/si1226.wav 548 | train/dr3/mtjm0/si1856.wav 549 | train/dr3/mtjm0/si655.wav 550 | train/dr3/mtkp0/si1023.wav 551 | train/dr3/mtkp0/si2283.wav 552 | train/dr3/mtkp0/si454.wav 553 | train/dr3/mtlb0/si1134.wav 554 | train/dr3/mtlb0/si1764.wav 555 | train/dr3/mtlb0/si504.wav 556 | train/dr3/mtpg0/si1383.wav 557 | train/dr3/mtpg0/si2013.wav 558 | train/dr3/mtpg0/si753.wav 559 | train/dr3/mtpp0/si1508.wav 560 | train/dr3/mtpp0/si2138.wav 561 | train/dr3/mtpp0/si878.wav 562 | train/dr3/mvjh0/si1556.wav 563 | train/dr3/mvjh0/si2186.wav 564 | train/dr3/mvjh0/si926.wav 565 | train/dr3/mwdk0/si1436.wav 566 | train/dr3/mwdk0/si2017.wav 567 | train/dr3/mwdk0/si806.wav 568 | train/dr3/mwgr0/si1606.wav 569 | train/dr3/mwgr0/si2236.wav 570 | train/dr3/mwgr0/si976.wav 571 | train/dr4/falr0/si1325.wav 572 | train/dr4/falr0/si1955.wav 573 | train/dr4/falr0/si695.wav 574 | train/dr4/fbas0/si1387.wav 575 | train/dr4/fbas0/si1472.wav 576 | train/dr4/fbas0/si2066.wav 577 | train/dr4/fbmj0/si1776.wav 578 | train/dr4/fbmj0/si516.wav 579 | train/dr4/fbmj0/si815.wav 580 | train/dr4/fcag0/si1503.wav 581 | train/dr4/fcag0/si1641.wav 582 | train/dr4/fcag0/si2133.wav 583 | train/dr4/fdkn0/si1081.wav 584 | train/dr4/fdkn0/si1202.wav 585 | train/dr4/fdkn0/si1711.wav 586 | train/dr4/feeh0/si1112.wav 587 | train/dr4/feeh0/si1742.wav 588 | train/dr4/feeh0/si471.wav 589 | train/dr4/fjwb1/si2055.wav 590 | train/dr4/fjwb1/si748.wav 591 | train/dr4/fjwb1/si795.wav 592 | train/dr4/fjxp0/si1122.wav 593 | train/dr4/fjxp0/si1752.wav 594 | train/dr4/fjxp0/si492.wav 595 | train/dr4/fkdw0/si1207.wav 596 | train/dr4/fkdw0/si1891.wav 597 | train/dr4/fkdw0/si577.wav 598 | train/dr4/fklc0/si1615.wav 599 | train/dr4/fklc0/si2245.wav 600 | train/dr4/fklc0/si985.wav 601 | train/dr4/flhd0/si1344.wav 602 | train/dr4/flhd0/si1827.wav 603 | train/dr4/flhd0/si1974.wav 604 | train/dr4/flkm0/si1880.wav 605 | train/dr4/flkm0/si620.wav 606 | train/dr4/flkm0/si686.wav 607 | train/dr4/fpaf0/si1054.wav 608 | train/dr4/fpaf0/si1684.wav 609 | train/dr4/fpaf0/si2314.wav 610 | train/dr4/fsak0/si1300.wav 611 | train/dr4/fsak0/si1930.wav 612 | train/dr4/fsak0/si670.wav 613 | train/dr4/fssb0/si1082.wav 614 | train/dr4/fssb0/si1712.wav 615 | train/dr4/fssb0/si2342.wav 616 | train/dr4/maeb0/si1411.wav 617 | train/dr4/maeb0/si2250.wav 618 | train/dr4/maeb0/si990.wav 619 | train/dr4/marw0/si1276.wav 620 | train/dr4/marw0/si1906.wav 621 | train/dr4/marw0/si646.wav 622 | train/dr4/mbma0/si1222.wav 623 | train/dr4/mbma0/si1852.wav 624 | train/dr4/mbma0/si592.wav 625 | train/dr4/mbwp0/si1531.wav 626 | train/dr4/mbwp0/si1969.wav 627 | train/dr4/mbwp0/si709.wav 628 | train/dr4/mcdr0/si1154.wav 629 | train/dr4/mcdr0/si1784.wav 630 | train/dr4/mcdr0/si524.wav 631 | train/dr4/mcss0/si1380.wav 632 | train/dr4/mcss0/si688.wav 633 | train/dr4/mcss0/si750.wav 634 | train/dr4/mdcd0/si1415.wav 635 | train/dr4/mdcd0/si2045.wav 636 | train/dr4/mdcd0/si785.wav 637 | train/dr4/mdma0/si1238.wav 638 | train/dr4/mdma0/si1430.wav 639 | train/dr4/mdma0/si2060.wav 640 | train/dr4/mesg0/si1332.wav 641 | train/dr4/mesg0/si1962.wav 642 | train/dr4/mesg0/si702.wav 643 | train/dr4/mfrm0/si1155.wav 644 | train/dr4/mfrm0/si1717.wav 645 | train/dr4/mfrm0/si1785.wav 646 | train/dr4/mfwk0/si1249.wav 647 | train/dr4/mfwk0/si1879.wav 648 | train/dr4/mfwk0/si619.wav 649 | train/dr4/mgag0/si1321.wav 650 | train/dr4/mgag0/si645.wav 651 | train/dr4/mgag0/si691.wav 652 | train/dr4/mgjc0/si1256.wav 653 | train/dr4/mgjc0/si1335.wav 654 | train/dr4/mgjc0/si1965.wav 655 | train/dr4/mgrp0/si1317.wav 656 | train/dr4/mgrp0/si1947.wav 657 | train/dr4/mgrp0/si687.wav 658 | train/dr4/mgxp0/si1087.wav 659 | train/dr4/mgxp0/si457.wav 660 | train/dr4/mgxp0/si525.wav 661 | train/dr4/mjac0/si1331.wav 662 | train/dr4/mjac0/si2148.wav 663 | train/dr4/mjac0/si701.wav 664 | train/dr4/mjdc0/si1161.wav 665 | train/dr4/mjdc0/si2165.wav 666 | train/dr4/mjdc0/si531.wav 667 | train/dr4/mjee0/si1237.wav 668 | train/dr4/mjee0/si1867.wav 669 | train/dr4/mjee0/si607.wav 670 | train/dr4/mjjj0/si1163.wav 671 | train/dr4/mjjj0/si1793.wav 672 | train/dr4/mjjj0/si533.wav 673 | train/dr4/mjlb0/si1616.wav 674 | train/dr4/mjlb0/si2246.wav 675 | train/dr4/mjlb0/si986.wav 676 | train/dr4/mjls0/si1096.wav 677 | train/dr4/mjls0/si1726.wav 678 | train/dr4/mjls0/si466.wav 679 | train/dr4/mjmm0/si1255.wav 680 | train/dr4/mjmm0/si1885.wav 681 | train/dr4/mjmm0/si625.wav 682 | train/dr4/mjpm1/si1897.wav 683 | train/dr4/mjpm1/si2280.wav 684 | train/dr4/mjpm1/si761.wav 685 | train/dr4/mjrh0/si1125.wav 686 | train/dr4/mjrh0/si1755.wav 687 | train/dr4/mjrh0/si1840.wav 688 | train/dr4/mjsr0/si1424.wav 689 | train/dr4/mjsr0/si2054.wav 690 | train/dr4/mjsr0/si794.wav 691 | train/dr4/mjws0/si1143.wav 692 | train/dr4/mjws0/si1773.wav 693 | train/dr4/mjws0/si513.wav 694 | train/dr4/mjxl0/si1172.wav 695 | train/dr4/mjxl0/si1795.wav 696 | train/dr4/mjxl0/si542.wav 697 | train/dr4/mkam0/si1250.wav 698 | train/dr4/mkam0/si1316.wav 699 | train/dr4/mkam0/si1465.wav 700 | train/dr4/mlbc0/si1239.wav 701 | train/dr4/mlbc0/si1869.wav 702 | train/dr4/mlbc0/si609.wav 703 | train/dr4/mlel0/si1246.wav 704 | train/dr4/mlel0/si1876.wav 705 | train/dr4/mlel0/si616.wav 706 | train/dr4/mljc0/si1225.wav 707 | train/dr4/mljc0/si1855.wav 708 | train/dr4/mljc0/si595.wav 709 | train/dr4/mljh0/si1324.wav 710 | train/dr4/mljh0/si1422.wav 711 | train/dr4/mljh0/si694.wav 712 | train/dr4/mlsh0/si1417.wav 713 | train/dr4/mlsh0/si2047.wav 714 | train/dr4/mlsh0/si787.wav 715 | train/dr4/mmbs0/si1151.wav 716 | train/dr4/mmbs0/si1781.wav 717 | train/dr4/mmbs0/si521.wav 718 | train/dr4/mmdm0/si1311.wav 719 | train/dr4/mmdm0/si1941.wav 720 | train/dr4/mmdm0/si681.wav 721 | train/dr4/mmgc0/si1305.wav 722 | train/dr4/mmgc0/si1935.wav 723 | train/dr4/mmgc0/si2184.wav 724 | train/dr4/mnet0/si1446.wav 725 | train/dr4/mnet0/si2076.wav 726 | train/dr4/mnet0/si816.wav 727 | train/dr4/mpeb0/si1034.wav 728 | train/dr4/mpeb0/si1860.wav 729 | train/dr4/mpeb0/si600.wav 730 | train/dr4/mprk0/si1097.wav 731 | train/dr4/mprk0/si1727.wav 732 | train/dr4/mprk0/si467.wav 733 | train/dr4/mprt0/si1210.wav 734 | train/dr4/mprt0/si495.wav 735 | train/dr4/mprt0/si580.wav 736 | train/dr4/mrab1/si1478.wav 737 | train/dr4/mrab1/si2108.wav 738 | train/dr4/mrab1/si848.wav 739 | train/dr4/mrfl0/si1156.wav 740 | train/dr4/mrfl0/si1786.wav 741 | train/dr4/mrfl0/si526.wav 742 | train/dr4/mrgm0/si1162.wav 743 | train/dr4/mrgm0/si1792.wav 744 | train/dr4/mrgm0/si532.wav 745 | train/dr4/mrsp0/si1429.wav 746 | train/dr4/mrsp0/si2059.wav 747 | train/dr4/mrsp0/si799.wav 748 | train/dr4/msfh0/si1216.wav 749 | train/dr4/msfh0/si1738.wav 750 | train/dr4/msfh0/si586.wav 751 | train/dr4/msmc0/si1907.wav 752 | train/dr4/msmc0/si509.wav 753 | train/dr4/msmc0/si647.wav 754 | train/dr4/msms0/si1433.wav 755 | train/dr4/msms0/si2063.wav 756 | train/dr4/msms0/si803.wav 757 | train/dr4/msrg0/si1221.wav 758 | train/dr4/msrg0/si1851.wav 759 | train/dr4/msrg0/si591.wav 760 | train/dr4/mstf0/si1396.wav 761 | train/dr4/mstf0/si766.wav 762 | train/dr4/mstf0/si852.wav 763 | train/dr4/mtas0/si1385.wav 764 | train/dr4/mtas0/si2015.wav 765 | train/dr4/mtas0/si755.wav 766 | train/dr4/mtqc0/si1441.wav 767 | train/dr4/mtqc0/si2071.wav 768 | train/dr4/mtqc0/si480.wav 769 | train/dr4/mtrc0/si1623.wav 770 | train/dr4/mtrc0/si589.wav 771 | train/dr4/mtrc0/si993.wav 772 | train/dr4/mtrt0/si1227.wav 773 | train/dr4/mtrt0/si1857.wav 774 | train/dr4/mtrt0/si597.wav 775 | train/dr5/fbjl0/si1552.wav 776 | train/dr5/fbjl0/si2182.wav 777 | train/dr5/fbjl0/si922.wav 778 | train/dr5/fbmh0/si1136.wav 779 | train/dr5/fbmh0/si1766.wav 780 | train/dr5/fbmh0/si970.wav 781 | train/dr5/fcdr1/si1186.wav 782 | train/dr5/fcdr1/si1816.wav 783 | train/dr5/fcdr1/si556.wav 784 | train/dr5/fdmy0/si1197.wav 785 | train/dr5/fdmy0/si567.wav 786 | train/dr5/fdmy0/si714.wav 787 | train/dr5/fdtd0/si1561.wav 788 | train/dr5/fdtd0/si2191.wav 789 | train/dr5/fdtd0/si931.wav 790 | train/dr5/fear0/si1252.wav 791 | train/dr5/fear0/si1882.wav 792 | train/dr5/fear0/si622.wav 793 | train/dr5/fexm0/si1101.wav 794 | train/dr5/fexm0/si1731.wav 795 | train/dr5/fexm0/si482.wav 796 | train/dr5/fgdp0/si1618.wav 797 | train/dr5/fgdp0/si2248.wav 798 | train/dr5/fgdp0/si988.wav 799 | train/dr5/fgmb0/si1145.wav 800 | train/dr5/fgmb0/si1775.wav 801 | train/dr5/fgmb0/si515.wav 802 | train/dr5/fjxm0/si1211.wav 803 | train/dr5/fjxm0/si1971.wav 804 | train/dr5/fjxm0/si581.wav 805 | train/dr5/fkkh0/si1290.wav 806 | train/dr5/fkkh0/si1920.wav 807 | train/dr5/fkkh0/si660.wav 808 | train/dr5/flja0/si1078.wav 809 | train/dr5/flja0/si1708.wav 810 | train/dr5/flja0/si2338.wav 811 | train/dr5/fljg0/si1611.wav 812 | train/dr5/fljg0/si2241.wav 813 | train/dr5/fljg0/si981.wav 814 | train/dr5/flmk0/si1035.wav 815 | train/dr5/flmk0/si1229.wav 816 | train/dr5/flmk0/si2295.wav 817 | train/dr5/flod0/si1287.wav 818 | train/dr5/flod0/si1917.wav 819 | train/dr5/flod0/si657.wav 820 | train/dr5/fmpg0/si1602.wav 821 | train/dr5/fmpg0/si2232.wav 822 | train/dr5/fmpg0/si972.wav 823 | train/dr5/fpmy0/si1153.wav 824 | train/dr5/fpmy0/si1783.wav 825 | train/dr5/fpmy0/si523.wav 826 | train/dr5/fsag0/si1323.wav 827 | train/dr5/fsag0/si1953.wav 828 | train/dr5/fsag0/si693.wav 829 | train/dr5/fsdc0/si1312.wav 830 | train/dr5/fsdc0/si1942.wav 831 | train/dr5/fsdc0/si2234.wav 832 | train/dr5/fsjg0/si1570.wav 833 | train/dr5/fsjg0/si2200.wav 834 | train/dr5/fsjg0/si940.wav 835 | train/dr5/fskp0/si1098.wav 836 | train/dr5/fskp0/si1728.wav 837 | train/dr5/fskp0/si468.wav 838 | train/dr5/fsmm0/si1314.wav 839 | train/dr5/fsmm0/si1944.wav 840 | train/dr5/fsmm0/si684.wav 841 | train/dr5/fsms1/si1504.wav 842 | train/dr5/fsms1/si2134.wav 843 | train/dr5/fsms1/si874.wav 844 | train/dr5/ftbw0/si1345.wav 845 | train/dr5/ftbw0/si1975.wav 846 | train/dr5/ftbw0/si715.wav 847 | train/dr5/ftlg0/si1743.wav 848 | train/dr5/ftlg0/si483.wav 849 | train/dr5/ftlg0/si840.wav 850 | train/dr5/mbgt0/si1341.wav 851 | train/dr5/mbgt0/si1841.wav 852 | train/dr5/mbgt0/si711.wav 853 | train/dr5/mchl0/si1347.wav 854 | train/dr5/mchl0/si1404.wav 855 | train/dr5/mchl0/si1977.wav 856 | train/dr5/mclm0/si1456.wav 857 | train/dr5/mclm0/si2086.wav 858 | train/dr5/mclm0/si826.wav 859 | train/dr5/mdas0/si1266.wav 860 | train/dr5/mdas0/si1896.wav 861 | train/dr5/mdas0/si636.wav 862 | train/dr5/mdhl0/si1439.wav 863 | train/dr5/mdhl0/si2069.wav 864 | train/dr5/mdhl0/si809.wav 865 | train/dr5/mdsj0/si1462.wav 866 | train/dr5/mdsj0/si2092.wav 867 | train/dr5/mdsj0/si832.wav 868 | train/dr5/mdwh0/si1168.wav 869 | train/dr5/mdwh0/si1925.wav 870 | train/dr5/mdwh0/si665.wav 871 | train/dr5/megj0/si1337.wav 872 | train/dr5/megj0/si1967.wav 873 | train/dr5/megj0/si707.wav 874 | train/dr5/mewm0/si1348.wav 875 | train/dr5/mewm0/si1978.wav 876 | train/dr5/mewm0/si718.wav 877 | train/dr5/mfer0/si1492.wav 878 | train/dr5/mfer0/si2122.wav 879 | train/dr5/mfer0/si862.wav 880 | train/dr5/mges0/si1481.wav 881 | train/dr5/mges0/si2111.wav 882 | train/dr5/mges0/si851.wav 883 | train/dr5/mgsh0/si1176.wav 884 | train/dr5/mgsh0/si1806.wav 885 | train/dr5/mgsh0/si546.wav 886 | train/dr5/mhit0/si1613.wav 887 | train/dr5/mhit0/si2243.wav 888 | train/dr5/mhit0/si983.wav 889 | train/dr5/mhmg0/si1365.wav 890 | train/dr5/mhmg0/si1995.wav 891 | train/dr5/mhmg0/si735.wav 892 | train/dr5/mjdm0/si1340.wav 893 | train/dr5/mjdm0/si1937.wav 894 | train/dr5/mjdm0/si974.wav 895 | train/dr5/mjfh0/si1107.wav 896 | train/dr5/mjfh0/si1737.wav 897 | train/dr5/mjfh0/si477.wav 898 | train/dr5/mjpg0/si1191.wav 899 | train/dr5/mjpg0/si1821.wav 900 | train/dr5/mjpg0/si561.wav 901 | train/dr5/mjrg0/si1366.wav 902 | train/dr5/mjrg0/si1996.wav 903 | train/dr5/mjrg0/si736.wav 904 | train/dr5/mjwg0/si2155.wav 905 | train/dr5/mjwg0/si813.wav 906 | train/dr5/mjwg0/si895.wav 907 | train/dr5/mjxa0/si1507.wav 908 | train/dr5/mjxa0/si2137.wav 909 | train/dr5/mjxa0/si877.wav 910 | train/dr5/mmab1/si1494.wav 911 | train/dr5/mmab1/si2124.wav 912 | train/dr5/mmab1/si864.wav 913 | train/dr5/mmcc0/si1338.wav 914 | train/dr5/mmcc0/si1968.wav 915 | train/dr5/mmcc0/si708.wav 916 | train/dr5/mmdm1/si1650.wav 917 | train/dr5/mmdm1/si2043.wav 918 | train/dr5/mmdm1/si783.wav 919 | train/dr5/mmvp0/si1284.wav 920 | train/dr5/mmvp0/si1914.wav 921 | train/dr5/mmvp0/si654.wav 922 | train/dr5/mmwb0/si1619.wav 923 | train/dr5/mmwb0/si2249.wav 924 | train/dr5/mmwb0/si989.wav 925 | train/dr5/mpmb0/si1501.wav 926 | train/dr5/mpmb0/si2131.wav 927 | train/dr5/mpmb0/si871.wav 928 | train/dr5/mram0/si1275.wav 929 | train/dr5/mram0/si1905.wav 930 | train/dr5/mram0/si1951.wav 931 | train/dr5/mrav0/si1008.wav 932 | train/dr5/mrav0/si1638.wav 933 | train/dr5/mrav0/si2268.wav 934 | train/dr5/mrew1/si1500.wav 935 | train/dr5/mrew1/si2130.wav 936 | train/dr5/mrew1/si870.wav 937 | train/dr5/mrkm0/si1267.wav 938 | train/dr5/mrkm0/si1391.wav 939 | train/dr5/mrkm0/si637.wav 940 | train/dr5/mrld0/si1594.wav 941 | train/dr5/mrld0/si2224.wav 942 | train/dr5/mrld0/si964.wav 943 | train/dr5/mrml0/si1421.wav 944 | train/dr5/mrml0/si2051.wav 945 | train/dr5/mrml0/si791.wav 946 | train/dr5/mrvg0/si1140.wav 947 | train/dr5/mrvg0/si1770.wav 948 | train/dr5/mrvg0/si510.wav 949 | train/dr5/msas0/si1376.wav 950 | train/dr5/msas0/si2006.wav 951 | train/dr5/msas0/si746.wav 952 | train/dr5/msdh0/si2113.wav 953 | train/dr5/msdh0/si2240.wav 954 | train/dr5/msdh0/si980.wav 955 | train/dr5/msem1/si1440.wav 956 | train/dr5/msem1/si2070.wav 957 | train/dr5/msem1/si810.wav 958 | train/dr5/msrr0/si1131.wav 959 | train/dr5/msrr0/si1761.wav 960 | train/dr5/msrr0/si501.wav 961 | train/dr5/mtat0/si1110.wav 962 | train/dr5/mtat0/si1740.wav 963 | train/dr5/mtat0/si811.wav 964 | train/dr5/mtdp0/si1274.wav 965 | train/dr5/mtdp0/si1521.wav 966 | train/dr5/mtdp0/si2151.wav 967 | train/dr5/mtmt0/si1118.wav 968 | train/dr5/mtmt0/si1748.wav 969 | train/dr5/mtmt0/si488.wav 970 | train/dr5/mvlo0/si1147.wav 971 | train/dr5/mvlo0/si1777.wav 972 | train/dr5/mvlo0/si517.wav 973 | train/dr5/mwac0/si1601.wav 974 | train/dr5/mwac0/si2231.wav 975 | train/dr5/mwac0/si971.wav 976 | train/dr5/mwch0/si1622.wav 977 | train/dr5/mwch0/si1895.wav 978 | train/dr5/mwch0/si2252.wav 979 | train/dr5/mwem0/si1320.wav 980 | train/dr5/mwem0/si1393.wav 981 | train/dr5/mwem0/si1950.wav 982 | train/dr5/mwsh0/si1426.wav 983 | train/dr5/mwsh0/si2266.wav 984 | train/dr5/mwsh0/si796.wav 985 | train/dr6/fapb0/si1063.wav 986 | train/dr6/fapb0/si1693.wav 987 | train/dr6/fapb0/si2323.wav 988 | train/dr6/fbch0/si1586.wav 989 | train/dr6/fbch0/si956.wav 990 | train/dr6/fbch0/si959.wav 991 | train/dr6/fhxs0/si1075.wav 992 | train/dr6/fhxs0/si2302.wav 993 | train/dr6/fhxs0/si2335.wav 994 | train/dr6/fjdm2/si1582.wav 995 | train/dr6/fjdm2/si1964.wav 996 | train/dr6/fjdm2/si2212.wav 997 | train/dr6/fklc1/si1048.wav 998 | train/dr6/fklc1/si1678.wav 999 | train/dr6/fklc1/si2308.wav 1000 | train/dr6/flag0/si1464.wav 1001 | train/dr6/flag0/si2094.wav 1002 | train/dr6/flag0/si834.wav 1003 | train/dr6/fmju0/si1389.wav 1004 | train/dr6/fmju0/si2019.wav 1005 | train/dr6/fmju0/si759.wav 1006 | train/dr6/fpad0/si1346.wav 1007 | train/dr6/fpad0/si1976.wav 1008 | train/dr6/fpad0/si716.wav 1009 | train/dr6/frjb0/si1427.wav 1010 | train/dr6/frjb0/si1470.wav 1011 | train/dr6/frjb0/si1794.wav 1012 | train/dr6/fsbk0/si1069.wav 1013 | train/dr6/fsbk0/si1699.wav 1014 | train/dr6/fsbk0/si2329.wav 1015 | train/dr6/fsdj0/si1115.wav 1016 | train/dr6/fsdj0/si1745.wav 1017 | train/dr6/fsdj0/si485.wav 1018 | train/dr6/fsgf0/si1557.wav 1019 | train/dr6/fsgf0/si2187.wav 1020 | train/dr6/fsgf0/si927.wav 1021 | train/dr6/ftaj0/si1329.wav 1022 | train/dr6/ftaj0/si474.wav 1023 | train/dr6/ftaj0/si699.wav 1024 | train/dr6/mabc0/si1620.wav 1025 | train/dr6/mabc0/si2041.wav 1026 | train/dr6/mabc0/si781.wav 1027 | train/dr6/majp0/si1074.wav 1028 | train/dr6/majp0/si1704.wav 1029 | train/dr6/majp0/si2334.wav 1030 | train/dr6/mbma1/si2207.wav 1031 | train/dr6/mbma1/si2214.wav 1032 | train/dr6/mbma1/si954.wav 1033 | train/dr6/mcae0/si1447.wav 1034 | train/dr6/mcae0/si2077.wav 1035 | train/dr6/mcae0/si817.wav 1036 | train/dr6/mdrd0/si1382.wav 1037 | train/dr6/mdrd0/si2012.wav 1038 | train/dr6/mdrd0/si752.wav 1039 | train/dr6/meal0/si1547.wav 1040 | train/dr6/meal0/si2177.wav 1041 | train/dr6/meal0/si917.wav 1042 | train/dr6/mejl0/si1592.wav 1043 | train/dr6/mejl0/si1654.wav 1044 | train/dr6/mejl0/si962.wav 1045 | train/dr6/mesj0/si2039.wav 1046 | train/dr6/mesj0/si2257.wav 1047 | train/dr6/mesj0/si997.wav 1048 | train/dr6/mjrk0/si1662.wav 1049 | train/dr6/mjrk0/si2103.wav 1050 | train/dr6/mjrk0/si880.wav 1051 | train/dr6/mkes0/si1253.wav 1052 | train/dr6/mkes0/si1883.wav 1053 | train/dr6/mkes0/si623.wav 1054 | train/dr6/mkln0/si1598.wav 1055 | train/dr6/mkln0/si2228.wav 1056 | train/dr6/mkln0/si968.wav 1057 | train/dr6/mmdb0/si1358.wav 1058 | train/dr6/mmdb0/si1617.wav 1059 | train/dr6/mmdb0/si987.wav 1060 | train/dr6/mpgr1/si1269.wav 1061 | train/dr6/mpgr1/si1499.wav 1062 | train/dr6/mpgr1/si2129.wav 1063 | train/dr6/mrmb0/si1581.wav 1064 | train/dr6/mrmb0/si2211.wav 1065 | train/dr6/mrmb0/si951.wav 1066 | train/dr6/mrxb0/si1585.wav 1067 | train/dr6/mrxb0/si2215.wav 1068 | train/dr6/mrxb0/si955.wav 1069 | train/dr6/msat1/si1073.wav 1070 | train/dr6/msat1/si1703.wav 1071 | train/dr6/msat1/si2333.wav 1072 | train/dr6/msds0/si1077.wav 1073 | train/dr6/msds0/si1707.wav 1074 | train/dr6/msds0/si2337.wav 1075 | train/dr6/msjk0/si1596.wav 1076 | train/dr6/msjk0/si2226.wav 1077 | train/dr6/msjk0/si966.wav 1078 | train/dr6/msmr0/si1150.wav 1079 | train/dr6/msmr0/si1405.wav 1080 | train/dr6/msmr0/si775.wav 1081 | train/dr6/msvs0/si1568.wav 1082 | train/dr6/msvs0/si2198.wav 1083 | train/dr6/msvs0/si938.wav 1084 | train/dr6/mtju0/si2020.wav 1085 | train/dr6/mtju0/si2269.wav 1086 | train/dr6/mtju0/si760.wav 1087 | train/dr6/mtxs0/si1060.wav 1088 | train/dr6/mtxs0/si1690.wav 1089 | train/dr6/mtxs0/si2320.wav 1090 | train/dr7/fblv0/si1058.wav 1091 | train/dr7/fblv0/si1688.wav 1092 | train/dr7/fblv0/si2318.wav 1093 | train/dr7/fcjs0/si1607.wav 1094 | train/dr7/fcjs0/si2237.wav 1095 | train/dr7/fcjs0/si977.wav 1096 | train/dr7/fcrz0/si1913.wav 1097 | train/dr7/fcrz0/si2053.wav 1098 | train/dr7/fcrz0/si793.wav 1099 | train/dr7/fjen0/si1047.wav 1100 | train/dr7/fjen0/si1677.wav 1101 | train/dr7/fjen0/si2307.wav 1102 | train/dr7/fjhk0/si1022.wav 1103 | train/dr7/fjhk0/si1652.wav 1104 | train/dr7/fjhk0/si2282.wav 1105 | train/dr7/fjrp1/si1432.wav 1106 | train/dr7/fjrp1/si2062.wav 1107 | train/dr7/fjrp1/si802.wav 1108 | train/dr7/fjsk0/si1052.wav 1109 | train/dr7/fjsk0/si1682.wav 1110 | train/dr7/fjsk0/si2312.wav 1111 | train/dr7/fkde0/si1141.wav 1112 | train/dr7/fkde0/si1771.wav 1113 | train/dr7/fkde0/si2221.wav 1114 | train/dr7/fksr0/si1117.wav 1115 | train/dr7/fksr0/si1747.wav 1116 | train/dr7/fksr0/si487.wav 1117 | train/dr7/fleh0/si1051.wav 1118 | train/dr7/fleh0/si1681.wav 1119 | train/dr7/fleh0/si2311.wav 1120 | train/dr7/flet0/si1137.wav 1121 | train/dr7/flet0/si1767.wav 1122 | train/dr7/flet0/si507.wav 1123 | train/dr7/fmah1/si1509.wav 1124 | train/dr7/fmah1/si2139.wav 1125 | train/dr7/fmah1/si879.wav 1126 | train/dr7/fmkc0/si1041.wav 1127 | train/dr7/fmkc0/si1072.wav 1128 | train/dr7/fmkc0/si1702.wav 1129 | train/dr7/fpab1/si1471.wav 1130 | train/dr7/fpab1/si2101.wav 1131 | train/dr7/fpab1/si841.wav 1132 | train/dr7/fpac0/si1921.wav 1133 | train/dr7/fpac0/si2011.wav 1134 | train/dr7/fpac0/si661.wav 1135 | train/dr7/freh0/si1315.wav 1136 | train/dr7/freh0/si1945.wav 1137 | train/dr7/freh0/si685.wav 1138 | train/dr7/fspm0/si1241.wav 1139 | train/dr7/fspm0/si1871.wav 1140 | train/dr7/fspm0/si611.wav 1141 | train/dr7/fvkb0/si1159.wav 1142 | train/dr7/fvkb0/si1789.wav 1143 | train/dr7/fvkb0/si529.wav 1144 | train/dr7/madd0/si1295.wav 1145 | train/dr7/madd0/si1798.wav 1146 | train/dr7/madd0/si538.wav 1147 | train/dr7/maeo0/si1326.wav 1148 | train/dr7/maeo0/si1655.wav 1149 | train/dr7/maeo0/si1956.wav 1150 | train/dr7/mafm0/si1569.wav 1151 | train/dr7/mafm0/si2199.wav 1152 | train/dr7/mafm0/si939.wav 1153 | train/dr7/mbar0/si1319.wav 1154 | train/dr7/mbar0/si1949.wav 1155 | train/dr7/mbar0/si689.wav 1156 | train/dr7/mbbr0/si1055.wav 1157 | train/dr7/mbbr0/si1685.wav 1158 | train/dr7/mbbr0/si2315.wav 1159 | train/dr7/mbml0/si1169.wav 1160 | train/dr7/mbml0/si1799.wav 1161 | train/dr7/mbml0/si539.wav 1162 | train/dr7/mbom0/si1014.wav 1163 | train/dr7/mbom0/si1644.wav 1164 | train/dr7/mbom0/si2274.wav 1165 | train/dr7/mbth0/si2102.wav 1166 | train/dr7/mbth0/si505.wav 1167 | train/dr7/mbth0/si757.wav 1168 | train/dr7/mclk0/si1660.wav 1169 | train/dr7/mclk0/si2290.wav 1170 | train/dr7/mclk0/si650.wav 1171 | train/dr7/mcre0/si1121.wav 1172 | train/dr7/mcre0/si1725.wav 1173 | train/dr7/mcre0/si1751.wav 1174 | train/dr7/mcth0/si1209.wav 1175 | train/dr7/mcth0/si1839.wav 1176 | train/dr7/mcth0/si579.wav 1177 | train/dr7/mdcm0/si1480.wav 1178 | train/dr7/mdcm0/si2110.wav 1179 | train/dr7/mdcm0/si850.wav 1180 | train/dr7/mded0/si1170.wav 1181 | train/dr7/mded0/si1800.wav 1182 | train/dr7/mded0/si540.wav 1183 | train/dr7/mdks0/si1066.wav 1184 | train/dr7/mdks0/si1696.wav 1185 | train/dr7/mdks0/si2326.wav 1186 | train/dr7/mdlc1/si1435.wav 1187 | train/dr7/mdlc1/si2065.wav 1188 | train/dr7/mdlc1/si2144.wav 1189 | train/dr7/mdlm0/si1234.wav 1190 | train/dr7/mdlm0/si1864.wav 1191 | train/dr7/mdlm0/si604.wav 1192 | train/dr7/mdlr0/si1233.wav 1193 | train/dr7/mdlr0/si1863.wav 1194 | train/dr7/mdlr0/si603.wav 1195 | train/dr7/mdlr1/si1299.wav 1196 | train/dr7/mdlr1/si1929.wav 1197 | train/dr7/mdlr1/si669.wav 1198 | train/dr7/mdpb0/si1760.wav 1199 | train/dr7/mdpb0/si2126.wav 1200 | train/dr7/mdpb0/si866.wav 1201 | train/dr7/mfxs0/si1674.wav 1202 | train/dr7/mfxs0/si2225.wav 1203 | train/dr7/mfxs0/si2304.wav 1204 | train/dr7/mfxv0/si1005.wav 1205 | train/dr7/mfxv0/si1342.wav 1206 | train/dr7/mfxv0/si1635.wav 1207 | train/dr7/mgak0/si1036.wav 1208 | train/dr7/mgak0/si1666.wav 1209 | train/dr7/mgak0/si2296.wav 1210 | train/dr7/mgar0/si1212.wav 1211 | train/dr7/mgar0/si1694.wav 1212 | train/dr7/mgar0/si1842.wav 1213 | train/dr7/mgaw0/si1165.wav 1214 | train/dr7/mgaw0/si1802.wav 1215 | train/dr7/mgaw0/si535.wav 1216 | train/dr7/mgsl0/si1164.wav 1217 | train/dr7/mgsl0/si534.wav 1218 | train/dr7/mgsl0/si797.wav 1219 | train/dr7/mhbs0/si1575.wav 1220 | train/dr7/mhbs0/si2205.wav 1221 | train/dr7/mhbs0/si945.wav 1222 | train/dr7/mhxl0/si1772.wav 1223 | train/dr7/mhxl0/si512.wav 1224 | train/dr7/mhxl0/si612.wav 1225 | train/dr7/mjai0/si1604.wav 1226 | train/dr7/mjai0/si682.wav 1227 | train/dr7/mjai0/si710.wav 1228 | train/dr7/mjdg0/si1042.wav 1229 | train/dr7/mjdg0/si1672.wav 1230 | train/dr7/mjdg0/si1705.wav 1231 | train/dr7/mjfr0/si1605.wav 1232 | train/dr7/mjfr0/si2235.wav 1233 | train/dr7/mjfr0/si975.wav 1234 | train/dr7/mjjm0/si1251.wav 1235 | train/dr7/mjjm0/si1457.wav 1236 | train/dr7/mjjm0/si827.wav 1237 | train/dr7/mjra0/si1236.wav 1238 | train/dr7/mjra0/si1866.wav 1239 | train/dr7/mjra0/si606.wav 1240 | train/dr7/mkag0/si1609.wav 1241 | train/dr7/mkag0/si2239.wav 1242 | train/dr7/mkag0/si979.wav 1243 | train/dr7/mkdb0/si2132.wav 1244 | train/dr7/mkdb0/si588.wav 1245 | train/dr7/mkdb0/si872.wav 1246 | train/dr7/mklr0/si1059.wav 1247 | train/dr7/mklr0/si1689.wav 1248 | train/dr7/mklr0/si2319.wav 1249 | train/dr7/mmdg0/si1780.wav 1250 | train/dr7/mmdg0/si2035.wav 1251 | train/dr7/mmdg0/si520.wav 1252 | train/dr7/mmws1/si1071.wav 1253 | train/dr7/mmws1/si1701.wav 1254 | train/dr7/mmws1/si2331.wav 1255 | train/dr7/mntw0/si1068.wav 1256 | train/dr7/mntw0/si1698.wav 1257 | train/dr7/mntw0/si2328.wav 1258 | train/dr7/mpar0/si1576.wav 1259 | train/dr7/mpar0/si2206.wav 1260 | train/dr7/mpar0/si946.wav 1261 | train/dr7/mpfu0/si1258.wav 1262 | train/dr7/mpfu0/si1888.wav 1263 | train/dr7/mpfu0/si628.wav 1264 | train/dr7/mrem0/si1591.wav 1265 | train/dr7/mrem0/si511.wav 1266 | train/dr7/mrem0/si961.wav 1267 | train/dr7/mrlj1/si1671.wav 1268 | train/dr7/mrlj1/si2301.wav 1269 | train/dr7/mrlj1/si2332.wav 1270 | train/dr7/mrmg0/si1080.wav 1271 | train/dr7/mrmg0/si1710.wav 1272 | train/dr7/mrmg0/si2340.wav 1273 | train/dr7/mrmh0/si1021.wav 1274 | train/dr7/mrmh0/si1349.wav 1275 | train/dr7/mrmh0/si2281.wav 1276 | train/dr7/mrpc1/si1482.wav 1277 | train/dr7/mrpc1/si2026.wav 1278 | train/dr7/mrpc1/si2112.wav 1279 | train/dr7/msah1/si1049.wav 1280 | train/dr7/msah1/si1679.wav 1281 | train/dr7/msah1/si2309.wav 1282 | train/dr7/msdb0/si1007.wav 1283 | train/dr7/msdb0/si1637.wav 1284 | train/dr7/msdb0/si2267.wav 1285 | train/dr7/mses0/si1589.wav 1286 | train/dr7/mses0/si2216.wav 1287 | train/dr7/mses0/si2219.wav 1288 | train/dr7/mtab0/si1572.wav 1289 | train/dr7/mtab0/si2202.wav 1290 | train/dr7/mtab0/si942.wav 1291 | train/dr7/mter0/si1157.wav 1292 | train/dr7/mter0/si1787.wav 1293 | train/dr7/mter0/si527.wav 1294 | train/dr7/mtkd0/si1187.wav 1295 | train/dr7/mtkd0/si1817.wav 1296 | train/dr7/mtkd0/si630.wav 1297 | train/dr7/mtlc0/si1313.wav 1298 | train/dr7/mtlc0/si1477.wav 1299 | train/dr7/mtlc0/si847.wav 1300 | train/dr7/mtml0/si1065.wav 1301 | train/dr7/mtml0/si1695.wav 1302 | train/dr7/mtml0/si2325.wav 1303 | train/dr7/mtmn0/si1064.wav 1304 | train/dr7/mtmn0/si2324.wav 1305 | train/dr7/mtmn0/si582.wav 1306 | train/dr7/mtpr0/si1600.wav 1307 | train/dr7/mtpr0/si2230.wav 1308 | train/dr7/mtpr0/si506.wav 1309 | train/dr7/mtwh1/si1512.wav 1310 | train/dr7/mtwh1/si2142.wav 1311 | train/dr7/mtwh1/si882.wav 1312 | train/dr7/mvrw0/si1485.wav 1313 | train/dr7/mvrw0/si2115.wav 1314 | train/dr7/mvrw0/si855.wav 1315 | train/dr7/mwre0/si1057.wav 1316 | train/dr7/mwre0/si1687.wav 1317 | train/dr7/mwre0/si2317.wav 1318 | train/dr7/mwrp0/si1443.wav 1319 | train/dr7/mwrp0/si1525.wav 1320 | train/dr7/mwrp0/si2073.wav 1321 | train/dr8/fbcg1/si1612.wav 1322 | train/dr8/fbcg1/si2242.wav 1323 | train/dr8/fbcg1/si982.wav 1324 | train/dr8/fceg0/si1248.wav 1325 | train/dr8/fceg0/si1878.wav 1326 | train/dr8/fceg0/si618.wav 1327 | train/dr8/fclt0/si1438.wav 1328 | train/dr8/fclt0/si2068.wav 1329 | train/dr8/fclt0/si808.wav 1330 | train/dr8/fjrb0/si1302.wav 1331 | train/dr8/fjrb0/si1932.wav 1332 | train/dr8/fjrb0/si672.wav 1333 | train/dr8/fklh0/si1257.wav 1334 | train/dr8/fklh0/si1887.wav 1335 | train/dr8/fklh0/si627.wav 1336 | train/dr8/fmbg0/si1160.wav 1337 | train/dr8/fmbg0/si1790.wav 1338 | train/dr8/fmbg0/si2264.wav 1339 | train/dr8/fnkl0/si1522.wav 1340 | train/dr8/fnkl0/si2152.wav 1341 | train/dr8/fnkl0/si892.wav 1342 | train/dr8/fpls0/si1590.wav 1343 | train/dr8/fpls0/si2220.wav 1344 | train/dr8/fpls0/si960.wav 1345 | train/dr8/mbcg0/si2217.wav 1346 | train/dr8/mbcg0/si486.wav 1347 | train/dr8/mbcg0/si957.wav 1348 | train/dr8/mbsb0/si1353.wav 1349 | train/dr8/mbsb0/si1983.wav 1350 | train/dr8/mbsb0/si723.wav 1351 | train/dr8/mcxm0/si1351.wav 1352 | train/dr8/mcxm0/si1981.wav 1353 | train/dr8/mcxm0/si721.wav 1354 | train/dr8/mejs0/si1240.wav 1355 | train/dr8/mejs0/si1870.wav 1356 | train/dr8/mejs0/si610.wav 1357 | train/dr8/mkdd0/si1567.wav 1358 | train/dr8/mkdd0/si2197.wav 1359 | train/dr8/mkdd0/si937.wav 1360 | train/dr8/mkrg0/si1491.wav 1361 | train/dr8/mkrg0/si2121.wav 1362 | train/dr8/mkrg0/si861.wav 1363 | train/dr8/mmea0/si1388.wav 1364 | train/dr8/mmea0/si2018.wav 1365 | train/dr8/mmea0/si758.wav 1366 | train/dr8/mmlm0/si1527.wav 1367 | train/dr8/mmlm0/si2150.wav 1368 | train/dr8/mmlm0/si897.wav 1369 | train/dr8/mmpm0/si1061.wav 1370 | train/dr8/mmpm0/si1691.wav 1371 | train/dr8/mmpm0/si2321.wav 1372 | train/dr8/mmws0/si1518.wav 1373 | train/dr8/mmws0/si559.wav 1374 | train/dr8/mmws0/si888.wav 1375 | train/dr8/mrdm0/si1044.wav 1376 | train/dr8/mrdm0/si1595.wav 1377 | train/dr8/mrdm0/si965.wav 1378 | train/dr8/mrlk0/si1468.wav 1379 | train/dr8/mrlk0/si2140.wav 1380 | train/dr8/mrlk0/si843.wav 1381 | train/dr8/mrre0/si1334.wav 1382 | train/dr8/mrre0/si704.wav 1383 | train/dr8/mrre0/si952.wav 1384 | train/dr8/mtcs0/si1972.wav 1385 | train/dr8/mtcs0/si2265.wav 1386 | train/dr8/mtcs0/si712.wav 1387 | -------------------------------------------------------------------------------- /dnn_models.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | import torch 3 | import torch.nn.functional as F 4 | import torch.nn as nn 5 | import sys 6 | from torch.autograd import Variable 7 | import math 8 | 9 | def flip(x, dim): 10 | xsize = x.size() 11 | dim = x.dim() + dim if dim < 0 else dim 12 | x = x.contiguous() 13 | x = x.view(-1, *xsize[dim:]) 14 | x = x.view(x.size(0), x.size(1), -1)[:, getattr(torch.arange(x.size(1)-1, 15 | -1, -1), ('cpu','cuda')[x.is_cuda])().long(), :] 16 | return x.view(xsize) 17 | 18 | 19 | def sinc(band,t_right): 20 | y_right= torch.sin(2*math.pi*band*t_right)/(2*math.pi*band*t_right) 21 | y_left= flip(y_right,0) 22 | 23 | y=torch.cat([y_left,Variable(torch.ones(1)).cuda(),y_right]) 24 | 25 | return y 26 | 27 | class sinc_conv(nn.Module): 28 | 29 | def __init__(self, N_filt,Filt_dim,fs): 30 | super(sinc_conv,self).__init__() 31 | 32 | # Mel Initialization of the filterbanks 33 | low_freq_mel = 80 34 | high_freq_mel = (2595 * np.log10(1 + (fs / 2) / 700)) # Convert Hz to Mel 35 | mel_points = np.linspace(low_freq_mel, high_freq_mel, N_filt) # Equally spaced in Mel scale 36 | f_cos = (700 * (10**(mel_points / 2595) - 1)) # Convert Mel to Hz 37 | b1=np.roll(f_cos,1) 38 | b2=np.roll(f_cos,-1) 39 | b1[0]=30 40 | b2[-1]=(fs/2)-100 41 | 42 | self.freq_scale=fs*1.0 43 | self.filt_b1 = nn.Parameter(torch.from_numpy(b1/self.freq_scale)) 44 | self.filt_band = nn.Parameter(torch.from_numpy((b2-b1)/self.freq_scale)) 45 | 46 | 47 | self.N_filt=N_filt 48 | self.Filt_dim=Filt_dim 49 | self.fs=fs 50 | 51 | 52 | def forward(self, x): 53 | 54 | filters=Variable(torch.zeros((self.N_filt,self.Filt_dim))).cuda() 55 | N=self.Filt_dim 56 | t_right=Variable(torch.linspace(1, (N-1)/2, steps=int((N-1)/2))/self.fs).cuda() 57 | 58 | 59 | min_freq=50.0; 60 | min_band=50.0; 61 | 62 | filt_beg_freq=torch.abs(self.filt_b1)+min_freq/self.freq_scale 63 | filt_end_freq=filt_beg_freq+(torch.abs(self.filt_band)+min_band/self.freq_scale) 64 | 65 | n=torch.linspace(0, N, steps=N) 66 | 67 | # Filter window (hamming) 68 | window=0.54-0.46*torch.cos(2*math.pi*n/N); 69 | window=Variable(window.float().cuda()) 70 | 71 | 72 | for i in range(self.N_filt): 73 | 74 | low_pass1 = 2*filt_beg_freq[i].float()*sinc(filt_beg_freq[i].float()*self.freq_scale,t_right) 75 | low_pass2 = 2*filt_end_freq[i].float()*sinc(filt_end_freq[i].float()*self.freq_scale,t_right) 76 | band_pass=(low_pass2-low_pass1) 77 | 78 | band_pass=band_pass/torch.max(band_pass) 79 | 80 | filters[i,:]=band_pass.cuda()*window 81 | 82 | out=F.conv1d(x, filters.view(self.N_filt,1,self.Filt_dim)) 83 | 84 | return out 85 | 86 | 87 | def act_fun(act_type): 88 | 89 | if act_type=="relu": 90 | return nn.ReLU() 91 | 92 | if act_type=="tanh": 93 | return nn.Tanh() 94 | 95 | if act_type=="sigmoid": 96 | return nn.Sigmoid() 97 | 98 | if act_type=="leaky_relu": 99 | return nn.LeakyReLU(0.2) 100 | 101 | if act_type=="elu": 102 | return nn.ELU() 103 | 104 | if act_type=="softmax": 105 | return nn.LogSoftmax(dim=1) 106 | #return AdditiveMarginSoftmax() 107 | 108 | if act_type=="linear": 109 | return nn.LeakyReLU(1) # initializzed like this, but not used in forward! 110 | 111 | class LayerNorm(nn.Module): 112 | 113 | def __init__(self, features, eps=1e-6): 114 | super(LayerNorm,self).__init__() 115 | self.gamma = nn.Parameter(torch.ones(features)) 116 | self.beta = nn.Parameter(torch.zeros(features)) 117 | self.eps = eps 118 | 119 | def forward(self, x): 120 | mean = x.mean(-1, keepdim=True) 121 | std = x.std(-1, keepdim=True) 122 | return self.gamma * (x - mean) / (std + self.eps) + self.beta 123 | 124 | 125 | class MLP(nn.Module): 126 | def __init__(self, options): 127 | super(MLP, self).__init__() 128 | 129 | self.input_dim=int(options['input_dim']) 130 | self.fc_lay=options['fc_lay'] 131 | self.fc_drop=options['fc_drop'] 132 | self.fc_use_batchnorm=options['fc_use_batchnorm'] 133 | self.fc_use_laynorm=options['fc_use_laynorm'] 134 | self.fc_use_laynorm_inp=options['fc_use_laynorm_inp'] 135 | self.fc_use_batchnorm_inp=options['fc_use_batchnorm_inp'] 136 | self.fc_act=options['fc_act'] 137 | 138 | 139 | self.wx = nn.ModuleList([]) 140 | self.bn = nn.ModuleList([]) 141 | self.ln = nn.ModuleList([]) 142 | self.act = nn.ModuleList([]) 143 | self.drop = nn.ModuleList([]) 144 | 145 | 146 | 147 | # input layer normalization 148 | if self.fc_use_laynorm_inp: 149 | self.ln0=LayerNorm(self.input_dim) 150 | 151 | # input batch normalization 152 | if self.fc_use_batchnorm_inp: 153 | self.bn0=nn.BatchNorm1d([self.input_dim],momentum=0.05) 154 | 155 | 156 | self.N_fc_lay=len(self.fc_lay) 157 | 158 | current_input=self.input_dim 159 | 160 | # Initialization of hidden layers 161 | 162 | for i in range(self.N_fc_lay): 163 | 164 | # dropout 165 | self.drop.append(nn.Dropout(p=self.fc_drop[i])) 166 | 167 | # activation 168 | self.act.append(act_fun(self.fc_act[i])) 169 | 170 | 171 | add_bias=True 172 | 173 | # layer norm initialization 174 | self.ln.append(LayerNorm(self.fc_lay[i])) 175 | self.bn.append(nn.BatchNorm1d(self.fc_lay[i],momentum=0.05)) 176 | 177 | if self.fc_use_laynorm[i] or self.fc_use_batchnorm[i]: 178 | add_bias=False 179 | 180 | 181 | # Linear operations 182 | self.wx.append(nn.Linear(current_input, self.fc_lay[i],bias=add_bias)) 183 | 184 | # weight initialization 185 | self.wx[i].weight = torch.nn.Parameter(torch.Tensor(self.fc_lay[i],current_input).uniform_(-np.sqrt(0.01/(current_input+self.fc_lay[i])),np.sqrt(0.01/(current_input+self.fc_lay[i])))) 186 | self.wx[i].bias = torch.nn.Parameter(torch.zeros(self.fc_lay[i])) 187 | 188 | current_input=self.fc_lay[i] 189 | 190 | 191 | def forward(self, x): 192 | 193 | # Applying Layer/Batch Norm 194 | if bool(self.fc_use_laynorm_inp): 195 | x=self.ln0((x)) 196 | 197 | if bool(self.fc_use_batchnorm_inp): 198 | x=self.bn0((x)) 199 | 200 | for i in range(self.N_fc_lay): 201 | 202 | if self.fc_act[i]!='linear': 203 | 204 | if self.fc_use_laynorm[i]: 205 | x = self.drop[i](self.act[i](self.ln[i](self.wx[i](x)))) 206 | 207 | if self.fc_use_batchnorm[i]: 208 | x = self.drop[i](self.act[i](self.bn[i](self.wx[i](x)))) 209 | 210 | if self.fc_use_batchnorm[i]==False and self.fc_use_laynorm[i]==False: 211 | x = self.drop[i](self.act[i](self.wx[i](x))) 212 | 213 | else: 214 | if self.fc_use_laynorm[i]: 215 | x = self.drop[i](self.ln[i](self.wx[i](x))) 216 | 217 | if self.fc_use_batchnorm[i]: 218 | x = self.drop[i](self.bn[i](self.wx[i](x))) 219 | 220 | if self.fc_use_batchnorm[i]==False and self.fc_use_laynorm[i]==False: 221 | x = self.drop[i](self.wx[i](x)) 222 | 223 | return x 224 | 225 | 226 | 227 | class SincNet(nn.Module): 228 | 229 | def __init__(self,options): 230 | super(SincNet,self).