"okay, but I want GPT to perform 10x for my specific use case" - Here is how

[@trickybarrel444]

No clickbait, no long intro, no over-excitement. Just pure, straight-to-the-point content, with real-world examples and neatly divided sections. You are a hero

[@668771]

You understood that? I wish.

[@Bashar3A]

YES! Getting tired of those time filler AI videos, almost didn't watch this one. Now subscribed

[@myob3074]

Aaaah just like the good ol days 😅😂

[@myob3074]

​@@668771knuckle down and study practice study practice, it's not that difficult 🤷🏽‍♂️

[@Andyonrice]

specific use case: filter videos for overexcitement and clickbait, then tag them

[@spoonikle]

holy heck dude, you just taught us magic. This is the magic! not even the fine tuning, but the data set generation! data sources! all the pieces I have been searching for!

[@AIJasonZ]

Thanks so much mate!! That means a lot, DM me if you get any question pls!

[@EGGHEADIUS]

Literally brother you have taught us prompt engineering wizardry, there’s no other way to put it.

[@DeepDiveNFT]

I’ve been generating massive data sets and manipulating in python then bringing it back to fine tune. Shits getting wild out here

[@Zer-ei4co]

@@DeepDiveNFTlol sounds like fun, what kind of general use cases do you have?

[@stan9682]

Be wary that OpenAI prohibits the use of their models to generate training data (no competing closure). It's hard to detect, obviously, but it's useful knowledge imo regardless.

[@sayantandas7544]

One thing I really like about your videos is that it always comes with really quality examples. And the examples are pretty much practical oriented. Do make the embedding video.

[@AIJasonZ]

Thanks man, that means a lot! 😬

[@treadaar]

@@AIJasonZ Absolutely agree. Very helpful and i like the speed you go, staying on topic. Please make the embedding video! Very interested to see your take on it.

[@668771]

I thought the opposite, this was like watching someone assemble a nuclear reactor. I'm more confused about AI after watching this. You need to be a master programmer to properly view this video.

[@antonschillen2633]

@@AIJasonZ +1 on make the embedding video. Really great stuff :)

[@pikleman5880]

Honestly, I love how you don't put a clickbaity title or thumbnail, and how you explain it well but also fast.

[@じゅげむ-s6b]

honestly, I was not expecting jin yang to teach us but I'm all for it straight to the point, loved it

[@hansjochen2193]

One of the best and most straight forward tutorials i have Seen after 5 years of ComputerScience! Keep it up now i am honna binge eatvh your other videos

[@unicornist]

most underrated channel in AI. Thank you for the effort you are putting into this.

[@loktheswe6768]

Omg this is the best one I have seen so far!!!I’m planning to train one of my own model soon!!

[@typrider]

This is an amazing video that helps me make sense of the Azure Open AI playground to train my model. Love this.

[@raivolution]

hey man, thanks so much for this. Because of you, I managed to build my first model. Not looking exactly how I want it to, but let me know if you would have some time to troubleshoot.

[@lucyn.7501]

Fantastic video once again, can’t wait for the next embedding one 😊

[@loudorz]

Great job !

[@sweetmelon3365]

Very good teaching style. I also liked that you actually went through the process ourself. Thank you very much

[@unshadowlabs]

Great Video! I would love to see a video on how to create an embedded knowledge base and use it along side this fine-tuning.

[@BananaHammyForYou]

I'm commenting before I watch the video because that thumbnail is so god dammed hilarious. Edit: Now that I've watched this, I want to like it again.

