Skip to content

Personal AI, On Personal Devices

Python

test: update test_all_specs_present with 7 new GPUs

Abhinav Cherukuru9c9c883f24ac
diff --git a/tests/telemetry/test_gpu_monitor.py b/tests/telemetry/test_gpu_monitor.py
index e4b0f1dc..40aa2b1a 100644
--- a/tests/telemetry/test_gpu_monitor.py
+++ b/tests/telemetry/test_gpu_monitor.py
@@ -10,11 +10,11 @@ from unittest.mock import MagicMock, patch
import pytest
+
# ---------------------------------------------------------------------------
# Helpers: build a fake pynvml module
# ---------------------------------------------------------------------------
-
@dataclass
class _FakeUtilization:
gpu: int = 80
@@ -60,7 +60,6 @@ def _snap(power=300, util=80, mem=12, temp=65, dev=0):
# Tests: GpuHardwareSpec lookup
# ---------------------------------------------------------------------------
-
class TestGpuHardwareSpec:
def test_lookup_exact_key(self):
from openjarvis.telemetry.gpu_monitor import lookup_gpu_spec
@@ -107,6 +106,13 @@ class TestGpuHardwareSpec:
"MI250X",
"M4 Max",
"M2 Ultra",
+ "Arc B580",
+ "Arc B570",
+ "Jetson Orin NX 16GB",
+ "Jetson Orin NX 8GB",
+ "Jetson AGX Orin",
+ "Snapdragon X Elite",
+ "Snapdragon X Plus",
}
assert set(GPU_SPECS.keys()) == expected
@@ -122,7 +128,6 @@ class TestGpuHardwareSpec:
# Tests: energy integration math (trapezoidal rule)
# ---------------------------------------------------------------------------
-
class TestEnergyIntegration:
def test_constant_power(self):
"""Constant 300W for 10 seconds = 3000 J."""
@@ -130,7 +135,6 @@ class TestEnergyIntegration:
snapshots = [[_snap(power=300)] for _ in range(11)]
timestamps = [float(i) for i in range(11)]
-
sample = GpuMonitor._aggregate(snapshots, timestamps, wall_duration=10.0)
assert sample.energy_joules == pytest.approx(3000.0, rel=1e-6)
assert sample.mean_power_watts == pytest.approx(300.0, rel=1e-6)
@@ -146,7 +150,6 @@ class TestEnergyIntegration:
[_snap(power=100.0 * i, util=50, mem=8, temp=60)] for i in range(5)
]
timestamps = [float(i) for i in range(5)]
-
sample = GpuMonitor._aggregate(snapshots, timestamps, wall_duration=4.0)
assert sample.energy_joules == pytest.approx(800.0, rel=1e-6)
assert sample.mean_power_watts == pytest.approx(200.0, rel=1e-6)
@@ -167,7 +170,6 @@ class TestEnergyIntegration:
snapshots = [[_snap(power=250, util=90, mem=16, temp=70)]]
timestamps = [0.0]
-
sample = GpuMonitor._aggregate(snapshots, timestamps, wall_duration=0.05)
assert sample.energy_joules == 0.0
assert sample.num_snapshots == 1
@@ -178,7 +180,6 @@ class TestEnergyIntegration:
# Tests: GpuSample aggregation
# ---------------------------------------------------------------------------
-
class TestGpuSampleAggregation:
def test_peak_values(self):
from openjarvis.telemetry.gpu_monitor import GpuMonitor
@@ -189,7 +190,6 @@ class TestGpuSampleAggregation:
[_snap(power=300, util=70, mem=15, temp=65)],
]
timestamps = [0.0, 1.0, 2.0]
-
sample = GpuMonitor._aggregate(snapshots, timestamps, wall_duration=2.0)
assert sample.peak_power_watts == pytest.approx(400.0)
assert sample.peak_utilization_pct == pytest.approx(95.0)
@@ -205,7 +205,6 @@ class TestGpuSampleAggregation:
[_snap(power=300, util=80, mem=16, temp=70)],
]
timestamps = [0.0, 1.0, 2.0]
-
sample = GpuMonitor._aggregate(snapshots, timestamps, wall_duration=2.0)
assert sample.mean_power_watts == pytest.approx(200.0)
assert sample.mean_utilization_pct == pytest.approx(60.0)
@@ -217,7 +216,6 @@ class TestGpuSampleAggregation:
# Tests: Multi-GPU aggregation
# ---------------------------------------------------------------------------
-
class TestMultiGpu:
def test_multi_device_power_sum(self):
"""Power summed across devices; util/temp averaged."""
@@ -229,7 +227,6 @@ class TestMultiGpu:
]
snapshots = [tick, tick]
timestamps = [0.0, 1.0]
-
sample = GpuMonitor._aggregate(snapshots, timestamps, wall_duration=1.0)
assert sample.energy_joules == pytest.approx(500.0)
assert sample.mean_power_watts == pytest.approx(500.0)
@@ -252,7 +249,6 @@ class TestMultiGpu:
],
]
timestamps = [0.0, 2.0]
-
sample = GpuMonitor._aggregate(snapshots, timestamps, wall_duration=2.0)
# Tick powers: 200, 600 => 0.5*(200+600)*2 = 800
assert sample.energy_joules == pytest.approx(800.0)
@@ -263,7 +259,6 @@ class TestMultiGpu:
# Tests: context manager flow (with mocked pynvml)
# ---------------------------------------------------------------------------
-
class TestContextManager:
def test_sample_context_manager(self):
"""sample() starts/stops polling and populates result."""
@@ -280,10 +275,8 @@ class TestContextManager:
monitor._initialized = True
monitor._device_count = 1
monitor._handles = ["handle-0"]
-
with monitor.sample() as result:
time.sleep(0.1)
-
assert result.duration_seconds > 0
assert result.num_snapshots > 0
assert result.mean_power_watts > 0
@@ -301,10 +294,8 @@ class TestContextManager:
monitor._handles = []
monitor._device_count = 0
monitor._initialized = False
-
with monitor.sample() as result:
pass
-
assert result.num_snapshots == 0
assert result.energy_joules == 0.0
assert result.duration_seconds >= 0
@@ -314,7 +305,6 @@ class TestContextManager:
# Tests: available()
# ---------------------------------------------------------------------------
-
class TestAvailable:
def test_available_false_when_pynvml_missing(self):
"""available() returns False when pynvml not importable."""
@@ -330,7 +320,6 @@ class TestAvailable:
def test_available_true_with_fake_pynvml(self):
"""available() returns True when pynvml can init."""
fake_pynvml = _make_fake_pynvml()
-
with patch.dict(sys.modules, {"pynvml": fake_pynvml}):
import openjarvis.telemetry.gpu_monitor as mod
@@ -348,7 +337,6 @@ class TestAvailable:
"""available() returns False when nvmlInit raises."""
fake_pynvml = _make_fake_pynvml()
fake_pynvml.nvmlInit.side_effect = RuntimeError("no driver")
-
with patch.dict(sys.modules, {"pynvml": fake_pynvml}):
import openjarvis.telemetry.gpu_monitor as mod
@@ -365,7 +353,6 @@ class TestAvailable:
# Tests: dataclass defaults
# ---------------------------------------------------------------------------
-
class TestDataclasses:
def test_gpu_snapshot_defaults(self):
from openjarvis.telemetry.gpu_monitor import GpuSnapshot