Summary
The user is exploring the Flyte Sandbox for executing parallel tasks that require 1 GPU and 8 cores each, given their resources of 2 GPUs and 16 cores. They are unsure if the sandbox can handle concurrent job execution or if a real Kubernetes installation is needed for effective orchestration. The user appreciates Flyte's support for parallel executions and the use of ArrayNode for concurrent tasks. They are considering using Resources(gpu=1) in a task to run two jobs concurrently in the sandbox with map_task(). The user seeks clarification on the sandbox's ability to run tasks concurrently and whether tasks in the same container can access different GPUs. They acknowledge the sandbox operates on k3s and confirm its capability to execute all Flyte features, including map_tasks. The user finds the sandbox's start-up scripts and overall implementation helpful, expressing gratitude for the information provided.