Krishna Mohan

Flash Attention 1 forward pass Cuda kernel

Solved PDE using PINNs

Time for LLM inference

IOU -> Intersection Over Union NMS -> Non Maximum Suppression mAP -> mean Average Precision

Object detection is a tough task

How's the GPT 5 hype?

Same thoughts 👍

C++ >>>> python

My prof when I show him my assignment

You are one prompt away from the answer

When I can't get it, let's build it from scratch

Wrote a new post about Flow Matching, where I try to explain my intuitions about it as concisely as possible. Hope it's useful :)

what should I name this

Thanks @GoogleIndia Got it 👍

ok I need this

Kimi K2 is basically DeepSeek V3 but with fewer heads and more experts:

🚀 Hello, Kimi K2! Open-Source Agentic Model! 🔹 1T total / 32B active MoE model 🔹 SOTA on SWE Bench Verified, Tau2 & AceBench among open models 🔹Strong in coding and agentic tasks 🐤 Multimodal & thought-mode not supported for now With Kimi K2, advanced agentic intelligence…

I think there is a model named llama 4 behemoth Where is it ? Still training...

an unsung hero of the Transformer's success is not the attention mechanism per se, but *causal masking* it let us simplify two for loops to just one: > RNN for t in [1..S]: for i in [1..L]: x[t], h = layers[i](x[t], h) > transformer for i in [1..L]: x = layers[i](x)

T5Gemma is here

There is so much to learn, if only I had enough time