that's really good I'm from India and I love to here that our people also contributing in AI opensource community
@electric_mind
5 ай бұрын
Fine tuning a pre trained LLM is no big deal. Training one from scratch is
@straightforward8928
5 ай бұрын
Ohh thanks for the info, but I bet for beginners it would be a great way to start I'm a beginner
@electric_mind
3 ай бұрын
Indeed, for a beginner it's a great feat to fine tune a LLM but the problem is as a beginner it's difficult to have beefy GPUs
@anirudh514
5 ай бұрын
Amazing!!
@1littlecoder
5 ай бұрын
Nice work it is !
@devagarwal3250
5 ай бұрын
good to see open source model from India
@rathindrasaran
5 ай бұрын
Very promising development for Indian languages
@mahavir8148
5 ай бұрын
krutim ai was huge scam happened to indian ai ecosystem. its pleasing to see that the development of ai doesn't end there, and Navarasa is really game changer. it can be used as an awareness tool in rural parts of india to tell them about efficiency of ai and how powerfull tool it is and many more stuffs
@1littlecoder
5 ай бұрын
I wouldn't write them off, but it's good to see indie developers building great things!
@mahavir8148
5 ай бұрын
@@1littlecoder yeah, currently In highschool or else even I would have gotten full time invested in AI especially In India as India market has got much much more potential for AI usecases
@1littlecoder
5 ай бұрын
Woah, I've got a sub from Highschool 🥹!!!
@mahavir8148
5 ай бұрын
@@1littlecoderYes Anna 😅 from Rajasthan ❤ you are doing great job
@-rate6326
5 ай бұрын
@@mahavir8148 it wasn't a scam. As someone who is trying to learn about llms and fine tuning. They must have used chatgpt for feedback. Thats how they were able to train it so fast. If krutrim generates gibberish Chatgpt can tell if it's gibberish or not. OPENAI used humans to do that. Chatgpt might have been used in other ways too. It's not scam you are just uninformed.
@mahavir8148
5 ай бұрын
Off topic :- Recently I was scrolling through reddit and came across a post which said that by using claude 3 he has optimised code of Animate Diffusion by 100x, so i was thinking is it possible right ? To use large context window length LLMs to be used to optimise code of existing models by much greater efficiency I tried sharing link, but comment is getting removed automatically
@sahilmeena485
5 ай бұрын
But did he share it?
@mahavir8148
5 ай бұрын
@@sahilmeena485 share what ?
@sahilmeena485
5 ай бұрын
@@mahavir8148 i meant the GitHub repo. But i Googled it and got to know that it's not an open source model, so just ignore my comment now
@PallaviChauhan91
5 ай бұрын
Great, being a north Indian I always wanted to learn some south Indian languages. Recently started learning Tamil. Is there a tutorial of yours for beginners, that teaches how to run this llm locally? I am a complete beginner.
@1littlecoder
5 ай бұрын
Congratulations on your attempt to speak Tamil . I hope you get better at it. I'm watching shark tank India to improve my Hindi 😆😆😆 on this particular LLM, I'll soon drop one!
@julien-hf
5 ай бұрын
awesome video as usual 🔥
@AI4Ever4Life
5 ай бұрын
Abdul... Make one beginners guide to unsloth. 😊 Along with a few hands on examples
@varunkalyan3135
5 ай бұрын
hey, I'm working on a semantic search for an ecommerce using elastic search. Is it possible to include regional element using this?(say i search adrak & i need the engine to output ginger)
@BhanudeepSimhadri
5 ай бұрын
Dude try making videos in fireship channel style.
@abhaygholap3613
5 ай бұрын
All this is fine but make a video on how to use this in Rvc as we don't know anything about coding
@SubhagatoDutta
5 ай бұрын
I am interested to see a video where we can use the released dataset to fine tune mistral-7b as well along with Gemma
@gunasekhar8440
5 ай бұрын
Thanks for making the video. But majority times google translator doesn't gives exact translation which may leas to a poor training. My idea is to build robust tokenization model and then use that tokenization for embedding and use that to fine tuned. I know whether im going on right direction or not. If not Correct me? Thanks in advance @1littlecoder
@kotcraftchannelukraine6118
5 ай бұрын
I hope this hyperfixation on textual modality will fade away soon. It is sometimes good for coding or Internet searching, but this is not how normal intelligence works. I know that Transformer architecture was used to create ViT and ImageGPT and they are working fine beyond text, but motor control, real-time audio and video processing, and normal memory module are yet to be created.
@mshonle
5 ай бұрын
Tell me (us) more about unsloth!
@sahilmeena485
5 ай бұрын
Can you make a roadmap about what to learn for getting started building generative AI in 2024
@devagarwal3250
5 ай бұрын
pls make a video on unsloth framework
@siriyakcr
5 ай бұрын
There was a open challenge on kodava language for bringing it's own alphabet, this work might help
@stark258
5 ай бұрын
unsloth video please !!!
@priyanshu2655
5 ай бұрын
I want tutorial on unsloth 😊
@mathavansg9227
5 ай бұрын
i want unsloth tutorial
@chetanmali231
5 ай бұрын
bro make tuto how to use it
@haricharanvalleru4411
5 ай бұрын
Make one video on fine tuning
@velugucharan8096
5 ай бұрын
Bro can explain about training party code line by line??
@kmano2915
5 ай бұрын
When a Tamil guy meets another Tamil guy: 🥰
@1littlecoder
5 ай бұрын
Thanks bro!
@Ai_Engineer
5 ай бұрын
video on unstoth
@gokulraj6633
5 ай бұрын
Hi , bro neenga tamil ah ?
@muhammedajmalg6426
5 ай бұрын
please make unsloth videos
@Pradip_Deshmukh
5 ай бұрын
I want tutorial on unsloth
@mathavansg9227
5 ай бұрын
i never know your from tamil nadu nanba
@PP-pe3fs
3 ай бұрын
bruh you just explained one...
@Macorelppa
5 ай бұрын
JAI HIND.
@KaranChecker
5 ай бұрын
What do you mean normal guys 😂😂
@mahavir8148
5 ай бұрын
People with job's, and casual life. Without a big name in startup world
@Ravisidharthan
5 ай бұрын
മലയാളം വന്നല്ലോ... പക്ഷേ dataset pora, വിഷ്ണുവിൻ്റെ ആണ്... നിലവിൽ ഉള്ള ബെസ്റ്റ്... Pls do a video on Unsloth
Пікірлер: 50