Tensoic

🚀 Hindi / Hinglish CoT datasets made available: Let us know your thoughts!

Vision and AI Lab (VAL), IISc has been recognized as the top AI lab in India by @CSrankings 🥇🎉, reflecting a decade of dedicated research. IISc is also ranked #1 in AI research nationwide 🥇. Thanks to our amazing team for their hard work and commitment🙏 #AI #CV #ML #IISc

🚨🚨 Paper Alert!! 🚨🚨 New #LLMs supporting Indian Languages come out each day, yet no reliable benchmarks exist to evaluate them. Introducing our work: MILU - A Multi-task Indic Language Understanding Benchmark - a comprehensive benchmark for evaluating LLMs for 11 Indian Languages! 🇮🇳 A collaborative effort between @ai4bharat & @IBMResearch India under the AI Alliance @thealliance_ai initiative. 🌏🤝 Paper 📄: arxiv.org/abs/2411.02538 Code 💻: github.com/AI4Bharat/MILU Dataset 🤗: huggingface.co/datasets/ai4bh… [1/N]

🚨 Speaker Alert! 🚨 Join Raghav Ravishankar and Adarsh Arunkumar @tensoic at #Devfest2024Chennai as they dive into multimodality across Indic languages! 🌍 Learn how multimodal AI is breaking new ground in language understanding. 💡 🎟️ Tickets: gdg.community.dev/events/details… #AI #Multimodality #LanguageTech #TechInnovation

@giffmana There were some bugs initially. Had to write a different separator for gemma as the usual llava default did not work. Also @danielhanchen fixed a couple bugs on HF for gemma. But overall the model is great! Kudos to you guys 🥳

@tensoic Yup, i saw it. I actually expected even more to appear after gemma release :) Some nontechnical shenanigans delayed our release a lot :(

I guess we did it first @giffmana 😉 hehe huggingface.co/Tensoic/Cerule…

@XPhyxer1 @reach_vb We were indeed🫡😉

@reach_vb My bros @tensoic were just quicker 😉

PaliGemma - Open Vision Model from Google! 💎 > 3B parameter model - SigLiP + Gemma 2B > Supports images upto 896 x 896 resolution > Capable of Document understanding, Image detection, visual question answering, captioning and more > In addition to general purpose checkpoints they also release specialised models - Diagram understanding, science question answering, COCO captions, etc > Models on the Hub & Integrated with Transformers! 🤗 > Overall 160 checkpoints across JAX, PyTorch (are being released) Good day for GPU Poors! 🔥 - Thank you Google and Big Vision group!

🚨Training code out! We welcome Contributions! Codes are Apache 2.0, Models(Gemma) and dataset subject to original licenses. github.com/tensoic/Cerule

@realmrfakename @ClementDelangue @Google Uhm yeah! The model will adhere to gemma terms of use and the dataset used(LAION and SVIT). But the training codes will be apache 2.0

Cerule - A Tiny Mighty Vision Model Based on @Google's Gemma + SigLIP we release an exciting class of vision models. Multimodal + Multilingual soon? huggingface.co/Tensoic/Cerule…

github.com/Tensoic-AI/Cer… Inference code!

Prompt: What's funny about this image? Cerule: The image is quite humorous as it depicts a man ironing clothes on the back of a yellow taxi cab. This is not a typical sight you'd expect to see in everyday life.

Prompt: Who are the characters in the image? Cerule: The image features three characters, two of them are Mario and Luigi, and the third one is Yoshi.

Gear up for an exciting #AIUnconference! We'll have parallel sessions featuring insightful tech talks, demos, and panel discussions. 📅Date: Sunday, April 28, 2024 🕐Time: 11:00 AM - 5:00 PM 📍Location: Indian Institute of Management, Bangalore Register now: lu.ma/x9zvujud Agenda: 11:00 am: Intro by Partners 11:30 am: AI In Education by @junafinity - HaiVE 12:30 pm: Open / Lunch 01:30 pm: 2 Parallel talks by @banerjee_atreyo - OpenNyAI & Adarsh Shirawalmath - @tensoic 02:30 pm: 2 Parallel talks by @upperwal - @soketlabs & @ravithejads - @llama_index 3:30 pm: 2 Parallel talks by L. Janardhan Rajan - @techmahindracsr & @divyatakart - joyus studio 4:30 pm: Open Floor/ Q&A Event Experience Connect with industry leaders Share knowledge and insights Explore the latest trends in AI Engage in Q&A sessions @iimb_official #AIConference #AI #ArtificialIntelligence #E2E #Tech

@fakjfhhfdaoij Not sure at all. All they had was some news articles covering the release. No trademark or copyright that I could find. It's a single word name of a Greek goddess. Not sure how legal that is.

@tensoic Do they have trademark for covering that use case?

@vspeeeeee @Google yeah! we are 2 college undergrads from India yes

@tensoic @Google Indian company?

Link: huggingface.co/datasets/Tenso… ps: This dataset was generated, please note that some content may not be entirely precise or reflect expert consensus. Users are encouraged to verify information independently for critical purposes. #indic #India #Hindi #LLMs #AI #data

🚨🚀We release "Guftagoo" a Hindi + Hinglish multi-turn conversational dataset comprised of 16k high quality examples⚡️ Guftagoo contains multi-turn conversations on multiple topics usually revolving around daily life experiences, along with COT, and coding. Author: @Adi_kmt

I guess this post is getting views! Please go out and support some amazing Indie and Academic work from India, for India: Sarvam's OpenHaathi: huggingface.co/sarvamai/OpenH… @ravithejads x @ramsri_goutham's Telugu LLM Labs: huggingface.co/Telugu-LLM-Labs @tensoic's KanLlama: huggingface.co/collections/Te… @4evaBehindSOTA and his quest for HumanEval yet general models

@Google's Brand new model - Gemma dropped yesterday! And today we have a full fine-tune of Gemma 2B (7B coming soon!) on @SarvamAI's Samvaad-hi-v1🚀🚀 huggingface.co/Tensoic/Gemma-…