Metadata-Version: 2.4
Name: krrood_compute_resources
Version: 26.10.0
Summary: The processor cores, memory and GPUs a computation runs on, and how a machine is divided among processes.
Author-email: Tom Schierenbeck <tom_sch@uni-bremen.de>
License-Expression: LGPL-3.0-only
Project-URL: Source, https://github.com/cram2/cognitive_robot_abstract_machine
Project-URL: Bug Tracker, https://github.com/cram2/cognitive_robot_abstract_machine/issues
Project-URL: Changelog, https://github.com/cram2/cognitive_robot_abstract_machine/blob/main/CHANGELOG.md
Keywords: compute resources,memory,gpu,scheduling
Classifier: Programming Language :: Python :: 3
Classifier: Operating System :: POSIX :: Linux
Requires-Python: <3.13,>=3.12
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: krrood>=26.10.0
Requires-Dist: typing_extensions
Provides-Extra: jax
Requires-Dist: jax>=0.6.0; extra == "jax"
Provides-Extra: torch
Requires-Dist: torch>=2.4; extra == "torch"
Provides-Extra: dev
Requires-Dist: pytest<8.0.0,>=7.0.0; extra == "dev"
Dynamic: license-file

# krrood_compute_resources

The processor cores, memory and GPUs a computation runs on, and how a machine is divided
among processes that run side by side.

- `krrood_compute_resources.devices`: `Device`, the kinds of device a computation runs on.
- `krrood_compute_resources.memory`: `DeviceMemory`, the free memory of a device;
  `HostMemory`, the main memory of a Linux host within its control group; and
  `BudgetedMemory`, the share of a device's memory one process may use.
- `krrood_compute_resources.machine`: `Machine`, the cores and memory of the machine a
  process runs on; `MachineDivision`, which divides it into `MachineShare`s; and the
  environment and core affinity that keep a process within its share.
- `krrood_compute_resources.jax_memory`, with the extra `krrood_compute_resources[jax]`:
  `JaxDeviceMemory`, the memory of an accelerator jax stores arrays on.
- `krrood_compute_resources.torch_memory`, with the extra `krrood_compute_resources[torch]`:
  `TorchDeviceMemory`, the memory of a CUDA device torch stores tensors on.
