adrienpetralia / appliancedetectionbenchmark Goto Github PK

Good afternoon,

First of all, I would like to extend my gratitude to you for making available your artifact available for the community. I thoroughly enjoyed reading your paper, and really appreciate the additional effort you have taken to make your codebase available.

I am currently working to replicate your experiments, and I wanted to send a quick update with some observations, and also some questions for you.

• I encountered some issues, and some reading online suggests that PyTorch requires a 3.9.x version number and is not compatible with later versions. The README.md suggests Python >=3.7, and I wonder if this restriction needs to be tightened up a little bit. I primarily work with R rather than Python, so I am working with relatively little prior knowledge here.
• I had to update requirements.txt to change pytorch==1.8.1 to torch==1.8.1.

I'm now currently seeing the following error when I try to run

/usr/local/bin/python3.7 /Users/myusername/ApplianceDetectionBenchmark-main/src/REFIT_Benchmark.py 
Traceback (most recent call last):
  File "/Users/myusername/ApplianceDetectionBenchmark-main/src/REFIT_Benchmark.py", line 20, in <module>
    from Utils.Models.ConvNet import ConvNet
  File "/Users/myusername/ApplianceDetectionBenchmark-main/src/Utils/Models/ConvNet.py", line 4, in <module>
    from Utils.Layers.ConvLayers import Conv1dSame
ModuleNotFoundError: No module named 'Utils.Layers'

Process finished with exit code 1

• I am unable to locate Conv1dSame which is imported from Utils.Layers.ConvLayers in ConvNet.py:L4. Are there some files which are missing from the repository, perhaps?

I'm seeing other things which have not yet become an issue at runtime, but I wanted to make you available of these too.

• I am unable to resolve the reference to classif_trainer mentioned in UKDALE_Benchmark.py
• In CER_Benchmark.py, should path_data = os.getcwd()+'/Datasets/CER_Electricity/' instead read path_data = os.getcwd()+'/data/CER/'
• Two further filenames in _utils_preprocessing_.py appear to reference a /Datasets location.

I really appreicate your support and hope you may be able to help me overcome a couple of these hurdles to reproduce your work.

Best wishes,
Matt

Hi,

In PyTorch before trunk/89695, torch.jit.annotations.parse_type_line can cause arbitrary code execution because eval is used unsafely. The fix for this issue is available in version 1.13.1. There is a release checker in pytorch/pytorch#89855.

The torch version in the requirement.txt is torch==1.8.1. We should ensure the benchmark works with torch==1.13.1 and update the requirement.txt.

Paul