[Platform] Replace torch.cuda.mem_get_info with torch.accelerator.get_memory_info (#44825)

Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
Signed-off-by: Kunshang Ji <jikunshang95@gmail.com>
This commit is contained in:
Kunshang Ji
2026-06-30 14:39:52 +08:00
committed by GitHub
parent 0feca7ffa8
commit fb42e5219e
17 changed files with 56 additions and 52 deletions
+9
View File
@@ -38,6 +38,14 @@ class _StreamPlaceholder:
pass
from vllm.utils.cpu_resource_utils import get_memory_node_info
def get_memory_info(*args: Any, **kwargs: Any) -> tuple[int, int]:
meminfo = get_memory_node_info()
return meminfo.available_memory, meminfo.total_memory
torch.Event = _EventPlaceholder
torch.cuda.Event = _EventPlaceholder
torch.cuda.Stream = _StreamPlaceholder
@@ -46,6 +54,7 @@ torch.cuda.current_stream = lambda *args, **kwargs: _StreamPlaceholder()
torch.accelerator.synchronize = noop
torch.accelerator.empty_cache = noop
torch.Tensor.pin_memory = fake_pin_memory
torch.accelerator.get_memory_info = get_memory_info
# Patch vLLM torch utils
import vllm.utils.torch_utils as torch_utils