Blogs

DeepCoder is an open-source AI model designed for high-performance code generation, leveraging 14 billion parameters. Developed by UC Berkeley's Agentica team and Together AI, it is fine-tuned using Reinforcement Learning (RL) techniques and provides an innovative approach to training coding models. Training & Data Details DeepCoder was trained over 2.5 weeks on 32 Nvidia A100 GPUs, utilizing a curated dataset of 24,000 verified coding problems. This dataset includes: 7.5K pr