![]() ![]() ? Transformers provides access to thousands of pretrained models for a wide range of tasks. It reduces computation costs, your carbon footprint, and allows you to use state-of-the-art models without having to train one from scratch. Using GPT-4 to build the training set would probably be even better, but the usage is too restricted right now to quickly do a job of that size.Īnyway, I hope this helps someone out there get a raise.There are significant benefits to using a pretrained model. ![]() have it rephrase questions in different ways to generate the same answer) to do even better. I'm sure you can keep iterating on this concept (i.e. The training is pretty expensive (relative to the monthly limits), but spamming ChatGPT to build the prompts was dirt cheap. I have to wait until next month to retrain on the new data, but I'm very optimistic. It took a little trivial code to adjust some of the responses to the json I actually wanted, de-dup the results, but the 8000 new training prompts look FANTASTIC. Create the questions and answers in the reverse order they appear in the original text.\n." Then, out of that hierarchical text file, I built a jsonl file like \nCreate as many question/answer pairs relevant to this prompt/completion pair as you can. Luckily this turned out to be mostly reasonably sized chunks for feeding to gpt-3.5-turbo. I wrote some code using iText7 to extract text from a 1500-page PDF, and then infer a hierarchy from it. Empty prompts and empty completions were all over the place, it wasn't going to work. It looked OK at first, but after 6 hours of processing and a lot of hand-holding for invalid json, I realized the results were generally terrible. Generate prompt/completion pairs for training based on this information. ![]() I saved a 1500-page PDF to text, and fed it in roughly 4000-character chunks to ChatGPT, advancing roughly 2000 characters at a time, and fed those chunks to ChatGPT with something like "You're building GPT-3 training data based on chunks of a PDF. I saw this and thought I'd share my experiences working on something like that. Please don't claim openai api fine_tunes.create -t "model_prepared.jsonl" -m "davinci" will create a model based on text-davinci-003, it is not true, it uses base davinci. This is not how GPT works and is explained by GPT blog posts and support posts: Is there any way to fine-tune Davinci up to the point where it can model some of the things Instruct does? I don't need full capabilities, but if I can make it narrowed down to my use case it would be ideal.īy the way there is a common misconception that fine-tuning a GPT-3 model on a base (davinci, ada, babbage, etc.) will train it on the latest, eg: text-davinci-003. I've attempted to increase epochs from 4 to 10 but the quality is really nowhere near as good). (Please let me know if there's a third option. or B: Find a way to fine-tune Davinci into something capable of simpler InstructGPT behaviours.A: Find a way to harvest 10x more data (I don't see an easy option here).I have a data set (n~20) which I'd like to train the model with more but there is no way to fine-tune these InstructGPT models, only base GPT models. I have a few-shot GPT-3 text-davinci-003 prompt that produces "pretty good" results, but I quickly run out of tokens per request for interesting use cases.
0 Comments
Leave a Reply. |