[model-gateway] add GPU allocator and model pool infrastructure for parallel E2E tests (#16460)
This commit is contained in:
@@ -0,0 +1,31 @@
|
||||
"""Infrastructure for parallel GPU test execution."""
|
||||
|
||||
from .gpu_allocator import (
|
||||
GPUAllocator,
|
||||
GPUInfo,
|
||||
GPUSlot,
|
||||
get_gpu_memory_usage,
|
||||
get_open_port,
|
||||
get_physical_device_indices,
|
||||
nvml_context,
|
||||
wait_for_gpu_memory_to_clear,
|
||||
)
|
||||
from .model_pool import ModelInstance, ModelPool
|
||||
from .model_specs import MODEL_SPECS
|
||||
|
||||
__all__ = [
|
||||
# GPU allocation
|
||||
"GPUAllocator",
|
||||
"GPUInfo",
|
||||
"GPUSlot",
|
||||
# GPU utilities
|
||||
"nvml_context",
|
||||
"get_open_port",
|
||||
"get_physical_device_indices",
|
||||
"get_gpu_memory_usage",
|
||||
"wait_for_gpu_memory_to_clear",
|
||||
# Model management
|
||||
"ModelInstance",
|
||||
"ModelPool",
|
||||
"MODEL_SPECS",
|
||||
]
|
||||
Reference in New Issue
Block a user