Search Results for "idle"

...The repository contains JSON trace files like train.json, prefill.json, decode.json, and associated assets. Users can load them into tools like Chrome tracing to inspect GPU idle times, overlapping operations, and scheduling alignment. The idea is to bring transparency to internal efficiency tradeoffs, enabling researchers to reproduce, analyze, or improve on DeepSeek’s parallelism strategies. The README explains how trace data corresponds to forward/backward chunks, settings (e.g. EP64, TP1, 4K sequence length), and notes that pipeline communication is excluded for simplicity.

...The system simulates a population of birds that each possess their own neural network, which acts as a decision-making controller during gameplay. The neural network receives input features representing the bird’s position relative to the next obstacle and determines whether the bird should flap or remain idle. Over successive generations, a genetic algorithm evolves the neural networks by selecting high-performing agents and recombining their parameters to produce improved offspring. This process allows the AI agents to gradually learn better strategies for navigating the obstacles and surviving longer in the game environment.