Nitish Reddy (@nitishreddy116) 's Twitter Profile
Nitish Reddy

@nitishreddy116

Co-Founder @alphabakeIO ( AI Visualisation for Apparel) /
IIT Bombay, Elec 2008-12 /
Previously WorldQuant

ID: 345521823

linkhttp://tri3d.in calendar_today30-07-2011 18:52:35

774 Tweet

112 Takipçi

798 Takip Edilen

Anshuman Mishra (@heyyanshuman) 's Twitter Profile Photo

You're in a ML Engineer interview at Groq, and the interviewer asks: "How do you measure LLM inference performance? What metrics matter most for production systems?" Here's how you can answer

Sayak Paul (@risingsayak) 's Twitter Profile Photo

Flux.2 said that perceptual and semantic qualities in a representation space can be conflicting. Optimizing both can be difficult, and they studied it beautifully. And then comes: The Prism Hypothesis: Harmonizing Semantic and Pixel Representations via Unified Autoencoding

Flux.2 said that perceptual and semantic qualities in a representation space can be conflicting. 

Optimizing both can be difficult, and they studied it beautifully.

And then  comes:

The Prism Hypothesis: Harmonizing Semantic and Pixel Representations via Unified Autoencoding
⚡AI Search⚡ (@aisearchio) 's Twitter Profile Photo

This AI allows you to stream as anyone in realtime. Free & open source, works with regular GPU. It was a pain to install but I eventually got it to work. Full tutorial: youtu.be/fzgwKJzAcVs

Gauri Tripathi (@gauri_the_great) 's Twitter Profile Photo

Let's get into the depth of why Let's take Qwen 2.5 (32B parameters) running in BF16 on a single NVIDIA A100 80 GB (SXM) GPU. --> Parameters: 32B --> BF16 storage: 32×2=64 ; 64 GB --> GPU memory: 80 GB --> Free memory: 80−64=16 GB perfect right ? No 1. There are two major

Let's get into the depth of why

Let's take Qwen 2.5 (32B parameters) running in BF16 on a single NVIDIA A100 80 GB (SXM) GPU.

 --> Parameters: 32B
--> BF16 storage:  32×2=64 ; 64 GB
--> GPU memory: 80 GB
--> Free memory: 80−64=16 GB
perfect right ?
No 

1.  There are two major
⚡AI Search⚡ (@aisearchio) 's Twitter Profile Photo

This AI can make 1080p videos in seconds. HiStream speeds up Wan video generations by 107x 🤯 Code under review haonanqiu.com/projects/HiStr…

⚡AI Search⚡ (@aisearchio) 's Twitter Profile Photo

Finally! Long continuous AI video is here Free & open source. Infinite length Prompts: she looks at the camera and smiles she drinks the glass of water she stands up and walks away Made with SVI 2.0 Pro Full tutorial: youtu.be/-3DVJu72VhE

Pau Labarta Bajo (@paulabartabajo_) 's Twitter Profile Photo

Real-time audio transcription. Entirely on-device. This hands-on tutorial shows you how to build it from scratch. No cloud dependencies, no API calls, complete privacy, using LFM2-Audio-1.5B by Liquid AI Enjoy it ↓ github.com/Liquid4All/coo…

Rohan Paul (@rohanpaul_ai) 's Twitter Profile Photo

The core concept of DeepSeek's Engram is just so beautiful. A typical AI model treats “knowing” like a workout. Even for basic facts, it flexes a bunch of compute, like someone who has to do a complicated mental trick just to spell their own name each time. The reason is that

The core concept of DeepSeek's Engram is just so beautiful.

A typical AI model treats “knowing” like a workout. Even for basic facts, it flexes a bunch of compute, like someone who has to do a complicated mental trick just to spell their own name each time. 

The reason is that
Matt Workman (@cinedatabase) 's Twitter Profile Photo

I made tutorial on the "180 Degree Rule" for AI Filmmaking when creating dialogue scenes. Watch here: youtu.be/0VxTJYZ2Hcc?si…

Wildminder (@wildmindai) 's Twitter Profile Photo

No Python, just pure wild C: FLUX.2-klein-4B can now run on systems with just 8GB of RAM; - zero-dependency C lib, uses memory-mapped weights to slash peak memory from 16GB to 4GB; - matches PyTorch speed; 1024x1024 image in 29s. github.com/antirez/flux2.c

No Python, just pure wild C: FLUX.2-klein-4B can now run on systems with just 8GB of RAM; 
- zero-dependency C lib, uses memory-mapped weights to slash peak memory from 16GB to 4GB; 
- matches PyTorch speed; 1024x1024 image in 29s.
github.com/antirez/flux2.c
Oliver Cameron (@olivercameron) 's Twitter Profile Photo

This simulation (generated by Odyssey-2 Pro in real-time) felt profound to me. It's demonstrating the model is learning early cause-and-effect. When pressure from hands is applied to wet clay, the clay is moulded. You can now integrate this model into your app with our API!

Nitish Reddy (@nitishreddy116) 's Twitter Profile Photo

being numb vs being mindful which makes you happier on average ? I swam with my old goggles, hardly visible, aware but was okay with it, I have -5 myopia by mistake I picked my wife's goggles one day and realised it's super clear.