Today was a breakthrough day! We were able to take our first step in optimizing our machine model. And we have begun to see promising results when training. As none of us have very minimal experience when it comes to deep learning, seeing the progress we have made with the model is fascinating. However, we still have to make modifications to the model through parameter adjustments, adding & removing relevant features and so on.
The major issue we have was difficulty setting a cloud VM to run our model on. But just today, my teammate was able to gain access to OSU’s cloud that uses the infamous Nvidia tesla GPUs. Machine learning is an expensive operation and having capable hardware is a must to save time. Personally, being the proud owner of a GTX 1070 (sarcasm), I am unable to do expensive computational tasks on my local device.
Moving forward, we plan on running our model on OSU’s cloud when training and testing with datasets. We would also ideally like to host our project as well, but since it is not one of the hard requirements, it is not a priority at the moment. I am looking forward to see how just how much faster our model runs with high end hardware and to learn more about deep learning in general!
My team and I continue to chip away at this project. Although, we have had some difficulties along the way, today was a breakthrough daysimply due to the fact that we overcame many of these issues. Furthermore, we have been working fairly well as a team and help each other resolve with bugs/problems we come across.
Although we have not gotten to actually constructing our machine learning model yet, we have completed the portion where we prepare the input that will eventually server as the ML input. The biggest hurdle we faced was working with continuous integration on Git. We set up CI to help us in the long run, so that we can spot bugs before we make permanent branch changes but it proved to be time consuming in debugging the CI itself. But we solved the issue today and began integrating our branches.
Overall, we have been good progress. We plan to complete constructing a base machine learning model by end of this sprint, after which we can begin training using our converted inputs. Machine learning is proving to be exciting and I am hoping to learn more on the topic going into the next sprint!
This week marks my 4th week in this Capstone course and so far things are going pretty good. My team, with whom I am creating this project with, all seem to be excited about our project: a music genre classifier that using a machine trained model!
My team and I are uninitiated when it comes to machine learning so we had some trouble setting up a functional project plan. However, it turns out that is nothing a little research cant solve! I am excited to get a Minimum Viable Product up and running to see the fruits of our labor. We have decided to go with pytorch for the ML framework and that seems to have been the right call in the short time that I have used it. This is primarily due to the fact that we will be utilizing python as our primary language in this project So seeing as though pytorch was a ML library built specifically for python, we abandoned our initial decision to use tensor flow for the more friendly framework.
Being as this will be my final class in this CS program, I am excited to add machine learning to my list of versed technologies. Although there is ways to go before we can began seeing results, so until then, our work continues!
Computers have been something I have always been fascinated by. The possibilities are endless when you have a functional PC and a set of peripherals. Even before starting the Computer Science program at Oregon State University, I had basic knowledge on how to operate a computer but no more that that.
The interest in wanting to learn more was always there but the issue was the lack of knowledge and experience. As I began taking classes at OSU (especially the first introductory classes), I was amazed at the new world of opportunities that have been opened up to me through the gateway that was programming. It was both beautiful and a bit frightening at the same time. The former because I had so many ideas of things I wanted to create up but could not up until that point and the latter because of being afraid I would fail in the process.
Now, having progressed through the program to currently taking my final class, I can say I have greater confidence in myself and a deeper understanding regarding programming and development as well as a hunger to discover what lies beyond this program.