__init__() 231 | 232 | self.cnn_N_filt=options['cnn_N_filt'] 233 | self.cnn_len_filt=options['cnn_len_filt'] 234 | self.cnn_max_pool_len=options['cnn_max_pool_len'] 235 | 236 | 237 | self.cnn_act=options['cnn_act'] 238 | self.cnn_drop=options['cnn_drop'] 239 | 240 | self.cnn_use_laynorm=options['cnn_use_laynorm'] 241 | self.cnn_use_batchnorm=options['cnn_use_batchnorm'] 242 | self.cnn_use_laynorm_inp=options['cnn_use_laynorm_inp'] 243 | self.cnn_use_batchnorm_inp=options['cnn_use_batchnorm_inp'] 244 | 245 | self.input_dim=int(options['input_dim']) 246 | 247 | self.fs=options['fs'] 248 | 249 | self.N_cnn_lay=len(options['cnn_N_filt']) 250 | self.conv = nn.ModuleList([]) 251 | self.bn = nn.ModuleList([]) 252 | self.ln = nn.ModuleList([]) 253 | self.act = nn.ModuleList([]) 254 | self.drop = nn.ModuleList([]) 255 | 256 | 257 | if self.cnn_use_laynorm_inp: 258 | self.ln0=LayerNorm(self.input_dim) 259 | 260 | if self.cnn_use_batchnorm_inp: 261 | self.bn0=nn.BatchNorm1d([self.input_dim],momentum=0.05) 262 | 263 | current_input=self.input_dim 264 | 265 | for i in range(self.N_cnn_lay): 266 | 267 | N_filt=int(self.cnn_N_filt[i]) 268 | len_filt=int(self.cnn_len_filt[i]) 269 | 270 | # dropout 271 | self.drop.append(nn.Dropout(p=self.cnn_drop[i])) 272 | 273 | # activation 274 | self.act.append(act_fun(self.cnn_act[i])) 275 | 276 | # layer norm initialization 277 | self.ln.append(LayerNorm([N_filt,int((current_input-self.cnn_len_filt[i]+1)/self.cnn_max_pool_len[i])])) 278 | 279 | self.bn.append(nn.BatchNorm1d(N_filt,int((current_input-self.cnn_len_filt[i]+1)/self.cnn_max_pool_len[i]),momentum=0.05)) 280 | 281 | 282 | if i==0: 283 | self.conv.append(sinc_conv(self.cnn_N_filt[0],self.cnn_len_filt[0],self.fs)) 284 | 285 | else: 286 | self.conv.append(nn.Conv1d(self.cnn_N_filt[i-1], self.cnn_N_filt[i], self.cnn_len_filt[i])) 287 | 288 | current_input=int((current_input-self.cnn_len_filt[i]+1)/self.cnn_max_pool_len[i]) 289 | 290 | 291 | self.out_dim=current_input*N_filt 292 | 293 | 294 | 295 | def forward(self, x): 296 | batch=x.shape[0] 297 | seq_len=x.shape[1] 298 | 299 | if bool(self.cnn_use_laynorm_inp): 300 | x=self.ln0((x)) 301 | 302 | if bool(self.cnn_use_batchnorm_inp): 303 | x=self.bn0((x)) 304 | 305 | x=x.view(batch,1,seq_len) 306 | 307 | 308 | for i in range(self.N_cnn_lay): 309 | 310 | if self.cnn_use_laynorm[i]: 311 | if i==0: 312 | x = self.drop[i](self.act[i](self.ln[i](F.max_pool1d(torch.abs(self.conv[i](x)), self.cnn_max_pool_len[i])))) 313 | else: 314 | x = self.drop[i](self.act[i](self.ln[i](F.max_pool1d(self.conv[i](x), self.cnn_max_pool_len[i])))) 315 | 316 | if self.cnn_use_batchnorm[i]: 317 | x = self.drop[i](self.act[i](self.bn[i](F.max_pool1d(self.conv[i](x), self.cnn_max_pool_len[i])))) 318 | 319 | if self.cnn_use_batchnorm[i]==False and self.cnn_use_laynorm[i]==False: 320 | x = self.drop[i](self.act[i](F.max_pool1d(self.conv[i](x), self.cnn_max_pool_len[i]))) 321 | 322 | 323 | x = x.view(batch,-1) 324 | 325 | return x 326 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=5.471452 err_tr=0.980264 loss_te=5.352653 err_te=0.972546 err_te_snt=0.945887 2 | epoch 8, loss_tr=1.643855 err_tr=0.419707 loss_te=3.316838 err_te=0.682025 err_te_snt=0.233766 3 | epoch 16, loss_tr=0.905011 err_tr=0.239795 loss_te=2.749034 err_te=0.553273 err_te_snt=0.059163 4 | epoch 24, loss_tr=0.571821 err_tr=0.154033 loss_te=2.871697 err_te=0.528443 err_te_snt=0.046176 5 | epoch 32, loss_tr=0.399153 err_tr=0.110293 loss_te=2.903849 err_te=0.502956 err_te_snt=0.036075 6 | epoch 40, loss_tr=0.303347 err_tr=0.085322 loss_te=3.067187 err_te=0.495896 err_te_snt=0.030303 7 | epoch 48, loss_tr=0.241376 err_tr=0.069150 loss_te=2.861865 err_te=0.466745 err_te_snt=0.025974 8 | epoch 56, loss_tr=0.192656 err_tr=0.056562 loss_te=2.916151 err_te=0.452041 err_te_snt=0.013709 9 | epoch 64, loss_tr=0.165501 err_tr=0.048838 loss_te=3.027639 err_te=0.454021 err_te_snt=0.017316 10 | epoch 72, loss_tr=0.138672 err_tr=0.040674 loss_te=3.675381 err_te=0.500879 err_te_snt=0.040404 11 | epoch 80, loss_tr=0.119559 err_tr=0.036348 loss_te=2.960571 err_te=0.434944 err_te_snt=0.012266 12 | epoch 88, loss_tr=0.108751 err_tr=0.032480 loss_te=3.201554 err_te=0.446222 err_te_snt=0.013709 13 | epoch 96, loss_tr=0.098862 err_tr=0.030479 loss_te=3.326365 err_te=0.448302 err_te_snt=0.010823 14 | epoch 104, loss_tr=0.088793 err_tr=0.027568 loss_te=3.583155 err_te=0.469383 err_te_snt=0.017316 15 | epoch 112, loss_tr=0.084270 err_tr=0.025645 loss_te=3.720177 err_te=0.471092 err_te_snt=0.018759 16 | epoch 120, loss_tr=0.077579 err_tr=0.023613 loss_te=3.312567 err_te=0.438810 err_te_snt=0.010101 17 | epoch 128, loss_tr=0.074391 err_tr=0.022998 loss_te=3.742078 err_te=0.457128 err_te_snt=0.013709 18 | epoch 136, loss_tr=0.068480 err_tr=0.020957 loss_te=3.830033 err_te=0.465246 err_te_snt=0.019481 19 | epoch 144, loss_tr=0.066762 err_tr=0.020400 loss_te=3.642031 err_te=0.448976 err_te_snt=0.011544 20 | epoch 152, loss_tr=0.062507 err_tr=0.019424 loss_te=3.508821 err_te=0.437030 err_te_snt=0.014430 21 | epoch 160, loss_tr=0.061322 err_tr=0.018828 loss_te=3.789496 err_te=0.465430 err_te_snt=0.017316 22 | epoch 168, loss_tr=0.059032 err_tr=0.018018 loss_te=3.674577 err_te=0.425368 err_te_snt=0.007215 23 | epoch 176, loss_tr=0.055441 err_tr=0.016406 loss_te=3.668556 err_te=0.435626 err_te_snt=0.012987 24 | epoch 184, loss_tr=0.052643 err_tr=0.016162 loss_te=3.638472 err_te=0.435609 err_te_snt=0.012266 25 | epoch 192, loss_tr=0.050823 err_tr=0.015654 loss_te=3.994011 err_te=0.468326 err_te_snt=0.009380 26 | epoch 200, loss_tr=0.047492 err_tr=0.014971 loss_te=3.761250 err_te=0.436925 err_te_snt=0.012987 27 | epoch 208, loss_tr=0.046534 err_tr=0.014229 loss_te=3.780094 err_te=0.434492 err_te_snt=0.008658 28 | epoch 216, loss_tr=0.045753 err_tr=0.014150 loss_te=3.754180 err_te=0.434183 err_te_snt=0.010823 29 | epoch 224, loss_tr=0.046134 err_tr=0.013994 loss_te=3.912216 err_te=0.425586 err_te_snt=0.008658 30 | epoch 232, loss_tr=0.042789 err_tr=0.012959 loss_te=3.708867 err_te=0.428197 err_te_snt=0.007937 31 | epoch 240, loss_tr=0.047336 err_tr=0.014209 loss_te=3.897845 err_te=0.441917 err_te_snt=0.012266 32 | epoch 248, loss_tr=0.042010 err_tr=0.013203 loss_te=4.051550 err_te=0.442765 err_te_snt=0.012266 33 | epoch 256, loss_tr=0.041905 err_tr=0.012715 loss_te=4.006833 err_te=0.453208 err_te_snt=0.010823 34 | epoch 264, loss_tr=0.039690 err_tr=0.012100 loss_te=4.303536 err_te=0.452281 err_te_snt=0.011544 35 | epoch 272, loss_tr=0.039691 err_tr=0.012070 loss_te=4.216856 err_te=0.436458 err_te_snt=0.010101 36 | epoch 280, loss_tr=0.037749 err_tr=0.011758 loss_te=4.084363 err_te=0.426902 err_te_snt=0.007937 37 | epoch 288, loss_tr=0.037360 err_tr=0.011572 loss_te=4.100496 err_te=0.436431 err_te_snt=0.009380 38 | epoch 296, loss_tr=0.038303 err_tr=0.011934 loss_te=4.204782 err_te=0.438481 err_te_snt=0.010823 39 | epoch 304, loss_tr=0.035441 err_tr=0.011074 loss_te=4.260917 err_te=0.443717 err_te_snt=0.010823 40 | epoch 312, loss_tr=0.034102 err_tr=0.010449 loss_te=4.109076 err_te=0.420779 err_te_snt=0.007937 41 | epoch 320, loss_tr=0.035728 err_tr=0.010850 loss_te=4.572606 err_te=0.463935 err_te_snt=0.012987 42 | epoch 328, loss_tr=0.033551 err_tr=0.010527 loss_te=4.424119 err_te=0.437086 err_te_snt=0.009380 43 | epoch 336, loss_tr=0.033747 err_tr=0.010088 loss_te=4.726429 err_te=0.479633 err_te_snt=0.013709 44 | epoch 344, loss_tr=0.033517 err_tr=0.009971 loss_te=4.717675 err_te=0.449958 err_te_snt=0.010823 45 | epoch 352, loss_tr=0.032505 err_tr=0.009766 loss_te=4.476956 err_te=0.446461 err_te_snt=0.011544 46 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m005/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=7.165319 err_tr=0.992510 loss_te=7.696419 err_te=0.983470 err_te_snt=0.968975 2 | epoch 8, loss_tr=5.608348 err_tr=0.553232 loss_te=7.197257 err_te=0.729832 err_te_snt=0.299423 3 | epoch 16, loss_tr=5.363404 err_tr=0.330703 loss_te=7.038791 err_te=0.608959 err_te_snt=0.113997 4 | epoch 24, loss_tr=5.246703 err_tr=0.205488 loss_te=6.910500 err_te=0.519088 err_te_snt=0.049062 5 | epoch 32, loss_tr=5.179927 err_tr=0.131348 loss_te=6.899113 err_te=0.511497 err_te_snt=0.050505 6 | epoch 40, loss_tr=5.137461 err_tr=0.083984 loss_te=6.829139 err_te=0.463778 err_te_snt=0.030303 7 | epoch 48, loss_tr=5.112072 err_tr=0.059434 loss_te=6.751622 err_te=0.412808 err_te_snt=0.014430 8 | epoch 56, loss_tr=5.092450 err_tr=0.040176 loss_te=6.701430 err_te=0.380235 err_te_snt=0.012987 9 | epoch 64, loss_tr=5.079013 err_tr=0.028574 loss_te=6.699046 err_te=0.379210 err_te_snt=0.012266 10 | epoch 72, loss_tr=5.068756 err_tr=0.021787 loss_te=6.700624 err_te=0.381235 err_te_snt=0.010101 11 | epoch 80, loss_tr=5.060963 err_tr=0.016416 loss_te=6.747684 err_te=0.415720 err_te_snt=0.020924 12 | epoch 88, loss_tr=5.053737 err_tr=0.011582 loss_te=6.638336 err_te=0.339316 err_te_snt=0.007215 13 | epoch 96, loss_tr=5.048484 err_tr=0.008936 loss_te=6.639503 err_te=0.342589 err_te_snt=0.009380 14 | epoch 104, loss_tr=5.043802 err_tr=0.006699 loss_te=6.654626 err_te=0.350024 err_te_snt=0.012266 15 | epoch 112, loss_tr=5.040381 err_tr=0.005703 loss_te=6.609321 err_te=0.323130 err_te_snt=0.007215 16 | epoch 120, loss_tr=5.036947 err_tr=0.004160 loss_te=6.665843 err_te=0.362086 err_te_snt=0.013709 17 | epoch 128, loss_tr=5.034272 err_tr=0.003340 loss_te=6.656021 err_te=0.353028 err_te_snt=0.012987 18 | epoch 136, loss_tr=5.031900 err_tr=0.002832 loss_te=6.602369 err_te=0.316945 err_te_snt=0.009380 19 | epoch 144, loss_tr=5.029726 err_tr=0.002266 loss_te=6.596703 err_te=0.311080 err_te_snt=0.007215 20 | epoch 152, loss_tr=5.028069 err_tr=0.001807 loss_te=6.579475 err_te=0.302412 err_te_snt=0.004329 21 | epoch 160, loss_tr=5.026039 err_tr=0.001504 loss_te=6.581237 err_te=0.301527 err_te_snt=0.004329 22 | epoch 168, loss_tr=5.024954 err_tr=0.001494 loss_te=6.577566 err_te=0.302805 err_te_snt=0.006494 23 | epoch 176, loss_tr=5.023540 err_tr=0.001299 loss_te=6.606863 err_te=0.320844 err_te_snt=0.007937 24 | epoch 184, loss_tr=5.022149 err_tr=0.001006 loss_te=6.563942 err_te=0.291204 err_te_snt=0.005051 25 | epoch 192, loss_tr=5.021203 err_tr=0.001006 loss_te=6.590603 err_te=0.307035 err_te_snt=0.005051 26 | epoch 200, loss_tr=5.020251 err_tr=0.000928 loss_te=6.618120 err_te=0.327231 err_te_snt=0.010823 27 | epoch 208, loss_tr=5.019353 err_tr=0.000840 loss_te=6.552083 err_te=0.282368 err_te_snt=0.005051 28 | epoch 216, loss_tr=5.018186 err_tr=0.000605 loss_te=6.578920 err_te=0.298991 err_te_snt=0.007215 29 | epoch 224, loss_tr=5.017516 err_tr=0.000508 loss_te=6.564766 err_te=0.291021 err_te_snt=0.004329 30 | epoch 232, loss_tr=5.016509 err_tr=0.000391 loss_te=6.545503 err_te=0.279746 err_te_snt=0.003608 31 | epoch 240, loss_tr=5.016015 err_tr=0.000459 loss_te=6.584795 err_te=0.302023 err_te_snt=0.006494 32 | epoch 248, loss_tr=5.015352 err_tr=0.000312 loss_te=6.595364 err_te=0.309818 err_te_snt=0.007215 33 | epoch 256, loss_tr=5.014683 err_tr=0.000264 loss_te=6.624124 err_te=0.326484 err_te_snt=0.007937 34 | epoch 264, loss_tr=5.014106 err_tr=0.000322 loss_te=6.552195 err_te=0.283481 err_te_snt=0.003608 35 | epoch 272, loss_tr=5.013628 err_tr=0.000332 loss_te=6.581553 err_te=0.300330 err_te_snt=0.006494 36 | epoch 280, loss_tr=5.013073 err_tr=0.000273 loss_te=6.556212 err_te=0.284967 err_te_snt=0.004329 37 | epoch 288, loss_tr=5.012586 err_tr=0.000225 loss_te=6.536858 err_te=0.273982 err_te_snt=0.004329 38 | epoch 296, loss_tr=5.012124 err_tr=0.000322 loss_te=6.546865 err_te=0.277666 err_te_snt=0.004329 39 | epoch 304, loss_tr=5.011715 err_tr=0.000234 loss_te=6.543432 err_te=0.276772 err_te_snt=0.005051 40 | epoch 312, loss_tr=5.011271 err_tr=0.000234 loss_te=6.573205 err_te=0.297363 err_te_snt=0.006494 41 | epoch 320, loss_tr=5.010700 err_tr=0.000176 loss_te=6.544925 err_te=0.277900 err_te_snt=0.005772 42 | epoch 328, loss_tr=5.010506 err_tr=0.000146 loss_te=6.566723 err_te=0.292154 err_te_snt=0.005772 43 | epoch 336, loss_tr=5.010048 err_tr=0.000117 loss_te=6.530147 err_te=0.268077 err_te_snt=0.002165 44 | epoch 344, loss_tr=5.009772 err_tr=0.000186 loss_te=6.539469 err_te=0.272379 err_te_snt=0.004329 45 | epoch 352, loss_tr=5.009363 err_tr=0.000098 loss_te=6.537186 err_te=0.270637 err_te_snt=0.005051 46 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m010/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=8.684309 err_tr=0.992480 loss_te=9.207470 err_te=0.986193 err_te_snt=0.974747 2 | epoch 8, loss_tr=7.128023 err_tr=0.574648 loss_te=8.671103 err_te=0.711233 err_te_snt=0.197691 3 | epoch 16, loss_tr=6.867762 err_tr=0.340986 loss_te=8.515574 err_te=0.592181 err_te_snt=0.088745 4 | epoch 24, loss_tr=6.745337 err_tr=0.208555 loss_te=8.431546 err_te=0.532689 err_te_snt=0.054834 5 | epoch 32, loss_tr=6.676165 err_tr=0.131943 loss_te=8.365090 err_te=0.484394 err_te_snt=0.034632 6 | epoch 40, loss_tr=6.634643 err_tr=0.087520 loss_te=8.304257 err_te=0.448311 err_te_snt=0.023810 7 | epoch 48, loss_tr=6.608153 err_tr=0.058584 loss_te=8.314785 err_te=0.456807 err_te_snt=0.037518 8 | epoch 56, loss_tr=6.587384 err_tr=0.039658 loss_te=8.223672 err_te=0.395980 err_te_snt=0.015873 9 | epoch 64, loss_tr=6.573398 err_tr=0.027959 loss_te=8.216445 err_te=0.393995 err_te_snt=0.013709 10 | epoch 72, loss_tr=6.563238 err_tr=0.020811 loss_te=8.225517 err_te=0.399428 err_te_snt=0.014430 11 | epoch 80, loss_tr=6.554845 err_tr=0.015020 loss_te=8.161489 err_te=0.357195 err_te_snt=0.008658 12 | epoch 88, loss_tr=6.547846 err_tr=0.010908 loss_te=8.185444 err_te=0.374113 err_te_snt=0.010823 13 | epoch 96, loss_tr=6.542544 err_tr=0.008359 loss_te=8.144538 err_te=0.345429 err_te_snt=0.007937 14 | epoch 104, loss_tr=6.537790 err_tr=0.006143 loss_te=8.169119 err_te=0.361471 err_te_snt=0.013709 15 | epoch 112, loss_tr=6.534246 err_tr=0.005088 loss_te=8.126130 err_te=0.333242 err_te_snt=0.012987 16 | epoch 120, loss_tr=6.530890 err_tr=0.004121 loss_te=8.122930 err_te=0.330406 err_te_snt=0.007215 17 | epoch 128, loss_tr=6.528433 err_tr=0.003584 loss_te=8.128399 err_te=0.336577 err_te_snt=0.008658 18 | epoch 136, loss_tr=6.526052 err_tr=0.002783 loss_te=8.103005 err_te=0.316921 err_te_snt=0.008658 19 | epoch 144, loss_tr=6.523769 err_tr=0.001924 loss_te=8.076916 err_te=0.301499 err_te_snt=0.006494 20 | epoch 152, loss_tr=6.522068 err_tr=0.001836 loss_te=8.100381 err_te=0.316730 err_te_snt=0.007937 21 | epoch 160, loss_tr=6.520308 err_tr=0.001436 loss_te=8.098194 err_te=0.312496 err_te_snt=0.006494 22 | epoch 168, loss_tr=6.519077 err_tr=0.001201 loss_te=8.092541 err_te=0.312516 err_te_snt=0.007937 23 | epoch 176, loss_tr=6.517595 err_tr=0.001094 loss_te=8.076816 err_te=0.299129 err_te_snt=0.006494 24 | epoch 184, loss_tr=6.516472 err_tr=0.000977 loss_te=8.069084 err_te=0.296833 err_te_snt=0.005051 25 | epoch 192, loss_tr=6.515477 err_tr=0.000889 loss_te=8.062324 err_te=0.290642 err_te_snt=0.004329 26 | epoch 200, loss_tr=6.514348 err_tr=0.000723 loss_te=8.074215 err_te=0.299657 err_te_snt=0.004329 27 | epoch 208, loss_tr=6.513477 err_tr=0.000625 loss_te=8.058772 err_te=0.288290 err_te_snt=0.002886 28 | epoch 216, loss_tr=6.512595 err_tr=0.000732 loss_te=8.067327 err_te=0.293207 err_te_snt=0.004329 29 | epoch 224, loss_tr=6.511906 err_tr=0.000469 loss_te=8.049620 err_te=0.283101 err_te_snt=0.004329 30 | epoch 232, loss_tr=6.510880 err_tr=0.000361 loss_te=8.099556 err_te=0.313977 err_te_snt=0.007215 31 | epoch 240, loss_tr=6.510486 err_tr=0.000430 loss_te=8.074638 err_te=0.296679 err_te_snt=0.007215 32 | epoch 248, loss_tr=6.509695 err_tr=0.000244 loss_te=8.061823 err_te=0.288746 err_te_snt=0.004329 33 | epoch 256, loss_tr=6.509089 err_tr=0.000352 loss_te=8.076419 err_te=0.298085 err_te_snt=0.007215 34 | epoch 264, loss_tr=6.508498 err_tr=0.000264 loss_te=8.054988 err_te=0.285732 err_te_snt=0.004329 35 | epoch 272, loss_tr=6.508002 err_tr=0.000283 loss_te=8.057830 err_te=0.287793 err_te_snt=0.006494 36 | epoch 280, loss_tr=6.507598 err_tr=0.000244 loss_te=8.076240 err_te=0.295780 err_te_snt=0.007937 37 | epoch 288, loss_tr=6.507060 err_tr=0.000215 loss_te=8.043901 err_te=0.277317 err_te_snt=0.003608 38 | epoch 296, loss_tr=6.506594 err_tr=0.000215 loss_te=8.046594 err_te=0.278780 err_te_snt=0.003608 39 | epoch 304, loss_tr=6.506148 err_tr=0.000215 loss_te=8.042494 err_te=0.278081 err_te_snt=0.004329 40 | epoch 312, loss_tr=6.505764 err_tr=0.000186 loss_te=8.054591 err_te=0.284271 err_te_snt=0.007215 41 | epoch 320, loss_tr=6.505363 err_tr=0.000186 loss_te=8.053020 err_te=0.281943 err_te_snt=0.002886 42 | epoch 328, loss_tr=6.504943 err_tr=0.000176 loss_te=8.046471 err_te=0.277771 err_te_snt=0.005051 43 | epoch 336, loss_tr=6.504746 err_tr=0.000107 loss_te=8.033590 err_te=0.270851 err_te_snt=0.003608 44 | epoch 344, loss_tr=6.504409 err_tr=0.000186 loss_te=8.043655 err_te=0.276882 err_te_snt=0.003608 45 | epoch 352, loss_tr=6.503916 err_tr=0.000176 loss_te=8.065159 err_te=0.289667 err_te_snt=0.005051 46 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m015/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=10.167981 err_tr=0.992588 loss_te=10.704318 err_te=0.991073 err_te_snt=0.988456 2 | epoch 8, loss_tr=8.603772 err_tr=0.549834 loss_te=10.124459 err_te=0.673905 err_te_snt=0.129149 3 | epoch 16, loss_tr=8.348659 err_tr=0.320811 loss_te=10.068061 err_te=0.631732 err_te_snt=0.162338 4 | epoch 24, loss_tr=8.231871 err_tr=0.196748 loss_te=9.865870 err_te=0.484859 err_te_snt=0.024531 5 | epoch 32, loss_tr=8.165555 err_tr=0.122217 loss_te=9.819831 err_te=0.457118 err_te_snt=0.025974 6 | epoch 40, loss_tr=8.125192 err_tr=0.078760 loss_te=9.821972 err_te=0.460690 err_te_snt=0.025974 7 | epoch 48, loss_tr=8.099962 err_tr=0.053643 loss_te=9.745125 err_te=0.409508 err_te_snt=0.014430 8 | epoch 56, loss_tr=8.080460 err_tr=0.035986 loss_te=9.712397 err_te=0.388385 err_te_snt=0.010823 9 | epoch 56, loss_tr=8.080943 err_tr=0.036172 loss_te=9.765769 err_te=0.425717 err_te_snt=0.034632 10 | epoch 64, loss_tr=8.068219 err_tr=0.026748 loss_te=9.758327 err_te=0.423475 err_te_snt=0.023810 11 | epoch 72, loss_tr=8.057103 err_tr=0.017783 loss_te=9.667294 err_te=0.358319 err_te_snt=0.009380 12 | epoch 80, loss_tr=8.049166 err_tr=0.012598 loss_te=9.653979 err_te=0.351508 err_te_snt=0.009380 13 | epoch 88, loss_tr=8.043018 err_tr=0.009482 loss_te=9.670422 err_te=0.363674 err_te_snt=0.010823 14 | epoch 96, loss_tr=8.037906 err_tr=0.006924 loss_te=9.660546 err_te=0.357324 err_te_snt=0.012266 15 | epoch 104, loss_tr=8.035010 err_tr=0.006406 loss_te=9.621144 err_te=0.330869 err_te_snt=0.009380 16 | epoch 112, loss_tr=8.031785 err_tr=0.005557 loss_te=9.614733 err_te=0.325675 err_te_snt=0.010101 17 | epoch 120, loss_tr=8.028734 err_tr=0.004082 loss_te=9.601480 err_te=0.317305 err_te_snt=0.008658 18 | epoch 128, loss_tr=8.026419 err_tr=0.003477 loss_te=9.631982 err_te=0.339467 err_te_snt=0.007937 19 | epoch 136, loss_tr=8.023536 err_tr=0.002383 loss_te=9.597156 err_te=0.315085 err_te_snt=0.007215 20 | epoch 144, loss_tr=8.021737 err_tr=0.002422 loss_te=9.579190 err_te=0.305646 err_te_snt=0.004329 21 | epoch 152, loss_tr=8.020322 err_tr=0.002109 loss_te=9.561045 err_te=0.292443 err_te_snt=0.003608 22 | epoch 160, loss_tr=8.018994 err_tr=0.001826 loss_te=9.588040 err_te=0.308510 err_te_snt=0.007937 23 | epoch 168, loss_tr=8.017251 err_tr=0.001299 loss_te=9.559177 err_te=0.289165 err_te_snt=0.005772 24 | epoch 176, loss_tr=8.016410 err_tr=0.001396 loss_te=9.580623 err_te=0.304004 err_te_snt=0.007937 25 | epoch 184, loss_tr=8.014762 err_tr=0.000898 loss_te=9.572680 err_te=0.297527 err_te_snt=0.006494 26 | epoch 192, loss_tr=8.014073 err_tr=0.000898 loss_te=9.569499 err_te=0.295183 err_te_snt=0.004329 27 | epoch 200, loss_tr=8.013089 err_tr=0.000889 loss_te=9.560384 err_te=0.291890 err_te_snt=0.006494 28 | epoch 208, loss_tr=8.012019 err_tr=0.000664 loss_te=9.569747 err_te=0.292149 err_te_snt=0.005051 29 | epoch 216, loss_tr=8.011354 err_tr=0.000752 