[@齐洋-o2s]

keep the good job Jin Yang，your top follower from shanghai

[@hughesadam87]

This is such a relevant and timely video. Thanks bud. PS I love all your silicon valley video titles

[@sashariaboshapka8706]

Thought it was another clickbait meme video, and here am I, writing a comment about how cool it is, straight to the point, stuffed with info, super interesting, informative and useful in case someone needed to make his language model and was searching for info on it

[@sirrr.9961]

Greaeeeeeeeeeeettttttt ! I have been surfing through internet for last 6 months for this type of workflow tutorial and here you have just made it. Thank you so much for this. My biggest question was how to prepare my own data to finetune a specific AI model for my use case. You have explained that part very easily. Now I can start making my own chatbots. P.S I am not a tech guy .... just an accountant.😅

[@Theineluctable_SOME_CANT]

You should ditch your boring profession.

[@tahamohd1409]

I'm glad i found this channal! wonderful work!!

[@dogemaster6361]

Excellent video Jason. Well organised and easy to understand.

[@lucaswei9133]

I like it. Because your video i can follow it every step. It's important for a newbine .

[@serispeaks]

Bro you've done something wonderful for the world with sharing this video 🎉👏 As an app dev I learned something I didn't know I needed to know

[@theAIsearch]

Let's be honest. Who here clicked because you thought he was JIMMY YANG?

[@shuzhu6633]

Who’s Jimmy Yang? 😂

[@NESDUB]

@@shuzhu6633 the comedian in the thumbnail

[@aiutksemua]

jian yang

[@bakasenpaidesu]

I'm not 😂

[@tevx9670]

That's 😂

[@wiktormigaszewski8684]

v good tutorial, that goes through a lot of stuff, from overview, datasets, even dataset creation, downloading, finetuning and even uploading a model; wow

[@AIJasonZ]

Thank you! Let me know if anything else missing!

[@wiktormigaszewski8684]

I am now waiting for your tutorial on the reborn Guidance :) Regards! @@AIJasonZ

[@ayushnayak6138]

I have been looking for you for so long. Thank you so much for this. Now I can ditch open AI api key and use this for free for my usecase.

[@MisterDivineAdVenture]

Really appreciate your free thought. Will try to follow you closely.

[@joe12344281]

Thanks!

[@AIJasonZ]

Thank YOU!

[@AdrianMark]

Learning so much from you. Thank you for making this.

[@DEFACTO9]

Bad audio Dude. However, this is Exceptional content... finally... a workable explanation. Thank You.

[@soyitiel]

Your audio gradually went from studio recording to cathedral ambiant. Hey thanks for the info. Love it ♥

[@awesomeGuss]

brooo....God bless you for this...gained a ton

[@ericmackrodt9441]

Holy Crap, JZ? Oh my God! I stumbled upon your channel because I've been looking into AI stuff and I was like "I know this voice and this face". This is a great video!

[@AIJasonZ]

Hahahah hey Eric! Long time dude

[@Yadu-bv1kn]

One of the best, to the point, goldmine of a tutorial

[@CloudWithUgur]

give this man a cookie, dude thank you !!!!

[@LC-lj5kd]

well done! looking forward to the embedding video too

[@Steindustin1010]

Thanks!

[@squibtechnologies]

I'm floored by your content. It is amazing 👏👏👏👏👏👏

[@sparebatteries]

great video Jason - also my first ever comment on KZbin. Most practical tutorials on AI so far imo - I would be keen to see a video where you set up a vector DB and load your own data

[@AIJasonZ]

Hah thanks, first comment means a lot to me! Will get more content :)

[@arfh2759]

All thriller no filler. Wicked vid well done sir

[@davidconnelly]

This channel is going to be huge.

[@ryanfarran]

I have been spinning my wheels on a project and this video unstuck me - thanks a million!

