💻 DEVELOPERS, I'M HIRING! Apply here: forms.gle/Y8yNSpCapcDPuTXt6 If you want a personalized AI strategy to future-proof yourself and your business, join my community: www.skool.com/new-society
Ive been a wordpress developer for the past 20 years and then became the lead search engine optimization manager for an agency, I see a lot of things in AI that are replacing the need for me, I’d like to learn a new skill set in AI that makes me irreplaceable. Maybe I can run an ai automation agency
why are some here trashing david ondrej? he is imparting knowledge in an easy to understand way for peeps that do not know. i wrote my first neural net from scratch in 1993 and i have been an ML practitioner since then. i can tell u that info back then was hard to come by. be grateful that u have easy access to it. if u dont like it better to move along rather than disperse caustic.
vicious circle. give llm a little data and say simulate. llm uses trained data to simulate. user takes simulated data and does the same in another llm. very little new data has been added to the system.
I have seen a lot of videos on fine-tuning and read a lot, and I have to say this is one of the most lucid, explanations. By making it very concrete and showing the code and, importantly, the training data you make very clear what is going on in fine tuning great job!
He talks about this in his example where he made an app similar to Blinkist and uploaded a free ebook as his example and he mentioned a section where he added a pdf script somewhere in his Python code so that it would filter certain things in the pdf like the useless cover pages and page numbers etc etc
each problem you will need to adjust your dataset, this is why data engineers and data scientists work together. If i were a beginner i would start reading public IA notebooks in kaggle, in order to be able to create my own dataset.
This should be relatively simple either using Python or a SQL Script. Let's imagine a dataset containing movies and actors in the movies. You could write a script to generate several prompts describing each line such as "Tom Cruise starred in Top Gun", "Tom Cruise was 28 years old when he starred in Top Gun", "Tom Cruise's character was named "Maverick" in the Top Gun Movie"... and so on and so forth using variables from the dataset to define the various prompts, then simply do the same for each movie/character line in the original dataset. Each prompt may sound similar but they are slightly different in order for the model to fully understand the context behind each. Similarly, generating prompts for financial data is very similar as I do this on a daily basis. My scripts gather the sales for the day, the yearly revenues up until that point in time and generate nightly prompts to train our local LLM. In the morning the data is available to our executive team to query. They can then ask questions such as "What are our average daily sales for December of 2024?". You can also get assistance from Data Scientists on Fiverr to assist with your specific use case.
Thanks for the video, David. Really appreciated seeing such a detailed E2E example as I evaluate options for training a custom LLM. Keep up the great work!
Nice video, down to the point and good overview. And cool your honest about how much you know. I like that. Would be interesting to have a video about creating dataset and using Agents for training.
I love how enthusiastic you seem. You seem super interested in this stuff. AI is not my specialty, im utilizing it more as a tool for the projects I work on (security research and other stuff). Cool to see someone just being themselves :)
This is a great video David, you've got yourself a new subscriber. I've been looking for some guidance on this for a while. Don't sweat on not being a complete expert on the topic; you don't need to be 100% across every aspect of a topic to point someone in the right direction. People can fill in the gaps as required.
Awesome video, glad I found your channel. Shout out from South Africa. A video on how to prepare datasets for training, like finetuning an 8b model for a MySQL database. Peace brotherman.
10:29 that would be great tbh, using agents to make dataset to finetune the model is just like inception, you can also make agents to prepare dataset for other agents to create dataset to finetune the model (inception level 2) or make agents to prepare dataset for agents to prepare dataset for finetuning the model which will be used to prepare dataset for agents to prepare dataset............
This is a very useful topic, in the future we can train our datasets to specifically use them for different applications, particularly in healthcare or other institutions, benefiting people. Hopefully, a next topic will be about how to create your own datasets. Thanks for the explanation
Hey I could be wrong but I believe you'll have better luck fine tuning with a less quantized version of the model. At least 8 or 16bit would be preferable to 4. I'm not an expert on quantized models, but you lose a lot of resolution when you quantize that much and that likely makes it more challenging for the LoRA to train. Definitely correct me if I'm wrong, folks, but I think this is the case.
