swagshaw / asc-cl

Official Pytorch Implementation for Continual Learning For On-Device Environmental Sound Classification

Hi there, I'm Yang! 👋

Welcome to my GitHub. I'm Yang Xiao or you can call me Austin. I research speech and audio for the Fortemedia. Before that, I graduated with a Master of Artificial Intelligence at Nanyang Technological University. My research is building an efficient, robust, data-centric machine learning system. At this moment, I mainly work in data-centric speech systems that are continual, real-time, and efficient.

• 📫 How to reach me: [Email], [LinkedIn], [Homepage].

asc-cl's Issues

How to train on a custom dataset

I am performing audio classification on a custom dataset. COuld you please provide me some information on training the model on a custom dataset.

Look For Help

Dear Author:
When I reproduce your code, this error occurs. The online search does not solve this error. Can you help me to check it. Thank You !

Error during the "Apply after_task" during the training step > 0 for "shift", "noise", and "noisytune"

Please let me know what to be done here:

I am using the ESC50 dataset and the train.sh file for reproducing the results that you have published. I have not changed any configurations except for batch size.

2023-07-07 19:31:50,057 - INFO - Epoch 7 | Evaluation Accuracy: 1.0
 2023-07-07 19:31:50,057 - INFO - EarlyStopping counter: 7 out of 7.
 2023-07-07 19:31:50,058 - INFO - [2-4] Update the information for the current task
 2023-07-07 19:31:50,058 - INFO - Apply after_task
 2023-07-07 19:31:50,058 - INFO - Update memory over 4 classes by uncertainty
 2023-07-07 19:31:50,058 - INFO - Compute uncertainty by noise!
 2023-07-07 19:32:00,750 - WARNING - Fill the unused slots by breaking the equilibrium.
Traceback (most recent call last):
  File "main.py", line 196, in <module>
    method.after_task(cur_iter)
  File "/home/shikkalven/SELMA/ContinualLearning/ASC_CL_spkid/ASC-CL/methods/base.py", line 108, in after_task
    self.update_memory(cur_iter)
  File "/home/shikkalven/SELMA/ContinualLearning/ASC_CL_spkid/ASC-CL/methods/base.py", line 140, in update_memory
    num_class=num_class,
  File "/home/shikkalven/SELMA/ContinualLearning/ASC_CL_spkid/ASC-CL/methods/base.py", line 379, in uncertainty_sampling
    .sample(n=num_rest_slots)
  File "/home/shikkalven/jupyter-conda-base-environment/conda/envs/CL_asc/lib/python3.7/site-packages/pandas/core/generic.py", line 5365, in sample
    locs = rs.choice(axis_length, size=n, replace=replace, p=weights)
  File "mtrand.pyx", line 903, in numpy.random.mtrand.RandomState.choice
ValueError: a must be greater than 0 unless no samples are taken