Vatsadev

Working on Robotics/ML/CV, 3D mode

Articles

• Getting better models by dropping floating point sizes and increasing parameters
• The effects of Grad_Acc and Batch_Size
• Whats better: Neural nets wider with less layers or thinner with more layers
• Monkey thot compression: musings on the neuralink challenge
• Hallucinations and Memory leaks: Transformers
• Transformers learn patterns, math is patterns
• NanoPhi: replicating parts of Phi-1.5

Threads/Posts

• FRC 2024 Crescendo (...) (...) (...)
• Lilith (...) (...) (...)
• Clip Similarity and Sora (...) (...)
• TinyGrad dependencies (...)

Resources

• CSHS: Notes on competitions we've been to
• How a neuron learns: Linear layers
• Thoughts on FDVR