- 187
- 75 217
Tensordroid
India
เข้าร่วมเมื่อ 13 ส.ค. 2014
I am a Machine Learning Engineer who creates videos about Machine Learning and Generative AI. I strive to keep you updated on the latest ML developments and guide you on becoming an ML Developer."
Not all Attention is Needed in Transformers ?!
Paper Link: arxiv.org/abs/2406.15786
My Links 🔗
👉🏻 Subscribe: youtube.com/@Tensordroid
👉🏻 Twitter: vishesh_t27
👉🏻 LinkedIn: www.linkedin.com/in/vishesh-tripathi/
My Links 🔗
👉🏻 Subscribe: youtube.com/@Tensordroid
👉🏻 Twitter: vishesh_t27
👉🏻 LinkedIn: www.linkedin.com/in/vishesh-tripathi/
มุมมอง: 0
วีดีโอ
Understanding Cross Entropy and Perplexity
มุมมอง 3312 ชั่วโมงที่ผ่านมา
My Links 🔗 👉🏻 Subscribe: youtube.com/@Tensordroid 👉🏻 Twitter: vishesh_t27 👉🏻 LinkedIn: www.linkedin.com/in/vishesh-tripathi/
But what is Differential Transformer ?
มุมมอง 6519 ชั่วโมงที่ผ่านมา
Paper Link: arxiv.org/abs/2410.02703 My Links 🔗 👉🏻 Subscribe: youtube.com/@Tensordroid 👉🏻 Twitter: vishesh_t27 👉🏻 LinkedIn: www.linkedin.com/in/vishesh-tripathi/
But what is selective Attention ?
มุมมอง 33วันที่ผ่านมา
Hey guys, sorry for the horizontal view, was trying something and realised it during final editing. Paper link: arxiv.org/abs/2410.02703 My Links 🔗 👉🏻 Subscribe: youtube.com/@Tensordroid 👉🏻 Twitter: vishesh_t27 👉🏻 LinkedIn: www.linkedin.com/in/vishesh-tripathi/
Google presents Astute RAG !!
มุมมอง 14614 วันที่ผ่านมา
Paper Link: arxiv.org/abs/2410.07176 My Links 🔗 👉🏻 Subscribe: youtube.com/@Tensordroid 👉🏻 Twitter: vishesh_t27 👉🏻 LinkedIn: www.linkedin.com/in/vishesh-tripathi/
ColPali: Indexing Documents in RAG made easy using Vision Language Models !!
มุมมอง 18914 วันที่ผ่านมา
Paper Link: arxiv.org/abs/2407.01449 Blog: huggingface.co/blog/manu/colpali My Links 🔗 👉🏻 Subscribe: youtube.com/@Tensordroid 👉🏻 Twitter: vishesh_t27 👉🏻 LinkedIn: www.linkedin.com/in/vishesh-tripathi/
Best Paper for Retrieval Augmented Generation Pain Points !!
มุมมอง 85หลายเดือนก่อน
Retrieval Augmented Generation (RAG) and Beyond: A Comprehensive Survey on How to Make your LLMs use External Data More Wisely: arxiv.org/pdf/2409.14924 Searching for Best Practices in Retrieval-Augmented Generation: arxiv.org/pdf/2407.01219 My Links 🔗 👉🏻 Subscribe: youtube.com/@Tensordroid 👉🏻 Twitter: vishesh_t27 👉🏻 LinkedIn: www.linkedin.com/in/vishesh-tripathi/
But OpenAI o1 is here !!
มุมมอง 260หลายเดือนก่อน
OpenAI o1: openai.com/index/learning-to-reason-with-llms/ My Links 🔗 👉🏻 Subscribe: youtube.com/@Tensordroid 👉🏻 Twitter: vishesh_t27 👉🏻 LinkedIn: www.linkedin.com/in/vishesh-tripathi/
Smart India Hackathon Guide 2024
มุมมอง 2062 หลายเดือนก่อน
Smart India Hackathon Website: www.sih.gov.in/ My Links 🔗 👉🏻 Subscribe: youtube.com/@Tensordroid 👉🏻 Twitter: vishesh_t27 👉🏻 LinkedIn: www.linkedin.com/in/vishesh-tripathi/
But what is DeepSpeed ? DeepSpeed vs VLLM
มุมมอง 1522 หลายเดือนก่อน
Research paper: arxiv.org/abs/2401.08671 My Links 🔗 👉🏻 Subscribe: youtube.com/@Tensordroid 👉🏻 Twitter: vishesh_t27 👉🏻 LinkedIn: www.linkedin.com/in/vishesh-tripathi/
Preparing for Machine Learning Interviews in 2024
มุมมอง 653 หลายเดือนก่อน
Are you gearing up for your first Machine Learning interview and wondering where to start? In this video, we'll guide you through the essential steps to ace your ML interview as a fresher. Whether you're just starting your journey or looking to refine your skills, this video covers everything you need to know. Unstop Mentorship: unstop.com/mentor/vishesh?ref=kzVULcm TopMate Mentorship: topmate....
