manish @manish_iitg

AI/ML, Programming, Finance, Coding github.com/manishiitg New Delhi Joined April 2009

Tweets

423
Followers

175
Following

173
Likes

339

Prem Soni @ValueWithPrem

2 weeks ago

I admitted my son to a hospital today. I have a ₹1.2 crore Acko Platinum Health Plan the one with no room rent limit. @ACKOIndia @duavarun Guess what? The hospital flat out denied me a suite room. 👇🏻

1K 4K 17K 2.3M 3K

BURKOV @burkov

6 months ago

A vibe coder, ladies and gentlemen

51 43 858 106K 132

Download Image

Oliver @remoteoliver

7 months ago

Weekends are made up. It’s just another day to do what you love.

3 2 25 1K 0

rohit @knowrohit07

7 months ago

shiva burns ignorance with fire—ai burns data with loss: L = -log p(y|x). each epoch’s a tandava, shredding dumb. happy mahashivratri

2 1 51 1K 2

Download Image

Jyoti Bhatt @JyotiBh22825118

a year ago

Large heap of garbage is accumulating behind Bhutani Alphathum Noida (Near Sector 137). Please get it removed @noida_authority @CeoNoida @myogiadityanath

16 33 10 1K 0

Download Video

Shiv Aroor @ShivAroor

a year ago

And so it begins. The Pune Porsche crash accused, including the underage son, have ALL named the Agarwal family driver. He is being questioned today.

938 2K 7K 1.1M 331

Download Video

Tired of MMLU? The current models already hit the ceiling? It's time to upgrade MMLU! Introducing our new benchmark MMLU-Pro, a more robust and challenging massive multi-task language understanding benchmark with 12K questions. What's New? 1. MMLU-Pro uses 10 options instead of…

46 125 659 173K 295

Download Image

Roflindian @Roflindian

a year ago

The EU has found cancer-causing chemicals in 527 Indian products reaching their borders. Last night, after an hour or two of Internet search, I believe I succeeded in finding out the likely source of that list, and what I saw is really worrying. I will share all my findings in…

338 2K 5K 843K 3K

Download Image

manish @manish_iitg

2 years ago

Releasing aditi synthetic dataset containing high quality multi turn/instruct conversation on hinglish/hindi huggingface.co/datasets/manis… Also open sourcing the data pipeline used to generate these conversations so you can extend the same to your use case github.com/manishiitg/adi…

0 0 1 72 0

manish @manish_iitg

2 years ago

Want to learn how to generate synthetic data for fine-tuning large language models? Check out the airoboros repo: github.com/jondurbin/airo… This awesome open-source tool provides a full framework to create high-quality synthetic datasets tailored to your LLM use case.

0 1 0 78 0

manish @manish_iitg

2 years ago

When fine tuning LLM's what's the best to reduce the size of the dataset to filter out the lower quality document? Should word embedding be used to remove similar documents?

0 0 0 41 0

manish @manish_iitg

2 years ago

Not able to get dpo working at all.. what's the secret? What am I missing

0 0 0 23 0

JJ @JosephJacks_

2 years ago

We’ve backed several of the fastest growing and largest open source founders in India over the last few years via @OSSCapital … Many millions of users. If this does not get reversed, we will reverse our historical position of encouraging founders to stay put and build their…

Bindu Reddy @bindureddy

2 years ago

623 890 5K 1.5M 1K

Download Image

28 36 246 73K 47

manish @manish_iitg

2 years ago

🚀 Exciting News! Introducing Open-Aditi-Hi-v2, our cutting-edge Language Model for Hindi! 🌟 With superior performance over existing models like Airavata and OpenHati, it's perfect for content generation, chatbots, and more. huggingface.co/manishiitg/ope…

1 0 0 44 0

RWKV @RWKV_AI

2 years ago

Introducing Eagle-7B Based on the RWKV-v5 architecture, bringing into opensource space, the strongest - multi-lingual model (beating even mistral) - attention-free transformer today (10-100x+ lower inference) With comparable English performance with the best 1T 7B models