search

Le-Xiaohuai-speech / DPCRN_DNS3

Implementation of paper "DPCRN: Dual-Path Convolution Recurrent Network for Single Channel Speech Enhancement"
186 stars 41 forks source link

About Test #23

Open panhu opened 2 years ago

panhu commented 2 years ago

Hi, when i want to test the model: "conv_1 = Conv2D(32, (2,5),(1,2),name = name+'_conv_1',padding = [[0,0],[1,0],[0,2],[0,0]])(input_complex_spec) File "/root/anaconda3/lib/python3.9/site-packages/keras/utils/traceback_utils.py", line 67, in error_handler raise e.with_traceback(filtered_tb) from None File "/root/anaconda3/lib/python3.9/site-packages/tensorflow/python/training/tracking/data_structures.py", line 677, in hash raise TypeError("unhashable type: 'ListWrapper'") TypeError: unhashable type: 'ListWrapper' " Can you help me?

Le-Xiaohuai-speech commented 2 years ago

change the padding set to a tuple like((0,0),(0,1),(0,2)(0,0))?

---Original--- From: @.> Date: Wed, Jun 15, 2022 15:44 PM To: @.>; Cc: @.***>; Subject: [Le-Xiaohuai-speech/DPCRN_DNS3] About Test (Issue #23)

Hi, when i want to test the model: "conv_1 = Conv2D(32, (2,5),(1,2),name = name+'_conv_1',padding = [[0,0],[1,0],[0,2],[0,0]])(input_complex_spec) File "/root/anaconda3/lib/python3.9/site-packages/keras/utils/traceback_utils.py", line 67, in error_handler raise e.with_traceback(filtered_tb) from None File "/root/anaconda3/lib/python3.9/site-packages/tensorflow/python/training/tracking/data_structures.py", line 677, in hash raise TypeError("unhashable type: 'ListWrapper'") TypeError: unhashable type: 'ListWrapper' " Can you help me?

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.***>

panhu commented 2 years ago

i have changed the padding like (((0,0),(0,1),(0,2),(0,0))).but got the same error: conv_1 = Conv2D(32, (2,5),(1,2),name = name+'_conv_1',padding = [[0,0],[0,1],[0,2],[0,0]])(input_complex_spec) File "/root/anaconda3/lib/python3.9/site-packages/keras/utils/traceback_utils.py", line 67, in error_handler raise e.with_traceback(filtered_tb) from None File "/root/anaconda3/lib/python3.9/site-packages/tensorflow/python/training/tracking/data_structures.py", line 677, in hash raise TypeError("unhashable type: 'ListWrapper'") TypeError: unhashable type: 'ListWrapper'

Le-Xiaohuai-speech commented 2 years ago

please check the versions of tensorflow and keras

panhu commented 2 years ago

Tanks,beacuase of versions of tensorflow! Now i want to train the data of myself,but i don't konw how to organize my data.i have many clean and noise data.Using the RIRs,but when i want to train,happend some errors. like this: File "/home/DPCRN_DNS3/data_loader.py", line 191, in generator clean_s,noise_s,noisys, = mk_mixture(clean_s,noise_s,SNR,eps = 1e-8) File "/home/DPCRN_DNS3/data_loader.py", line 45, in mk_mixture mix = norm_sig2 + alpha*norm_sig1 ValueError: operands could not be broadcast together with shapes (0,) (39908,)

panhu commented 2 years ago

this is my data structure: datasets: clean: clnsp_1.wav .... noise: noise1_SNRdb_10.0.wav ... rirs: pointsource_noises: ... real_rirs_isotropic_noises: ... simulated_rirs: ...

Le-Xiaohuai-speech commented 2 years ago

the audio may not be loaded correctly

---Original--- From: @.> Date: Tue, Jun 28, 2022 16:26 PM To: @.>; Cc: @.**@.>; Subject: Re: [Le-Xiaohuai-speech/DPCRN_DNS3] About Test (Issue #23)

Tanks,beacuase of versions of tensorflow! Now i want to train the data of myself,but i don't konw how to organize my data.i have many clean and noise data.Using the RIRs,but when i want to train,happend some errors. like this: File "/home/DPCRN_DNS3/data_loader.py", line 191, in generator clean_s,noise_s,noisys, = mk_mixture(clean_s,noise_s,SNR,eps = 1e-8) File "/home/DPCRN_DNS3/data_loader.py", line 45, in mk_mixture mix = norm_sig2 + alpha*norm_sig1 ValueError: operands could not be broadcast together with shapes (0,) (39908,)

