Fine-Tuning GPT-3.5 on Custom Dataset: A Step-by-Step Guide | Code

Подписаться 32 тыс.

Просмотров 31 тыс.

50% 1

Dive into the world of GPT-3.5 fine-tuning in this tutorial. Learn how to customize AI models for better accuracy and performance, covering dataset formatting, fine-tuning with OpenAI's API, and practical testing of your model.
Code and Dataset: blog.futuresmart.ai/fine-tuni...
If you're curious about the latest in AI technology, I invite you to visit my project, AI Demos, at www.aidemos.com/. It's a rich resource offering a wide array of video demos showcasing the most advanced AI tools. My goal with AI Demos is to educate and illuminate the diverse possibilities of AI.
For even more in-depth exploration, be sure to visit my RU-vid channel at / @aidemos.futuresmart . Here, you'll find a wealth of content that delves into the exciting future of AI and its various applications.
🚀 Top Rated Plus Data Science Freelancer with 8+ years of experience, specializing in NLP and Back-End Development. Founder of FutureSmart AI, helping clients build custom AI NLP applications using cutting-edge models and techniques. Former Lead Data Scientist at Oracle, primarily working on NLP and MLOps.
💡 As a Freelancer on Upwork, I have earned over $100K with a 100% Job Success rate, creating custom NLP solutions using GPT-3, ChatGPT, GPT-4, and Hugging Face Transformers. Expert in building applications involving semantic search, sentence transformers, vector databases, and more.

Опубликовано:

12 ноя 2023

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 51

@applyailikeapro7191 4 дня назад

very helpful. thank you!

@bhanujinaidu 4 месяца назад

Very helpful, thank you so much

@Tiger-Tippu 8 месяцев назад

Thank you so much

@kundaniguru Месяц назад

Amazing content, Thanks

@jigneshpatel7239 8 месяцев назад

Hi Pradip, thanks for such great tutorial. i want to know if i want to create data analyst chatbot with Chat Assistant apis, does finetuned model works with it? if yes does it supports Code interpreter and retrieval capabilities same as Assistant?

@user-nl3fk8fh2v 4 месяца назад

Hi Pradip, thanks for your video. Please, I would like to know how I can fine-tune ChatGPT for threat and anomaly detection in information systems without using an API, directly from its source code or core.

@duetplay4551 8 месяцев назад

hi, teacher! First thanks for helping me find my first job🙂. Nothing, just drop by and say hi

@fadyabdo124 8 месяцев назад

congratulation , can you tell us what is the first job ?

@duetplay4551 8 месяцев назад

modeling risk analyst @@fadyabdo124

@NileshThonte Месяц назад

If I want to use this fine tunned model in other scripts of code which is running on server than how can i do it

@deepaktripathi4094 8 месяцев назад

Hi Pradip, Very good video. Can you please also make a video on how did you generate this data. Did you just went to GPT and used the required prompt to generate it or you used any specific plugin.

@FutureSmartAI 8 месяцев назад

Hi , I asked GPT to first generate 10 topics of support. Then I ask generate 10 examples for each topic. I simply use ChatGPT UI.

@Coralhere 2 месяца назад

Hey can i fine tune gpt to create custom images? With prompt

@fatihornek8295 7 месяцев назад

Hi Pradip, thanks for video. So how can we fine-tune our custom models that are not on huggingface or openai etc. other than LLM models? If we consider a private model that is not public. What would you suggest on this subject?

@FutureSmartAI 7 месяцев назад

you mean finetuning other ML or transformers model? you can check : ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-9he4XKqqzvE.htmlsi=SomWQ2HBgH5wJofx Also you can think of hugginggface models as private also. once you finetune model you can keep it private

@fatihornek8295 7 месяцев назад

@@FutureSmartAI Same logic as this, but I want to fine tune a custom model that has been pre-trained for the customer, not a ready-made public model.

@Kira-fj8el 8 месяцев назад

Hey pradip nichite, thank u so much for your videos. It is very helpful. Can u give a video for was deploying the chatgpt assistant chatbot using gpt4 new updated API. I tried to do that with help of your previous videos. But there is module error. Can u do new video for that??? please

@FutureSmartAI 8 месяцев назад

Yes, soon

@pratikahire2375 5 месяцев назад

Hi require some help, i am not able to fine tune due to openai limitation with azure that our company uses. Is there any alternative way that i can use using hugging face open source libraries to get classification output on the data, my train data is very low

@FutureSmartAI 5 месяцев назад

Not sure what is your problem statement? if its simple classification try finetuning models like BERT first

@aravindraamasamy9453 7 месяцев назад

Hi Pradip, thank you so much for the video it was very helpful . I have a doubt , I am preparing a customised chatbot for a construction company so how to format the input data for this , there are some huge specifications document and how to create input using that . I am looking forward to hearing from you