But GPT-4o-mini is here !!
มุมมอง 373 หลายเดือนก่อน
GPT-4o-mini: openai.com/index/gpt-4o-mini-advancing-cost-efficient-intelligence/ Leaderboard: chat.lmsys.org/?leaderboard Post on X: x.com/lmsysorg/status/1813999088758673875 My Links 🔗 👉🏻 Subscribe: youtube.com/@Tensordroid 👉🏻 Twitter: vishesh_t27 👉🏻 LinkedIn: www.linkedin.com/in/vishesh-tripathi/
Weekly ML News Episode - 8
มุมมอง 423 หลายเดือนก่อน
AI API Analysis: artificialanalysis.ai/ Groq: Quickstart: console.groq.com/docs/quickstart Groqbook: github.com/Bklieger/groqbook GroqNotes: github.com/Bklieger/groqnotes Research Papers: Learning to (Learn at Test Time): RNNs with Expressive Hidden States: arxiv.org/abs/2407.04620 Searching for Best Practices in Retrieval-Augmented Generation: arxiv.org/pdf/2407.01219 RouteLLM: Learning to Rou...
Are Large Language Models really learning something or Not ?
มุมมอง 663 หลายเดือนก่อน
In this video, we dive into the paper "When Benchmarks are Targets: Revealing the Sensitivity of Large Language Model Leaderboards." We'll explore how minor changes in benchmark setups can drastically alter the rankings of LLMs, why this happens, and what best practices can be implemented for more robust evaluations. Don't miss out on understanding the pitfalls of relying solely on leaderboard ...
But Google's Gemma-2 is here !!
มุมมอง 484 หลายเดือนก่อน
Gemma-2 Research Paper: storage.googleapis.com/deepmind-media/gemma/gemma-2-report.pdf Gemma-2 Blog: blog.google/technology/developers/google-gemma-2/ Google's Gemma HugigngFace: huggingface.co/google/gemma-2-9b-it Try Gemma on HuggingFace Chat: huggingface.co/chat My Links 🔗 👉🏻 Subscribe: youtube.com/@Tensordroid 👉🏻 Twitter: vishesh_t27 👉🏻 LinkedIn: www.linkedin.com/in/vishesh-trip...
Training and adding new tokens in a Pre-trained Tokenizer !!
มุมมอง 2154 หลายเดือนก่อน
Training and adding new tokens in a Pre-trained Tokenizer !!
Weekly Machine Learning News Episode - 5
มุมมอง 475 หลายเดือนก่อน
Weekly Machine Learning News Episode - 5
Getting a Machine Learning Job/Internship in 2024 !!
มุมมอง 5915 หลายเดือนก่อน
Getting a Machine Learning Job/Internship in 2024 !!
All AI Updates from Google I/O 2024 🤯
มุมมอง 1545 หลายเดือนก่อน
All AI Updates from Google I/O 2024 🤯
Weekly Machine Learning News Episode - 4
มุมมอง 315 หลายเดือนก่อน
Weekly Machine Learning News Episode - 4
Key Value Cache in Large Language Models Explained
มุมมอง 2K5 หลายเดือนก่อน
Key Value Cache in Large Language Models Explained
Machine Learning Engineer Roadmap 2024 !!
มุมมอง 8825 หลายเดือนก่อน
Machine Learning Engineer Roadmap 2024 !!
Weekly Machine Learning News Episode - 3
มุมมอง 745 หลายเดือนก่อน
Weekly Machine Learning News Episode - 3
Hi, Great video, i am trying to build a model that will do OCR for driving license and rc book using PALI, how do you think i should approach this model
You are basically reading from the slides, it could have been much better if you could have refreshed/brushed upon the concept of paging and then explained how paged attention works
Straight to the point and insightful
please provide link for transformers video
where is paper published to?
This was informative 💯
Thanks !
great achievement thank you
Hey Bro Is there any Chance of increase in No. of Submission(Right Now It's only 2)
I did the same for 22 Indian languages. But when I searched a kannada language character in the tokens for a test purpose, it was not showing anything. Also, tokenizer separates punctuation as well. Your method of splitting is not optimal.
whats the minimum spec to run this??
