Видео

There are apis in langchain through which you can ask the llm to cite the source if you index the knowledge base with citations. Alternatively you can enter this exact prompt as it is and ask the llm to do it. Refer my simple rag example. LLMs might surprise you with the results 😃. If you need more help pls reach out to me either in telegram or discord. I'll share some links. Unfortunately sharing external links are not supported in YT comments

Thanks for your kind words 😇

Sure bro. I already have a video on a similar topic json vs yaml. Do check that out

🙏

Thank you for making my day 😊🤩

Ollama by itself doesn’t have any specific requirements except windows 10 or later. However, you need at least 2x model size RAM to run the models. For eg, to run 4B model, you need at least 8 GB ram, to run 7B model, you need at least 14GB ram etc

ChromaDB is an implementation of vector DB that I used in this example. Similarly you can connect to other implementations like Redis, Weviate, Pinecone etc

Sure bro. Fine tuning video is on the pipeline 😁

Are you referring to CNNs ?

Kandippa bro! Pls suggest any particular topic you want me to cover. I’ll definitely try 👍

Ollama should be running bro. Which OS are you using ? I cannot paste external links here . Can you join either CIT telegram or discord and ask your Q? Happy to help!

@@conceptsintamil bro discord ketan bro replay please

Thanks for your kind words and making my day 😇

Glad you liked it. I do my own animation bro using Apple keynote. All drawings I do on my own and I animate it

Yes bro. Ollama recently added windows support. Once you installed ollama and downloaded any local llm, you can easily fine tune it either via prompt engg system prompt or via LoRa methods. Most of the latest llms like llama2, mistral, mixtral, dolphin, gemma etc are very much versatile for all domains. You can easily make it work with little effort on prompt engg

@@conceptsintamil Thank you so much bro.

@@conceptsintamil Bro, next put a tutorial for it. This will really help non-technical people in fine-tuning LLM for particular domain

@@mithunkumar25557 yes it’s on the d way 😊

@@conceptsintamil Thanks a lot bro. Will be very helpful. I am also currently trying to fine-tune an LLM for my domain in Tamil.

You don't need a API key at all 😃. That's the beauty of locally hosting the llms. If you use OpenAi's LLM (like chat gpt) then you need to pay $ to open AI and get the API key. With local LLM, it's free, private and secure

bro unga linkedin id soluga

Hi 👋. There are various reasons for fine tuning. Recent research do suggest that prompt engg based fine tuning is preferable than training based fine tuning. One example why need fine tuning is domain based LLMs . For eg if you wanna personalize a general LLM for medical domain , then prompt Engg based fine tuning may not yield best results. It needs a traditional approach. Look at Microsoft’s phi small language model. Same use case

@@conceptsintamil Got it. Thanks

Thanks for your warm words 🙏 . You 🫵really made my day 🤩

Ollama doesn’t support windows bro. One option is to run a Linux vm and run ollama inside . The chat bot can still run in windows

Thank you bro. Glad it helps. I’m making more on AI and LLMs. Will get back to cloud native soon

Davinci resolve

Sure bro! I’ll try my best

🙏🙏🙏🙏🙏🙏

@@conceptsintamil cloud computing exam thala inaikku kannulaam yeriuthu... notes pathu... 🥹