mirror of
https://github.com/vale981/ray
synced 2025-03-05 10:01:43 -05:00
513881584d
There are mysterious memory usage growth in Ray clusters that disappear when running with jemalloc. Before we are able to figure out the root cause, it seems using jemalloc by default can be a good walkaround. Because of its efficiency, using jemalloc by default can be beneficial, but we need to run more benchmarks to verify. |
||
|---|---|---|
| .. | ||
| autoscaler | ||
| base-deps | ||
| development | ||
| examples | ||
| ray | ||
| ray-deps | ||
| ray-ml | ||
| ray-worker-container | ||
| retag-lambda | ||
| fix-docker-latest.sh | ||
| README.md | ||
Overview of how the ray images are built:
Images without a "-cpu" or "-gpu" tag are built on ubuntu/focal. They are just an alias for -cpu (e.g. ray:latest is the same as ray:latest-cpu).
ubuntu/focal
└── base-deps:cpu
└── ray-deps:cpu
└── ray:cpu
└── ray-ml:cpu
nvidia/cuda
└── base-deps:gpu
└── ray-deps:gpu
└── ray:gpu
└── ray-ml:gpu