loss_te=9.553350 err_te=0.285671 err_te_snt=0.004329 30 | epoch 224, loss_tr=8.010604 err_tr=0.000576 loss_te=9.553522 err_te=0.286376 err_te_snt=0.005051 31 | epoch 232, loss_tr=8.010016 err_tr=0.000664 loss_te=9.547487 err_te=0.281305 err_te_snt=0.004329 32 | epoch 240, loss_tr=8.009106 err_tr=0.000508 loss_te=9.548047 err_te=0.280827 err_te_snt=0.005051 33 | epoch 248, loss_tr=8.008578 err_tr=0.000508 loss_te=9.575784 err_te=0.299515 err_te_snt=0.005772 34 | epoch 256, loss_tr=8.008050 err_tr=0.000332 loss_te=9.570248 err_te=0.294732 err_te_snt=0.005772 35 | epoch 264, loss_tr=8.007357 err_tr=0.000283 loss_te=9.560033 err_te=0.291328 err_te_snt=0.004329 36 | epoch 272, loss_tr=8.007026 err_tr=0.000381 loss_te=9.558858 err_te=0.288204 err_te_snt=0.006494 37 | epoch 280, loss_tr=8.006484 err_tr=0.000215 loss_te=9.578096 err_te=0.299687 err_te_snt=0.005772 38 | epoch 288, loss_tr=8.006071 err_tr=0.000283 loss_te=9.530560 err_te=0.273535 err_te_snt=0.004329 39 | epoch 296, loss_tr=8.005646 err_tr=0.000234 loss_te=9.528425 err_te=0.270740 err_te_snt=0.005051 40 | epoch 304, loss_tr=8.005243 err_tr=0.000312 loss_te=9.535586 err_te=0.274702 err_te_snt=0.002886 41 | epoch 312, loss_tr=8.004684 err_tr=0.000166 loss_te=9.525249 err_te=0.268919 err_te_snt=0.004329 42 | epoch 320, loss_tr=8.004283 err_tr=0.000195 loss_te=9.540863 err_te=0.274731 err_te_snt=0.002165 43 | epoch 328, loss_tr=8.004045 err_tr=0.000215 loss_te=9.550922 err_te=0.282861 err_te_snt=0.004329 44 | epoch 336, loss_tr=8.003577 err_tr=0.000186 loss_te=9.540681 err_te=0.277159 err_te_snt=0.004329 45 | epoch 344, loss_tr=8.003511 err_tr=0.000107 loss_te=9.536006 err_te=0.272290 err_te_snt=0.002165 46 | epoch 352, loss_tr=8.003027 err_tr=0.000127 loss_te=9.540941 err_te=0.276222 err_te_snt=0.002165 47 | epoch 360, loss_tr=8.002729 err_tr=0.000146 loss_te=9.561654 err_te=0.289171 err_te_snt=0.005772 48 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m020/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=11.683121 err_tr=0.992852 loss_te=12.208253 err_te=0.990967 err_te_snt=0.987734 2 | epoch 8, loss_tr=10.062078 err_tr=0.512871 loss_te=11.651130 err_te=0.691000 err_te_snt=0.223665 3 | epoch 16, loss_tr=9.820509 err_tr=0.290986 loss_te=11.541439 err_te=0.611895 err_te_snt=0.120491 4 | epoch 24, loss_tr=9.714058 err_tr=0.175908 loss_te=11.367737 err_te=0.487838 err_te_snt=0.034632 5 | epoch 32, loss_tr=9.652985 err_tr=0.108672 loss_te=11.293821 err_te=0.439856 err_te_snt=0.022367 6 | epoch 40, loss_tr=9.617125 err_tr=0.071426 loss_te=11.290998 err_te=0.439417 err_te_snt=0.023810 7 | epoch 48, loss_tr=9.594190 err_tr=0.048154 loss_te=11.241060 err_te=0.404585 err_te_snt=0.018759 8 | epoch 56, loss_tr=9.576721 err_tr=0.034121 loss_te=11.271969 err_te=0.431792 err_te_snt=0.032468 9 | epoch 56, loss_tr=9.576880 err_tr=0.033740 loss_te=11.270598 err_te=0.427270 err_te_snt=0.028860 10 | epoch 64, loss_tr=9.564696 err_tr=0.023838 loss_te=11.176303 err_te=0.365255 err_te_snt=0.010101 11 | epoch 72, loss_tr=9.554777 err_tr=0.016484 loss_te=11.189351 err_te=0.375312 err_te_snt=0.012987 12 | epoch 80, loss_tr=9.546655 err_tr=0.011768 loss_te=11.145478 err_te=0.345582 err_te_snt=0.013709 13 | epoch 88, loss_tr=9.540641 err_tr=0.008477 loss_te=11.156441 err_te=0.355615 err_te_snt=0.011544 14 | epoch 96, loss_tr=9.536345 err_tr=0.006768 loss_te=11.119982 err_te=0.330508 err_te_snt=0.010101 15 | epoch 104, loss_tr=9.533257 err_tr=0.005674 loss_te=11.100846 err_te=0.317834 err_te_snt=0.006494 16 | epoch 112, loss_tr=9.530125 err_tr=0.004834 loss_te=11.092860 err_te=0.312897 err_te_snt=0.005772 17 | epoch 120, loss_tr=9.527232 err_tr=0.003574 loss_te=11.119978 err_te=0.327076 err_te_snt=0.008658 18 | epoch 128, loss_tr=9.525078 err_tr=0.003281 loss_te=11.094512 err_te=0.314701 err_te_snt=0.006494 19 | epoch 136, loss_tr=9.522354 err_tr=0.002266 loss_te=11.095909 err_te=0.315732 err_te_snt=0.005772 20 | epoch 144, loss_tr=9.520540 err_tr=0.002188 loss_te=11.090819 err_te=0.310356 err_te_snt=0.006494 21 | epoch 152, loss_tr=9.519281 err_tr=0.001943 loss_te=11.078888 err_te=0.304649 err_te_snt=0.005772 22 | epoch 160, loss_tr=9.517797 err_tr=0.001611 loss_te=11.067382 err_te=0.297053 err_te_snt=0.003608 23 | epoch 168, loss_tr=9.516322 err_tr=0.001279 loss_te=11.076224 err_te=0.301114 err_te_snt=0.004329 24 | epoch 176, loss_tr=9.515359 err_tr=0.001172 loss_te=11.087433 err_te=0.307338 err_te_snt=0.007215 25 | epoch 184, loss_tr=9.513715 err_tr=0.001074 loss_te=11.107254 err_te=0.320213 err_te_snt=0.006494 26 | epoch 192, loss_tr=9.513050 err_tr=0.000918 loss_te=11.090931 err_te=0.310686 err_te_snt=0.007215 27 | epoch 200, loss_tr=9.512083 err_tr=0.000703 loss_te=11.057369 err_te=0.287800 err_te_snt=0.005051 28 | epoch 208, loss_tr=9.510980 err_tr=0.000459 loss_te=11.051806 err_te=0.284145 err_te_snt=0.003608 29 | epoch 216, loss_tr=9.510611 err_tr=0.000762 loss_te=11.048875 err_te=0.282268 err_te_snt=0.002886 30 | epoch 224, loss_tr=9.509553 err_tr=0.000479 loss_te=11.051225 err_te=0.284817 err_te_snt=0.002165 31 | epoch 232, loss_tr=9.508972 err_tr=0.000488 loss_te=11.043619 err_te=0.281005 err_te_snt=0.005051 32 | epoch 240, loss_tr=9.508394 err_tr=0.000322 loss_te=11.049537 err_te=0.282234 err_te_snt=0.005051 33 | epoch 248, loss_tr=9.507674 err_tr=0.000420 loss_te=11.046480 err_te=0.280303 err_te_snt=0.005772 34 | epoch 256, loss_tr=9.507180 err_tr=0.000391 loss_te=11.039262 err_te=0.277039 err_te_snt=0.002165 35 | epoch 264, loss_tr=9.506676 err_tr=0.000312 loss_te=11.045479 err_te=0.278754 err_te_snt=0.004329 36 | epoch 272, loss_tr=9.506297 err_tr=0.000410 loss_te=11.053243 err_te=0.285074 err_te_snt=0.003608 37 | epoch 280, loss_tr=9.505725 err_tr=0.000254 loss_te=11.049723 err_te=0.281646 err_te_snt=0.005772 38 | epoch 288, loss_tr=9.505179 err_tr=0.000303 loss_te=11.035223 err_te=0.273733 err_te_snt=0.002886 39 | epoch 296, loss_tr=9.504791 err_tr=0.000166 loss_te=11.043822 err_te=0.278428 err_te_snt=0.005051 40 | epoch 304, loss_tr=9.504364 err_tr=0.000176 loss_te=11.041775 err_te=0.275296 err_te_snt=0.003608 41 | epoch 312, loss_tr=9.503921 err_tr=0.000156 loss_te=11.041041 err_te=0.278007 err_te_snt=0.002886 42 | epoch 320, loss_tr=9.503572 err_tr=0.000107 loss_te=11.029708 err_te=0.270081 err_te_snt=0.001443 43 | epoch 328, loss_tr=9.503238 err_tr=0.000117 loss_te=11.025208 err_te=0.266962 err_te_snt=0.002165 44 | epoch 336, loss_tr=9.502934 err_tr=0.000312 loss_te=11.029082 err_te=0.270856 err_te_snt=0.004329 45 | epoch 344, loss_tr=9.502666 err_tr=0.000166 loss_te=11.031834 err_te=0.269893 err_te_snt=0.004329 46 | epoch 352, loss_tr=9.502450 err_tr=0.000137 loss_te=11.065407 err_te=0.289732 err_te_snt=0.004329 47 | epoch 360, loss_tr=9.502085 err_tr=0.000166 loss_te=11.029808 err_te=0.269885 err_te_snt=0.003608 48 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m025/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=13.138774 err_tr=0.989521 loss_te=13.691639 err_te=0.985154 err_te_snt=0.981963 2 | epoch 8, loss_tr=11.561216 err_tr=0.514160 loss_te=13.145939 err_te=0.688081 err_te_snt=0.200577 3 | epoch 16, loss_tr=11.326384 err_tr=0.295752 loss_te=12.987169 err_te=0.571068 err_te_snt=0.087302 4 | epoch 24, loss_tr=11.218107 err_tr=0.178896 loss_te=12.888105 err_te=0.501886 err_te_snt=0.042569 5 | epoch 32, loss_tr=11.156673 err_tr=0.112227 loss_te=12.890818 err_te=0.509383 err_te_snt=0.051227 6 | epoch 40, loss_tr=11.119387 err_tr=0.072197 loss_te=12.770693 err_te=0.423655 err_te_snt=0.015873 7 | epoch 48, loss_tr=11.096473 err_tr=0.049961 loss_te=12.732988 err_te=0.401052 err_te_snt=0.013709 8 | epoch 48, loss_tr=11.095749 err_tr=0.049004 loss_te=12.744494 err_te=0.405675 err_te_snt=0.017316 9 | epoch 56, loss_tr=11.078966 err_tr=0.034434 loss_te=12.734629 err_te=0.402679 err_te_snt=0.011544 10 | epoch 64, loss_tr=11.066153 err_tr=0.024004 loss_te=12.820859 err_te=0.466758 err_te_snt=0.044733 11 | epoch 72, loss_tr=11.056009 err_tr=0.017207 loss_te=12.685168 err_te=0.375158 err_te_snt=0.010823 12 | epoch 80, loss_tr=11.048767 err_tr=0.012275 loss_te=12.677275 err_te=0.369027 err_te_snt=0.010823 13 | epoch 88, loss_tr=11.044851 err_tr=0.010859 loss_te=12.663502 err_te=0.356274 err_te_snt=0.008658 14 | epoch 96, loss_tr=11.039476 err_tr=0.008232 loss_te=12.656884 err_te=0.353408 err_te_snt=0.010101 15 | epoch 104, loss_tr=11.036096 err_tr=0.007334 loss_te=12.661606 err_te=0.354176 err_te_snt=0.008658 16 | epoch 112, loss_tr=11.032117 err_tr=0.005186 loss_te=12.616957 err_te=0.326640 err_te_snt=0.005051 17 | epoch 120, loss_tr=11.029541 err_tr=0.004453 loss_te=12.684483 err_te=0.370197 err_te_snt=0.015873 18 | epoch 128, loss_tr=11.026399 err_tr=0.003281 loss_te=12.616687 err_te=0.330874 err_te_snt=0.006494 19 | epoch 136, loss_tr=11.024214 err_tr=0.002969 loss_te=12.596954 err_te=0.316314 err_te_snt=0.005051 20 | epoch 144, loss_tr=11.022742 err_tr=0.002764 loss_te=12.614827 err_te=0.325803 err_te_snt=0.010101 21 | epoch 152, loss_tr=11.020971 err_tr=0.002168 loss_te=12.599216 err_te=0.315106 err_te_snt=0.005051 22 | epoch 160, loss_tr=11.019444 err_tr=0.001748 loss_te=12.591361 err_te=0.310088 err_te_snt=0.005051 23 | epoch 168, loss_tr=11.018287 err_tr=0.001650 loss_te=12.609571 err_te=0.320175 err_te_snt=0.006494 24 | epoch 176, loss_tr=11.016256 err_tr=0.001113 loss_te=12.590322 err_te=0.309646 err_te_snt=0.006494 25 | epoch 184, loss_tr=11.015578 err_tr=0.001201 loss_te=12.586614 err_te=0.308334 err_te_snt=0.005772 26 | epoch 192, loss_tr=11.014815 err_tr=0.001143 loss_te=12.586162 err_te=0.304258 err_te_snt=0.005772 27 | epoch 200, loss_tr=11.013468 err_tr=0.000820 loss_te=12.572260 err_te=0.295407 err_te_snt=0.004329 28 | epoch 208, loss_tr=11.012733 err_tr=0.000830 loss_te=12.565554 err_te=0.293950 err_te_snt=0.002886 29 | epoch 216, loss_tr=11.011755 err_tr=0.000664 loss_te=12.587970 err_te=0.306313 err_te_snt=0.005051 30 | epoch 224, loss_tr=11.011136 err_tr=0.000547 loss_te=12.555874 err_te=0.286607 err_te_snt=0.007215 31 | epoch 232, loss_tr=11.010334 err_tr=0.000430 loss_te=12.573319 err_te=0.297101 err_te_snt=0.005051 32 | epoch 240, loss_tr=11.009685 err_tr=0.000439 loss_te=12.557748 err_te=0.288742 err_te_snt=0.005772 33 | epoch 248, loss_tr=11.009013 err_tr=0.000361 loss_te=12.559904 err_te=0.288283 err_te_snt=0.003608 34 | epoch 256, loss_tr=11.008668 err_tr=0.000400 loss_te=12.566543 err_te=0.292325 err_te_snt=0.005051 35 | epoch 264, loss_tr=11.008096 err_tr=0.000439 loss_te=12.569366 err_te=0.293635 err_te_snt=0.006494 36 | epoch 272, loss_tr=11.007565 err_tr=0.000400 loss_te=12.561090 err_te=0.290009 err_te_snt=0.005772 37 | epoch 280, loss_tr=11.006984 err_tr=0.000254 loss_te=12.572070 err_te=0.295337 err_te_snt=0.003608 38 | epoch 288, loss_tr=11.006498 err_tr=0.000273 loss_te=12.554837 err_te=0.282270 err_te_snt=0.002886 39 | epoch 296, loss_tr=11.006139 err_tr=0.000283 loss_te=12.546182 err_te=0.279874 err_te_snt=0.002886 40 | epoch 304, loss_tr=11.005502 err_tr=0.000225 loss_te=12.554942 err_te=0.284237 err_te_snt=0.003608 41 | epoch 312, loss_tr=11.005266 err_tr=0.000264 loss_te=12.554085 err_te=0.285072 err_te_snt=0.002886 42 | epoch 320, loss_tr=11.004796 err_tr=0.000195 loss_te=12.556213 err_te=0.286412 err_te_snt=0.002886 43 | epoch 328, loss_tr=11.004539 err_tr=0.000215 loss_te=12.539425 err_te=0.276493 err_te_snt=0.004329 44 | epoch 336, loss_tr=11.004196 err_tr=0.000137 loss_te=12.560452 err_te=0.286932 err_te_snt=0.005051 45 | epoch 344, loss_tr=11.003824 err_tr=0.000205 loss_te=12.560912 err_te=0.287683 err_te_snt=0.004329 46 | epoch 352, loss_tr=11.003458 err_tr=0.000176 loss_te=12.538014 err_te=0.274741 err_te_snt=0.003608 47 | epoch 360, loss_tr=11.003131 err_tr=0.000186 loss_te=12.566143 err_te=0.293050 err_te_snt=0.006494 48 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m030/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=14.681456 err_tr=0.992480 loss_te=15.208365 err_te=0.988352 err_te_snt=0.982684 2 | epoch 8, loss_tr=13.071479 err_tr=0.522822 loss_te=14.658555 err_te=0.699416 err_te_snt=0.230159 3 | epoch 16, loss_tr=12.832218 err_tr=0.302979 loss_te=14.481071 err_te=0.567732 err_te_snt=0.088745 4 | epoch 24, loss_tr=12.720451 err_tr=0.182852 loss_te=14.409398 err_te=0.519628 err_te_snt=0.067100 5 | epoch 32, loss_tr=12.657666 err_tr=0.114424 loss_te=14.293962 err_te=0.438791 err_te_snt=0.018038 6 | epoch 40, loss_tr=12.619868 err_tr=0.074043 loss_te=14.259212 err_te=0.415978 err_te_snt=0.012987 7 | epoch 48, loss_tr=12.596425 err_tr=0.050918 loss_te=14.231331 err_te=0.400277 err_te_snt=0.015152 8 | epoch 48, loss_tr=12.595761 err_tr=0.050107 loss_te=14.266138 err_te=0.424873 err_te_snt=0.023088 9 | epoch 56, loss_tr=12.578827 err_tr=0.035088 loss_te=14.247141 err_te=0.414561 err_te_snt=0.018038 10 | epoch 64, loss_tr=12.565794 err_tr=0.024082 loss_te=14.214602 err_te=0.391964 err_te_snt=0.012266 11 | epoch 72, loss_tr=12.555547 err_tr=0.016436 loss_te=14.169896 err_te=0.361350 err_te_snt=0.012987 12 | epoch 80, loss_tr=12.548080 err_tr=0.012510 loss_te=14.141914 err_te=0.345520 err_te_snt=0.012266 13 | epoch 88, loss_tr=12.544434 err_tr=0.010693 loss_te=14.147778 err_te=0.348376 err_te_snt=0.009380 14 | epoch 96, loss_tr=12.538944 err_tr=0.008066 loss_te=14.128584 err_te=0.335549 err_te_snt=0.008658 15 | epoch 104, loss_tr=12.535051 err_tr=0.006367 loss_te=14.131093 err_te=0.335838 err_te_snt=0.010101 16 | epoch 112, loss_tr=12.531675 err_tr=0.005439 loss_te=14.132043 err_te=0.337709 err_te_snt=0.008658 17 | epoch 120, loss_tr=12.528814 err_tr=0.004551 loss_te=14.108083 err_te=0.324097 err_te_snt=0.007937 18 | epoch 128, loss_tr=12.525944 err_tr=0.003242 loss_te=14.123873 err_te=0.333591 err_te_snt=0.011544 19 | epoch 136, loss_tr=12.523390 err_tr=0.002549 loss_te=14.100324 err_te=0.315902 err_te_snt=0.005772 20 | epoch 144, loss_tr=12.522059 err_tr=0.002344 loss_te=14.109303 err_te=0.320953 err_te_snt=0.008658 21 | epoch 152, loss_tr=12.520431 err_tr=0.002080 loss_te=14.085230 err_te=0.305898 err_te_snt=0.005051 22 | epoch 160, loss_tr=12.518931 err_tr=0.001660 loss_te=14.082409 err_te=0.304424 err_te_snt=0.005051 23 | epoch 168, loss_tr=12.517639 err_tr=0.001387 loss_te=14.090055 err_te=0.308300 err_te_snt=0.007215 24 | epoch 176, loss_tr=12.516023 err_tr=0.001221 loss_te=14.077078 err_te=0.301905 err_te_snt=0.005051 25 | epoch 184, loss_tr=12.515079 err_tr=0.001006 loss_te=14.076969 err_te=0.302003 err_te_snt=0.007215 26 | epoch 192, loss_tr=12.514192 err_tr=0.001035 loss_te=14.072797 err_te=0.297994 err_te_snt=0.004329 27 | epoch 200, loss_tr=12.513116 err_tr=0.000732 loss_te=14.069839 err_te=0.295870 err_te_snt=0.004329 28 | epoch 208, loss_tr=12.512251 err_tr=0.000752 loss_te=14.074738 err_te=0.300292 err_te_snt=0.005772 29 | epoch 216, loss_tr=12.511398 err_tr=0.000547 loss_te=14.065916 err_te=0.291792 err_te_snt=0.005051 30 | epoch 224, loss_tr=12.510851 err_tr=0.000732 loss_te=14.071936 err_te=0.296030 err_te_snt=0.003608 31 | epoch 232, loss_tr=12.510091 err_tr=0.000547 loss_te=14.061568 err_te=0.289670 err_te_snt=0.004329 32 | epoch 240, loss_tr=12.509187 err_tr=0.000469 loss_te=14.064492 err_te=0.290669 err_te_snt=0.004329 33 | epoch 248, loss_tr=12.508672 err_tr=0.000479 loss_te=14.133069 err_te=0.331126 err_te_snt=0.008658 34 | epoch 256, loss_tr=12.508324 err_tr=0.000449 loss_te=14.080459 err_te=0.301455 err_te_snt=0.005051 35 | epoch 264, loss_tr=12.507621 err_tr=0.000332 loss_te=14.078759 err_te=0.300581 err_te_snt=0.005051 36 | epoch 272, loss_tr=12.506976 err_tr=0.000244 loss_te=14.072167 err_te=0.296790 err_te_snt=0.005051 37 | epoch 280, loss_tr=12.506754 err_tr=0.000332 loss_te=14.057206 err_te=0.286935 err_te_snt=0.004329 38 | epoch 288, loss_tr=12.506135 err_tr=0.000283 loss_te=14.056434 err_te=0.285263 err_te_snt=0.006494 39 | epoch 296, loss_tr=12.505715 err_tr=0.000264 loss_te=14.066737 err_te=0.290573 err_te_snt=0.004329 40 | epoch 304, loss_tr=12.505173 err_tr=0.000195 loss_te=14.075700 err_te=0.299782 err_te_snt=0.005772 41 | epoch 312, loss_tr=12.504968 err_tr=0.000205 loss_te=14.069112 err_te=0.292549 err_te_snt=0.004329 42 | epoch 320, loss_tr=12.504644 err_tr=0.000225 loss_te=14.060661 err_te=0.287772 err_te_snt=0.005051 43 | epoch 328, loss_tr=12.504128 err_tr=0.000137 loss_te=14.082790 err_te=0.301605 err_te_snt=0.002886 44 | epoch 336, loss_tr=12.503749 err_tr=0.000137 loss_te=14.059607 err_te=0.287295 err_te_snt=0.005772 45 | epoch 344, loss_tr=12.503518 err_tr=0.000146 loss_te=14.062671 err_te=0.289460 err_te_snt=0.005772 46 | epoch 352, loss_tr=12.503032 err_tr=0.000107 loss_te=14.045504 err_te=0.279070 err_te_snt=0.005772 47 | epoch 360, loss_tr=12.502886 err_tr=0.000186 loss_te=14.054032 err_te=0.284400 err_te_snt=0.004329 48 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m035/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=16.191803 err_tr=0.993437 loss_te=16.724869 err_te=0.987723 err_te_snt=0.979798 2 | epoch 8, loss_tr=14.601464 err_tr=0.548584 loss_te=16.142902 err_te=0.684763 err_te_snt=0.166667 3 | epoch 16, loss_tr=14.349505 err_tr=0.321006 loss_te=15.983832 err_te=0.567026 err_te_snt=0.069264 4 | epoch 24, loss_tr=14.233440 err_tr=0.195156 loss_te=15.922633 err_te=0.527664 err_te_snt=0.062049 5 | epoch 32, loss_tr=14.166851 err_tr=0.121865 loss_te=15.803791 err_te=0.442062 err_te_snt=0.025253 6 | epoch 40, loss_tr=14.127918 err_tr=0.079971 loss_te=15.843160 err_te=0.471469 err_te_snt=0.038240 7 | epoch 48, loss_tr=14.103033 err_tr=0.055703 loss_te=15.762350 err_te=0.419914 err_te_snt=0.023088 8 | epoch 56, loss_tr=14.083611 err_tr=0.037822 loss_te=15.742949 err_te=0.410273 err_te_snt=0.015873 9 | epoch 64, loss_tr=14.070601 err_tr=0.027236 loss_te=15.745046 err_te=0.415125 err_te_snt=0.018759 10 | epoch 72, loss_tr=14.060490 err_tr=0.020459 loss_te=15.702863 err_te=0.379828 err_te_snt=0.015873 11 | epoch 80, loss_tr=14.052592 err_tr=0.014795 loss_te=15.677136 err_te=0.363094 err_te_snt=0.012987 12 | epoch 88, loss_tr=14.046136 err_tr=0.011533 loss_te=15.647695 err_te=0.344902 err_te_snt=0.010101 13 | epoch 96, loss_tr=14.041190 err_tr=0.008604 loss_te=15.643970 err_te=0.343701 err_te_snt=0.008658 14 | epoch 104, loss_tr=14.036732 