Yes, this is consistent with the programming adage 'optimize last'. The trade off is speed for accuracy, but refined-model accuracy will be more important in the longer term than the speed of the refinement process itself.
I didn't understand almost anything, but the video is excellent. You really tried to help. The problem is mine, which needs to have an entire course on "who knows what" to understand that. Perhaps beginning to learn Python, Linux, and data engineering.😅
You can definitely fine-tune ChatGPT 3.5 and you can also ask open AI to invite you to their private waitlist to be able to fine tune GPT 4. So it is definitely possible.
Actually GPT-4 can be fine-tuned by the user, it's done within the openai API and of course used by it's API later on. It obviously has downsides, like the model is still invoked on the OpenAI servers and they are collecting all the data which goes through it (no privacy), but it is possible :)
Hey David, great video and great explanation. Please make a tutorial on how to generate dataset using LLM. For my use case, I have a classification problem and the class imbalance is severe. for the minority classes, I want to generate more meaningful samples using LLM and then build an LLM model to do text classification on the dataset. Any suggestion on achieving this would be great!
Hi David I think it would be a fantastic idea to do a video on how to make a dataset for llama 3.1 on one complex API. I repeat, complex, long API with multiple inputs and outputs. Every time I see an example the APIs are too simple but real world apis are long and complex they might not fit in a prompt
3:37 He is 100% correct that already fine-tuned LLMs like GPT, Claude, and even Gemini 1.5 Pro with 1m+ context, are freaking awful at trying to emulate writing styles. Worst part about ChatGPT for this purpose is that no matter how much you tell it not to, it's filled with clauses like "On the other hand,", "Finally, ", or "As a consequence" and I'll explain to it again all the reasons those phrases don't belong in a rap song.
it's a good video, thanks. But there are a lot of videos about fine tuning. It would be perfect if you would create a video on how to create own data sets for fine tuning. 👍
I really enjoyed watching this! Your energy gives me hope my project will be fine! I was also wondering what the process is called of finetuning in an iterative way. For instance if I were to finetune a model to make a specific style summary, and I am not fully happy with the output with regards to it not capturing specific information from the input text. How would I provide this as feedback / or turn it into more training data for a next round of finetuning?
Hey Ondrej! I think this might be a stretch of the topic, but is it possible to use an llm like llama 3 and fine tune it to respond in another language or would it be necessary to train an llm from scratch for this?
This video is exactly what I was looking for. Thank you. Now, I wish to know which hardware configuration I will need to install and use Llama 3 models locally in my own machine. Can you help me?
Oh wow you managed to fix the fine-tuning issue? Its been a headache for the entire open source rn, because Llama 3 trained their models differently so every fine tune would end up way worse than the original base model.
If you watch the video, you will see that I openly admit that I am not an expert when it comes to fine-tuning. In fact, making this video definitely was outside of my comfort zone.
@@DavidOndrejay respect man. In fact I think you should include more fine tuning to your videos in the future. You can’t run away from fine tuning if you want A.I to move to commercial use. Llama 3 is probably the only exception in the industry rn that has everyone stumped
Its is possible to create something local with llama to run on a raspberry pi and have it check spelling and grammar and rephrase things like Grammarly
@@DavidOndrej What I'm looking for is an unregistered Dalle model The reason for it is that Dalle can convert simple text into extended prompts, unlike many other engines like sdxl
10:36 -- Yes, please! Thats the most interesting thing to me. Lets assume I have a company with tons of internal information (Intranet posts, PDF files etc.) and I want an easy way to fine tune LLaMA 3 on all this info. How to create a dataset out of this? To create questions and answers based on those information would take a humans life. 😅
This seems like a lot of work in forming the data prep rather than the RAG approach (eg. custom GPTs) where you embed N documents to “fine tune”. Thoughts on each approach?
can you offer any advise about importing the ggufs into ollama, mine just spit out gibberish, I presume it has something to do with the modelcard but no idea
Hi, I didn't see how this is domain specific, as I tested the 2 prompts on llama3-8b-instruct itself, the ouput is correct as well. I've tried to fine-tune the instruct model with some Q&A dataset, and it can correctly answer original questions from the dataset, but cannot answer them correctly if I paraphrase the question. Would be great if you can share more on domain specific perspective.
when are you going to make datasets for fine-tuning, I have currently data in mysql that I need to extract and create the datasets for fine-tuning llama.
