trinity.perf.resource_backends module

System resource collection backends for performance tooling.

class trinity.perf.resource_backends.GPUSample(gpu_id: int, name: str, gpu_util_percent: float, gpu_memory_used_mb: float, gpu_memory_total_mb: float)

Bases: object

One GPU sample at one point in time.

gpu_id: int

name: str

gpu_util_percent: float

gpu_memory_used_mb: float

gpu_memory_total_mb: float

to_dict() → dict

Serialize the GPU sample to a dictionary.

__init__(gpu_id: int, name: str, gpu_util_percent: float, gpu_memory_used_mb: float, gpu_memory_total_mb: float) → None

class trinity.perf.resource_backends.ResourceSample(timestamp: float, cpu_percent: float, memory_rss_mb: float, memory_percent: float, gpu_metrics: list[GPUSample])

Bases: object

One system resource sample at one point in time.

timestamp: float

cpu_percent: float

memory_rss_mb: float

memory_percent: float

gpu_metrics: list[GPUSample]

to_dict() → dict

Serialize the resource sample to a dictionary.

__init__(timestamp: float, cpu_percent: float, memory_rss_mb: float, memory_percent: float, gpu_metrics: list[GPUSample]) → None

class trinity.perf.resource_backends.SystemResourceBackend(gpu_subsample_count: int = 5, gpu_subsample_interval_seconds: float = 0.2)

Bases: object

Collect system-level CPU, memory and per-GPU metrics.

__init__(gpu_subsample_count: int = 5, gpu_subsample_interval_seconds: float = 0.2) → None

open() → None

Initialize the GPU management library and validate the environment.

close() → None

Release NVML resources.

sample() → ResourceSample

Collect one resource sample.