[@markmonfort29]

Brilliant mate. Keep these coming

[@KunjaBihariKrishna]

In my experience, with embedding you have to be careful with the data you use. It's text only, so you can't use a source that makes references to any images or graphs, which means that you will have to rewrite those portions manually so that the information is embedded properly. If you don't do that, then your knowledge base makes no sense. How much that is a problem really depends on your use-case. Also, anything that has to do with calculations can be a problem, such as different methods of capital gains calculation. The embedded knowledge base can be good enough to allow for the model to explain how a calculation method works, but it's not necessarily going to understand them or be able to talk about them in a dynamic way. Which in the case of AI, means it's going to hallucinate, rendering it useless for those topics. I'm still waiting a bit until we have better access to plugins, so that we can create a system that's more like a body of multiple models, which each handle specific aspects of the data. I think there's a possibility there

[@googleyoutubechannel8554]

I think the Bard team also realized this....

[@rexdraconis1703]

But you can create embeddings to the images too. Or it's to hard (I use text-ada-002, and don't know about some techniques(models) for image embeddings. But I'm sure google has one, so it's possible...

[@DorkDork69]

Man this is exactly the sort of guide I was looking for. Thank you very much.

[@indiecowan33]

Wow, what a sick video. Thanks so much!

[@royxss]

This is just amazing! waiting for your embedding model

[@jasonfinance]

Oh man, this is so good, i always confused about when should i finetune LLM, this makes things super clear; Also falcon 40b seem to be an awesome choice!

[@chrismachabee3128]

Excellent lesson on Fine tuning. You are a good teacher. Thanks. Please make he next video I am now subbed.

[@martinskov5155]

Great video mate! Keep up the work (Y)

[@antonschillen2633]

Great content! A how-to without any fluff what so ever that is easy to follow.

[@kingmharbayato643]

Finally someone show it in the most simplest way

[@Triycyan]

Thanks

[@АндрейБогаев-ы2я]

Good Job Jason 👨🏽‍💻

[@JayanaKalansuriya]

What an amazing to the point video! You're an amazing human being Jason. Keep doing you!

[@dreamingtulpa]

Thank you for sharing this. This will definitely help. Also epic thumbnail 😂

[@joejiang8353]

Excellent video Jason! Step by step examples are really helpful

[@bayesian7404]

Excellent job, thank you. This is just what I needed.

[@karol_04]

Finally I've found an AI video that I don't understand at all! It's funny that there is actually so few experts willing to share their knowledge

[@NeuroNinjaX]

Awesome! I'm totally new to fine-tuning. I'm looking forward to fine-tuning a model with thousands of private 4K images!

[@michaelnurse9089]

I appreciate this. Subscribed.

[@MasterBrain182]

Astonishing content Man 👍👍👍

[@yashsoni5437]

Great Video! Keep up the good work

[@th3j0t46]

This was super informative and I learned a lot. Great video!! Not to be rude, see this as a constructive comment, be aware of some typos you have along the video. It's not like it bothers me much or anyone, but correcting those little details add up to make your content even better! Good luck and I hope to watch more of your videos :D

[@jasonwang3539]

Real thanks from another Jason man, clear and concise video! As a high schooler, I'd like to create an LLM project, so I hope I can follow through after I do my research :)

[@BiranchiNarayanNayak]

Excellent tutorial on finetuning Falcon 7B

[@mytechnotalent]

Great job Jason! Very helpful!

[@Blue-bd4vs]

my man great content ever, keep it up

[@dubey_ji]

that was really good glad this was recommended to me

[@SafetyLabsInc_ca]

Great video. Helped create a much tighter solution for our client.

[@wizarddon]

Thank you Jin yang

[@andrewperez1912]

Amazing content.. thank you for putting together such great guides. I used Falcon but before and after training, the output was just one line that kept repeating over and over. Did I do something wrong or could there be an issue with the model? Also, is it worth doing the same method you described but with Llama instead?

[@AIJasonZ]

Thank you so much mate 🙏 means a lot! What was the training dataset? And what is the expected behaviour?

[@DigitalMirrorComputing]

Absolutely loved the video mate! super focused! You inspired me for my own channel! thanks and keep it up! :))

[@kazimafzal]

Bloody awesome video - concise yet detailed! Looking for upcoming vids

[@maxgeorgopoulos3516]

Awesome video! So many great resources. Demystified a lot for me. Thanks!