I'm finding the data-set stuff very confusing. What if I want to create a data-set that's just my writing? I want the model to emulate my writing perfectly. I don't have question/answer pairs.
I am looking for tutorial how to generate dataset using Agents. There is no such tutorial (or I am not able to find it). It would be great to generate chat format (conversation) dataset as a response of task. So as an input you have list of task, question and then agents generate conversation to this topic.
perfect timing i was just thinking about having multiple llama 3 versions fine tuned for specific coding projects instead of a broad coding language base. is this just a waste of time and im better off having a general coding version instead? i was considering having a few fine tuned models to imitate a development team with crew.
Did anyone get the example running? The copied notbook results in an error when starting the training. I already fixed the missing comma and set the max_steps to 60.
Is the trained model able to be used with "ollama run trained_model_name"? Do I have to download it directly and put it some where for that to work? I currently have a python program setup that uses the ollama module and runs llama3. But I would like to use a fine tuned model instead as I am trying to make a Jarvis like personal assistant!
Could I fine tune my private llama 3 llm but also connect it to chat GPT 4 API such that it can reach out for additional information beyond what the pre-trained model and private knowledge base May have? If so would it then be using chat GPT for API almost like some people can figure rag to look at their private data sets?
Hello Sir !!! Please reply to my question. I have finetunes llama on my custom dataset but now I am having problem in Deployment I am using Flask and each time I run the flask app it load the lora adapters and then load the while base model architecture which end up loading all the ram and then end the session but I don't want that. I just want to use the lora adapters Also the problem I am facing is also with model.save_pretrained() it says that quantised model can't be stored like this. Please tell me what to do
💻 DEVELOPERS, I'M HIRING! Apply here: forms.gle/Y8yNSpCapcDPuTXt6
If you want a personalized AI strategy to future-proof yourself and your business, join my community: www.skool.com/new-society
I highly recommend it, the community is fabulous!
Ive been a wordpress developer for the past 20 years and then became the lead search engine optimization manager for an agency, I see a lot of things in AI that are replacing the need for me, I’d like to learn a new skill set in AI that makes me irreplaceable. Maybe I can run an ai automation agency
why are some here trashing david ondrej? he is imparting knowledge in an easy to understand way for peeps that do not know. i wrote my first neural net from scratch in 1993 and i have been an ML practitioner since then. i can tell u that info back then was hard to come by. be grateful that u have easy access to it. if u dont like it better to move along rather than disperse caustic.
Agree, I just want to confirm that you gerundized 'disperse caustic.'
I like it.
Both channel guys are amazing guys Ondrej and Fahid 👍👍
yes dataset made by agents ! Thx for all your content !
vicious circle.
give llm a little data and say simulate.
llm uses trained data to simulate.
user takes simulated data and does the same in another llm.
very little new data has been added to the system.
I have seen a lot of videos on fine-tuning and read a lot, and I have to say this is one of the most lucid, explanations. By making it very concrete and showing the code and, importantly, the training data you make very clear what is going on in fine tuning great job!
hello, do you know if it's possible to fine tune with ebook pdf on a specific domain (financial, medical...) ?
He talks about this in his example where he made an app similar to Blinkist and uploaded a free ebook as his example and he mentioned a section where he added a pdf script somewhere in his Python code so that it would filter certain things in the pdf like the useless cover pages and page numbers etc etc
Yes, please make a video on how to create the datasets for fine tuning AI - and Thanks for all you do.
It would be great if you could make a video on how to create datasets for fine tuning using LLM's/Agents!
It can be so helpful!
each problem you will need to adjust your dataset, this is why data engineers and data scientists work together. If i were a beginner i would start reading public IA notebooks in kaggle, in order to be able to create my own dataset.
did u find any way
Interested as well!
