Spaces:

---

IndianAIDevs

/

README

like 2

Running

Has anyone pre-trained LLM model from scratch ? If yes then share your experience, things to consider while training, notes, tips etc.

Hi i am also intrested into LLM Model , i am about to start this reserach from next week please give any inputs

Hi i am also intrested into LLM Model , i am about to start this reserach from next week please give any inputs

Hey @Shashank2k3 , if you want your own LLM model, first you need huge data. You can start with fine tuning already available good LLM models like Gemma, Phi, LLAMA, mistral etc with your dataset. Start with small models of sizes like 4 to 7B parameters. For pre-training LLM from scratch you need enormous data, good resources like heavy duty GPUs and CPUs and also have knowledge of training techniques, NLP, etc . You can always brainstorm with ChatGPT to get more knowledge.

Hey @kalashshah19 , thanks for the input! I already have a solid foundation in these areas from my Bachelor's degree in AIML, and now I’m looking to dive deeper into the world of LLMs.

Hey @kalashshah19 , thanks for the input! I already have a solid foundation in these areas from my Bachelor's degree in AIML, and now I’m looking to dive deeper into the world of LLMs.

Great !

Yupp so what you guys do, i mean profession!!!

Yupp so what you guys do, i mean profession!!!

I am an Associate Data Scientist at Casepoint.
What about you ?

I am an Associate Software Developer at Brillius Technologies

I am an Associate Software Developer at Brillius Technologies

Great, where are you from and where is the company ?

I am from hyderabad , company origin is calfornia pleasonton , but i am working at hyderabad branch.

I am from hyderabad , company origin is calfornia pleasonton , but i am working at hyderabad branch.

Okay, so do u work in AI ML there ?

Hello guys

Hello guys

Yo whatsup ?

Hello guys

Yo whatsup ?

fine and what about u?

Hello guys

Yo whatsup ?

fine and what about u?

I am great. Where do u work ?

Hello guys

Yo whatsup ?

fine and what about u?

I am great. Where do u work ?

I work at HelpingAI

Yeah previously my company is in it staffing side , just 6 months ago they have started in IT solutions to build product called brillius AI tutor where we are developing an online learning platform for it professionals, in that I'm taking care of many things where AIML is one of those things.

Hello guys

Yo whatsup ?

fine and what about u?

I am great. Where do u work ?

I work at HelpingAI

Are you talking about this ? - https://huggingface.co/HelpingAI
Is it a company or Open Source Community ?

Yeah previously my company is in it staffing side , just 6 months ago they have started in IT solutions to build product called brillius AI tutor where we are developing an online learning platform for it professionals, in that I'm taking care of many things where AIML is one of those things.

Wow, great. Is it a startup or big company ?

Hello guys

Yo whatsup ?

fine and what about u?

I am great. Where do u work ?

I work at HelpingAI

Are you talking about this ? - https://huggingface.co/HelpingAI
Is it a company or Open Source Community ?

A small startup

Hello guys

Yo whatsup ?

fine and what about u?

I am great. Where do u work ?

I work at HelpingAI

Are you talking about this ? - https://huggingface.co/HelpingAI
Is it a company or Open Source Community ?

A small startup

Okay and is the link for the Company correct ?

yes

yes

Is it your company (startup)? I mean are you owner or founder ?

yes

Is it your company (startup)? I mean are you owner or founder ?

Yes, I am its founder.

yes

Is it your company (startup)? I mean are you owner or founder ?

Yes, I am its founder.

Great, keep up the good work and whats your vision ?

@kalashshah19 its a medium scale company , but for me its a startup cause we are the first team of IT , and we dont have any seniors to guide us

yes

Is it your company (startup)? I mean are you owner or founder ?

Yes, I am its founder.

Great, keep up the good work and whats your vision ?

Making agi token and time efficient

Btw try Dhanishtha 2.0 preview
HelpingAI.co

@kalashshah19 its a medium scale company , but for me its a startup cause we are the first team of IT , and we dont have any seniors to guide us

ohh, I see

yes

Is it your company (startup)? I mean are you owner or founder ?

Yes, I am its founder.

Great, keep up the good work and whats your vision ?

Making agi token and time efficient

Cool

Btw try Dhanishtha 2.0 preview
HelpingAI.co

Sure I will !

hey bros

hey bros

yo whatsup !

Can we create an interface for our community for easy chatting

Yeah Sure, good idea !

