File size: 9,680 Bytes
82ea528
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
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
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
import torch
import pynvml
import comfy.model_management
from ..core import logger
# from ctypes import *
# from pyrsmi import rocml

class CGPUInfo:
    """
    This class is responsible for getting information from GPU (ONLY).
    """
    cuda = False
    pynvmlLoaded = False
    # pyamdLoaded = False
    # anygpuLoaded = False
    cudaAvailable = False
    torchDevice = 'cpu'
    cudaDevice = 'cpu'
    cudaDevicesFound = 0
    switchGPU = True
    switchVRAM = True
    switchTemperature = True
    gpus = []
    gpusUtilization = []
    gpusVRAM = []
    gpusTemperature = []

    def __init__(self):
        try:
            pynvml.nvmlInit()
            self.pynvmlLoaded = True
            logger.info('Pynvml (Nvidia) initialized.')
        except Exception as e:
            logger.error('Could not init pynvml (Nvidia).' + str(e))

        # if not self.pynvmlLoaded:
        #     try:
        #         rocml.smi_initialize()
        #         self.pyamdLoaded = True
        #         logger.info('Pyrsmi (AMD) initialized.')
        #     except Exception as e:
        #         logger.error('Could not init pyrsmi (AMD).' + str(e))

        # self.anygpuLoaded = self.pynvmlLoaded or self.pyamdLoaded
        self.anygpuLoaded = self.pynvmlLoaded

        try:
            self.torchDevice = comfy.model_management.get_torch_device_name(comfy.model_management.get_torch_device())
        except Exception as e:
            logger.error('Could not pick default device.' + str(e))

        # ZLUDA Check, self.torchDevice has 'ZLUDA' in it.
        if 'zluda' in self.torchDevice or 'ZLUDA' in self.torchDevice or 'Zluda' in self.torchDevice:
            logger.warn('ZLUDA detected. GPU monitoring will be disabled.')
            self.anygpuLoaded = False
            # self.pyamdLoaded = False
            self.pynvmlLoaded = False

        if self.anygpuLoaded and self.deviceGetCount() > 0:
            self.cudaDevicesFound = self.deviceGetCount()

            logger.info(f"GPU/s:")

            # for simulate multiple GPUs (for testing) interchange these comments:
            # for deviceIndex in range(3):
            #     deviceHandle = pynvml.nvmlDeviceGetHandleByIndex(0)
            for deviceIndex in range(self.cudaDevicesFound):
                deviceHandle = self.deviceGetHandleByIndex(deviceIndex)

                gpuName = self.deviceGetName(deviceHandle, deviceIndex)

                logger.info(f"{deviceIndex}) {gpuName}")

                self.gpus.append({
                    'index': deviceIndex,
                    'name': gpuName,
                })

                # same index as gpus, with default values
                self.gpusUtilization.append(True)
                self.gpusVRAM.append(True)
                self.gpusTemperature.append(True)

            self.cuda = True
            logger.info(self.systemGetDriverVersion())
        else:
            logger.warn('No GPU with CUDA detected.')

        self.cudaDevice = 'cpu' if self.torchDevice == 'cpu' else 'cuda'
        self.cudaAvailable = torch.cuda.is_available()

        if self.cuda and self.cudaAvailable and self.torchDevice == 'cpu':
            logger.warn('CUDA is available, but torch is using CPU.')

    def getInfo(self):
        logger.debug('Getting GPUs info...')
        return self.gpus

    def getStatus(self):
        # logger.debug('CGPUInfo getStatus')
        gpuUtilization = -1
        gpuTemperature = -1
        vramUsed = -1
        vramTotal = -1
        vramPercent = -1

        gpuType = ''
        gpus = []

        if self.cudaDevice == 'cpu':
            gpuType = 'cpu'
            gpus.append({
                'gpu_utilization': -1,
                'gpu_temperature': -1,
                'vram_total': -1,
                'vram_used': -1,
                'vram_used_percent': -1,
            })
        else:
            gpuType = self.cudaDevice

            if self.anygpuLoaded and self.cuda and self.cudaAvailable:
                # for simulate multiple GPUs (for testing) interchange these comments:
                # for deviceIndex in range(3):
                #     deviceHandle = self.deviceGetHandleByIndex(0)
                for deviceIndex in range(self.cudaDevicesFound):
                    deviceHandle = self.deviceGetHandleByIndex(deviceIndex)

                    gpuUtilization = -1
                    vramPercent = -1
                    vramUsed = -1
                    vramTotal = -1
                    gpuTemperature = -1

                    # GPU Utilization
                    if self.switchGPU and self.gpusUtilization[deviceIndex]:
                        try:
                            gpuUtilization = self.deviceGetUtilizationRates(deviceHandle)
                        except Exception as e:
                            if str(e) == "Unknown Error":
                                logger.error('For some reason, pynvml is not working in a laptop with only battery, try to connect and turn on the monitor')
                            else:
                                logger.error('Could not get GPU utilization.' + str(e))

                            logger.error('Monitor of GPU is turning off (not on UI!)')
                            self.switchGPU = False

                    # VRAM
                    if self.switchVRAM and self.gpusVRAM[deviceIndex]:
                        # Torch or pynvml?, pynvml is more accurate with the system, torch is more accurate with comfyUI
                        memory = self.deviceGetMemoryInfo(deviceHandle)
                        vramUsed = memory['used']
                        vramTotal = memory['total']

                        # device = torch.device(gpuType)
                        # vramUsed = torch.cuda.memory_allocated(device)
                        # vramTotal = torch.cuda.get_device_properties(device).total_memory

                        # check if vramTotal is not zero or None
                        if vramTotal and vramTotal != 0:
                            vramPercent = vramUsed / vramTotal * 100

                    # Temperature
                    if self.switchTemperature and self.gpusTemperature[deviceIndex]:
                        try:
                            gpuTemperature = self.deviceGetTemperature(deviceHandle)
                        except Exception as e:
                            logger.error('Could not get GPU temperature. Turning off this feature. ' + str(e))
                            self.switchTemperature = False

                    gpus.append({
                        'gpu_utilization': gpuUtilization,
                        'gpu_temperature': gpuTemperature,
                        'vram_total': vramTotal,
                        'vram_used': vramUsed,
                        'vram_used_percent': vramPercent,
                    })

        return {
            'device_type': gpuType,
            'gpus': gpus,
        }

    def deviceGetCount(self):
        if self.pynvmlLoaded:
            return pynvml.nvmlDeviceGetCount()
        # elif self.pyamdLoaded:
        #     return rocml.smi_get_device_count()
        else:
            return 0

    def deviceGetHandleByIndex(self, index):
        if self.pynvmlLoaded:
            return pynvml.nvmlDeviceGetHandleByIndex(index)
        # elif self.pyamdLoaded:
        #     return index
        else:
            return 0

    def deviceGetName(self, deviceHandle, deviceIndex):
        if self.pynvmlLoaded:
            gpuName = 'Unknown GPU'

            try:
                gpuName = pynvml.nvmlDeviceGetName(deviceHandle)
                try:
                    gpuName = gpuName.decode('utf-8', errors='ignore')
                except AttributeError as e:
                    pass

            except UnicodeDecodeError as e:
                gpuName = 'