Use instructions to create a machine learning python notebook to train and call GPT-J model

I'm tired of negotiating infrastructure and TPUs and GPU and CUDA errors. I want code that works. The deliverable for this project may be either of the following: A Google Colab notebook (shared with me so I can make a copy) A Paperspace Gradient python notebook (shared with me so I can make a copy) A .ipynb notebook I can upload to one of those platforms. (Specify which platform. It has to work there.) I have Google Colab Pro and/or am fine using paid instances on Paperspace. Your solution does not have to run for free. It just has to run. Either as the first cell in the notebook or as a separate document, provide step-by-step instructions for setting up the environment. What do I need to select for the connection? Etc. I will also accept ipynb code and instructions for some other paid service (AWS, Azure, Google Cloud, etc) but the instructions need to start from square one with the platform: What clusters do I need to set up? Where do I put this python code? How do I connect it to the training data and python code? Etc. The notebook should be the code to create an instance of the GPT-J transformer-based text generation model, located here: [login to view URL] Program functionality: Download/install GPT-J model and dependencies Prepare the data (attached csv) for training Train the model on the data Save the tuned model Call the tuned model Using any sample prompt The creators of the model also have a sample Collab notebook showing how to install and run the pre-trained model here: [login to view URL] Your notebook should then follow the fine-tuning instructions here: [login to view URL] (The github training example above uses Google Cloud TPUs. You can use any of the above platforms.) Your instructions on how to perform the steps like creating the proper project, installing dependencies should be more detailed than the link above. Document every step so it is very clear.) In text cells of the notebook, include detailed instructions on where to upload the training data. Then your code should prepare the data for training. Your code should then execute fine-tuning on the attached data set for 2 epochs. Acceptance criteria: The deliverable is not the fine-tuned model. The deliverable is the notebook and any instructions, so that I can do this repeatably myself.. I will test your code by running the notebook code on the platform you indicate, using the instructions you give me. If I am successfully able to execute the steps above: Download/install GPT-J model and dependencies Prepare the data (attached csv) for training Train the model on the data Save the tuned model Call the tuned model on a sample prompt. Then the project is a success. I will not be evaluating the project based on the quality of the response generated in step 5, only that I can successfully run the code. Note: The attached file is utf-8 format, but I think it may have been saved with a Byte Order Mark (BOM) for easy reading into Windows apps. Opening it with encoding='utf-8-sig' will ensure python can read it properly. Additional resources: You may find the notebook linked here helpful as well. [login to view URL]