Shashank2k3/Fake-Profile-Detection-Instagram, please check out my project Fake Profile Detection, where I have trained a random forest model using my own dataset consisting of metadata from various fake and real profiles on Instagram.

Can we create an interface for our community for easy chatting

Great idea! We should create a Discord server

Shashank2k3/Fake-Profile-Detection-Instagram, please check out my project Fake Profile Detection, where I have trained a random forest model using my own dataset consisting of metadata from various fake and real profiles on Instagram.

Nice, good effort !

Can we create an interface for our community for easy chatting

Great idea! We should create a Discord server

So you are suggesting that the HuggingFace community should chat in Discord community 😂
I think he told to create an inference on HuggingFace for chatting.

Can we create an interface for our community for easy chatting

Great idea! We should create a Discord server

So you are suggesting that the HuggingFace community should chat in Discord community 😂
I think he told to create an inference on HuggingFace for chatting.

😂

Hello everyone, I'm a 2nd year DSAI student at IIT Guwahati, i have fine-tuned few models and i have published few articles on ResearchGate. I am determined to build state-of-the-art AI in India for the world.
Its really great to be a part of this community.
Thank you

Hello everyone, I'm a 2nd year DSAI student at IIT Guwahati, i have fine-tuned few models and i have published few articles on ResearchGate. I am determined to build state-of-the-art AI in India for the world.
Its really great to be a part of this community.
Thank you

Hey Ashish, its great to have you in our community. We all will grow as Indian AI Devs and learn new things from each other. Keep in loop !
And also remember that we all are friends so no need to be formal !

Btw try Dhanishtha 2.0 preview
HelpingAI.co

i have used it (spaces), i even shared it with my friends

has anyone contributed datasets or models on AIkosh?

https://huggingface.co/Neural-Hacker/Qwen-BharatBench-Legal
please try this and share your feedback (i'm working on another version to make it even better)

sure !

sure !

please add it in the collection if you like it

Hi everyone :)
I created a dataset on Bhagavad Gita which happens to be the most liked dataset and most downloaded dataset on HuggingFace. I'd appreciate if you make something out of it.
Link :https://huggingface.co/datasets/JDhruv14/Bhagavad-Gita_Dataset

Hi everyone :)
I created a dataset on Bhagavad Gita which happens to be the most liked dataset and most downloaded dataset on HuggingFace. I'd appreciate if you make something out of it.
Link :https://huggingface.co/datasets/JDhruv14/Bhagavad-Gita_Dataset

already liked it and will use in a planned project

sure !

please add it in the collection if you like it

Done !

Hi everyone :)
I created a dataset on Bhagavad Gita which happens to be the most liked dataset and most downloaded dataset on HuggingFace. I'd appreciate if you make something out of it.
Link :https://huggingface.co/datasets/JDhruv14/Bhagavad-Gita_Dataset

Woah great !

🚀 Big news from XenArcAI!

We’ve just released our new dataset: Bhagwat‑Gita‑Infinity 🌸📖

✨ What’s inside:

• Verse‑aligned Sanskrit, Hindi, and English
• Clean, structured, and ready for ML/AI projects
• Perfect for research, education, and open‑source exploration

🔗 Hugging Face: https://huggingface.co/datasets/XenArcAI/Bhagwat-Gita-Infinity

Let’s bring timeless wisdom into modern AI together 🙌

I hope you all love this dataset and contribute positively to AI/ML Research

🚀 Big news from XenArcAI!

We’ve just released our new dataset: Bhagwat‑Gita‑Infinity 🌸📖

✨ What’s inside:

• Verse‑aligned Sanskrit, Hindi, and English
• Clean, structured, and ready for ML/AI projects
• Perfect for research, education, and open‑source exploration

🔗 Hugging Face: https://huggingface.co/datasets/XenArcAI/Bhagwat-Gita-Infinity

Let’s bring timeless wisdom into modern AI together 🙌

I hope you all love this dataset and contribute positively to AI/ML Research

Great, congratulations !

Hey everyone 👋

I've been working on this dataset since last 15 days and it's finally done. I'm pleased to announce that I have created the first QnA dataset for Bhagavad Gita not only in English but also in Hindi and Gujarati.

What’s inside:

• Verse‑aligned English, Hindi and Gujarati questions
• Each verse is paired with 5 question types exploring different aspects.
• Perfect for blending Spirituality and Technology

Link : https://huggingface.co/datasets/JDhruv14/Bhagavad-Gita-QA

(P.S: I'm the one who created the most liked and most downloaded dataset for Bhagavad Gita)

Hey everyone👋

I fine-tuned Qwen3-0.6B on the soketlabs/bhasha-wiki-indic dataset ( i used only ~50k Hindi samples ). The training went fine, loss was around 1.46 and accuracy about 59% but the outputs are completely wrong and make no sense.