@FutureSmartAI 7 месяцев назад

Hi formating will be same as conversation. Since its chatbot you will need to create conversation as message array. platform.openai.com/docs/guides/fine-tuning/example-format Each example will be messages array and it should have user message and expected assistant repsonse. {"messages": [{"role": "system", "content": "Marv is a factual chatbot that is also sarcastic."}, {"role": "user", "content": "What's the capital of France?"}, {"role": "assistant", "content": "Paris, as if everyone doesn't know that already."}]} {"messages": [{"role": "system", "content": "Marv is a factual chatbot that is also sarcastic."}, {"role": "user", "content": "Who wrote 'Romeo and Juliet'?"}, {"role": "assistant", "content": "Oh, just some guy named William Shakespeare. Ever heard of him?"}]} {"messages": [{"role": "system", "content": "Marv is a factual chatbot that is also sarcastic."}, {"role": "user", "content": "How far is the Moon from Earth?"}, {"role": "assistant", "content": "Around 384,400 kilometers. Give or take a few, like that really matters."}]}

@aravindraamasamy9453 7 месяцев назад

Thanks for the reply@@FutureSmartAI . I have further doubt in this can I contact you through email and can I get your email id. My doubt is creating training data for numerous documents and each document consists up to 50 pages. Should I manually create question and answers or what are the other approaches. I am looking forward to hearing from you.

@thoughtchit 4 месяца назад

I have a very long form prompt that I pass every time tk generate a valid json out of it. Structure of this prompt also changes little bit on the basis available parameters... If I fine-tune the model then will it start generating that json data without being too explicit in prompt?

@FutureSmartAI 4 месяца назад

Are you using JSON mode? platform.openai.com/docs/guides/text-generation/json-mode

@thoughtchit 4 месяца назад

Yep I do, my main concern is that even when i use json mode.... I have to pass those large sets of instructions as a prompt+ defining structure of json as well... That's what I want to get rid of... As I want it to generate the same kind of json with least information that it should be trained on it's past response, that's what I think?

@sagarikaghosh6532 Месяц назад

How to buy the API key because for fine tune we need a paid version?

@kajalkumari-il3ok 2 месяца назад

Please, I would like to know how I can save fine-tune gpt-3.5-turbo model locally, and load that model to predict with new input.during prediction no need to use client open api key. is this possible ?

@FutureSmartAI 2 месяца назад

No we cant store fientune model locally

@bimaljohn4583 5 месяцев назад

Hi Pradip, Thats a great tutorial. I'm curious about whether ChatGPT will be trained on the data we provide. Some of the data might be confidential, so if it's used for training, it could potentially be harmful. Additionally, is it possible for us to obtain a blend of results? For instance, if we're aiming to identify the top 5 schools in our area and our database which we used to finetune only contains information on 3 schools, can we supplement it with additional data from the OpenAI Database?

@FutureSmartAI 5 месяцев назад

Hi first of all you need to decide whether you should finetune or use RAG. If your model requires access to more knowledge you should use RAG

@harddikbafna1792 4 месяца назад

Hey, i got this error: BadRequestError: Error code: 400 - {'error': {'message': 'Invalid file format. Line 1, key "messages": At least one message must be from the assistant', 'type': 'invalid_request_error', 'param': None, 'code': None}} and cant seem to find any solution online. i followed step by step as what you did but got this while executing 12:17 . can u explain me what should i do now?

@ankes.1338 4 месяца назад

I had the same error. I figured, it is because in the code examples provided in the link, it says "system" instead of " assistant". For me it worked when I exchanged it in the code.

@ansharora5698 8 месяцев назад

Can you do a video on Chatbot using Google Bard API, its free to use and I am currently doing it right now but I can't complete it. It does not retain messages.

@FutureSmartAI 7 месяцев назад

Sure let me explore it

@asiddiqi123 5 месяцев назад

Imagine I want to create a LinkedIn post generator that follows certain style.. which option is best: - fine tuned LLM - customGPT/assistant API - system prompt?

@FutureSmartAI 5 месяцев назад

You should start with few shot examples prompt. And if it still doesnt give satisfactory results you can finetune model on more examples

@TechWithAbee 5 месяцев назад

liked ✅

@FutureSmartAI 5 месяцев назад

Thanks 💯

@rameshh3821 3 месяца назад

I'm getting NotFoundError: Error code: 404 - {'error': {'code': '404', 'message': 'Resource not found'}} while trying to run this code: response = client.fine_tuning.jobs.create( training_file=training_file_id, validation_file=validation_file_id, model="gpt-35-turbo" ) I'm using Azure OpenAI