[@mickelodiansurname9578]

Thanks for that man, very instructional... and yeah the 40b really bombs out colab for me... but the 7b is sort of piss poor. Having said that, the day when coders and users stop complaining about the power of a given application will be a fairly interesting day indeed!

[@668771]

There is ZERO chance all of you here understand what he is saying. Fine tuning models, downloading models, installing models......?! I thought AI was an improvement, this is too advanced for the smartest humans.

[@skateking8]

Looking forward to the video on embedding a knowledge base

[@AIJasonZ]

Coming soon, stay tuned!

[@chonmon]

Good work dude! This is super useful!

[@myhificloud]

@9:45 Yes, please and thank you. Also thank you for this.

[@chaseclingman]

Would love to see the video on the knowledge database. Thank you for this!

[@TheAiClub]

Good stuff Jason🤙🏻

[@me-cm8or]

That’s some high quality content right here

[@svetimfm]

Let’s fundraise this fella an A100 ASAP!!!!

[@AIJasonZ]

hahahah LOVE 💕 Thank you so much mate, means a lot to me!

[@chopradimpy]

Thanks for sharing. It was helpful for understanding the fine tuning process. Can you also make a video on KnowledgeBase embedding? Thanks

[@wilfredomartel7781]

Well explained my friend👏👏

[@mhc4124]

This is fantastic info, thanks! Subscribed.

[@AVYogesh]

Great video. Small tip though - Inform the users which tools are free and which ones are paid.

[@Omega-mr1jg]

What i have been wanting to do for the longest while, might just do it when im free

[@ahmadjunaidi21-l6l]

really helpful and straight forward. Subbed !

[@yelectric1893]

Beyond excellent. Damn. Reminds me of Yolo and Stable diffusion in a way. I have got to read the papers sometimes and try tot write a C implementation. This is beautiful!!!!!

[@Hmelino]

Yes please, waiting for knowledge base embedding video ! 🎉

[@aiordie]

This man is the source.

[@mitchellsigley8574]

Great video Jason.

[@sup5356]

fantastically clear

[@Muuip]

Great concise tutorial, thanks!👍

[@RailSuleymanov]

Sometimes I encounter "RuntimeError: Inference tensors cannot be saved for backward. To work around you can make a clone to get a normal tensor and use it in autograd." in the "Finetune the model" cell, any ideas what that could be?

[@SanuThoppil]

Hey, I just got the same error as well. Did you manage to solve it??

[@padmarajkore1555]

Hey i am also facing same error, got any solution?@@SanuThoppil

[@OrewaNULLz]

@@SanuThoppil also got same ERROR :(

[@trunkmak_xo]

🎯 Key Takeaways for quick navigation: 00:00 *🤖 Methods for GPT Usage* - Two methods discussed: fine-tuning and knowledge base. - Fine-tuning: Suitable for modifying behavior, like mimicking specific individuals. - Knowledge base: Utilized for providing accurate domain-specific data, such as legal or financial information. 01:38 *💡 Case Study: Fine-Tuning for Military Purpose* - Selection of a powerful large language model (Falcon) for fine-tuning. - Importance of quality dataset for fine-tuning model effectiveness. - Utilization of GPT to generate training data and platforms like Randomness AI for bulk processing. 05:08 *🛠️ Fine-Tuning Process* - Preparation steps for fine-tuning on Google Colab, including library installation and model loading. - Use of Low-Rank Adapters for efficient fine-tuning. - Saving and uploading the trained model to Hugging Face for sharing and further use.

[@nithinchinni]

Great information. I am interested in the embedding style. Would love to see a video of that.

[@lordofshadows8510]

30 seconds in and my brain is already hurting

[@nsoe8990]

thanks for this amzaing videos. looking forwards to your next videos for embedded knowledge base and creating a chatbot for UI using these trained models.