I think the issue might be with the dataset format since it’s plain text and not instruction-based so the model probably didn’t learn proper Q&A or instruction-following.

Has anyone faced this before? Should I switch to something like ai4bharat/indic-instruct-data-v0.1 for better results? Any suggestions? 🙏

Same bro, I ft qwen2.5 3b with my gita dataset and the acc was 60% something with loss around 1.42 and the responses were okayish but still it was writing correct hindi but when I made a space out of it, it's just writing anything

Same bro, I ft qwen2.5 3b with my gita dataset and the acc was 60% something with loss around 1.42 and the responses were okayish but still it was writing correct hindi but when I made a space out of it, it's just writing anything

Yeah exactly bro, I thought good loss & accuracy meant it was learning properly but turns out it's more about the data structure. My model also writes proper Hindi but gives totally unrelated or random outputs. I think because plain text data only teaches it next-word prediction and not how to follow instructions or answer properly.

I’m planning to try instruction-style data next (ai4bharat/indic-instruct-data-v0.1). Maybe that’ll fix it. Have you tried using instruction-tuned datasets or formatting your data as Q&A pairs? How you fixed that issue?

SarathiAI v1.0
Eternal Gita wisdom, guided by AI.

I tried fine-tuning Qwen2.5-3B model on a custom Bhagavad Gita dataset for the first time. The model now understands user queries more accurately and responds with answers grounded in Gita teachings. It's not perfect but I'm happy with everything I learned building it. Please give it a try and you can share what should I improve.

Link to Space: https://huggingface.co/spaces/JDhruv14/Sarathi.AI

Link to Model: https://huggingface.co/JDhruv14/Qwen2.5-3B-Gita-FT

Link to Dataset: https://huggingface.co/datasets/JDhruv14/Bhagavad-Gita-QA

Well, I am feeling good seeing your work guys.

@JDhruv14 , @Neural-Hacker

guys, i am building a model (fine-tuning), i want a lot of compute and vram please tell where i can find free or cheap gpu because aws, azure, etc are very costly and maybe kaggle won't work here but i will try it tomorrow. i am thinking to use ola cloud but if there's any platform better than ola then please suggest.

Hello everyone,

I’m excited to share NEET_BioBERT , a fine-tuned lightweight transformer model trained specifically on NEET-style biology multiple-choice questions.
It’s designed for educational AI assistants, practice exam bots and MCQ reasoning systems.

Explore it here: https://huggingface.co/Neural-Hacker/NEET_BioBERT

use and upvote it

Is it fine tuned on question papers only or all content of NEET like books, PDFs, etc ?

it is fine tuned on a dataset consisting ~800 questions including practice questions and pyqs

Nice !

I'm pleased to share that after putting in a lot of efforts and hard work, I have curated the first high quality and clean audio dataset of Shrimad Bhagavad Gita.

I hope this dataset proves to be helpful to all🌸

Link to Dataset : https://huggingface.co/datasets/JDhruv14/Bhagavad-Gita_Audio

ॐ नमो भगवते वासुदेवाय 🙏🙏

@JDhruv14 amazing brother, keep it up

I'm pleased to share that after putting in a lot of efforts and hard work, I have curated the first high quality and clean audio dataset of Shrimad Bhagavad Gita.

I hope this dataset proves to be helpful to all🌸

Link to Dataset : https://huggingface.co/datasets/JDhruv14/Bhagavad-Gita_Audio

ॐ नमो भगवते वासुदेवाय 🙏🙏

Great man, will check it out.

tried many arenas and LLM battles but couldn't find the best LLM for Indian use cases? try Indic LLM Arena by AI4Bharat (IIT Madras) to find the most suitable LLM for Indian use cases.
link: https://arena.ai4bharat.org/#/chat

Will try !

I'm pleased to share that after putting in a lot of efforts and hard work, I have curated the high quality and clean dataset of Mahabharata.

I hope this dataset proves to be helpful to all🌸

Link to Dataset : https://huggingface.co/datasets/JDhruv14/Mahabharata

I'm pleased to share that after putting in a lot of efforts and hard work, I have curated the high quality and clean dataset of Mahabharata.