This should be relatively simple either using Python or a SQL Script. Let's imagine a dataset containing movies and actors in the movies. You could write a script to generate several prompts describing each line such as "Tom Cruise starred in Top Gun", "Tom Cruise was 28 years old when he starred in Top Gun", "Tom Cruise's character was named "Maverick" in the Top Gun Movie"... and so on and so forth using variables from the dataset to define the various prompts, then simply do the same for each movie/character line in the original dataset. Each prompt may sound similar but they are slightly different in order for the model to fully understand the context behind each. Similarly, generating prompts for financial data is very similar as I do this on a daily basis. My scripts gather the sales for the day, the yearly revenues up until that point in time and generate nightly prompts to train our local LLM. In the morning the data is available to our executive team to query. They can then ask questions such as "What are our average daily sales for December of 2024?". You can also get assistance from Data Scientists on Fiverr to assist with your specific use case.
Thank you for this video. The topic of fine-tuning was very interesting to me.
Thanks for the video, David. Really appreciated seeing such a detailed E2E example as I evaluate options for training a custom LLM.
Keep up the great work!
Nice video, down to the point and good overview. And cool your honest about how much you know. I like that. Would be interesting to have a video about creating dataset and using Agents for training.
I love how enthusiastic you seem. You seem super interested in this stuff. AI is not my specialty, im utilizing it more as a tool for the projects I work on (security research and other stuff). Cool to see someone just being themselves :)
This is the kind of content that I've been wanting to see that I haven't been able to find in an easily digestible form.
Dude, just found you, you rule! This was an incredibly great analysis!
This is a great video David, you've got yourself a new subscriber. I've been looking for some guidance on this for a while. Don't sweat on not being a complete expert on the topic; you don't need to be 100% across every aspect of a topic to point someone in the right direction. People can fill in the gaps as required.
the best fine tuning video for ever , thx David
Awesome video, glad I found your channel. Shout out from South Africa. A video on how to prepare datasets for training, like finetuning an 8b model for a MySQL database. Peace brotherman.
Yes please. Team of Agents to create a fine tuning data set from your proprietary data.
+++
Keep up the amazing work bro.
You provide us valuable knowledge.
TY I finally trained my first model. Here's another vote for the how to create the fine tuning using LLM agents.
10:29 that would be great tbh, using agents to make dataset to finetune the model is just like inception, you can also make agents to prepare dataset for other agents to create dataset to finetune the model (inception level 2) or make agents to prepare dataset for agents to prepare dataset for finetuning the model which will be used to prepare dataset for agents to prepare dataset............
This is a very useful topic, in the future we can train our datasets to specifically use them for different applications, particularly in healthcare or other institutions, benefiting people.
Hopefully, a next topic will be about how to create your own datasets. Thanks for the explanation
Hey I could be wrong but I believe you'll have better luck fine tuning with a less quantized version of the model. At least 8 or 16bit would be preferable to 4. I'm not an expert on quantized models, but you lose a lot of resolution when you quantize that much and that likely makes it more challenging for the LoRA to train. Definitely correct me if I'm wrong, folks, but I think this is the case.
Nice to see you here!
Yes, this is consistent with the programming adage 'optimize last'. The trade off is speed for accuracy, but refined-model accuracy will be more important in the longer term than the speed of the refinement process itself.
LOVE these empowering videos, thanks for sharing 🙏
love the videos thanks a lot for taking the time to put them out
Excellent tutorials
It would be an interesting topic for a video on how to use agents to generate data for fine-tuning.
I didn't understand almost anything, but the video is excellent. You really tried to help. The problem is mine, which needs to have an entire course on "who knows what" to understand that. Perhaps beginning to learn Python, Linux, and data engineering.😅
your right headphone will be over your eye soon ;D Thanks for the great content David!
Thank you very much, really informative video. Keep it up!
You can definitely fine-tune ChatGPT 3.5 and you can also ask open AI to invite you to their private waitlist to be able to fine tune GPT 4. So it is definitely possible.
This sounds like exactly the thing that I need, thanks
thank you! this helped me so much.. I just want to know how to tune hyper-parameters? tips and tricks and so on
Great vid, it would be really awesome if you could make a video on how to make data sets for fine tuning! That would help a lot
Actually GPT-4 can be fine-tuned by the user, it's done within the openai API and of course used by it's API later on.