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>

VaneYb commented 2 years ago

Hi, when i want to test the model: "conv_1 = Conv2D(32, (2,5),(1,2),name = name+'_conv_1',padding = [[0,0],[1,0],[0,2],[0,0]])(input_complex_spec) File "/root/anaconda3/lib/python3.9/site-packages/keras/utils/traceback_utils.py", line 67, in error_handler raise e.with_traceback(filtered_tb) from None File "/root/anaconda3/lib/python3.9/site-packages/tensorflow/python/training/tracking/data_structures.py", line 677, in hash raise TypeError("unhashable type: 'ListWrapper'") TypeError: unhashable type: 'ListWrapper' " Can you help me?

It may because the parameter 'padding' of Conv2D in keras.layer must be 'same' or 'valid', not the explicit format as in here. I tried conv_1 = Conv2D(32, (2,5),(1,2),name = name+'_conv_1')(tf.pad(input_complex_spec, [[0,0],[1,0],[0,2],[0,0]])) instead of conv_1 = Conv2D(32, (2,5),(1,2),name = name+'_conv_1',padding = [[0,0],[1,0],[0,2],[0,0]])(input_complex_spec) for each conv_x, and run successfully.

bigcash commented 2 years ago

Tanks,beacuase of versions of tensorflow! Now i want to train the data of myself,but i don't konw how to organize my data.i have many clean and noise data.Using the RIRs,but when i want to train,happend some errors. like this: File "/home/DPCRN_DNS3/data_loader.py", line 191, in generator clean_s,noise_s,noisys, = mk_mixture(clean_s,noise_s,SNR,eps = 1e-8) File "/home/DPCRN_DNS3/data_loader.py", line 45, in mk_mixture mix = norm_sig2 + alpha*norm_sig1 ValueError: operands could not be broadcast together with shapes (0,) (39908,)

which version of tensorflow is correct? i tryed 2.8 and 2.9, raise this error at all

bigcash commented 2 years ago

Tanks,beacuase of versions of tensorflow! Now i want to train the data of myself,but i don't konw how to organize my data.i have many clean and noise data.Using the RIRs,but when i want to train,happend some errors. like this: File "/home/DPCRN_DNS3/data_loader.py", line 191, in generator clean_s,noise_s,noisys, = mk_mixture(clean_s,noise_s,SNR,eps = 1e-8) File "/home/DPCRN_DNS3/data_loader.py", line 45, in mk_mixture mix = norm_sig2 + alpha*norm_sig1 ValueError: operands could not be broadcast together with shapes (0,) (39908,)

which version of tensorflow is correct? i tryed 2.8 and 2.9, raise this error at all

i try tensorflow-gpu==1.15, and success, hope can help others.(tensorflow==2.4.4, failed with this error too)

Le-Xiaohuai-speech commented 2 years ago

it looks like a audio sample was not loaded correctly

------------------ 原始邮件 ------------------ 发件人: "Le-Xiaohuai-speech/DPCRN_DNS3" @.>; 发送时间: 2022年9月20日(星期二) 下午4:33 @.>; @.**@.>; 主题: Re: [Le-Xiaohuai-speech/DPCRN_DNS3] About Test (Issue #23)

Tanks,beacuase of versions of tensorflow! Now i want to train the data of myself,but i don't konw how to organize my data.i have many clean and noise data.Using the RIRs,but when i want to train,happend some errors. like this: File "/home/DPCRN_DNS3/data_loader.py", line 191, in generator clean_s,noise_s,noisys, = mk_mixture(clean_s,noise_s,SNR,eps = 1e-8) File "/home/DPCRN_DNS3/data_loader.py", line 45, in mk_mixture mix = norm_sig2 + alpha*norm_sig1 ValueError: operands could not be broadcast together with shapes (0,) (39908,)

which version of tensorflow is correct? i tryed 2.8 and 2.9, raise this error at all

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>

Le-Xiaohuai-speech commented 2 years ago

clean_1.wav ... noise_1.wav...

---Original--- From: @.> Date: Thu, Jun 30, 2022 09:36 AM To: @.>; Cc: @.**@.>; Subject: Re: [Le-Xiaohuai-speech/DPCRN_DNS3] About Test (Issue #23)

Tanks,Can i know your naming rules of clean and noise files?

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>

wangsheng3 commented 1 year ago

请检查 TensorFlow 和 Keras 的版本

Can you release a complete version of these packages?