I hope this dataset proves to be helpful to all🌸

Link to Dataset : https://huggingface.co/datasets/JDhruv14/Mahabharata

Next stop Ramayan💪

I'm pleased to share that after putting in a lot of efforts and hard work, I have curated the high quality and clean dataset of Mahabharata.

I hope this dataset proves to be helpful to all🌸

Link to Dataset : https://huggingface.co/datasets/JDhruv14/Mahabharata

Great ! Will try it out.

Guys, what do u think, which is better ?

1. Finetuning LLM model for specific data
2. RAG system for the same specific data

What and why ?

Guys, what do u think, which is better ?

1. Finetuning LLM model for specific data
2. RAG system for the same specific data

What and why ?

i believe it depends on the user/client what they want.
Fine-tuning works best with large labeled, consistent data that teaches behavior like format, rules, decisions. A major con is hallucination, in law, healthcare, etc accuracy is everything.
For large factual or knowledge-heavy data, RAG is better because it can scale, stays up to date, hallucinates less and avoids retraining.

tried many arenas and LLM battles but couldn't find the best LLM for Indian use cases? try Indic LLM Arena by AI4Bharat (IIT Madras) to find the most suitable LLM for Indian use cases.
link: https://arena.ai4bharat.org/#/chat

I tried it and it is quite handy. 👍

After investing a lot of days and nights and passionately researching about our Sanatana Dharma, I'm excited to tell that I have over 150k Sanskrit Verses of different books with Translation, Transliteration and Shlokas in my collection.

My collection has around 10 books including Mahabharat, Ramayana, Gita, Markadeye Purana, Devi Mahatmya, Yoga Vasistha and so on.

Please share it with everyone and please upvote me for my work. More to come soon.

Link to Collection : https://huggingface.co/collections/JDhruv14/sanatana-dharma

After investing a lot of days and nights and passionately researching about our Sanatana Dharma, I'm excited to tell that I have over 150k Sanskrit Verses of different books with Translation, Transliteration and Shlokas in my collection.

My collection has around 10 books including Mahabharat, Ramayana, Gita, Markadeye Purana, Devi Mahatmya, Yoga Vasistha and so on.

Please share it with everyone and please upvote me for my work. More to come soon.

Link to Collection : https://huggingface.co/collections/JDhruv14/sanatana-dharma

Amazing brother, this is really important for our Indic LLMs

After investing a lot of days and nights and passionately researching about our Sanatana Dharma, I'm excited to tell that I have over 150k Sanskrit Verses of different books with Translation, Transliteration and Shlokas in my collection.

My collection has around 10 books including Mahabharat, Ramayana, Gita, Markadeye Purana, Devi Mahatmya, Yoga Vasistha and so on.

Please share it with everyone and please upvote me for my work. More to come soon.

Link to Collection : https://huggingface.co/collections/JDhruv14/sanatana-dharma

Amazing brother, this is really important for our Indic LLMs

Thanks man. It means a lot

under Mahabharatha collection, could you please change the 5th column name from translation to english_text ? and 6th column is actually translation which you have named it correctly.
A great work indeed.. thanks a lot. much needed

under Mahabharatha collection, could you please change the 5th column name from translation to english_text ? and 6th column is actually translation which you have named it correctly.
A great work indeed.. thanks a lot. much needed

I remember I have named one column transliteration in each dataset. Also, thank you for your kind words

Very soon my HF account will complete the firsr year. So, will it retain the 100 GB private storage, zero-GPU access, and inference capabilities after completing my first year as a user?

After investing a lot of days and nights and passionately researching about our Sanatana Dharma, I'm excited to tell that I have over 150k Sanskrit Verses of different books with Translation, Transliteration and Shlokas in my collection.

My collection has around 10 books including Mahabharat, Ramayana, Gita, Markadeye Purana, Devi Mahatmya, Yoga Vasistha and so on.

Please share it with everyone and please upvote me for my work. More to come soon.

Link to Collection : https://huggingface.co/collections/JDhruv14/sanatana-dharma

Awesome job bro. You can fine tune models based on this and also publish them on HF referencing the same dataset.
Also publish everything on Kaggle.

Very soon my HF account will complete the firsr year. So, will it retain the 100 GB private storage, zero-GPU access, and inference capabilities after completing my first year as a user?

Don't know, haven't completed 1 year yet.

hey everyone,
read my blog on NPU and openVINO toolkit

https://huggingface.co/blog/Neural-Hacker/openvino

Happy New year