Видео

@@user-hl1nf8jo7i yes for Claude you have to pay via anthropic

True it was super cool

@@paulyflynn thank you so much

I think so. With Ollama it should be possible. Good idea I will try it too

@@pavankumarreddy9871 yes it does

Thank you so much 🙏

@@NanoGi-lt5fc not sure. Did you click the bell icon ?

@@superlazycoder1984 yes di I hv clicked the bell icon. If u say I will double click it 😂

@@NanoGi-lt5fc hehe. I understand. I will raise it to the RUclips support. Thank you for letting me know

Not sure if U meant the logic. But there could be multiple ways to do this . I just chose the one which suited the best then

@@superlazycoder1984 It's okay that is fine. We are thankful for the efforts that you put.

Thank you so much

Not yet. right now it can only be used for image analysis

@@superlazycoder1984 thanks

@@bagguji1455 thank you

@@yogesh-ru7uu that's a great idea . Definitely will try to make a video on it

@@NanoGi-lt5fc thank you so much

Oh yes will try it out

Thank you

I think if you use OpenCV to read video frames then U can use the image frames to send to LLama3.2 vision. Llm can do real time analysis on it. Still might be a bit slow. But that's how I would do it

@@superlazycoder1984 Yes please

Hi . I havent tried it but hopefully this can help you sort it docs.llamaindex.ai/en/stable/examples/vector_stores/chroma_metadata_filter/

If I understand for unsupervised data maybe it's better to make a rag pipeline and answer questions from the document data

@@superlazycoder1984 No your understanding is not correct because when rag fails we choose fine tune option,it means that already what ever the llm model we have it is already trained on huge data as well as we have rag (knowledge base on custom data) when we ask question first it goes to db(chroma,pinecone,weaviet,fiass) then fetch top k from db based on similarity search(cosine,eculdian,dotproduct) and then the response are going to the llm here the llm is analyse the these answers and his knowledge after this it sends answer,here the response is not relevant to the question what we are asking then in this situation rag fails so here we definitely goes to fine-tune on custom data and also here we are doing two ways to fine-tune the model one is supervised fine tune and another one is unsupervised fine-tuning, supervised means labelled data and unsupervised means unlabel data. I hope you understand.

As of now I have only came across supervised fine tuning. Unsupervised LLM training is done but for fine-tuning I have only seen supervised. That is why I suggested for your use case RAG so you can read data from any type of document or text.

@@Emran-i8u thank you so much

Thank you so much

Thank you so much

The 8B model is available. But I couldn't find the 405B one huggingface.co/docs/transformers/en/model_doc/llama3

Company behind it is Sambanova systems. It was founded in 2017 sambanova.ai/enterprise-ai-company

Unfortunately I don't know either. A100 is paid on Google Colab pro subscription. Worth checking if they have some discount

Thank you so much :)

Sure will try to explore that too. For now I made an introduction video of jetpack with WearOS ruclips.net/video/wII1J9VrK2A/видео.html . If u wanna check that.

@@user-wr4yl7tx3w as we will be using aider on local it's pretty safe in my opinion. If U don't trust the LLM from groq you can even have them on local with Ollama

@@NanoGi-lt5fc thank you for your message

@@VenkatesanRamasamy-g4x thank you so much

I made a video to fine tune in hugging face with trainer library ruclips.net/video/OCNraV2Toa0/видео.html. You can check it out

Hi Robert, seems like an OS issue to me. I didn't face this before but did U try this code in Linux environment github.com/milvus-io/milvus/issues/34854

@@superlazycoder1984 Thank you so much for your quick response. I already fixed the problem by installing pip install langchain_community

Thank you so much

Apologies could be the accent. Please try using subtitles it might make it clear. Will also try improving audio

So you can find tune model using Transformers trainer module or auto train for free if your dataset is small as it uses CPU for free. After that to use that model you can create a Collab notebook where you can load that model hugging face transformers library again. However if you want to create it's rest API and host it then you need web hosting service which might not be free. But still you can create apps with that model and host it on hugging face spaces

@@superlazycoder1984 I want to find tune chat gpt 2.0 on my dataset which is about 102 mb and than host it on hugging face is it possible to do that auto train doesn't support text generation

Hmm I don't think gpt2.0 is open source? Are U sure that's the model

And hugging face does support text generation models U can check the drop-down for options

@@superlazycoder1984 yupp di gpt 2 is open source . You are right auto train support text generation model but it didn't support auto training for text generation task

@@NanoGi-lt5fc thank you so much

Agreed I just mentioned the config for anyone curious. Have removed the 8GB config. However still everything here is done on local with no installation.

​@@superlazycoder1984 yeah I know i watched the video but i click the video expecting something vastly different. Thus the clickbait.

Sorry about that experience. Unfortunately cannot change the thumbnail now. But will be careful in future

@@superlazycoder1984 You can change the title and thumbnail anytime you want? I do it all the time on my channel I upload content to daily. In fact, most channels try multiple title and thumbnail combinations to have the best results and best clickthrough rate for their viewers... RUclips itself allows you to test 3 right from the start. Just a lil RUclips protip for you there ;) it might be in your channel name but no need to be lazy all the time..

Hi, Yes it's possible you can download the model on local with Ollama and then use it with the Rag

Hey sorry didn't understand the full question? Did U mean how to integrate human feedback with LlamaFactory?

@@superlazycoder1984 Yea I was hoping for a feedback learning video too. Love your content btw.

@@PtYt24 I agree guess just a habit. But will try make it better

For now it doesn't support spring. But because it supports Java, I would believe it would be still possible to code in spring

@@superlazycoder1984 okay di happy rhaki

Thank you so much

Would recommend to follow this track github.com/mlabonne/llm-course

Thank you

Thank you so much

Yes we created a video on that as well. Please refer ruclips.net/video/7eRKObFFe7A/видео.htmlsi=u2LzUk68uqrk0nO-

@@superlazycoder1984 Thank you

Thank you so much

Hey sorry won't be able to do it for you but if you face an error please post here. I can suggest fixes

Hehe thank you

Thank you so much

Thank you so much