Atleast 2 A10
thanks for the video, from an inferencing hardware point of view, it would be good to see basic results comparing performance on a Multi-GPU system with NVLINK and without, meaning the GPUs using the PCIe bus on the host system
Is kv cache in every LLM? How about the small models
Some quality content here. So the new tokens just get appended to the current tokenizer right?
Yes
awesome presentation and make sense .❤
Glad you liked it
Very well explained !!! Good work 👍 Keep it up 🤝
Thanks !!
from systemxpert???
Can I learn java for ML ?
Java is good for DSA, but for ML, you need to learn ML
can you give the ppt is shown in the video please?
here you go: docs.google.com/presentation/d/1hfK_k0jotCNwZWMDuHsbeL3R4r5q7hCwQXsJWbE3o6k/edit?usp=sharing
Thanks for sharing:)
Very good research and well explained 🙌
Thank you so much !!
dsa in python or java. Coz I saw your leetcode and majority of the problems are solved using java
it's a choice, I did in java, because DSA is always better in C++ or Java. but python also works, as I have shifted to work in python full time, I am going to do DSA in python only
Nice one, Vishesh
Thanks
good shit bro
thanks
thanks g, you the best.
👏👏
🎉
What about cp and can we do dsa in python or it is modatory to do in cpp or java
Yes python me bhi DSA ho sakti hai par resources kam hai
always a goddamn indian
Volume is abit low
Sorry, some mic issue
do you earn enough to pay back the loan for vit? im not being condescending , im okay and happy with a very little disposable income but if u earn really good that is just cherry on top
I actually did not take any loan, my income I would say is descent, I am able to afford stuff I want rn
@@Tensordroid but its fees is so high ?
Sorry for asking, at 1 year of distance. But if I want to detect something outside the YAMnet labels, like creating a new label, then YAMnet is not whorty any more right? Or is it whorty anyway for the power of the whole dataset, and training it with new audio to create new label would be a good idea? And then how to do it in a proper way?
No problem ! ummm, if you want to detect something outside YAMNet's existing labels, you can still use YAMNet for its powerful pre-trained features. So actually you can use transfer learning: extract features from your new audio data using YAMNet, then train a new model on these features with your custom labels. this approach will help you from YAMNet's dataset while adapting it to your specific needs.
Can you share more respurces for maths
You can also study from khan-academy Rest Ig these 2 are more than enough for ML
Skarparised you bro ❤ 6:56
Hi bhaiya I am going to join lnct bhopal btech ai ml this year . Please can you help me I am really really confused abt my clg but as much as I researched it is the only decent clg I am getting according to my rank in jee mains (89.35 %ile,1.59lk ). Q1. The doubt in my mind is will the faculties be good ? Q2. Will I get any good job ? (my goal is to learn ml ,neural networking, how to teach mechine through data ,how to build networks ,etc stuff in detail) Q3. Will I need to do mtech? From where? For Which exam I have to prepare for? Q4. What skill should I start learning ? (I know little bit of python like class 12th level I had created a billing program as my cs boards practical project using python and mysql In it you can login ,create productl ist,quantity, edit quantity , add ,remove(only admin can other staff can create bill only) , add user info , add his list,he will get discount on basis of his visit frequency ,it also provide demo data for user to examin software ) Q5. Should I start learning c++ or first complete python? Q6.What skills should I start learning? Q7. From where should I learn skills like coding, communication skills (I am horrible at it,also lnct clg do not have much activity ), etc? Q8. is possible to improve skill while maintaining 75% attendance ?(I do not participate in much events,trips I can spare time from it) . Q9. How helpful are clg lectures? And much more advice from you please help.
Hey brother please reply i need help
Nice Video, I'll start folliwing this 🌟
Thanks
the volume is very low, can’t hear clearly
Sorry, there was some mic issue
Insightful
Thanks !
🙌
Amazing
Thanks
What hardware do you need to run this locally?
You will need at-least 32 GB space, it you use torch.bfloat16, then 16 GB is required
@@Tensordroid Is this is System RAM or GPU VRAM?
@@novmikvis So you can set it right, on which you want to se using torch device, else device_map="auto", but the speed you are going to get at GPU is no where close to System RAM, you can use VLLM also for inferencing faster if you have multiple GPUs
@@Tensordroid Great! Thanks for clarifying that.
Amazing ❤
Thanks 😄
Can it help me in Data Science field? How?
It can help you get an interview atleast, because it is proof that you are doing good in tech stacks and hackathons
👍
Thank you
Is there any internship opportunity or can we request the HR there to refer us?
any way to make a website out of it?
Well done 👍
Congratulations 🎉
bhaiya is there any platform I can contact you like discord , I want to ask something about one project I am working on
BRO PLEASE TAKE A BIT BIG SIZE OF DATASET, THEN PERFORM THIS MODEL