err_tr=0.006621 loss_te=15.640619 err_te=0.341109 err_te_snt=0.009380 15 | epoch 112, loss_tr=14.033079 err_tr=0.005205 loss_te=15.629821 err_te=0.335734 err_te_snt=0.005772 16 | epoch 120, loss_tr=14.030107 err_tr=0.004473 loss_te=15.612963 err_te=0.324339 err_te_snt=0.004329 17 | epoch 128, loss_tr=14.027699 err_tr=0.003467 loss_te=15.626942 err_te=0.330322 err_te_snt=0.006494 18 | epoch 136, loss_tr=14.025082 err_tr=0.002939 loss_te=15.611290 err_te=0.321636 err_te_snt=0.005772 19 | epoch 144, loss_tr=14.023034 err_tr=0.002344 loss_te=15.601734 err_te=0.315526 err_te_snt=0.004329 20 | epoch 152, loss_tr=14.021301 err_tr=0.001748 loss_te=15.587961 err_te=0.305916 err_te_snt=0.004329 21 | epoch 160, loss_tr=14.019643 err_tr=0.001660 loss_te=15.590463 err_te=0.307851 err_te_snt=0.005051 22 | epoch 168, loss_tr=14.018536 err_tr=0.001543 loss_te=15.607316 err_te=0.320971 err_te_snt=0.007215 23 | epoch 176, loss_tr=14.017033 err_tr=0.001230 loss_te=15.604437 err_te=0.318442 err_te_snt=0.006494 24 | epoch 184, loss_tr=14.015934 err_tr=0.001191 loss_te=15.600755 err_te=0.314523 err_te_snt=0.007215 25 | epoch 192, loss_tr=14.014906 err_tr=0.001064 loss_te=15.586446 err_te=0.305904 err_te_snt=0.006494 26 | epoch 200, loss_tr=14.013992 err_tr=0.000996 loss_te=15.580591 err_te=0.302520 err_te_snt=0.002886 27 | epoch 208, loss_tr=14.013083 err_tr=0.000908 loss_te=15.597700 err_te=0.311789 err_te_snt=0.004329 28 | epoch 216, loss_tr=14.012069 err_tr=0.000703 loss_te=15.622425 err_te=0.326326 err_te_snt=0.009380 29 | epoch 224, loss_tr=14.011389 err_tr=0.000508 loss_te=15.603099 err_te=0.316439 err_te_snt=0.005772 30 | epoch 232, loss_tr=14.010413 err_tr=0.000518 loss_te=15.605697 err_te=0.315680 err_te_snt=0.005772 31 | epoch 240, loss_tr=14.010112 err_tr=0.000420 loss_te=15.565483 err_te=0.291342 err_te_snt=0.005772 32 | epoch 248, loss_tr=14.009213 err_tr=0.000430 loss_te=15.568596 err_te=0.292861 err_te_snt=0.005051 33 | epoch 256, loss_tr=14.008748 err_tr=0.000439 loss_te=15.589338 err_te=0.304100 err_te_snt=0.004329 34 | epoch 264, loss_tr=14.008046 err_tr=0.000273 loss_te=15.624151 err_te=0.323786 err_te_snt=0.005772 35 | epoch 272, loss_tr=14.007592 err_tr=0.000381 loss_te=15.570316 err_te=0.292832 err_te_snt=0.006494 36 | epoch 280, loss_tr=14.007202 err_tr=0.000312 loss_te=15.558095 err_te=0.283979 err_te_snt=0.003608 37 | epoch 288, loss_tr=14.006772 err_tr=0.000381 loss_te=15.561639 err_te=0.287080 err_te_snt=0.002886 38 | epoch 296, loss_tr=14.006234 err_tr=0.000371 loss_te=15.559149 err_te=0.286427 err_te_snt=0.005051 39 | epoch 304, loss_tr=14.005661 err_tr=0.000137 loss_te=15.617585 err_te=0.325992 err_te_snt=0.008658 40 | epoch 312, loss_tr=14.005452 err_tr=0.000166 loss_te=15.562146 err_te=0.286445 err_te_snt=0.005051 41 | epoch 320, loss_tr=14.004883 err_tr=0.000234 loss_te=15.561239 err_te=0.287684 err_te_snt=0.004329 42 | epoch 328, loss_tr=14.004395 err_tr=0.000205 loss_te=15.566426 err_te=0.290233 err_te_snt=0.005051 43 | epoch 336, loss_tr=14.004208 err_tr=0.000234 loss_te=15.552260 err_te=0.279258 err_te_snt=0.004329 44 | epoch 344, loss_tr=14.003960 err_tr=0.000186 loss_te=15.564229 err_te=0.289547 err_te_snt=0.005051 45 | epoch 352, loss_tr=14.003410 err_tr=0.000137 loss_te=15.569575 err_te=0.292241 err_te_snt=0.005051 46 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m040/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=17.646955 err_tr=0.991826 loss_te=18.189207 err_te=0.987626 err_te_snt=0.983405 2 | epoch 0, loss_tr=17.662252 err_tr=0.992559 loss_te=18.197947 err_te=0.987655 err_te_snt=0.980519 3 | epoch 8, loss_tr=16.081934 err_tr=0.531875 loss_te=17.724751 err_te=0.750284 err_te_snt=0.388167 4 | epoch 16, loss_tr=15.835681 err_tr=0.305762 loss_te=17.499638 err_te=0.579332 err_te_snt=0.096681 5 | epoch 24, loss_tr=15.724241 err_tr=0.187803 loss_te=17.397839 err_te=0.510644 err_te_snt=0.054834 6 | epoch 32, loss_tr=15.661330 err_tr=0.118242 loss_te=17.328390 err_te=0.463705 err_te_snt=0.028139 7 | epoch 40, loss_tr=15.623908 err_tr=0.078193 loss_te=17.320789 err_te=0.460326 err_te_snt=0.034632 8 | epoch 48, loss_tr=15.599311 err_tr=0.052764 loss_te=17.227921 err_te=0.398815 err_te_snt=0.013709 9 | epoch 56, loss_tr=15.580780 err_tr=0.036348 loss_te=17.264669 err_te=0.425288 err_te_snt=0.030303 10 | epoch 64, loss_tr=15.568013 err_tr=0.026348 loss_te=17.194605 err_te=0.380587 err_te_snt=0.011544 11 | epoch 72, loss_tr=15.558234 err_tr=0.018691 loss_te=17.184986 err_te=0.372963 err_te_snt=0.013709 12 | epoch 80, loss_tr=15.550757 err_tr=0.014561 loss_te=17.172665 err_te=0.363763 err_te_snt=0.010101 13 | epoch 88, loss_tr=15.544481 err_tr=0.011162 loss_te=17.125786 err_te=0.334212 err_te_snt=0.008658 14 | epoch 96, loss_tr=15.539618 err_tr=0.008428 loss_te=17.136055 err_te=0.341170 err_te_snt=0.006494 15 | epoch 104, loss_tr=15.535223 err_tr=0.006299 loss_te=17.109716 err_te=0.324815 err_te_snt=0.006494 16 | epoch 112, loss_tr=15.531937 err_tr=0.005400 loss_te=17.099194 err_te=0.318649 err_te_snt=0.005051 17 | epoch 120, loss_tr=15.528576 err_tr=0.004111 loss_te=17.105919 err_te=0.321257 err_te_snt=0.008658 18 | epoch 128, loss_tr=15.526193 err_tr=0.003066 loss_te=17.147350 err_te=0.352085 err_te_snt=0.011544 19 | epoch 136, loss_tr=15.523940 err_tr=0.002695 loss_te=17.082342 err_te=0.307223 err_te_snt=0.006494 20 | epoch 144, loss_tr=15.522044 err_tr=0.002275 loss_te=17.076342 err_te=0.303745 err_te_snt=0.006494 21 | epoch 152, loss_tr=15.520386 err_tr=0.002021 loss_te=17.084345 err_te=0.304542 err_te_snt=0.005051 22 | epoch 160, loss_tr=15.518626 err_tr=0.001533 loss_te=17.126125 err_te=0.335202 err_te_snt=0.009380 23 | epoch 168, loss_tr=15.517612 err_tr=0.001592 loss_te=17.085407 err_te=0.309462 err_te_snt=0.008658 24 | epoch 176, loss_tr=15.516044 err_tr=0.001299 loss_te=17.063795 err_te=0.294995 err_te_snt=0.004329 25 | epoch 184, loss_tr=15.514886 err_tr=0.001133 loss_te=17.066391 err_te=0.296018 err_te_snt=0.007215 26 | epoch 192, loss_tr=15.514047 err_tr=0.000977 loss_te=17.107904 err_te=0.321082 err_te_snt=0.008658 27 | epoch 200, loss_tr=15.512926 err_tr=0.000781 loss_te=17.074905 err_te=0.303449 err_te_snt=0.008658 28 | epoch 208, loss_tr=15.512268 err_tr=0.000898 loss_te=17.065863 err_te=0.294555 err_te_snt=0.003608 29 | epoch 216, loss_tr=15.511285 err_tr=0.000811 loss_te=17.063446 err_te=0.292358 err_te_snt=0.004329 30 | epoch 224, loss_tr=15.510714 err_tr=0.000576 loss_te=17.123451 err_te=0.329885 err_te_snt=0.011544 31 | epoch 232, loss_tr=15.509741 err_tr=0.000469 loss_te=17.048658 err_te=0.284042 err_te_snt=0.005051 32 | epoch 240, loss_tr=15.509254 err_tr=0.000439 loss_te=17.053879 err_te=0.286814 err_te_snt=0.005051 33 | epoch 248, loss_tr=15.508662 err_tr=0.000361 loss_te=17.079100 err_te=0.302900 err_te_snt=0.006494 34 | epoch 256, loss_tr=15.508009 err_tr=0.000381 loss_te=17.104187 err_te=0.319159 err_te_snt=0.007937 35 | epoch 264, loss_tr=15.507533 err_tr=0.000352 loss_te=17.066978 err_te=0.292961 err_te_snt=0.003608 36 | epoch 272, loss_tr=15.506984 err_tr=0.000459 loss_te=17.076401 err_te=0.299031 err_te_snt=0.004329 37 | epoch 280, loss_tr=15.506495 err_tr=0.000244 loss_te=17.055597 err_te=0.287950 err_te_snt=0.004329 38 | epoch 288, loss_tr=15.506102 err_tr=0.000293 loss_te=17.032141 err_te=0.272369 err_te_snt=0.005051 39 | epoch 296, loss_tr=15.505508 err_tr=0.000254 loss_te=17.034121 err_te=0.271758 err_te_snt=0.006494 40 | epoch 304, loss_tr=15.505040 err_tr=0.000176 loss_te=17.054232 err_te=0.284325 err_te_snt=0.005051 41 | epoch 312, loss_tr=15.504649 err_tr=0.000146 loss_te=17.034697 err_te=0.272443 err_te_snt=0.005051 42 | epoch 320, loss_tr=15.504334 err_tr=0.000264 loss_te=17.048426 err_te=0.282188 err_te_snt=0.004329 43 | epoch 328, loss_tr=15.503891 err_tr=0.000205 loss_te=17.060427 err_te=0.287222 err_te_snt=0.006494 44 | epoch 336, loss_tr=15.503654 err_tr=0.000234 loss_te=17.060728 err_te=0.287350 err_te_snt=0.003608 45 | epoch 344, loss_tr=15.503329 err_tr=0.000156 loss_te=17.032114 err_te=0.271204 err_te_snt=0.003608 46 | epoch 352, loss_tr=15.502844 err_tr=0.000088 loss_te=17.035181 err_te=0.275708 err_te_snt=0.004329 47 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m045/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=19.168388 err_tr=0.991836 loss_te=19.692604 err_te=0.987180 err_te_snt=0.979077 2 | epoch 8, loss_tr=17.572832 err_tr=0.526514 loss_te=19.138887 err_te=0.680261 err_te_snt=0.182540 3 | epoch 16, loss_tr=17.329014 err_tr=0.299395 loss_te=18.986752 err_te=0.572949 err_te_snt=0.078644 4 | epoch 24, loss_tr=17.222267 err_tr=0.183398 loss_te=18.884552 err_te=0.500485 err_te_snt=0.038240 5 | epoch 32, loss_tr=17.160246 err_tr=0.115303 loss_te=18.803715 err_te=0.445797 err_te_snt=0.022367 6 | epoch 40, loss_tr=17.122950 err_tr=0.076855 loss_te=18.771273 err_te=0.425241 err_te_snt=0.018759 7 | epoch 48, loss_tr=17.099054 err_tr=0.052578 loss_te=18.804321 err_te=0.454351 err_te_snt=0.038240 8 | epoch 56, loss_tr=17.080378 err_tr=0.035713 loss_te=18.709194 err_te=0.387062 err_te_snt=0.010101 9 | epoch 64, loss_tr=17.068075 err_tr=0.027119 loss_te=18.755362 err_te=0.420545 err_te_snt=0.034632 10 | epoch 72, loss_tr=17.060078 err_tr=0.020957 loss_te=18.716843 err_te=0.395432 err_te_snt=0.016595 11 | epoch 80, loss_tr=17.051151 err_tr=0.014590 loss_te=18.672188 err_te=0.365789 err_te_snt=0.010101 12 | epoch 88, loss_tr=17.044537 err_tr=0.010586 loss_te=18.634270 err_te=0.338552 err_te_snt=0.007937 13 | epoch 96, loss_tr=17.039534 err_tr=0.008281 loss_te=18.624872 err_te=0.335025 err_te_snt=0.005772 14 | epoch 104, loss_tr=17.035295 err_tr=0.006250 loss_te=18.620487 err_te=0.333509 err_te_snt=0.006494 15 | epoch 112, loss_tr=17.032084 err_tr=0.005254 loss_te=18.627117 err_te=0.336169 err_te_snt=0.007215 16 | epoch 120, loss_tr=17.029114 err_tr=0.004180 loss_te=18.595579 err_te=0.314732 err_te_snt=0.005772 17 | epoch 128, loss_tr=17.026506 err_tr=0.003271 loss_te=18.589041 err_te=0.311026 err_te_snt=0.007215 18 | epoch 136, loss_tr=17.024288 err_tr=0.002646 loss_te=18.592651 err_te=0.314081 err_te_snt=0.006494 19 | epoch 144, loss_tr=17.022400 err_tr=0.002568 loss_te=18.578209 err_te=0.303478 err_te_snt=0.005051 20 | epoch 152, loss_tr=17.020800 err_tr=0.002080 loss_te=18.565744 err_te=0.294395 err_te_snt=0.003608 21 | epoch 160, loss_tr=17.018980 err_tr=0.001650 loss_te=18.583017 err_te=0.306420 err_te_snt=0.004329 22 | epoch 168, loss_tr=17.017887 err_tr=0.001465 loss_te=18.618549 err_te=0.332050 err_te_snt=0.006494 23 | epoch 176, loss_tr=17.016460 err_tr=0.001465 loss_te=18.580542 err_te=0.304562 err_te_snt=0.005772 24 | epoch 184, loss_tr=17.015192 err_tr=0.001094 loss_te=18.561436 err_te=0.292489 err_te_snt=0.004329 25 | epoch 192, loss_tr=17.014200 err_tr=0.000928 loss_te=18.556385 err_te=0.289944 err_te_snt=0.005051 26 | epoch 200, loss_tr=17.013401 err_tr=0.000889 loss_te=18.561876 err_te=0.294025 err_te_snt=0.003608 27 | epoch 208, loss_tr=17.012569 err_tr=0.000898 loss_te=18.548828 err_te=0.286080 err_te_snt=0.005772 28 | epoch 216, loss_tr=17.011597 err_tr=0.000781 loss_te=18.597763 err_te=0.313420 err_te_snt=0.007215 29 | epoch 224, loss_tr=17.010841 err_tr=0.000674 loss_te=18.565041 err_te=0.296808 err_te_snt=0.005772 30 | epoch 232, loss_tr=17.009848 err_tr=0.000498 loss_te=18.539675 err_te=0.280488 err_te_snt=0.002165 31 | epoch 240, loss_tr=17.009424 err_tr=0.000518 loss_te=18.539663 err_te=0.278903 err_te_snt=0.005772 32 | epoch 248, loss_tr=17.008608 err_tr=0.000342 loss_te=18.549328 err_te=0.283911 err_te_snt=0.003608 33 | epoch 256, loss_tr=17.008207 err_tr=0.000488 loss_te=18.571001 err_te=0.296436 err_te_snt=0.005051 34 | epoch 264, loss_tr=17.007578 err_tr=0.000303 loss_te=18.550964 err_te=0.282889 err_te_snt=0.003608 35 | epoch 272, loss_tr=17.007044 err_tr=0.000332 loss_te=18.551025 err_te=0.284960 err_te_snt=0.004329 36 | epoch 280, loss_tr=17.006565 err_tr=0.000322 loss_te=18.540251 err_te=0.275849 err_te_snt=0.003608 37 | epoch 288, loss_tr=17.006067 err_tr=0.000342 loss_te=18.536406 err_te=0.274574 err_te_snt=0.006494 38 | epoch 296, loss_tr=17.005587 err_tr=0.000293 loss_te=18.529560 err_te=0.269862 err_te_snt=0.002165 39 | epoch 304, loss_tr=17.005249 err_tr=0.000234 loss_te=18.545118 err_te=0.278656 err_te_snt=0.004329 40 | epoch 312, loss_tr=17.004835 err_tr=0.000264 loss_te=18.539103 err_te=0.278201 err_te_snt=0.007215 41 | epoch 320, loss_tr=17.004341 err_tr=0.000156 loss_te=18.544758 err_te=0.278207 err_te_snt=0.004329 42 | epoch 328, loss_tr=17.004007 err_tr=0.000264 loss_te=18.566341 err_te=0.291792 err_te_snt=0.006494 43 | epoch 336, loss_tr=17.003763 err_tr=0.000176 loss_te=18.560884 err_te=0.290001 err_te_snt=0.003608 44 | epoch 344, loss_tr=17.003412 err_tr=0.000215 loss_te=18.541599 err_te=0.276303 err_te_snt=0.005051 45 | epoch 352, loss_tr=17.002949 err_tr=0.000117 loss_te=18.533457 err_te=0.270753 err_te_snt=0.004329 46 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m050/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=20.670527 err_tr=0.992373 loss_te=21.211617 err_te=0.990626 err_te_snt=0.985570 2 | epoch 8, loss_tr=19.071844 err_tr=0.522852 loss_te=20.635164 err_te=0.680344 err_te_snt=0.184704 3 | epoch 16, loss_tr=18.827572 err_tr=0.299121 loss_te=20.504557 err_te=0.583755 err_te_snt=0.098124 4 | epoch 24, loss_tr=18.715324 err_tr=0.177930 loss_te=20.387863 err_te=0.503030 err_te_snt=0.049784 5 | epoch 32, loss_tr=18.652544 err_tr=0.106914 loss_te=20.286491 err_te=0.434687 err_te_snt=0.013709 6 | epoch 40, loss_tr=18.615353 err_tr=0.069170 loss_te=20.380014 err_te=0.502490 err_te_snt=0.062771 7 | epoch 48, loss_tr=18.593130 err_tr=0.047490 loss_te=20.237856 err_te=0.405472 err_te_snt=0.015873 8 | epoch 56, loss_tr=18.575861 err_tr=0.032988 loss_te=20.227024 err_te=0.401621 err_te_snt=0.011544 9 | epoch 64, loss_tr=18.563589 err_tr=0.022715 loss_te=20.191673 err_te=0.380200 err_te_snt=0.008658 10 | epoch 72, loss_tr=18.554922 err_tr=0.017314 loss_te=20.184410 err_te=0.375066 err_te_snt=0.012987 11 | epoch 80, loss_tr=18.547533 err_tr=0.012422 loss_te=20.145796 err_te=0.348917 err_te_snt=0.009380 12 | epoch 88, loss_tr=18.541716 err_tr=0.009336 loss_te=20.134068 err_te=0.338020 err_te_snt=0.007937 13 | epoch 96, loss_tr=18.537113 err_tr=0.007363 loss_te=20.133242 err_te=0.336800 err_te_snt=0.005051 14 | epoch 104, loss_tr=18.532967 err_tr=0.005693 loss_te=20.186001 err_te=0.375990 err_te_snt=0.012987 15 | epoch 112, loss_tr=18.529909 err_tr=0.004863 loss_te=20.110970 err_te=0.324233 err_te_snt=0.006494 16 | epoch 120, loss_tr=18.526859 err_tr=0.003496 loss_te=20.100243 err_te=0.316575 err_te_snt=0.007937 17 | epoch 128, loss_tr=18.524670 err_tr=0.002861 loss_te=20.082413 err_te=0.304974 err_te_snt=0.005772 18 | epoch 136, loss_tr=18.522478 err_tr=0.002305 loss_te=20.099455 err_te=0.318407 err_te_snt=0.006494 19 | epoch 144, loss_tr=18.520653 err_tr=0.001963 loss_te=20.087200 err_te=0.309284 err_te_snt=0.003608 20 | epoch 152, loss_tr=18.519236 err_tr=0.001670 loss_te=20.074562 err_te=0.298547 err_te_snt=0.004329 21 | epoch 160, loss_tr=18.517609 err_tr=0.001484 loss_te=20.084480 err_te=0.306534 err_te_snt=0.005772 22 | epoch 168, loss_tr=18.516632 err_tr=0.001660 loss_te=20.086996 err_te=0.307494 err_te_snt=0.006494 23 | epoch 176, loss_tr=18.515141 err_tr=0.001113 loss_te=20.076845 err_te=0.301156 err_te_snt=0.005772 24 | epoch 184, loss_tr=18.514147 err_tr=0.001035 loss_te=20.086447 err_te=0.310832 err_te_snt=0.005051 25 | epoch 192, loss_tr=18.513075 err_tr=0.000723 loss_te=20.091150 err_te=0.307372 err_te_snt=0.007215 26 | epoch 200, loss_tr=18.512360 err_tr=0.000801 loss_te=20.062868 err_te=0.292265 err_te_snt=0.004329 27 | epoch 208, loss_tr=18.511362 err_tr=0.000674 loss_te=20.060173 err_te=0.292042 err_te_snt=0.006494 28 | epoch 216, loss_tr=18.510509 err_tr=0.000566 loss_te=20.060522 err_te=0.289228 err_te_snt=0.003608 29 | epoch 224, loss_tr=18.510036 err_tr=0.000527 loss_te=20.050055 err_te=0.281679 err_te_snt=0.004329 30 | epoch 232, loss_tr=18.508945 err_tr=0.000332 loss_te=20.053684 err_te=0.284131 err_te_snt=0.004329 31 | epoch 240, loss_tr=18.508579 err_tr=0.000430 loss_te=20.056845 err_te=0.286269 err_te_snt=0.003608 32 | epoch 248, loss_tr=18.507996 err_tr=0.000391 loss_te=20.044842 err_te=0.280039 err_te_snt=0.004329 33 | epoch 256, loss_tr=18.507450 err_tr=0.000439 loss_te=20.056276 err_te=0.286412 err_te_snt=0.005051 34 | epoch 264, loss_tr=18.506914 err_tr=0.000244 loss_te=20.073000 err_te=0.297898 err_te_snt=0.006494 35 | epoch 272, loss_tr=18.506538 err_tr=0.000312 loss_te=20.048300 err_te=0.283231 err_te_snt=0.005051 36 | epoch 280, loss_tr=18.506121 err_tr=0.000410 loss_te=20.044466 err_te=0.280246 err_te_snt=0.005051 37 | epoch 288, loss_tr=18.505632 err_tr=0.000352 loss_te=20.051775 err_te=0.284110 err_te_snt=0.005051 38 | epoch 296, loss_tr=18.505003 err_tr=0.000234 loss_te=20.039747 err_te=0.276666 err_te_snt=0.002886 39 | epoch 304, loss_tr=18.504694 err_tr=0.000146 loss_te=20.053530 err_te=0.285420 err_te_snt=0.005051 40 | epoch 312, loss_tr=18.504372 err_tr=0.000283 loss_te=20.038235 err_te=0.274364 err_te_snt=0.003608 41 | epoch 320, loss_tr=18.503859 err_tr=0.000215 loss_te=20.038788 err_te=0.273772 err_te_snt=0.002165 42 | epoch 328, loss_tr=18.503538 err_tr=0.000195 loss_te=20.035769 err_te=0.272006 err_te_snt=0.002886 43 | epoch 336, loss_tr=18.503229 err_tr=0.000186 loss_te=20.039082 err_te=0.274256 err_te_snt=0.004329 44 | epoch 344, loss_tr=18.502987 err_tr=0.000088 loss_te=20.052710 err_te=0.283297 err_te_snt=0.004329 45 | epoch 352, loss_tr=18.502556 err_tr=0.000098 loss_te=20.041336 err_te=0.278610 err_te_snt=0.005051 46 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m055/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=22.130636 err_tr=0.990195 loss_te=22.689425 err_te=0.980856 err_te_snt=0.956710 2 | epoch 8, loss_tr=20.568150 err_tr=0.521367 loss_te=22.130934 err_te=0.677157 err_te_snt=0.178932 3 | epoch 16, loss_tr=20.326235 err_tr=0.296035 loss_te=21.947004 err_te=0.540979 err_te_snt=0.054113 4 | epoch 24, loss_tr=20.214611 err_tr=0.175332 loss_te=21.878752 err_te=0.494983 err_te_snt=0.036075 5 | epoch 32, loss_tr=20.152908 