It obviously has downsides, like the model is still invoked on the OpenAI servers and they are collecting all the data which goes through it (no privacy), but it is possible :)
Great work... Thanks for the tutorial ❤❤❤
I would love to see a dataset fine tuning tutorial!
It would be amazing and helpful if you could make a video on how to create datasets for fine tuning using LLM's/Agents!
Hey David, great video and great explanation. Please make a tutorial on how to generate dataset using LLM. For my use case, I have a classification problem and the class imbalance is severe. for the minority classes, I want to generate more meaningful samples using LLM and then build an LLM model to do text classification on the dataset. Any suggestion on achieving this would be great!
Please do make a video on creating datasets, both with and without the use of agents!
You're putting out some great content
Informative video!
Hi David I think it would be a fantastic idea to do a video on how to make a dataset for llama 3.1 on one complex API. I repeat, complex, long API with multiple inputs and outputs. Every time I see an example the APIs are too simple but real world apis are long and complex they might not fit in a prompt
3:37 He is 100% correct that already fine-tuned LLMs like GPT, Claude, and even Gemini 1.5 Pro with 1m+ context, are freaking awful at trying to emulate writing styles.
Worst part about ChatGPT for this purpose is that no matter how much you tell it not to, it's filled with clauses like "On the other hand,", "Finally, ", or "As a consequence" and I'll explain to it again all the reasons those phrases don't belong in a rap song.
please make a video on how to create datasets!
love the 3 primary colors at 10:12 :)
Excellent information ❤
it's a good video, thanks. But there are a lot of videos about fine tuning. It would be perfect if you would create a video on how to create own data sets for fine tuning. 👍
awesome David
Yeah, it would be nice if you could set up a video showing how to automatically generate datasets for fine-tuning LLMs… Tks
Thank you so much
Very useful information!
I really enjoyed watching this! Your energy gives me hope my project will be fine!
I was also wondering what the process is called of finetuning in an iterative way. For instance if I were to finetune a model to make a specific style summary, and I am not fully happy with the output with regards to it not capturing specific information from the input text. How would I provide this as feedback / or turn it into more training data for a next round of finetuning?
thanks for the tutorial!!
Thanks alot !!! it is really helpful :)
your a savior
Hey Ondrej!
I think this might be a stretch of the topic, but is it possible to use an llm like llama 3 and fine tune it to respond in another language or would it be necessary to train an llm from scratch for this?
As I am , i need to know this
This video is exactly what I was looking for. Thank you. Now, I wish to know which hardware configuration I will need to install and use Llama 3 models locally in my own machine. Can you help me?
open AI has its own API for fine-tuning
Oh wow you managed to fix the fine-tuning issue? Its been a headache for the entire open source rn, because Llama 3 trained their models differently so every fine tune would end up way worse than the original base model.
If you watch the video, you will see that I openly admit that I am not an expert when it comes to fine-tuning. In fact, making this video definitely was outside of my comfort zone.
@@DavidOndrejay respect man. In fact I think you should include more fine tuning to your videos in the future.
You can’t run away from fine tuning if you want A.I to move to commercial use.
Llama 3 is probably the only exception in the industry rn that has everyone stumped
Has a video on making custom data sets using a team of agents been made?
Its is possible to create something local with llama to run on a raspberry pi and have it check spelling and grammar and rephrase things like Grammarly
is there a unrestricted dalle e generator? or a something simular like dalle. i like it but the restrictions around it is just crazy now these days.
of course, there are plenty of unrestricted SDXL models
@@DavidOndrej What I'm looking for is an unregistered Dalle model The reason for it is that Dalle can convert simple text into extended prompts, unlike many other engines like sdxl
Stable cascade @@jimmysrandomness
10:36 -- Yes, please! Thats the most interesting thing to me. Lets assume I have a company with tons of internal information (Intranet posts, PDF files etc.) and I want an easy way to fine tune LLaMA 3 on all this info. How to create a dataset out of this? To create questions and answers based on those information would take a humans life. 😅
Can you make a video performing fine tuning?
Thanks for the contents. How can we made a dataset by agents, for simulating an interview for example ?
i liked that you also don't get 100% of everything and accepts it.....cool btw nice tutorial helped a lot
More content with lama!!! 🙏❤️
This seems like a lot of work in forming the data prep rather than the RAG approach (eg. custom GPTs) where you embed N documents to “fine tune”. Thoughts on each approach?
