RSSAmplifier

Blog

Welcome! :tada: on Shital Shah's Chain of Thought

Recent content in Welcome! :tada: on Shital Shah's Chain of Thought

shital.com ↗RSS feed ↗916 posts

Latest posts

Releasing StreamHope - Merge Duplicate Playlists in YouTube Music, transfer from Spotify

How to Get and Interpret GPU Memory Profiling

o3-mini Level Hard Problem

DeepSeek's Recipe Evolution

Can You Spot Error Proving 1 = -1?

Phi-4 Now On HuggingFace with MIT Licence!

Phi-4: The Most Powerful smol Model!

Announcement: PhD Internship for AI for Reasoning

Some Hand Wavy Estimates on FLOPs and Computational Reducibility

Erdős Open Problems as ASI Benchmark

Tokenization and Decoding

Scaling Laws and Data Wall

OpenCoder

The WSD Learning Rate Schedule

Healthier Flossing with Silk!

Tricky Validation Plot

DeepMind's Chess Model

Solving Chicken-Egg Dilemma

New Optimizer: Muon

Vocab Scaling

Using Cursor

Terence Tao Grades o1

Things on Big Bang Theory Writer's Table

Inference Scaling

IMO 2024 Test

LLM Judges Prefers Fluff?

Announcement: Phi-3 14B Rivaling Llama3 70B

Power of Inference Scaling

Human Hours to FLOPs

Sample Efficiency

Phi-3: "ChatGPT" Natively on Your Phone

Inference Memory

Llama3 Leaps Ahead with Data Alone

Pretraining has Gotten 10x Faster in Past Year!

Great Papers

Marvels of ENIAC

Intelligence and FLOPs Efficiency

CIFAR-10 Training to 94% Accuracy in 3.29s

Matmul on CPUs

Using rusage

When Uniform Isn't Uniform: A Bug Hunt Tale

GPT-4 Training Time

Scarce to Plentiful

A Lifetime in 1T Tokens

Power of Two Choices

PyTorch Distributed Debugging Tips

EMA Tuning

Phixtral Fusion

Phi-2: The Most Powerful 2.7B Model Yet

Amazing Robotics Demo with Cheap Hardware