err_tr=0.108164 loss_te=21.802107 err_te=0.442365 err_te_snt=0.018759 6 | epoch 40, loss_tr=20.116455 err_tr=0.069648 loss_te=21.772699 err_te=0.426425 err_te_snt=0.017316 7 | epoch 48, loss_tr=20.094275 err_tr=0.048281 loss_te=21.739206 err_te=0.404967 err_te_snt=0.013709 8 | epoch 56, loss_tr=20.076710 err_tr=0.033281 loss_te=21.746204 err_te=0.409933 err_te_snt=0.013709 9 | epoch 64, loss_tr=20.065138 err_tr=0.024082 loss_te=21.699606 err_te=0.381376 err_te_snt=0.013709 10 | epoch 72, loss_tr=20.055702 err_tr=0.017012 loss_te=21.693150 err_te=0.376662 err_te_snt=0.013709 11 | epoch 80, loss_tr=20.048767 err_tr=0.013105 loss_te=21.669874 err_te=0.363432 err_te_snt=0.012266 12 | epoch 88, loss_tr=20.042864 err_tr=0.010146 loss_te=21.646151 err_te=0.345264 err_te_snt=0.007215 13 | epoch 96, loss_tr=20.038095 err_tr=0.007637 loss_te=21.678627 err_te=0.368259 err_te_snt=0.007937 14 | epoch 104, loss_tr=20.034296 err_tr=0.006211 loss_te=21.656757 err_te=0.353556 err_te_snt=0.007215 15 | epoch 112, loss_tr=20.030832 err_tr=0.005010 loss_te=21.664806 err_te=0.360525 err_te_snt=0.015152 16 | epoch 120, loss_tr=20.027924 err_tr=0.003955 loss_te=21.608404 err_te=0.320737 err_te_snt=0.005772 17 | epoch 128, loss_tr=20.025738 err_tr=0.003008 loss_te=21.611029 err_te=0.324137 err_te_snt=0.009380 18 | epoch 136, loss_tr=20.023598 err_tr=0.002695 loss_te=21.606113 err_te=0.321268 err_te_snt=0.005772 19 | epoch 144, loss_tr=20.021843 err_tr=0.002158 loss_te=21.596443 err_te=0.311578 err_te_snt=0.006494 20 | epoch 152, loss_tr=20.020426 err_tr=0.002070 loss_te=21.626902 err_te=0.334194 err_te_snt=0.010823 21 | epoch 160, loss_tr=20.018764 err_tr=0.001855 loss_te=21.572058 err_te=0.299243 err_te_snt=0.004329 22 | epoch 168, loss_tr=20.017603 err_tr=0.001650 loss_te=21.583168 err_te=0.303122 err_te_snt=0.004329 23 | epoch 176, loss_tr=20.016178 err_tr=0.001396 loss_te=21.576262 err_te=0.299520 err_te_snt=0.004329 24 | epoch 184, loss_tr=20.015093 err_tr=0.001113 loss_te=21.578722 err_te=0.302754 err_te_snt=0.004329 25 | epoch 192, loss_tr=20.014025 err_tr=0.001016 loss_te=21.633204 err_te=0.339224 err_te_snt=0.007937 26 | epoch 200, loss_tr=20.013226 err_tr=0.000928 loss_te=21.569771 err_te=0.295899 err_te_snt=0.007215 27 | epoch 208, loss_tr=20.012365 err_tr=0.000664 loss_te=21.570087 err_te=0.295184 err_te_snt=0.003608 28 | epoch 216, loss_tr=20.011543 err_tr=0.000693 loss_te=21.585882 err_te=0.304152 err_te_snt=0.007937 29 | epoch 224, loss_tr=20.011003 err_tr=0.000762 loss_te=21.565939 err_te=0.293915 err_te_snt=0.002886 30 | epoch 232, loss_tr=20.009939 err_tr=0.000400 loss_te=21.579107 err_te=0.299145 err_te_snt=0.005051 31 | epoch 240, loss_tr=20.009647 err_tr=0.000596 loss_te=21.558554 err_te=0.287436 err_te_snt=0.005772 32 | epoch 248, loss_tr=20.008867 err_tr=0.000381 loss_te=21.561380 err_te=0.289822 err_te_snt=0.007215 33 | epoch 256, loss_tr=20.008274 err_tr=0.000371 loss_te=21.562649 err_te=0.289686 err_te_snt=0.005051 34 | epoch 264, loss_tr=20.007744 err_tr=0.000332 loss_te=21.553391 err_te=0.285328 err_te_snt=0.002886 35 | epoch 272, loss_tr=20.007351 err_tr=0.000410 loss_te=21.573067 err_te=0.294920 err_te_snt=0.003608 36 | epoch 280, loss_tr=20.006933 err_tr=0.000352 loss_te=21.601892 err_te=0.314370 err_te_snt=0.007215 37 | epoch 288, loss_tr=20.006435 err_tr=0.000361 loss_te=21.559923 err_te=0.288449 err_te_snt=0.005051 38 | epoch 296, loss_tr=20.005978 err_tr=0.000312 loss_te=21.545202 err_te=0.279492 err_te_snt=0.004329 39 | epoch 304, loss_tr=20.005514 err_tr=0.000156 loss_te=21.568916 err_te=0.294223 err_te_snt=0.006494 40 | epoch 312, loss_tr=20.005236 err_tr=0.000293 loss_te=21.549580 err_te=0.282344 err_te_snt=0.005772 41 | epoch 320, loss_tr=20.004686 err_tr=0.000137 loss_te=21.562912 err_te=0.288209 err_te_snt=0.004329 42 | epoch 328, loss_tr=20.004297 err_tr=0.000195 loss_te=21.552143 err_te=0.281331 err_te_snt=0.005051 43 | epoch 336, loss_tr=20.004248 err_tr=0.000264 loss_te=21.538580 err_te=0.275007 err_te_snt=0.004329 44 | epoch 344, loss_tr=20.003801 err_tr=0.000186 loss_te=21.565237 err_te=0.290697 err_te_snt=0.005051 45 | epoch 352, loss_tr=20.003399 err_tr=0.000186 loss_te=21.544838 err_te=0.278123 err_te_snt=0.003608 46 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m060/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=23.683996 err_tr=0.992988 loss_te=24.212011 err_te=0.991331 err_te_snt=0.992063 2 | epoch 8, loss_tr=22.079113 err_tr=0.528447 loss_te=23.656118 err_te=0.696548 err_te_snt=0.208514 3 | epoch 16, loss_tr=21.834455 err_tr=0.305117 loss_te=23.480438 err_te=0.564443 err_te_snt=0.075758 4 | epoch 24, loss_tr=21.725529 err_tr=0.187441 loss_te=23.437529 err_te=0.538255 err_te_snt=0.065657 5 | epoch 32, loss_tr=21.662609 err_tr=0.119482 loss_te=23.322306 err_te=0.455613 err_te_snt=0.024531 6 | epoch 40, loss_tr=21.623466 err_tr=0.076914 loss_te=23.299143 err_te=0.443120 err_te_snt=0.024531 7 | epoch 48, loss_tr=21.598669 err_tr=0.052246 loss_te=23.221714 err_te=0.391709 err_te_snt=0.011544 8 | epoch 56, loss_tr=21.579720 err_tr=0.036211 loss_te=23.248789 err_te=0.413402 err_te_snt=0.020202 9 | epoch 64, loss_tr=21.566858 err_tr=0.025449 loss_te=23.191517 err_te=0.374535 err_te_snt=0.015873 10 | epoch 72, loss_tr=21.557602 err_tr=0.019033 loss_te=23.161268 err_te=0.355002 err_te_snt=0.010101 11 | epoch 80, loss_tr=21.549809 err_tr=0.013467 loss_te=23.177200 err_te=0.369972 err_te_snt=0.010823 12 | epoch 88, loss_tr=21.543322 err_tr=0.010078 loss_te=23.136032 err_te=0.340576 err_te_snt=0.006494 13 | epoch 96, loss_tr=21.538404 err_tr=0.007910 loss_te=23.129725 err_te=0.334178 err_te_snt=0.006494 14 | epoch 104, loss_tr=21.534258 err_tr=0.006445 loss_te=23.188515 err_te=0.378706 err_te_snt=0.016595 15 | epoch 112, loss_tr=21.530962 err_tr=0.005117 loss_te=23.150175 err_te=0.350068 err_te_snt=0.012266 16 | epoch 120, loss_tr=21.527691 err_tr=0.003730 loss_te=23.103682 err_te=0.318200 err_te_snt=0.004329 17 | epoch 128, loss_tr=21.525442 err_tr=0.003262 loss_te=23.083496 err_te=0.305474 err_te_snt=0.004329 18 | epoch 136, loss_tr=21.523115 err_tr=0.002646 loss_te=23.107168 err_te=0.320821 err_te_snt=0.006494 19 | epoch 144, loss_tr=21.521297 err_tr=0.002354 loss_te=23.109144 err_te=0.323033 err_te_snt=0.007215 20 | epoch 152, loss_tr=21.519724 err_tr=0.001953 loss_te=23.075106 err_te=0.299250 err_te_snt=0.005051 21 | epoch 160, loss_tr=21.517860 err_tr=0.001426 loss_te=23.075277 err_te=0.302034 err_te_snt=0.002886 22 | epoch 168, loss_tr=21.516985 err_tr=0.001631 loss_te=23.132999 err_te=0.337096 err_te_snt=0.008658 23 | epoch 176, loss_tr=21.515495 err_tr=0.001494 loss_te=23.069151 err_te=0.297693 err_te_snt=0.005051 24 | epoch 184, loss_tr=21.514385 err_tr=0.001006 loss_te=23.069675 err_te=0.295971 err_te_snt=0.005772 25 | epoch 192, loss_tr=21.513351 err_tr=0.000752 loss_te=23.056547 err_te=0.287446 err_te_snt=0.006494 26 | epoch 200, loss_tr=21.512314 err_tr=0.000645 loss_te=23.063753 err_te=0.294973 err_te_snt=0.005051 27 | epoch 208, loss_tr=21.511526 err_tr=0.000703 loss_te=23.052620 err_te=0.285462 err_te_snt=0.002886 28 | epoch 216, loss_tr=21.510607 err_tr=0.000654 loss_te=23.063532 err_te=0.292521 err_te_snt=0.005051 29 | epoch 224, loss_tr=21.510201 err_tr=0.000469 loss_te=23.080822 err_te=0.301268 err_te_snt=0.006494 30 | epoch 232, loss_tr=21.509197 err_tr=0.000508 loss_te=23.063358 err_te=0.290311 err_te_snt=0.004329 31 | epoch 240, loss_tr=21.508818 err_tr=0.000557 loss_te=23.067854 err_te=0.292165 err_te_snt=0.006494 32 | epoch 248, loss_tr=21.508064 err_tr=0.000439 loss_te=23.043053 err_te=0.278369 err_te_snt=0.003608 33 | epoch 256, loss_tr=21.507565 err_tr=0.000410 loss_te=23.042967 err_te=0.279056 err_te_snt=0.004329 34 | epoch 264, loss_tr=21.507023 err_tr=0.000312 loss_te=23.052889 err_te=0.281997 err_te_snt=0.002886 35 | epoch 272, loss_tr=21.506533 err_tr=0.000400 loss_te=23.086588 err_te=0.307053 err_te_snt=0.007215 36 | epoch 280, loss_tr=21.506113 err_tr=0.000283 loss_te=23.067612 err_te=0.291526 err_te_snt=0.005051 37 | epoch 288, loss_tr=21.505480 err_tr=0.000273 loss_te=23.040218 err_te=0.276907 err_te_snt=0.003608 38 | epoch 296, loss_tr=21.505056 err_tr=0.000283 loss_te=23.050409 err_te=0.280408 err_te_snt=0.007937 39 | epoch 304, loss_tr=21.504768 err_tr=0.000186 loss_te=23.047205 err_te=0.278895 err_te_snt=0.005051 40 | epoch 312, loss_tr=21.504396 err_tr=0.000166 loss_te=23.053291 err_te=0.281468 err_te_snt=0.004329 41 | epoch 320, loss_tr=21.503916 err_tr=0.000195 loss_te=23.036600 err_te=0.274099 err_te_snt=0.002165 42 | epoch 328, loss_tr=21.503521 err_tr=0.000176 loss_te=23.034229 err_te=0.272371 err_te_snt=0.005772 43 | epoch 336, loss_tr=21.503376 err_tr=0.000166 loss_te=23.035580 err_te=0.271820 err_te_snt=0.004329 44 | epoch 344, loss_tr=21.503073 err_tr=0.000156 loss_te=23.036047 err_te=0.272130 err_te_snt=0.003608 45 | epoch 352, loss_tr=21.502609 err_tr=0.000098 loss_te=23.055176 err_te=0.282845 err_te_snt=0.005051 46 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m065/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=25.126490 err_tr=0.990195 loss_te=25.698282 err_te=0.981430 err_te_snt=0.971140 2 | epoch 8, loss_tr=23.584196 err_tr=0.535498 loss_te=25.135757 err_te=0.676565 err_te_snt=0.135642 3 | epoch 16, loss_tr=23.329285 err_tr=0.299512 loss_te=24.953424 err_te=0.546991 err_te_snt=0.055556 4 | epoch 24, loss_tr=23.218367 err_tr=0.179443 loss_te=24.873926 err_te=0.493291 err_te_snt=0.028860 5 | epoch 32, loss_tr=23.156158 err_tr=0.111348 loss_te=24.880737 err_te=0.499818 err_te_snt=0.067100 6 | epoch 40, loss_tr=23.119537 err_tr=0.073750 loss_te=24.778748 err_te=0.430535 err_te_snt=0.025974 7 | epoch 48, loss_tr=23.096645 err_tr=0.051211 loss_te=24.747393 err_te=0.412542 err_te_snt=0.015873 8 | epoch 56, loss_tr=23.077915 err_tr=0.034414 loss_te=24.692213 err_te=0.374873 err_te_snt=0.010823 9 | epoch 64, loss_tr=23.066267 err_tr=0.024883 loss_te=24.679352 err_te=0.368396 err_te_snt=0.009380 10 | epoch 72, loss_tr=23.056726 err_tr=0.018096 loss_te=24.664648 err_te=0.360500 err_te_snt=0.009380 11 | epoch 80, loss_tr=23.049702 err_tr=0.013682 loss_te=24.641457 err_te=0.344707 err_te_snt=0.007937 12 | epoch 88, loss_tr=23.043400 err_tr=0.010166 loss_te=24.621075 err_te=0.330395 err_te_snt=0.007937 13 | epoch 96, loss_tr=23.038889 err_tr=0.008467 loss_te=24.617935 err_te=0.330741 err_te_snt=0.007215 14 | epoch 104, loss_tr=23.034885 err_tr=0.006787 loss_te=24.618498 err_te=0.330602 err_te_snt=0.005772 15 | epoch 112, loss_tr=23.031153 err_tr=0.004736 loss_te=24.592234 err_te=0.315047 err_te_snt=0.006494 16 | epoch 120, loss_tr=23.028368 err_tr=0.003965 loss_te=24.599699 err_te=0.318040 err_te_snt=0.004329 17 | epoch 128, loss_tr=23.026114 err_tr=0.003242 loss_te=24.614733 err_te=0.327911 err_te_snt=0.004329 18 | epoch 136, loss_tr=23.023909 err_tr=0.002949 loss_te=24.638393 err_te=0.344870 err_te_snt=0.010101 19 | epoch 144, loss_tr=23.021851 err_tr=0.002197 loss_te=24.575684 err_te=0.300167 err_te_snt=0.003608 20 | epoch 152, loss_tr=23.020372 err_tr=0.001641 loss_te=24.564299 err_te=0.294127 err_te_snt=0.003608 21 | epoch 160, loss_tr=23.018545 err_tr=0.001514 loss_te=24.593290 err_te=0.312436 err_te_snt=0.008658 22 | epoch 168, loss_tr=23.017452 err_tr=0.001426 loss_te=24.574720 err_te=0.302193 err_te_snt=0.005772 23 | epoch 176, loss_tr=23.016275 err_tr=0.001406 loss_te=24.560583 err_te=0.292387 err_te_snt=0.005051 24 | epoch 184, loss_tr=23.015087 err_tr=0.001162 loss_te=24.570473 err_te=0.298002 err_te_snt=0.005051 25 | epoch 192, loss_tr=23.014055 err_tr=0.000986 loss_te=24.557993 err_te=0.290518 err_te_snt=0.004329 26 | epoch 200, loss_tr=23.013308 err_tr=0.000986 loss_te=24.561478 err_te=0.293246 err_te_snt=0.003608 27 | epoch 208, loss_tr=23.012421 err_tr=0.000947 loss_te=24.583355 err_te=0.305759 err_te_snt=0.005051 28 | epoch 216, loss_tr=23.011457 err_tr=0.000674 loss_te=24.553843 err_te=0.284509 err_te_snt=0.006494 29 | epoch 224, loss_tr=23.010960 err_tr=0.000596 loss_te=24.553049 err_te=0.286003 err_te_snt=0.003608 30 | epoch 232, loss_tr=23.009975 err_tr=0.000527 loss_te=24.551449 err_te=0.284651 err_te_snt=0.004329 31 | epoch 240, loss_tr=23.009399 err_tr=0.000498 loss_te=24.543699 err_te=0.279592 err_te_snt=0.003608 32 | epoch 248, loss_tr=23.008686 err_tr=0.000420 loss_te=24.567415 err_te=0.293736 err_te_snt=0.003608 33 | epoch 256, loss_tr=23.008265 err_tr=0.000430 loss_te=24.546492 err_te=0.279597 err_te_snt=0.003608 34 | epoch 264, loss_tr=23.007626 err_tr=0.000420 loss_te=24.544516 err_te=0.281042 err_te_snt=0.003608 35 | epoch 272, loss_tr=23.007250 err_tr=0.000381 loss_te=24.559261 err_te=0.287959 err_te_snt=0.004329 36 | epoch 280, loss_tr=23.006660 err_tr=0.000342 loss_te=24.537241 err_te=0.273987 err_te_snt=0.002886 37 | epoch 288, loss_tr=23.006266 err_tr=0.000312 loss_te=24.528437 err_te=0.269719 err_te_snt=0.005772 38 | epoch 296, loss_tr=23.005705 err_tr=0.000205 loss_te=24.556452 err_te=0.286213 err_te_snt=0.004329 39 | epoch 304, loss_tr=23.005327 err_tr=0.000225 loss_te=24.530630 err_te=0.269958 err_te_snt=0.003608 40 | epoch 312, loss_tr=23.004965 err_tr=0.000225 loss_te=24.542469 err_te=0.276850 err_te_snt=0.006494 41 | epoch 320, loss_tr=23.004597 err_tr=0.000234 loss_te=24.540367 err_te=0.275470 err_te_snt=0.004329 42 | epoch 328, loss_tr=23.004118 err_tr=0.000176 loss_te=24.551825 err_te=0.281874 err_te_snt=0.005051 43 | epoch 336, loss_tr=23.003935 err_tr=0.000215 loss_te=24.541388 err_te=0.275470 err_te_snt=0.002165 44 | epoch 344, loss_tr=23.003559 err_tr=0.000146 loss_te=24.529316 err_te=0.268604 err_te_snt=0.002886 45 | epoch 352, loss_tr=23.003000 err_tr=0.000117 loss_te=24.544741 err_te=0.279205 err_te_snt=0.005051 46 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m070/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=26.770388 err_tr=0.990918 loss_te=27.174791 err_te=0.976557 err_te_snt=0.961039 2 | epoch 8, loss_tr=25.092051 err_tr=0.543760 loss_te=26.625462 err_te=0.675059 err_te_snt=0.139971 3 | epoch 16, loss_tr=24.839375 err_tr=0.310449 loss_te=26.491241 err_te=0.572397 err_te_snt=0.085859 4 | epoch 24, loss_tr=24.724174 err_tr=0.185469 loss_te=26.379467 err_te=0.495711 err_te_snt=0.037518 5 | epoch 32, loss_tr=24.660753 err_tr=0.117559 loss_te=26.309702 err_te=0.448449 err_te_snt=0.022367 6 | epoch 40, loss_tr=24.622911 err_tr=0.076641 loss_te=26.237833 err_te=0.400617 err_te_snt=0.012266 7 | epoch 48, loss_tr=24.599060 err_tr=0.052998 loss_te=26.216099 err_te=0.388753 err_te_snt=0.015152 8 | epoch 56, loss_tr=24.581255 err_tr=0.037197 loss_te=26.215912 err_te=0.391423 err_te_snt=0.014430 9 | epoch 64, loss_tr=24.568537 err_tr=0.026553 loss_te=26.211958 err_te=0.388691 err_te_snt=0.015152 10 | epoch 72, loss_tr=24.559544 err_tr=0.020361 loss_te=26.158012 err_te=0.354009 err_te_snt=0.009380 11 | epoch 80, loss_tr=24.552162 err_tr=0.015068 loss_te=26.139133 err_te=0.341171 err_te_snt=0.006494 12 | epoch 88, loss_tr=24.545580 err_tr=0.011016 loss_te=26.227453 err_te=0.400984 err_te_snt=0.025974 13 | epoch 96, loss_tr=24.541105 err_tr=0.009189 loss_te=26.121040 err_te=0.331364 err_te_snt=0.005051 14 | epoch 104, loss_tr=24.536940 err_tr=0.007295 loss_te=26.130825 err_te=0.336864 err_te_snt=0.007937 15 | epoch 112, loss_tr=24.533886 err_tr=0.006582 loss_te=26.141693 err_te=0.345097 err_te_snt=0.009380 16 | epoch 120, loss_tr=24.530460 err_tr=0.004766 loss_te=26.116619 err_te=0.326655 err_te_snt=0.005772 17 | epoch 128, loss_tr=24.528069 err_tr=0.003701 loss_te=26.117767 err_te=0.327723 err_te_snt=0.008658 18 | epoch 136, loss_tr=24.525993 err_tr=0.003369 loss_te=26.126122 err_te=0.335277 err_te_snt=0.009380 19 | epoch 144, loss_tr=24.523911 err_tr=0.002764 loss_te=26.141422 err_te=0.345847 err_te_snt=0.010823 20 | epoch 152, loss_tr=24.522509 err_tr=0.002793 loss_te=26.082045 err_te=0.303504 err_te_snt=0.005051 21 | epoch 160, loss_tr=24.520704 err_tr=0.002051 loss_te=26.084131 err_te=0.307767 err_te_snt=0.006494 22 | epoch 168, loss_tr=24.519501 err_tr=0.001924 loss_te=26.095490 err_te=0.313136 err_te_snt=0.005772 23 | epoch 176, loss_tr=24.518024 err_tr=0.001582 loss_te=26.111422 err_te=0.323866 err_te_snt=0.007215 24 | epoch 184, loss_tr=24.517050 err_tr=0.001562 loss_te=26.065067 err_te=0.294071 err_te_snt=0.004329 25 | epoch 192, loss_tr=24.515909 err_tr=0.001230 loss_te=26.086901 err_te=0.308354 err_te_snt=0.004329 26 | epoch 200, loss_tr=24.514950 err_tr=0.001084 loss_te=26.098341 err_te=0.312207 err_te_snt=0.005051 27 | epoch 208, loss_tr=24.513969 err_tr=0.001016 loss_te=26.059698 err_te=0.290256 err_te_snt=0.004329 28 | epoch 216, loss_tr=24.513170 err_tr=0.001055 loss_te=26.087769 err_te=0.305786 err_te_snt=0.008658 29 | epoch 224, loss_tr=24.512533 err_tr=0.000840 loss_te=26.083059 err_te=0.303632 err_te_snt=0.005051 30 | epoch 232, loss_tr=24.511301 err_tr=0.000596 loss_te=26.075499 err_te=0.299836 err_te_snt=0.007937 31 | epoch 240, loss_tr=24.511196 err_tr=0.000859 loss_te=26.067596 err_te=0.293356 err_te_snt=0.005772 32 | epoch 248, loss_tr=24.510386 err_tr=0.000596 loss_te=26.097338 err_te=0.311436 err_te_snt=0.005772 33 | epoch 256, loss_tr=24.509769 err_tr=0.000615 loss_te=26.102644 err_te=0.315501 err_te_snt=0.009380 34 | epoch 264, loss_tr=24.509199 err_tr=0.000566 loss_te=26.115561 err_te=0.322794 err_te_snt=0.008658 35 | epoch 272, loss_tr=24.508768 err_tr=0.000547 loss_te=26.088871 err_te=0.307494 err_te_snt=0.006494 36 | epoch 280, loss_tr=24.508196 err_tr=0.000566 loss_te=26.060453 err_te=0.288931 err_te_snt=0.003608 37 | epoch 288, loss_tr=24.507698 err_tr=0.000459 loss_te=26.063734 err_te=0.291028 err_te_snt=0.005051 38 | epoch 296, loss_tr=24.507236 err_tr=0.000420 loss_te=26.059450 err_te=0.288170 err_te_snt=0.004329 39 | epoch 304, loss_tr=24.506838 err_tr=0.000420 loss_te=26.054487 err_te=0.282422 err_te_snt=0.004329 40 | epoch 312, loss_tr=24.506445 err_tr=0.000391 loss_te=26.084192 err_te=0.302596 err_te_snt=0.005051 41 | epoch 320, loss_tr=24.506037 err_tr=0.000391 loss_te=26.046255 err_te=0.279048 err_te_snt=0.003608 42 | epoch 328, loss_tr=24.505537 err_tr=0.000273 loss_te=26.076414 err_te=0.296531 err_te_snt=0.006494 43 | epoch 