I turned my old rotary phone into a ChatGPT-based home assistant and now you can too

No description available

No description available

No description available

No description available

Implemented GPT with Rotary Embeddings and LSTM from scratch using Pytorch

Implemented a character-level GPT from scratch in PyTorch, architecting and comparing a novel implementation of Rotary Positional Embeddings (RoPE) against a vanilla transformer. Improved sequence generation accuracy from less than 10\% to over 30\% by leveraging RoPE and a full pre-training/fine-tuning pipeline.

Gpt model Implementation from scratch.Then performance enhancement using rotary embedding,group query attention etc.

Educational GPT implementation for exploring and comparing attention mechanisms (vanilla, linear, sparse, sliding window, and rotary) with built-in visualization and analysis tools.

Implementation of GPT-2 small model with 125M parameters along with changes in architecture such as Rotary position embeddings and grouped query attention.

GPT Phone is a Python-based project that simulates a phone call with various fictional characters on an old Bell 304 series rotary telephone. It uses speech recognition, an LLM for reasoning, and text-to-speech.

RLM — Rithan Language Model is a 150 million parameter transformer-based language model built from scratch, designed for general conversation and basic arithmetic inspired form Mr Anderj Karpathy MIcrogpt .architecture draws from modern language model research, incorporating Rotary Positional Embeddings (RoPE) and GPT 2 ecoding for consumer HRD