dding neviroment

dataset_loader.py

@@ -1,4 +1,3 @@
-from typing import Tuple, Dict
 import ast
 
 import tensorflow as tf
@@ -7,7 +6,6 @@ import os
 import numpy as np
 from scipy.fft import fft, ifft
 import soundfile as sf
-import librosa
 import math
 import pandas as pd
 import scipy as sp
@@ -32,10 +30,7 @@ def __noise_sample_generator(info_file,fs, length_seq, split):
             else:
                 num=np.random.randint(0,len(segments))
                 loaded_data, Fs=sf.read(os.path.join(head,load_data_split["recording"].loc[i],segments[num]))
-
+            assert(fs==Fs, "wrong sampling rate")
-            if fs!=Fs:
-                print("wrong fs, resampling...")
-                data=librosa.resample(loaded_data, Fs, fs)
 
             yield __extend_sample_by_repeating(loaded_data,fs,length_seq)
 
@@ -385,11 +380,7 @@ def generator_train(path_music, path_noises,split, fs=44100, seg_len_s=5, extend
         random.shuffle(train_samples)
         for file in train_samples:  
             data, samplerate = sf.read(file)
-            if samplerate!=fs: 
+            assert(samplerate==fs, "wrong sampling rate")
-                print("!!!!WRONG SAMPLE RATe!!!")
-                data=np.transpose(data)
-                data=librosa.resample(data, samplerate, 44100)
-                data=np.transpose(data)
             data_clean=data
             #Stereo to mono
             if len(data.shape)>1 :
@@ -456,7 +447,7 @@ def generator_train(path_music, path_noises,split, fs=44100, seg_len_s=5, extend
                 summed=summed.astype('float32')
                 yield tf.convert_to_tensor(summed), tf.convert_to_tensor(segment)
         
-def load_data(buffer_size, path_music_train, path_music_val,  path_noises,  fs=44100, seg_len_s=5,  extend=True, stereo=False) -> Tuple[tf.data.Dataset, tf.data.Dataset]:
+def load_data(buffer_size, path_music_train, path_music_val,  path_noises,  fs=44100, seg_len_s=5,  extend=True, stereo=False) :
     print("Generating train dataset")
     trainshape=int(fs*seg_len_s)
 
@@ -470,14 +461,14 @@ def load_data(buffer_size, path_music_train, path_music_val,  path_noises,  fs=4
 
     return dataset_train.shuffle(buffer_size), dataset_val
 
-def load_data_test(buffer_size, path_pianos_test,   path_noises,  **kwargs) -> Tuple[tf.data.Dataset]:
+def load_data_test(buffer_size, path_pianos_test,   path_noises,  **kwargs):
     print("Generating test dataset")
     segments_noisy, segments_clean=generate_test_data(path_pianos_test, path_noises, extend=True, **kwargs)
     dataset_test=tf.data.Dataset.from_tensor_slices((segments_noisy, segments_clean))
     #dataset_test=tf.data.Dataset.from_tensor_slices((segments_noisy[1:3], segments_clean[1:3]))
     #train_dataset = train.cache().shuffle(buffer_size).take(info.splits["train"].num_examples)
     return dataset_test
-def load_data_formal( path_pianos_test,   path_noises,  **kwargs) -> Tuple[tf.data.Dataset]:
+def load_data_formal( path_pianos_test,   path_noises,  **kwargs) :
     print("Generating test dataset")
     segments_noisy, segments_clean=generate_paired_data_test_formal(path_pianos_test, path_noises, extend=True, **kwargs)
     print("segments::")
@@ -487,7 +478,7 @@ def load_data_formal( path_pianos_test,   path_noises,  **kwargs) -> Tuple[tf.da
     #train_dataset = train.cache().shuffle(buffer_size).take(info.splits["train"].num_examples)
     return dataset_test
 
-def load_real_test_recordings(buffer_size, path_recordings,   **kwargs) -> Tuple[tf.data.Dataset]:
+def load_real_test_recordings(buffer_size, path_recordings,   **kwargs):
     print("Generating real test dataset")
         
     segments_noisy=generate_real_recordings_data(path_recordings, **kwargs)