Yeah, a video on agents for finetuning datasets with a fine tuned LLMs, and used by agents for a real world application.
can you offer any advise about importing the ggufs into ollama, mine just spit out gibberish, I presume it has something to do with the modelcard but no idea
Where is fine tuning models stored and how can I find and download it for use?
do you think that sora could be at the point to "read a book" and create a direct visual video of the story of the book?
I want to Create a Lama 3 legal Assistent. I would be happy in you can Show a data prep example
Hi, I didn't see how this is domain specific, as I tested the 2 prompts on llama3-8b-instruct itself, the ouput is correct as well. I've tried to fine-tune the instruct model with some Q&A dataset, and it can correctly answer original questions from the dataset, but cannot answer them correctly if I paraphrase the question. Would be great if you can share more on domain specific perspective.
Need help with this can you do a video how to use it on gpt4all after fine tuning I'm unable to do that. Also amazing video thank you soo much
what's the dataset like if train for conversations? for example: in a conversation, we have one instruction, multi inputs, and multi outputs
when are you going to make datasets for fine-tuning, I have currently data in mysql that I need to extract and create the datasets for fine-tuning llama.
I'm finding the data-set stuff very confusing. What if I want to create a data-set that's just my writing? I want the model to emulate my writing perfectly. I don't have question/answer pairs.
Nice video
What if I run this on a Linux Virtual Machine where there will not be a GPU ? Would it work? Cause in first place we check the gpu compatibility
I am looking for tutorial how to generate dataset using Agents. There is no such tutorial (or I am not able to find it). It would be great to generate chat format (conversation) dataset as a response of task. So as an input you have list of task, question and then agents generate conversation to this topic.
How about Fine-Tuning vs. RAG in those specific things?
Thank you...
What about running this locally? I'd really like to know how to do that, cause the documents aren't easy to follow.
perfect timing i was just thinking about having multiple llama 3 versions fine tuned for specific coding projects instead of a broad coding language base. is this just a waste of time and im better off having a general coding version instead? i was considering having a few fine tuned models to imitate a development team with crew.
NotImplementedError: No operator found for `memory_efficient_attention_forward` with inputs:
But what if I have tabular data containing different columns such as name age dob etc and I want to generate synthetic data from it
thanks
Do you ever sleep? Wow this is amazing 🎉👏
Did anyone get the example running? The copied notbook results in an error when starting the training. I already fixed the missing comma and set the max_steps to 60.
@David what do you suggest if you want to create javascript? and how do i train it?
Hi David
I have requirement to fine tune off file images. Any idea on how to do it?
Is the trained model able to be used with "ollama run trained_model_name"? Do I have to download it directly and put it some where for that to work? I currently have a python program setup that uses the ollama module and runs llama3. But I would like to use a fine tuned model instead as I am trying to make a Jarvis like personal assistant!
If I fine tune it about info about myself, is 7 facts enough? Like age, weight, hobbies, etc m?
I didn't understand, where you gave your dataset to fine-tune on?
Hey what about phi-3..?
Could I fine tune my private llama 3 llm but also connect it to chat GPT 4 API such that it can reach out for additional information beyond what the pre-trained model and private knowledge base May have? If so would it then be using chat GPT for API almost like some people can figure rag to look at their private data sets?
Hello Sir !!!
Please reply to my question.
I have finetunes llama on my custom dataset but now I am having problem in Deployment I am using Flask and each time I run the flask app it load the lora adapters and then load the while base model architecture which end up loading all the ram and then end the session but I don't want that. I just want to use the lora adapters
Also the problem I am facing is also with model.save_pretrained() it says that quantised model can't be stored like this.
Please tell me what to do
After fine tuning this model, how do I create a chatbot with the model.
Are local LLM really that local or just free ? Because I'm not really running it on my computer, more of a cloud base free and flexible llm ?
Yea they are. You can run ollama on your computer then pull down a model, such as llama3, Mistral or Dolphin and run everything, completely locally
Hi David could you please provide the video on Llama4