Collect information about the environment and display it.
- Python version
- Package availablility and versions: Zeus, PyTorch, JAX.
- NVIDIA GPU availability: Number of GPUs and moels.
- AMD GPU availability: Number of GPUs and models.
- Intel RAPL availability: Number of CPUs and whether DRAM measurements are available.
show_env
Collect information about the environment and display it.
Source code in zeus/show_env.py
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167 | def show_env():
"""Collect information about the environment and display it."""
print(SECTION_SEPARATOR)
print("## Package availability and versions\n")
print("Logging output:")
package_availability = " Python version: " + platform.python_version() + "\n"
package_availability += f" Zeus: {zeus.__version__}\n"
try:
torch_available = framework.torch_is_available()
torch_cuda_available = True
except RuntimeError:
torch_available = framework.torch_is_available(ensure_cuda=False)
torch_cuda_available = False
if torch_available and torch_cuda_available:
torch = framework.MODULE_CACHE["torch"]
package_availability += f" PyTorch: {torch.__version__} (with CUDA support)\n"
elif torch_available and not torch_cuda_available:
torch = framework.MODULE_CACHE["torch"]
package_availability += (
f" PyTorch: {torch.__version__} (without CUDA support)\n"
)
else:
package_availability += " PyTorch: not available\n"
try:
jax_available = framework.jax_is_available()
jax_cuda_available = True
except RuntimeError:
jax_available = framework.jax_is_available(ensure_cuda=False)
jax_cuda_available = False
if jax_available and jax_cuda_available:
jax = framework.MODULE_CACHE["jax"]
package_availability += f" JAX: {jax.__version__} (with CUDA support)\n"
elif jax_available and not jax_cuda_available:
jax = framework.MODULE_CACHE["jax"]
package_availability += f" JAX: {jax.__version__} (without CUDA support)\n"
else:
package_availability += " JAX: not available\n"
print("\nDetected:\n" + package_availability)
print(SECTION_SEPARATOR)
print("## Device availability\n")
print("Logging output:")
gpu_availability = ""
try:
gpus = get_gpus()
except ZeusGPUInitError:
gpus = EmptyGPUs()
except ZeusBaseGPUError as e:
gpu_availability += f" Error initializing GPUs: {e}\n"
gpus = EmptyGPUs()
if len(gpus) > 0:
for i in range(len(gpus)):
gpu_availability += f" GPU {i}: {gpus.getName(i)}\n"
else:
gpu_availability += " No GPUs available.\n"
print("\nDetected:\n" + gpu_availability)
print(SECTION_SEPARATOR)
print("## CPU availability\n")
print("Logging output:")
cpu_availability = ""
try:
cpus = get_cpus()
except ZeusCPUInitError:
cpus = EmptyCPUs()
except ZeusBaseCPUError as e:
cpu_availability += f" Error initializing CPUs: {e}\n"
cpus = EmptyCPUs()
if len(cpus) > 0:
assert isinstance(cpus, RAPLCPUs)
for i, cpu in enumerate(cpus.cpus):
if isinstance(cpu, ZeusdRAPLCPU):
cpu_availability += f" CPU {i}:\n CPU measurements available (Zeusd at {cpu.zeusd_sock_path})\n"
if cpu.supportsGetDramEnergyConsumption():
cpu_availability += f" DRAM measurements available (Zeusd at {cpu.zeusd_sock_path})\n"
else:
cpu_availability += " DRAM measurements unavailable\n"
elif isinstance(cpu, RAPLCPU):
cpu_availability += f" CPU {i}:\n CPU measurements available ({cpu.rapl_file.path})\n"
if cpu.supportsGetDramEnergyConsumption():
dram = cpu.dram
assert dram is not None
cpu_availability += (
f" DRAM measurements available ({dram.path})\n"
)
else:
cpu_availability += " DRAM measurements unavailable\n"
else:
raise TypeError("Unexpected CPU type: " + str(type(cpu)))
else:
cpu_availability += " No CPUs available.\n"
print("\nDetected:\n" + cpu_availability)
print(SECTION_SEPARATOR)
print("## SoC availability\n")
print("Logging output:")
soc_availability = ""
try:
soc = get_soc()
except ZeusSoCInitError:
soc = EmptySoC()
except ZeusBaseSoCError as e:
soc_availability += f" Error initializing SoC: {e}\n"
soc = EmptySoC()
if isinstance(soc, AppleSilicon):
metrics = soc.getAvailableMetrics()
soc_availability += " Apple Silicon SoC available.\n"
soc_availability += f" Available metrics: {', '.join(metrics)}\n"
elif isinstance(soc, Jetson):
metrics = soc.getAvailableMetrics()
soc_availability += " NVIDIA Jetson SoC available.\n"
soc_availability += f" Available metrics: {', '.join(metrics)}\n"
else:
soc_availability += " No SoC available.\n"
# If this is an Apple Silicon device but AppleSilicon was not detected,
# print out a warning.
if platform.system() == "Darwin" and platform.machine() in ["arm64", "aarch64"]:
soc_availability += (
"\nThis appears to be an Apple Silicon device, but it wasn't picked up by Zeus.\n"
"Have you installed Zeus with the `apple` extra?\n\n"
" pip install 'zeus[apple]'\n"
)
print("\nDetected:\n" + soc_availability)
print(SECTION_SEPARATOR)
|