336, loss_tr=24.505188 err_tr=0.000166 loss_te=26.080503 err_te=0.300013 err_te_snt=0.005051 44 | epoch 344, loss_tr=24.504957 err_tr=0.000234 loss_te=26.047831 err_te=0.278556 err_te_snt=0.004329 45 | epoch 352, loss_tr=24.504496 err_tr=0.000205 loss_te=26.078793 err_te=0.297638 err_te_snt=0.006494 46 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m075/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=28.153053 err_tr=0.989551 loss_te=28.689018 err_te=0.982145 err_te_snt=0.957431 2 | epoch 8, loss_tr=26.569099 err_tr=0.519482 loss_te=28.136696 err_te=0.680369 err_te_snt=0.192641 3 | epoch 16, loss_tr=26.321791 err_tr=0.289639 loss_te=28.037548 err_te=0.609853 err_te_snt=0.156566 4 | epoch 24, loss_tr=26.210789 err_tr=0.172246 loss_te=27.841843 err_te=0.468285 err_te_snt=0.027417 5 | epoch 32, loss_tr=26.149345 err_tr=0.104346 loss_te=27.800703 err_te=0.443290 err_te_snt=0.033189 6 | epoch 40, loss_tr=26.113997 err_tr=0.067559 loss_te=27.788668 err_te=0.438826 err_te_snt=0.020924 7 | epoch 48, loss_tr=26.091135 err_tr=0.045068 loss_te=27.698404 err_te=0.379593 err_te_snt=0.010823 8 | epoch 56, loss_tr=26.073891 err_tr=0.030674 loss_te=27.700239 err_te=0.380421 err_te_snt=0.010101 9 | epoch 64, loss_tr=26.062435 err_tr=0.021650 loss_te=27.683516 err_te=0.373630 err_te_snt=0.015873 10 | epoch 72, loss_tr=26.053139 err_tr=0.015176 loss_te=27.702723 err_te=0.386777 err_te_snt=0.011544 11 | epoch 80, loss_tr=26.046413 err_tr=0.011963 loss_te=27.648592 err_te=0.347263 err_te_snt=0.010823 12 | epoch 88, loss_tr=26.040545 err_tr=0.008437 loss_te=27.638330 err_te=0.341534 err_te_snt=0.009380 13 | epoch 96, loss_tr=26.036005 err_tr=0.006641 loss_te=27.635626 err_te=0.340066 err_te_snt=0.008658 14 | epoch 104, loss_tr=26.032104 err_tr=0.004902 loss_te=27.634354 err_te=0.339827 err_te_snt=0.005772 15 | epoch 112, loss_tr=26.029177 err_tr=0.004277 loss_te=27.604397 err_te=0.321325 err_te_snt=0.008658 16 | epoch 120, loss_tr=26.026142 err_tr=0.003203 loss_te=27.641922 err_te=0.344685 err_te_snt=0.007937 17 | epoch 128, loss_tr=26.024025 err_tr=0.002822 loss_te=27.620108 err_te=0.331377 err_te_snt=0.009380 18 | epoch 136, loss_tr=26.021912 err_tr=0.002217 loss_te=27.584499 err_te=0.307084 err_te_snt=0.003608 19 | epoch 144, loss_tr=26.020124 err_tr=0.001826 loss_te=27.587059 err_te=0.306003 err_te_snt=0.005772 20 | epoch 152, loss_tr=26.018707 err_tr=0.001729 loss_te=27.584858 err_te=0.306587 err_te_snt=0.006494 21 | epoch 160, loss_tr=26.017082 err_tr=0.001367 loss_te=27.573500 err_te=0.298311 err_te_snt=0.006494 22 | epoch 168, loss_tr=26.015982 err_tr=0.001221 loss_te=27.595425 err_te=0.310967 err_te_snt=0.006494 23 | epoch 176, loss_tr=26.014729 err_tr=0.001133 loss_te=27.582144 err_te=0.303880 err_te_snt=0.008658 24 | epoch 184, loss_tr=26.013859 err_tr=0.001045 loss_te=27.582970 err_te=0.305939 err_te_snt=0.008658 25 | epoch 192, loss_tr=26.012892 err_tr=0.000859 loss_te=27.627903 err_te=0.332149 err_te_snt=0.007937 26 | epoch 200, loss_tr=26.011984 err_tr=0.000732 loss_te=27.559088 err_te=0.289799 err_te_snt=0.003608 27 | epoch 208, loss_tr=26.011183 err_tr=0.000635 loss_te=27.557013 err_te=0.288368 err_te_snt=0.002886 28 | epoch 216, loss_tr=26.010441 err_tr=0.000635 loss_te=27.558214 err_te=0.288045 err_te_snt=0.005051 29 | epoch 224, loss_tr=26.009661 err_tr=0.000498 loss_te=27.548264 err_te=0.283481 err_te_snt=0.004329 30 | epoch 232, loss_tr=26.008757 err_tr=0.000332 loss_te=27.549431 err_te=0.280703 err_te_snt=0.004329 31 | epoch 240, loss_tr=26.008480 err_tr=0.000459 loss_te=27.552574 err_te=0.284366 err_te_snt=0.003608 32 | epoch 248, loss_tr=26.007788 err_tr=0.000361 loss_te=27.551683 err_te=0.283970 err_te_snt=0.003608 33 | epoch 256, loss_tr=26.007231 err_tr=0.000410 loss_te=27.683527 err_te=0.371611 err_te_snt=0.012266 34 | epoch 264, loss_tr=26.006718 err_tr=0.000312 loss_te=27.543749 err_te=0.279260 err_te_snt=0.002165 35 | epoch 272, loss_tr=26.006159 err_tr=0.000312 loss_te=27.691525 err_te=0.373524 err_te_snt=0.023810 36 | epoch 280, loss_tr=26.005768 err_tr=0.000273 loss_te=27.541985 err_te=0.276724 err_te_snt=0.002165 37 | epoch 288, loss_tr=26.005373 err_tr=0.000215 loss_te=27.539911 err_te=0.276444 err_te_snt=0.002165 38 | epoch 296, loss_tr=26.004900 err_tr=0.000176 loss_te=27.559622 err_te=0.285320 err_te_snt=0.005051 39 | epoch 304, loss_tr=26.004511 err_tr=0.000195 loss_te=27.549759 err_te=0.281368 err_te_snt=0.006494 40 | epoch 312, loss_tr=26.004179 err_tr=0.000176 loss_te=27.544083 err_te=0.276361 err_te_snt=0.003608 41 | epoch 320, loss_tr=26.003706 err_tr=0.000166 loss_te=27.574484 err_te=0.293938 err_te_snt=0.003608 42 | epoch 328, loss_tr=26.003380 err_tr=0.000107 loss_te=27.533012 err_te=0.270722 err_te_snt=0.001443 43 | epoch 336, loss_tr=26.003210 err_tr=0.000156 loss_te=27.543417 err_te=0.276093 err_te_snt=0.002886 44 | epoch 344, loss_tr=26.002758 err_tr=0.000166 loss_te=27.541979 err_te=0.276677 err_te_snt=0.005051 45 | epoch 352, loss_tr=26.002438 err_tr=0.000127 loss_te=27.531746 err_te=0.269554 err_te_snt=0.002886 46 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m080/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=29.662241 err_tr=0.992480 loss_te=30.207417 err_te=0.987834 err_te_snt=0.981241 2 | epoch 8, loss_tr=28.098215 err_tr=0.545400 loss_te=29.722664 err_te=0.745959 err_te_snt=0.411255 3 | epoch 16, loss_tr=27.850687 err_tr=0.322109 loss_te=29.465813 err_te=0.556569 err_te_snt=0.064935 4 | epoch 24, loss_tr=27.733288 err_tr=0.197773 loss_te=29.381256 err_te=0.498633 err_te_snt=0.043290 5 | epoch 32, loss_tr=27.666039 err_tr=0.123291 loss_te=29.360279 err_te=0.486815 err_te_snt=0.043290 6 | epoch 40, loss_tr=27.625864 err_tr=0.080098 loss_te=29.260355 err_te=0.419264 err_te_snt=0.022367 7 | epoch 48, loss_tr=27.600824 err_tr=0.054570 loss_te=29.267090 err_te=0.424572 err_te_snt=0.020924 8 | epoch 56, loss_tr=27.581503 err_tr=0.037764 loss_te=29.298346 err_te=0.450075 err_te_snt=0.035354 9 | epoch 64, loss_tr=27.568354 err_tr=0.026846 loss_te=29.188515 err_te=0.373402 err_te_snt=0.010101 10 | epoch 72, loss_tr=27.558846 err_tr=0.019824 loss_te=29.176489 err_te=0.367747 err_te_snt=0.005051 11 | epoch 80, loss_tr=27.551130 err_tr=0.014551 loss_te=29.142000 err_te=0.345110 err_te_snt=0.006494 12 | epoch 88, loss_tr=27.544813 err_tr=0.010605 loss_te=29.158020 err_te=0.353257 err_te_snt=0.008658 13 | epoch 96, loss_tr=27.539835 err_tr=0.008379 loss_te=29.135729 err_te=0.341465 err_te_snt=0.008658 14 | epoch 104, loss_tr=27.535444 err_tr=0.006396 loss_te=29.209509 err_te=0.392384 err_te_snt=0.018038 15 | epoch 112, loss_tr=27.532248 err_tr=0.005176 loss_te=29.115885 err_te=0.327478 err_te_snt=0.007937 16 | epoch 120, loss_tr=27.529419 err_tr=0.004609 loss_te=29.135120 err_te=0.337010 err_te_snt=0.006494 17 | epoch 128, loss_tr=27.526863 err_tr=0.003428 loss_te=29.138554 err_te=0.342086 err_te_snt=0.009380 18 | epoch 136, loss_tr=27.524496 err_tr=0.002891 loss_te=29.095333 err_te=0.313696 err_te_snt=0.006494 19 | epoch 144, loss_tr=27.522654 err_tr=0.002207 loss_te=29.090181 err_te=0.310755 err_te_snt=0.004329 20 | epoch 152, loss_tr=27.520901 err_tr=0.002031 loss_te=29.106438 err_te=0.318488 err_te_snt=0.005051 21 | epoch 160, loss_tr=27.519279 err_tr=0.001621 loss_te=29.090593 err_te=0.309890 err_te_snt=0.006494 22 | epoch 168, loss_tr=27.518051 err_tr=0.001514 loss_te=29.105558 err_te=0.317453 err_te_snt=0.003608 23 | epoch 176, loss_tr=27.516806 err_tr=0.001152 loss_te=29.091349 err_te=0.309243 err_te_snt=0.006494 24 | epoch 184, loss_tr=27.515633 err_tr=0.001172 loss_te=29.071806 err_te=0.295959 err_te_snt=0.007215 25 | epoch 192, loss_tr=27.514528 err_tr=0.000977 loss_te=29.071249 err_te=0.296124 err_te_snt=0.004329 26 | epoch 200, loss_tr=27.513613 err_tr=0.000820 loss_te=29.071392 err_te=0.299414 err_te_snt=0.004329 27 | epoch 208, loss_tr=27.512814 err_tr=0.000898 loss_te=29.087828 err_te=0.307156 err_te_snt=0.003608 28 | epoch 216, loss_tr=27.511858 err_tr=0.000811 loss_te=29.069839 err_te=0.296138 err_te_snt=0.004329 29 | epoch 224, loss_tr=27.511332 err_tr=0.000781 loss_te=29.066587 err_te=0.294012 err_te_snt=0.003608 30 | epoch 232, loss_tr=27.510378 err_tr=0.000508 loss_te=29.082279 err_te=0.302529 err_te_snt=0.006494 31 | epoch 240, loss_tr=27.509840 err_tr=0.000566 loss_te=29.074196 err_te=0.298029 err_te_snt=0.005051 32 | epoch 248, loss_tr=27.509169 err_tr=0.000527 loss_te=29.079334 err_te=0.299591 err_te_snt=0.003608 33 | epoch 256, loss_tr=27.508688 err_tr=0.000537 loss_te=29.058689 err_te=0.287636 err_te_snt=0.004329 34 | epoch 264, loss_tr=27.507999 err_tr=0.000312 loss_te=29.071255 err_te=0.292549 err_te_snt=0.002886 35 | epoch 272, loss_tr=27.507421 err_tr=0.000400 loss_te=29.068838 err_te=0.293344 err_te_snt=0.004329 36 | epoch 280, loss_tr=27.507133 err_tr=0.000488 loss_te=29.048180 err_te=0.281079 err_te_snt=0.002886 37 | epoch 288, loss_tr=27.506588 err_tr=0.000322 loss_te=29.054995 err_te=0.283459 err_te_snt=0.003608 38 | epoch 296, loss_tr=27.506117 err_tr=0.000439 loss_te=29.046722 err_te=0.279643 err_te_snt=0.004329 39 | epoch 304, loss_tr=27.505661 err_tr=0.000205 loss_te=29.058332 err_te=0.288057 err_te_snt=0.006494 40 | epoch 312, loss_tr=27.505394 err_tr=0.000244 loss_te=29.070637 err_te=0.293951 err_te_snt=0.005772 41 | epoch 320, loss_tr=27.504763 err_tr=0.000205 loss_te=29.053230 err_te=0.283250 err_te_snt=0.002886 42 | epoch 328, loss_tr=27.504452 err_tr=0.000195 loss_te=29.058105 err_te=0.285305 err_te_snt=0.004329 43 | epoch 336, loss_tr=27.504116 err_tr=0.000234 loss_te=29.056849 err_te=0.286871 err_te_snt=0.003608 44 | epoch 344, loss_tr=27.503794 err_tr=0.000195 loss_te=29.070412 err_te=0.293485 err_te_snt=0.005772 45 | epoch 352, loss_tr=27.503281 err_tr=0.000078 loss_te=29.093245 err_te=0.308555 err_te_snt=0.005051 46 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m085/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=31.182407 err_tr=0.992217 loss_te=31.708124 err_te=0.985807 err_te_snt=0.974747 2 | epoch 0, loss_tr=31.115841 err_tr=0.989092 loss_te=31.674290 err_te=0.977926 err_te_snt=0.940837 3 | epoch 8, loss_tr=29.559862 err_tr=0.516123 loss_te=31.102921 err_te=0.657236 err_te_snt=0.137085 4 | epoch 16, loss_tr=29.314537 err_tr=0.281230 loss_te=30.945471 err_te=0.542559 err_te_snt=0.061328 5 | epoch 24, loss_tr=29.205217 err_tr=0.165068 loss_te=30.831175 err_te=0.463775 err_te_snt=0.028860 6 | epoch 32, loss_tr=29.146418 err_tr=0.101611 loss_te=30.773666 err_te=0.425705 err_te_snt=0.017316 7 | epoch 40, loss_tr=29.111418 err_tr=0.065439 loss_te=30.720898 err_te=0.393551 err_te_snt=0.012987 8 | epoch 48, loss_tr=29.089684 err_tr=0.044658 loss_te=30.693295 err_te=0.375885 err_te_snt=0.012987 9 | epoch 56, loss_tr=29.072882 err_tr=0.030957 loss_te=30.703857 err_te=0.383767 err_te_snt=0.007937 10 | epoch 64, loss_tr=29.061052 err_tr=0.021221 loss_te=30.724230 err_te=0.402657 err_te_snt=0.018038 11 | epoch 72, loss_tr=29.052691 err_tr=0.016250 loss_te=30.644485 err_te=0.348196 err_te_snt=0.006494 12 | epoch 80, loss_tr=29.045500 err_tr=0.011309 loss_te=30.656036 err_te=0.353614 err_te_snt=0.011544 13 | epoch 88, loss_tr=29.039833 err_tr=0.008721 loss_te=30.630896 err_te=0.336030 err_te_snt=0.008658 14 | epoch 96, loss_tr=29.035179 err_tr=0.006133 loss_te=30.623390 err_te=0.333280 err_te_snt=0.008658 15 | epoch 104, loss_tr=29.031408 err_tr=0.005000 loss_te=30.588514 err_te=0.311365 err_te_snt=0.007937 16 | epoch 112, loss_tr=29.028242 err_tr=0.003877 loss_te=30.601198 err_te=0.319561 err_te_snt=0.007215 17 | epoch 120, loss_tr=29.025505 err_tr=0.003340 loss_te=30.637800 err_te=0.342327 err_te_snt=0.008658 18 | epoch 128, loss_tr=29.023067 err_tr=0.002275 loss_te=30.608908 err_te=0.322863 err_te_snt=0.007937 19 | epoch 136, loss_tr=29.021027 err_tr=0.002070 loss_te=30.593184 err_te=0.315098 err_te_snt=0.009380 20 | epoch 144, loss_tr=29.019327 err_tr=0.001602 loss_te=30.585913 err_te=0.308741 err_te_snt=0.004329 21 | epoch 152, loss_tr=29.018023 err_tr=0.001689 loss_te=30.573252 err_te=0.297864 err_te_snt=0.006494 22 | epoch 160, loss_tr=29.016310 err_tr=0.001367 loss_te=30.605213 err_te=0.321474 err_te_snt=0.011544 23 | epoch 168, loss_tr=29.015146 err_tr=0.001045 loss_te=30.557985 err_te=0.291187 err_te_snt=0.003608 24 | epoch 176, loss_tr=29.013824 err_tr=0.000928 loss_te=30.552748 err_te=0.287068 err_te_snt=0.005051 25 | epoch 184, loss_tr=29.012800 err_tr=0.000771 loss_te=30.543684 err_te=0.281702 err_te_snt=0.003608 26 | epoch 192, loss_tr=29.011892 err_tr=0.000586 loss_te=30.539141 err_te=0.279211 err_te_snt=0.005772 27 | epoch 200, loss_tr=29.011049 err_tr=0.000625 loss_te=30.553427 err_te=0.285541 err_te_snt=0.005772 28 | epoch 208, loss_tr=29.010271 err_tr=0.000430 loss_te=30.545141 err_te=0.279709 err_te_snt=0.002886 29 | epoch 216, loss_tr=29.009451 err_tr=0.000430 loss_te=30.541853 err_te=0.279681 err_te_snt=0.004329 30 | epoch 224, loss_tr=29.008791 err_tr=0.000410 loss_te=30.545046 err_te=0.280753 err_te_snt=0.005772 31 | epoch 232, loss_tr=29.007978 err_tr=0.000264 loss_te=30.541807 err_te=0.279208 err_te_snt=0.005772 32 | epoch 240, loss_tr=29.007498 err_tr=0.000400 loss_te=30.551218 err_te=0.283912 err_te_snt=0.004329 33 | epoch 248, loss_tr=29.006935 err_tr=0.000410 loss_te=30.549187 err_te=0.283272 err_te_snt=0.004329 34 | epoch 256, loss_tr=29.006365 err_tr=0.000303 loss_te=30.599998 err_te=0.311982 err_te_snt=0.006494 35 | epoch 264, loss_tr=29.005880 err_tr=0.000176 loss_te=30.520559 err_te=0.266424 err_te_snt=0.002886 36 | epoch 272, loss_tr=29.005312 err_tr=0.000225 loss_te=30.553562 err_te=0.284662 err_te_snt=0.005051 37 | epoch 280, loss_tr=29.004921 err_tr=0.000176 loss_te=30.539967 err_te=0.273504 err_te_snt=0.002886 38 | epoch 288, loss_tr=29.004519 err_tr=0.000195 loss_te=30.519636 err_te=0.264977 err_te_snt=0.002886 39 | epoch 296, loss_tr=29.004013 err_tr=0.000186 loss_te=30.530638 err_te=0.270594 err_te_snt=0.003608 40 | epoch 304, loss_tr=29.003595 err_tr=0.000117 loss_te=30.567476 err_te=0.292378 err_te_snt=0.006494 41 | epoch 312, loss_tr=29.003366 err_tr=0.000225 loss_te=30.541445 err_te=0.276557 err_te_snt=0.004329 42 | epoch 320, loss_tr=29.002895 err_tr=0.000146 loss_te=30.540991 err_te=0.275400 err_te_snt=0.002165 43 | epoch 328, loss_tr=29.002615 err_tr=0.000117 loss_te=30.530384 err_te=0.270332 err_te_snt=0.002886 44 | epoch 336, loss_tr=29.002211 err_tr=0.000127 loss_te=30.513950 err_te=0.261862 err_te_snt=0.003608 45 | epoch 344, loss_tr=29.001991 err_tr=0.000088 loss_te=30.521730 err_te=0.265537 err_te_snt=0.005051 46 | epoch 352, loss_tr=29.001595 err_tr=0.000156 loss_te=30.529800 err_te=0.269582 err_te_snt=0.004329 47 | epoch 360, loss_tr=29.001364 err_tr=0.000127 loss_te=30.550049 err_te=0.283203 err_te_snt=0.005051 48 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m090/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=32.688496 err_tr=0.993125 loss_te=33.222008 err_te=0.987994 err_te_snt=0.981241 2 | epoch 8, loss_tr=31.111864 err_tr=0.560088 loss_te=32.649090 err_te=0.694123 err_te_snt=0.185426 3 | epoch 16, loss_tr=30.856909 err_tr=0.329990 loss_te=32.473866 err_te=0.561868 err_te_snt=0.069264 4 | epoch 24, loss_tr=30.735878 err_tr=0.198936 loss_te=32.380829 err_te=0.498575 err_te_snt=0.038240 5 | epoch 32, loss_tr=30.667448 err_tr=0.124023 loss_te=32.339069 err_te=0.469341 err_te_snt=0.029582 6 | epoch 40, loss_tr=30.626013 err_tr=0.080596 loss_te=32.257633 err_te=0.415394 err_te_snt=0.019481 7 | epoch 48, loss_tr=30.600290 err_tr=0.054619 loss_te=32.216316 err_te=0.391821 err_te_snt=0.011544 8 | epoch 56, loss_tr=30.581617 err_tr=0.037559 loss_te=32.212776 err_te=0.391401 err_te_snt=0.013709 9 | epoch 64, loss_tr=30.568878 err_tr=0.027207 loss_te=32.160366 err_te=0.356732 err_te_snt=0.007937 10 | epoch 72, loss_tr=30.558773 err_tr=0.019492 loss_te=32.160538 err_te=0.356573 err_te_snt=0.010101 11 | epoch 80, loss_tr=30.550400 err_tr=0.014258 loss_te=32.162270 err_te=0.356720 err_te_snt=0.012987 12 | epoch 88, loss_tr=30.543859 err_tr=0.010498 loss_te=32.198490 err_te=0.382199 err_te_snt=0.014430 13 | epoch 96, loss_tr=30.538847 err_tr=0.007910 loss_te=32.126011 err_te=0.335011 err_te_snt=0.006494 14 | epoch 104, loss_tr=30.534708 err_tr=0.006279 loss_te=32.130531 err_te=0.337953 err_te_snt=0.005772 15 | epoch 112, loss_tr=30.531609 err_tr=0.005381 loss_te=32.114670 err_te=0.327619 err_te_snt=0.005772 16 | epoch 120, loss_tr=30.528032 err_tr=0.004043 loss_te=32.085564 err_te=0.308462 err_te_snt=0.004329 17 | epoch 128, loss_tr=30.525957 err_tr=0.003701 loss_te=32.111301 err_te=0.326970 err_te_snt=0.007215 18 | epoch 136, loss_tr=30.523083 err_tr=0.002354 loss_te=32.080894 err_te=0.307661 err_te_snt=0.005772 19 | epoch 144, loss_tr=30.521355 err_tr=0.002217 loss_te=32.080807 err_te=0.307709 err_te_snt=0.005051 20 | epoch 152, loss_tr=30.519814 err_tr=0.001855 loss_te=32.077709 err_te=0.302088 err_te_snt=0.002165 21 | epoch 160, loss_tr=30.518011 err_tr=0.001592 loss_te=32.054325 err_te=0.289147 err_te_snt=0.003608 22 | epoch 168, loss_tr=30.516838 err_tr=0.001514 loss_te=32.071301 err_te=0.299445 err_te_snt=0.005051 23 | epoch 176, loss_tr=30.515514 err_tr=0.001221 loss_te=32.090508 err_te=0.311254 err_te_snt=0.004329 24 | epoch 184, loss_tr=30.514267 err_tr=0.001006 loss_te=32.073700 err_te=0.298672 err_te_snt=0.005051 25 | epoch 192, loss_tr=30.513214 err_tr=0.000967 loss_te=32.044441 err_te=0.281218 err_te_snt=0.002886 26 | epoch 200, loss_tr=30.512251 err_tr=0.000850 loss_te=32.052277 err_te=0.285959 err_te_snt=0.003608 27 | epoch 208, loss_tr=30.511471 err_tr=0.000664 loss_te=32.039360 err_te=0.278696 err_te_snt=0.003608 28 | epoch 216, loss_tr=30.510580 err_tr=0.000703 loss_te=32.049126 err_te=0.283023 err_te_snt=0.004329 29 | epoch 224, loss_tr=30.509731 err_tr=0.000508 loss_te=32.060001 err_te=0.292044 err_te_snt=0.005051 30 | epoch 232, loss_tr=30.509073 err_tr=0.000479 loss_te=32.035744 err_te=0.276471 err_te_snt=0.002165 31 | epoch 240, loss_tr=30.508396 err_tr=0.000400 loss_te=32.041477 err_te=0.278904 err_te_snt=0.003608 32 | epoch 248, loss_tr=30.507833 err_tr=0.000498 loss_te=32.042252 err_te=0.279327 err_te_snt=0.002886 33 | epoch 