environment.yml

@@ -0,0 +1,81 @@
+name: historical_denoiser
+channels:
+  - anaconda
+  - pytorch
+  - rapidsai
+  - defaults
+  - conda-forge
+  - nvidia
+  - gurobi
+  - IBMDecisionOptimization
+  - mosek
+dependencies:
+  - _libgcc_mutex=0.1=main
+  - _openmp_mutex=4.5=1_gnu
+  - ca-certificates=2020.10.14=0
+  - cudatoolkit=10.1.243=h6bb024c_0
+  - cudnn=7.6.5=cuda10.1_0
+  - ld_impl_linux-64=2.35.1=h7274673_9
+  - libffi=3.3=he6710b0_2
+  - libgcc-ng=9.3.0=h5101ec6_17
+  - libgomp=9.3.0=h5101ec6_17
+  - libstdcxx-ng=9.3.0=hd4cf53a_17
+  - ncurses=6.2=he6710b0_1
+  - openssl=1.1.1k=h27cfd23_0
+  - pip=21.0.1=py37h06a4308_0
+  - python=3.7.11=h12debd9_0
+  - readline=8.1=h27cfd23_0
+  - setuptools=52.0.0=py37h06a4308_0
+  - sqlite=3.36.0=hc218d9a_0
+  - tk=8.6.10=hbc83047_0
+  - wheel=0.37.0=pyhd3eb1b0_0
+  - xz=5.2.5=h7b6447c_0
+  - zlib=1.2.11=h7b6447c_3
+  - pip:
+    - absl-py==0.13.0
+    - astunparse==1.6.3
+    - cachetools==4.2.2
+    - certifi==2021.5.30
+    - cffi==1.14.6
+    - charset-normalizer==2.0.4
+    - gast==0.3.3
+    - google-auth==1.35.0
+    - google-auth-oauthlib==0.4.5
+    - google-pasta==0.2.0
+    - grpcio==1.39.0
+    - h5py==2.10.0
+    - hydra-core==0.11.3
+    - idna==3.2
+    - importlib-metadata==4.8.1
+    - keras-preprocessing==1.1.2
+    - markdown==3.3.4
+    - numpy==1.18.5
+    - oauthlib==3.1.1
+    - omegaconf==1.4.1
+    - opt-einsum==3.3.0
+    - pandas==1.3.2
+    - protobuf==3.17.3
+    - pyasn1==0.4.8
+    - pyasn1-modules==0.2.8
+    - pycparser==2.20
+    - python-dateutil==2.8.2
+    - pytz==2021.1
+    - pyyaml==5.4.1
+    - requests==2.26.0
+    - requests-oauthlib==1.3.0
+    - rsa==4.7.2
+    - scipy==1.4.1
+    - six==1.16.0
+    - soundfile==0.10.3.post1
+    - tensorboard==2.5.0
+    - tensorboard-data-server==0.6.1
+    - tensorboard-plugin-wit==1.8.0
+    - tensorflow==2.3.0
+    - tensorflow-estimator==2.3.0
+    - termcolor==1.1.0
+    - tqdm==4.62.2
+    - typing-extensions==3.10.0.2
+    - urllib3==1.26.6
+    - werkzeug==2.0.1
+    - wrapt==1.12.1
+    - zipp==3.5.0

inference.py

@@ -10,7 +10,7 @@ def run(args):
     import soundfile as sf
     import numpy as np
     from tqdm import tqdm
-    import librosa
+    import scipy.signal
     
     path_experiment=str(args.path_experiment)
 
@@ -47,16 +47,17 @@ def run(args):
 
     audio=str(args.inference.audio)
     data, samplerate = sf.read(audio)
-    if samplerate!=44100: 
+    print(data.dtype)
-        print("Resampling")
-        data=np.transpose(data)
-        data=librosa.resample(data, samplerate, 44100)
-        data=np.transpose(data)
-
     #Stereo to mono
     if len(data.shape)>1:
         data=np.mean(data,axis=1)
     
+    if samplerate!=44100: 
+        print("Resampling")
+   
+        data=scipy.signal.resample(data, int((44100  / samplerate )*len(data))+1)  
+ 
+    
     
     segment_size=44101*20  #20s segments
 

inference.sh

@@ -1,8 +1,8 @@
 #!/bin/bash
 
-module load anaconda
+#module load anaconda
 
-source activate /scratch/work/molinee2/conda_envs/unet_env
+#source activate /scratch/work/molinee2/conda_envs/unet_env
 
 
 python inference.py path_experiment="experiments/trained_model" inference.audio=$1

requirements.txt

@@ -0,0 +1,47 @@
+absl-py==0.13.0
+astunparse==1.6.3
+cachetools==4.2.2
+certifi==2021.5.30
+cffi==1.14.6
+charset-normalizer==2.0.4
+gast==0.3.3
+google-auth==1.35.0
+google-auth-oauthlib==0.4.5
+google-pasta==0.2.0
+grpcio==1.39.0
+h5py==2.10.0
+hydra-core==0.11.3
+idna==3.2
+importlib-metadata==4.8.1
+Keras-Preprocessing==1.1.2
+Markdown==3.3.4
+numpy==1.18.5
+oauthlib==3.1.1
+omegaconf==1.4.1
+opt-einsum==3.3.0
+pandas==1.3.2
+protobuf==3.17.3
+pyasn1==0.4.8
+pyasn1-modules==0.2.8
+pycparser==2.20
+python-dateutil==2.8.2
+pytz==2021.1
+PyYAML==5.4.1
+requests==2.26.0
+requests-oauthlib==1.3.0
+rsa==4.7.2
+scipy==1.4.1
+six==1.16.0
+SoundFile==0.10.3.post1
+tensorboard==2.5.0
+tensorboard-data-server==0.6.1
+tensorboard-plugin-wit==1.8.0
+tensorflow==2.3.0
+tensorflow-estimator==2.3.0
+termcolor==1.1.0
+tqdm==4.62.2
+typing-extensions==3.10.0.2
+urllib3==1.26.6
+Werkzeug==2.0.1
+wrapt==1.12.1
+zipp==3.5.0

train.py

@@ -7,7 +7,6 @@ logger = logging.getLogger(__name__)
 def run(args):
     import unet
     import tensorflow as tf
-    import tensorflow_addons as tfa
     import  dataset_loader
     from tensorflow.keras.optimizers import Adam
     import soundfile as sf

train.sh

@@ -1,9 +1,9 @@
 #!/bin/bash
 
-module load anaconda
+#module load anaconda
 #module load matlab/r2020b
 #
-source activate /scratch/work/molinee2/conda_envs/unet_env
+#source activate /scratch/work/molinee2/conda_envs/unet_env
 
 n=1
 PATH_EXPERIMENT=/scratch/work/molinee2/unet_dir/unet_denoising_github/experiments/${n}

trainer.py

@@ -1,15 +1,10 @@
 
 import os
 import numpy as np
-import cv2
-import librosa
-import imageio
 import tensorflow as tf
 import soundfile as sf
-import subprocess
 from tqdm import tqdm
 import pandas as pd                                    
-from scipy.io import loadmat
 
 class Trainer():
     def __init__(self, model,  optimizer,loss,  strategy, path_experiment,  args):