sachalapins/DAS-N2N — reverse-engineered prompt

I want a working version of this DAS N2N project from the paper, so I can actually run it end to end instead of just having the source sitting there.

Please get the code into a usable state, make sure the included notebook runs and reproduces the main paper style results and figures, and make it clear how to use the model for denoising Distributed Acoustic Sensing signals when there is no clean target data. If the current code is old or a bit broken, fix whatever is needed so the basic training, inference, and results workflow works smoothly.

Also add simple guidance for someone who is not deeply technical, like where to put data, how to run the model, and what outputs to expect. If there are sample files in the repo, use those to prove it works. If you need extra context, check the paper and the newer implementation online, but keep this project faithful to what this repository is supposed to do.