256, loss_tr=30.507219 err_tr=0.000322 loss_te=32.126945 err_te=0.329180 err_te_snt=0.007937 34 | epoch 264, loss_tr=30.506777 err_tr=0.000283 loss_te=32.067574 err_te=0.294571 err_te_snt=0.005772 35 | epoch 272, loss_tr=30.506100 err_tr=0.000303 loss_te=32.090797 err_te=0.307156 err_te_snt=0.004329 36 | epoch 280, loss_tr=30.505768 err_tr=0.000283 loss_te=32.051224 err_te=0.281815 err_te_snt=0.003608 37 | epoch 288, loss_tr=30.505384 err_tr=0.000312 loss_te=32.040508 err_te=0.278635 err_te_snt=0.004329 38 | epoch 296, loss_tr=30.504791 err_tr=0.000264 loss_te=32.038681 err_te=0.276921 err_te_snt=0.003608 39 | epoch 304, loss_tr=30.504395 err_tr=0.000254 loss_te=32.027996 err_te=0.268793 err_te_snt=0.002886 40 | epoch 312, loss_tr=30.504047 err_tr=0.000234 loss_te=32.060490 err_te=0.290210 err_te_snt=0.004329 41 | epoch 320, loss_tr=30.503593 err_tr=0.000264 loss_te=32.021145 err_te=0.267086 err_te_snt=0.002886 42 | epoch 328, loss_tr=30.503105 err_tr=0.000215 loss_te=32.026390 err_te=0.268899 err_te_snt=0.003608 43 | epoch 336, loss_tr=30.503008 err_tr=0.000137 loss_te=32.041103 err_te=0.276474 err_te_snt=0.003608 44 | epoch 344, loss_tr=30.502554 err_tr=0.000127 loss_te=32.043091 err_te=0.278421 err_te_snt=0.005051 45 | epoch 352, loss_tr=30.502224 err_tr=0.000098 loss_te=32.022717 err_te=0.265415 err_te_snt=0.002886 46 | epoch 360, loss_tr=30.501984 err_tr=0.000205 loss_te=32.066032 err_te=0.292378 err_te_snt=0.004329 47 | -------------------------------------------------------------------------------- /exp/SincNet_TIMIT_m095/res.res: -------------------------------------------------------------------------------- 1 | epoch 0, loss_tr=34.166695 err_tr=0.987305 loss_te=34.661682 err_te=0.975016 err_te_snt=0.943723 2 | epoch 8, loss_tr=32.541798 err_tr=0.499023 loss_te=34.105118 err_te=0.658864 err_te_snt=0.150794 3 | epoch 16, loss_tr=32.306347 err_tr=0.272812 loss_te=33.943050 err_te=0.537214 err_te_snt=0.048341 4 | epoch 24, loss_tr=32.201214 err_tr=0.159512 loss_te=33.839233 err_te=0.467152 err_te_snt=0.027417 5 | epoch 32, loss_tr=32.144402 err_tr=0.100273 loss_te=33.761150 err_te=0.414835 err_te_snt=0.013709 6 | epoch 40, loss_tr=32.109703 err_tr=0.062793 loss_te=33.773487 err_te=0.428098 err_te_snt=0.025974 7 | epoch 48, loss_tr=32.088219 err_tr=0.042939 loss_te=33.753792 err_te=0.416012 err_te_snt=0.018759 8 | epoch 56, loss_tr=32.071377 err_tr=0.028506 loss_te=33.691055 err_te=0.373894 err_te_snt=0.010823 9 | epoch 64, loss_tr=32.060223 err_tr=0.020264 loss_te=33.678959 err_te=0.369150 err_te_snt=0.012266 10 | epoch 72, loss_tr=32.051956 err_tr=0.014736 loss_te=33.688293 err_te=0.373932 err_te_snt=0.015152 11 | epoch 80, loss_tr=32.044510 err_tr=0.010127 loss_te=33.623077 err_te=0.332953 err_te_snt=0.005772 12 | epoch 88, loss_tr=32.038895 err_tr=0.007607 loss_te=33.611538 err_te=0.322797 err_te_snt=0.008658 13 | epoch 96, loss_tr=32.034344 err_tr=0.005791 loss_te=33.619644 err_te=0.327830 err_te_snt=0.007937 14 | epoch 104, loss_tr=32.030819 err_tr=0.004375 loss_te=33.637627 err_te=0.340463 err_te_snt=0.008658 15 | epoch 112, loss_tr=32.027931 err_tr=0.004033 loss_te=33.603287 err_te=0.319166 err_te_snt=0.006494 16 | epoch 120, loss_tr=32.025002 err_tr=0.003145 loss_te=33.596046 err_te=0.313969 err_te_snt=0.005051 17 | epoch 128, loss_tr=32.022854 err_tr=0.002275 loss_te=33.600937 err_te=0.317293 err_te_snt=0.007937 18 | epoch 136, loss_tr=32.020878 err_tr=0.002061 loss_te=33.587288 err_te=0.308821 err_te_snt=0.005772 19 | epoch 144, loss_tr=32.018917 err_tr=0.001660 loss_te=33.579792 err_te=0.301157 err_te_snt=0.005051 20 | epoch 152, loss_tr=32.017624 err_tr=0.001406 loss_te=33.569389 err_te=0.294776 err_te_snt=0.005772 21 | epoch 160, loss_tr=32.016125 err_tr=0.001182 loss_te=33.579845 err_te=0.299588 err_te_snt=0.005772 22 | epoch 168, loss_tr=32.015179 err_tr=0.001260 loss_te=33.564739 err_te=0.294379 err_te_snt=0.005772 23 | epoch 176, loss_tr=32.013760 err_tr=0.000937 loss_te=33.562809 err_te=0.290231 err_te_snt=0.005772 24 | epoch 184, loss_tr=32.012711 err_tr=0.000850 loss_te=33.581863 err_te=0.304984 err_te_snt=0.005772 25 | epoch 192, loss_tr=32.011730 err_tr=0.000586 loss_te=33.557060 err_te=0.288271 err_te_snt=0.005051 26 | epoch 200, loss_tr=32.010921 err_tr=0.000479 loss_te=33.556137 err_te=0.286743 err_te_snt=0.004329 27 | epoch 208, loss_tr=32.010159 err_tr=0.000566 loss_te=33.552788 err_te=0.284373 err_te_snt=0.003608 28 | epoch 216, loss_tr=32.009388 err_tr=0.000547 loss_te=33.576401 err_te=0.300053 err_te_snt=0.005051 29 | epoch 224, loss_tr=32.008762 err_tr=0.000430 loss_te=33.575401 err_te=0.295315 err_te_snt=0.005051 30 | epoch 232, loss_tr=32.007889 err_tr=0.000342 loss_te=33.548889 err_te=0.281406 err_te_snt=0.004329 31 | epoch 240, loss_tr=32.007473 err_tr=0.000488 loss_te=33.546329 err_te=0.279760 err_te_snt=0.005772 32 | epoch 248, loss_tr=32.006844 err_tr=0.000234 loss_te=33.559708 err_te=0.288019 err_te_snt=0.005051 33 | epoch 256, loss_tr=32.006275 err_tr=0.000283 loss_te=33.565811 err_te=0.289250 err_te_snt=0.003608 34 | epoch 264, loss_tr=32.005856 err_tr=0.000234 loss_te=33.601986 err_te=0.314550 err_te_snt=0.010823 35 | epoch 272, loss_tr=32.005402 err_tr=0.000244 loss_te=33.574051 err_te=0.296557 err_te_snt=0.006494 36 | epoch 280, loss_tr=32.004932 err_tr=0.000264 loss_te=33.545403 err_te=0.275925 err_te_snt=0.003608 37 | epoch 288, loss_tr=32.004593 err_tr=0.000195 loss_te=33.530079 err_te=0.270748 err_te_snt=0.004329 38 | epoch 296, loss_tr=32.004131 err_tr=0.000234 loss_te=33.545940 err_te=0.276437 err_te_snt=0.002886 39 | epoch 304, loss_tr=32.003555 err_tr=0.000068 loss_te=33.529778 err_te=0.269931 err_te_snt=0.002886 40 | epoch 312, loss_tr=32.003269 err_tr=0.000107 loss_te=33.539898 err_te=0.274828 err_te_snt=0.004329 41 | epoch 320, loss_tr=32.002850 err_tr=0.000098 loss_te=33.541080 err_te=0.274765 err_te_snt=0.002165 42 | epoch 328, loss_tr=32.002533 err_tr=0.000166 loss_te=33.545811 err_te=0.278029 err_te_snt=0.005772 43 | epoch 336, loss_tr=32.002258 err_tr=0.000127 loss_te=33.528374 err_te=0.266960 err_te_snt=0.003608 44 | epoch 344, loss_tr=32.002003 err_tr=0.000146 loss_te=33.537498 err_te=0.272904 err_te_snt=0.002886 45 | epoch 352, loss_tr=32.001530 err_tr=0.000049 loss_te=33.530689 err_te=0.266488 err_te_snt=0.004329 46 | epoch 360, loss_tr=32.001461 err_tr=0.000068 loss_te=33.545307 err_te=0.278195 err_te_snt=0.005772 47 | -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | cffi==1.11.5 2 | numpy==1.15.4 3 | Pillow==5.3.0 4 | pycparser==2.19 5 | PySoundFile==0.9.0.post1 6 | six==1.11.0 7 | torch==0.4.1 8 | torchvision==0.2.1 9 | -------------------------------------------------------------------------------- /speaker_id.py: -------------------------------------------------------------------------------- 1 | # speaker_id.py 2 | # ------------------ 3 | # Created by 4 | # Mirco Ravanelli 5 | # Mila - University of Montreal 6 | 7 | # July 2018 8 | # ------------------ 9 | # Modified by 10 | # João Antônio Chagas Nunes 11 | # Universidade Federal de Pernambuco 12 | 13 | # Add AM-Softmax loss function 14 | # December 2018 15 | # ------------------ 16 | 17 | # Description: 18 | # This code performs a speaker_id experiments with SincNet. 19 | 20 | # How to run it: 21 | # python speaker_id.py --cfg=cfg/SincNet_TIMIT.cfg 22 | 23 | # 24 | 25 | import os 26 | #import scipy.io.wavfile 27 | import soundfile as sf 28 | import torch 29 | import torch.nn as nn 30 | import torch.nn.functional as F 31 | import torch.optim as optim 32 | from torch.autograd import Variable 33 | 34 | import sys 35 | import numpy as np 36 | from dnn_models import MLP,flip 37 | from dnn_models import SincNet as CNN 38 | from data_io import ReadList,read_conf,str_to_bool 39 | 40 | 41 | def create_batches_rnd(batch_size,data_folder,wav_lst,N_snt,wlen,lab_dict,fact_amp): 42 | 43 | # Initialization of the minibatch (batch_size,[0=>x_t,1=>x_t+N,1=>random_samp]) 44 | sig_batch=np.zeros([batch_size,wlen]) 45 | lab_batch=np.zeros(batch_size) 46 | 47 | snt_id_arr=np.random.randint(N_snt, size=batch_size) 48 | 49 | rand_amp_arr = np.random.uniform(1.0-fact_amp,1+fact_amp,batch_size) 50 | 51 | for i in range(batch_size): 52 | 53 | # select a random sentence from the list 54 | #[fs,signal]=scipy.io.wavfile.read(data_folder+wav_lst[snt_id_arr[i]]) 55 | #signal=signal.astype(float)/32768 56 | 57 | [signal, fs] = sf.read(data_folder+wav_lst[snt_id_arr[i]]) 58 | 59 | # accesing to a random chunk 60 | snt_len=signal.shape[0] 61 | snt_beg=np.random.randint(snt_len-wlen-1) #randint(0, snt_len-2*wlen-1) 62 | snt_end=snt_beg+wlen 63 | 64 | sig_batch[i,:]=signal[snt_beg:snt_end]*rand_amp_arr[i] 65 | lab_batch[i]=lab_dict[wav_lst[snt_id_arr[i]]] 66 | 67 | inp=Variable(torch.from_numpy(sig_batch).float().cuda().contiguous()) 68 | lab=Variable(torch.from_numpy(lab_batch).float().cuda().contiguous()) 69 | 70 | return inp,lab 71 | 72 | class AdditiveMarginSoftmax(nn.Module): 73 | # AMSoftmax 74 | def __init__(self, margin=0.35, s=30): 75 | super().__init__() 76 | 77 | self.m = margin # 78 | self.s = s 79 | self.epsilon = 0.000000000001 80 | print('AMSoftmax m = ' + str(margin)) 81 | 82 | def forward(self, predicted, target): 83 | 84 | # ------------ AM Softmax ------------ # 85 | predicted = predicted / (predicted.norm(p=2, dim=0) + self.epsilon) 86 | indexes = range(predicted.size(0)) 87 | cos_theta_y = predicted[indexes, target] 88 | cos_theta_y_m = cos_theta_y - self.m 89 | exp_s = np.e ** (self.s * cos_theta_y_m) 90 | 91 | sum_cos_theta_j = (np.e ** (predicted * self.s)).sum(dim=1) - (np.e ** (predicted[indexes, target] * self.s)) 92 | 93 | log = -torch.log(exp_s/(exp_s+sum_cos_theta_j+self.epsilon)).mean() 94 | 95 | return log 96 | 97 | 98 | 99 | 100 | # Reading cfg file 101 | options=read_conf() 102 | 103 | 104 | #[data] 105 | tr_lst=options.tr_lst 106 | te_lst=options.te_lst 107 | pt_file=options.pt_file 108 | class_dict_file=options.lab_dict 109 | data_folder=options.data_folder+'/' 110 | output_folder=options.output_folder 111 | 112 | #[windowing] 113 | fs=int(options.fs) 114 | cw_len=int(options.cw_len) 115 | cw_shift=int(options.cw_shift) 116 | 117 | #[cnn] 118 | cnn_N_filt=list(map(int, options.cnn_N_filt.split(','))) 119 | cnn_len_filt=list(map(int, options.cnn_len_filt.split(','))) 120 | cnn_max_pool_len=list(map(int, options.cnn_max_pool_len.split(','))) 121 | cnn_use_laynorm_inp=str_to_bool(options.cnn_use_laynorm_inp) 122 | cnn_use_batchnorm_inp=str_to_bool(options.cnn_use_batchnorm_inp) 123 | cnn_use_laynorm=list(map(str_to_bool, options.cnn_use_laynorm.split(','))) 124 | cnn_use_batchnorm=list(map(str_to_bool, options.cnn_use_batchnorm.split(','))) 125 | cnn_act=list(map(str, options.cnn_act.split(','))) 126 | cnn_drop=list(map(float, options.cnn_drop.split(','))) 127 | 128 | 129 | #[dnn] 130 | fc_lay=list(map(int, options.fc_lay.split(','))) 131 | fc_drop=list(map(float, options.fc_drop.split(','))) 132 | fc_use_laynorm_inp=str_to_bool(options.fc_use_laynorm_inp) 133 | fc_use_batchnorm_inp=str_to_bool(options.fc_use_batchnorm_inp) 134 | fc_use_batchnorm=list(map(str_to_bool, options.fc_use_batchnorm.split(','))) 135 | fc_use_laynorm=list(map(str_to_bool, options.fc_use_laynorm.split(','))) 136 | fc_act=list(map(str, options.fc_act.split(','))) 137 | 138 | #[class] 139 | class_lay=list(map(int, options.class_lay.split(','))) 140 | class_drop=list(map(float, options.class_drop.split(','))) 141 | class_use_laynorm_inp=str_to_bool(options.class_use_laynorm_inp) 142 | class_use_batchnorm_inp=str_to_bool(options.class_use_batchnorm_inp) 143 | class_use_batchnorm=list(map(str_to_bool, options.class_use_batchnorm.split(','))) 144 | class_use_laynorm=list(map(str_to_bool, options.class_use_laynorm.split(','))) 145 | class_act=list(map(str, options.class_act.split(','))) 146 | 147 | 148 | #[optimization] 149 | lr=float(options.lr) 150 | batch_size=int(options.batch_size) 151 | N_epochs=int(options.N_epochs) 152 | N_batches=int(options.N_batches) 153 | N_eval_epoch=int(options.N_eval_epoch) 154 | seed=int(options.seed) 155 | 156 | 157 | # training list 158 | wav_lst_tr=ReadList(tr_lst) 159 | snt_tr=len(wav_lst_tr) 160 | 161 | # test list 162 | wav_lst_te=ReadList(te_lst) 163 | snt_te=len(wav_lst_te) 164 | 165 | 166 | # Folder creation 167 | try: 168 | os.stat(output_folder) 169 | except: 170 | os.mkdir(output_folder) 171 | 172 | 173 | # setting seed 174 | torch.manual_seed(seed) 175 | np.random.seed(seed) 176 | 177 | # loss function 178 | if (options.AMSoftmax == 'True'): 179 | print('Using AMSoftmax loss function...') 180 | cost = AdditiveMarginSoftmax(margin=float(options.AMSoftmax_m)) 181 | 182 | else: 183 | print('Using Softmax loss function...') 184 | cost = nn.NLLLoss() 185 | 186 | 187 | 188 | # Converting context and shift in samples 189 | wlen=int(fs*cw_len/1000.00) 190 | wshift=int(fs*cw_shift/1000.00) 191 | 192 | # Batch_dev 193 | Batch_dev=batch_size 194 | 195 | 196 | # Feature extractor CNN 197 | CNN_arch = {'input_dim': wlen, 198 | 'fs': fs, 199 | 'cnn_N_filt': cnn_N_filt, 200 | 'cnn_len_filt': cnn_len_filt, 201 | 'cnn_max_pool_len':cnn_max_pool_len, 202 | 'cnn_use_laynorm_inp': cnn_use_laynorm_inp, 203 | 'cnn_use_batchnorm_inp': cnn_use_batchnorm_inp, 204 | 'cnn_use_laynorm':cnn_use_laynorm, 205 | 'cnn_use_batchnorm':cnn_use_batchnorm, 206 | 'cnn_act': cnn_act, 207 | 'cnn_drop':cnn_drop, 208 | } 209 | 210 | CNN_net=CNN(CNN_arch) 211 | CNN_net.cuda() 212 | 213 | # Loading label dictionary 214 | lab_dict=np.load(class_dict_file).item() 215 | 216 | 217 | 218 | DNN1_arch = {'input_dim': CNN_net.out_dim, 219 | 'fc_lay': fc_lay, 220 | 'fc_drop': fc_drop, 221 | 'fc_use_batchnorm': fc_use_batchnorm, 222 | 'fc_use_laynorm': fc_use_laynorm, 223 | 'fc_use_laynorm_inp': fc_use_laynorm_inp, 224 | 'fc_use_batchnorm_inp':fc_use_batchnorm_inp, 225 | 'fc_act': fc_act, 226 | } 227 | 228 | DNN1_net=MLP(DNN1_arch) 229 | DNN1_net.cuda() 230 | 231 | 232 | DNN2_arch = {'input_dim':fc_lay[-1] , 233 | 'fc_lay': class_lay, 234 | 'fc_drop': class_drop, 235 | 'fc_use_batchnorm': class_use_batchnorm, 236 | 'fc_use_laynorm': class_use_laynorm, 237 | 'fc_use_laynorm_inp': class_use_laynorm_inp, 238 | 'fc_use_batchnorm_inp':class_use_batchnorm_inp, 239 | 'fc_act': class_act, 240 | } 241 | 242 | 243 | DNN2_net=MLP(DNN2_arch) 244 | DNN2_net.cuda() 245 | 246 | 247 | if pt_file!='none': 248 | checkpoint_load = torch.load(pt_file) 249 | CNN_net.load_state_dict(checkpoint_load['CNN_model_par']) 250 | DNN1_net.load_state_dict(checkpoint_load['DNN1_model_par']) 251 | DNN2_net.load_state_dict(checkpoint_load['DNN2_model_par']) 252 | 253 | 254 | 255 | optimizer_CNN = optim.RMSprop(CNN_net.parameters(), lr=lr,alpha=0.95, eps=1e-8) 256 | optimizer_DNN1 = optim.RMSprop(DNN1_net.parameters(), lr=lr,alpha=0.95, eps=1e-8) 257 | optimizer_DNN2 = optim.RMSprop(DNN2_net.parameters(), lr=lr,alpha=0.95, eps=1e-8) 258 | 259 | 260 | 261 | for epoch in range(N_epochs): 262 | 263 | test_flag=0 264 | CNN_net.train() 265 | DNN1_net.train() 266 | DNN2_net.train() 267 | 268 | loss_sum=0 269 | err_sum=0 270 | 271 | for i in range(N_batches): 272 | 273 | [inp,lab]=create_batches_rnd(batch_size,data_folder,wav_lst_tr,snt_tr,wlen,lab_dict,0.2) 274 | pout=DNN2_net(DNN1_net(CNN_net(inp))) 275 | 276 | pred=torch.max(pout,dim=1)[1] 277 | loss = cost(pout, lab.long()) 278 | 279 | err = torch.mean((pred!=lab.long()).float()) 280 | 281 | optimizer_CNN.zero_grad() 282 | optimizer_DNN1.zero_grad() 283 | optimizer_DNN2.zero_grad() 284 | 285 | loss.backward() 286 | optimizer_CNN.step() 287 | optimizer_DNN1.step() 288 | optimizer_DNN2.step() 289 | 290 | loss_sum=loss_sum+loss.detach() 291 | err_sum=err_sum+err.detach() 292 | 293 | 294 | loss_tot=loss_sum/N_batches 295 | err_tot=err_sum/N_batches 296 | 297 | 298 | 299 | 300 | # Full Validation new 301 | if epoch%N_eval_epoch==0: 302 | 303 | CNN_net.eval() 304 | DNN1_net.eval() 305 | DNN2_net.eval() 306 | test_flag=1 307 | loss_sum=0 308 | err_sum=0 309 | err_sum_snt=0 310 | 311 | with torch.no_grad(): 312 | for i in range(snt_te): 313 | 314 | #[fs,signal]=scipy.io.wavfile.read(data_folder+wav_lst_te[i]) 315 | #signal=signal.astype(float)/32768 316 | 317 | [signal, fs] = sf.read(data_folder+wav_lst_te[i]) 318 | 319 | signal=torch.from_numpy(signal).float().cuda().contiguous() 320 | lab_batch=lab_dict[wav_lst_te[i]] 321 | 322 | # split signals into chunks 323 | beg_samp=0 324 | end_samp=wlen 325 | 326 | N_fr=int((signal.shape[0]-wlen)/(wshift)) 327 | 328 | sig_arr=np.zeros([Batch_dev,wlen]) 329 | lab= Variable((torch.zeros(N_fr+1)+lab_batch).cuda().contiguous().long()) 330 | pout=Variable(torch.zeros(N_fr+1,class_lay[-1]).float().cuda().contiguous()) 331 | count_fr=0 332 | count_fr_tot=0 333 | while end_samp0: 346 | inp=Variable(torch.from_numpy(sig_arr[0:count_fr]).float().cuda().contiguous()) 347 | pout[count_fr_tot-count_fr:count_fr_tot,:]=DNN2_net(DNN1_net(CNN_net(inp))) 348 | 349 | pred=torch.max(pout,dim=1)[1] 350 | loss = cost(pout, lab.long()) 351 | 352 | err = torch.mean((pred!=lab.long()).float()) 353 | 354 | [val,best_class]=torch.max(torch.sum(pout,dim=0),0) 355 | err_sum_snt=err_sum_snt+(best_class!=lab[0]).float() 356 | 357 | 358 | loss_sum=loss_sum+loss.detach() 359 | err_sum=err_sum+err.detach() 360 | 361 | err_tot_dev_snt=err_sum_snt/snt_te 362 | loss_tot_dev=loss_sum/snt_te 363 | err_tot_dev=err_sum/snt_te 364 | 365 | 366 | print("epoch %i, loss_tr=%f err_tr=%f loss_te=%f err_te=%f err_te_snt=%f" % (epoch, loss_tot,err_tot,loss_tot_dev,err_tot_dev,err_tot_dev_snt)) 367 | 368 | with open(output_folder+"/res.res", "a") as res_file: 369 | res_file.write("epoch %i, loss_tr=%f err_tr=%f loss_te=%f err_te=%f err_te_snt=%f\n" % (epoch, loss_tot,err_tot,loss_tot_dev,err_tot_dev,err_tot_dev_snt)) 370 | 371 | checkpoint={'CNN_model_par': CNN_net.state_dict(), 372 | 'DNN1_model_par': DNN1_net.state_dict(), 373 | 'DNN2_model_par': DNN2_net.state_dict(), 374 | } 375 | torch.save(checkpoint,output_folder+'/model_raw_'+ str(epoch) +'.pkl') 376 | 377 | else: 378 | print("epoch %i / %i, loss_tr=%f err_tr=%f" % (epoch, N_epochs, loss_tot,err_tot)) 379 | --------------------------------------------------------------------------------