Skip to content

Show NPU memory in winml sys #1425

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
xieofxie merged 1 commit into from
Sep 17, 2026
+32 −26
Merged

Show NPU memory in winml sys #1425

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 2 additions & 1 deletion docs/commands/sys.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@ display version, build, update build revision (UBR), build branch, and build lab
It then probes PyTorch for CUDA availability and GPU device names.
Backend availability checks use the installed runtime environment. GPU and NPU
enumeration uses DXCore as the source of adapter identity and LUID, then enriches
those native rows with WMI/PnP driver and manufacturer details. GPU rows also
those native rows with WMI/PnP driver and manufacturer details. NPU and GPU rows also
report DXCore's 64-bit dedicated adapter memory and shared system memory
capacities in MiB. CPU enumeration uses WMI. Devices remain in NPU > GPU > CPU
priority order, and EP enumeration
Expand Down Expand Up @@ -91,6 +91,7 @@ ML Libraries
Available Devices (priority order)
#1 NPU Qualcomm(R) Hexagon NPU
LUID: 0x00000000_0x00018393 | Driver: 1.0.0 | Manufacturer: Qualcomm
Dedicated memory: 0 MiB | Shared memory: 8192 MiB
#2 GPU Qualcomm(R) Adreno GPU
LUID: 0x00000000_0x00018394 | Driver: 1.0.0 | Manufacturer: Qualcomm
Dedicated memory: 1024 MiB | Shared memory: 8192 MiB
Expand Down
21 changes: 11 additions & 10 deletions src/winml/modelkit/commands/sys.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -693,17 +693,18 @@ def _gather_device_info(
"manufacturer": getattr(metadata_item, "manufacturer", None),
"luid": item.luid if system_adapters else None,
}
if device_label == "GPU":
if system_adapters:
entry["details"].update(
{
"dedicated_memory_mib": (
item.dedicated_memory_mib
if system_adapters
else getattr(item, "vram_mib", None)
),
"shared_memory_mib": (
item.shared_memory_mib if system_adapters else None
),
"dedicated_memory_mib": item.dedicated_memory_mib,
"shared_memory_mib": item.shared_memory_mib,
}
)
elif device_label == "GPU":
entry["details"].update(
{
"dedicated_memory_mib": getattr(item, "vram_mib", None),
"shared_memory_mib": None,
}
)
elif device_label == "CPU":
Expand Down Expand Up @@ -847,7 +848,7 @@ def _output_device_text(devices: list[dict[str, Any]]) -> None:
if arch := details.get("architecture"):
parts.append(f"Architecture: {arch}")
console.print(f" {' | '.join(parts)}")
if dev["type"] == "GPU":
if dev["type"] in ("NPU", "GPU"):
dedicated_memory = details.get("dedicated_memory_mib")
shared_memory = details.get("shared_memory_mib")
parts = [
Expand Down
4 changes: 2 additions & 2 deletions tests/e2e/test_sys_e2e.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -199,10 +199,10 @@ def test_default_json_devices_include_luid(self):
r"0x[0-9A-F]{8}_0x[0-9A-F]{8}", luid
)

def test_default_json_gpus_include_memory(self):
def test_default_json_accelerators_include_memory(self):
data = _run_sys_json()
for device in data["devices"]:
if device["type"] != "GPU":
if device["type"] not in ("NPU", "GPU"):
continue
details = device["details"]
for field in ("dedicated_memory_mib", "shared_memory_mib"):
Expand Down
30 changes: 17 additions & 13 deletions tests/unit/commands/test_sys_device_info.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -252,39 +252,41 @@ def test_dxcore_is_primary_for_accelerator_identity(

def test_dxcore_accelerator_survives_wmi_failure(self, monkeypatch: pytest.MonkeyPatch) -> None:
"""Native identity remains visible when descriptive enrichment fails."""
native_gpu = DXCoreAdapterInfo(
device_type="GPU",
name="Example GPU",
native_npu = DXCoreAdapterInfo(
device_type="NPU",
name="Example NPU",
luid="0x00000000_0x00000001",
vendor_id=0x1234,
device_id=0x5678,
dedicated_memory_mib=0,
shared_memory_mib=16384,
)
monkeypatch.setattr(
"winml.modelkit.sysinfo.enumerate_compute_adapters",
lambda: [native_gpu],
lambda: [native_npu],
)

with (
patch("winml.modelkit.sysinfo.NPU.get_all", return_value=[]),
patch(
"winml.modelkit.sysinfo.GPU.get_all",
"winml.modelkit.sysinfo.NPU.get_all",
side_effect=RuntimeError("WMI unavailable"),
),
patch("winml.modelkit.sysinfo.GPU.get_all", return_value=[]),
patch("winml.modelkit.sysinfo.CPU.get_all", return_value=[]),
):
result = _gather_device_info()

assert result == [
{
"priority": 1,
"type": "GPU",
"name": "Example GPU",
"type": "NPU",
"name": "Example NPU",
"details": {
"driver": None,
"manufacturer": None,
"luid": "0x00000000_0x00000001",
"dedicated_memory_mib": None,
"shared_memory_mib": None,
"dedicated_memory_mib": 0,
"shared_memory_mib": 16384,
},
}
]
Expand Down Expand Up @@ -417,7 +419,9 @@ def test_compact_device_output_includes_luid(
assert "CPU: Test CPU (LUID: N/A)" in output


def test_gpu_text_output_includes_dedicated_and_shared_memory(
@pytest.mark.parametrize("device_type", ["NPU", "GPU"])
def test_accelerator_text_output_includes_dedicated_and_shared_memory(
device_type: str,
capsys: pytest.CaptureFixture[str],
) -> None:
from winml.modelkit.commands.sys import _output_device_text
Expand All @@ -426,8 +430,8 @@ def test_gpu_text_output_includes_dedicated_and_shared_memory(
[
{
"priority": 1,
"type": "GPU",
"name": "Test GPU",
"type": device_type,
"name": f"Test {device_type}",
"details": {
"luid": "0x00000000_0x00018393",
"driver": "1.0",
Expand Down