Rush project: Pytorch Machine Learning audio classification

I have a pyTorch CNN that accepts raw audio data (1D vector from wav file) and classifies the genre of music (from 6 different genres) with about %60 accuracy I would like three major modifications 1) For this, I would like the CNN to accept a 3d input, where each dimension is a different genre. Two of them have the same genre, and the third is different. 2) each input has a different length, e.g. number of features. (I think this means the batch size would have to be one) The data has been prepared as a pickle file. Essential there is a list, each entry is one 3 element list representing one 3d input (of course each of those 3 elements are the same feature size). A separate list has the classification, [0,0,1] or [2,2,4], which represents two genres of class zero and 1 of class 1, etc 3) This is the third major adjustment. Working this these vectors. I could change them to one-hot, but it would have to be a 12 element vector- 1 position if there are 2 of that type ofgenre and one position if there is 1 of that type of genre times 6 types of genres = 12 slots. If that is helpful, I will do it It is an interesting project and and I only have about 24 hours to get it done. I am looking at about $100, given the starter code and data that I have. I think a very knowledgeable person could do it quickly, and given the short time frame, an expert is required. I will leave a very good review if you can do itl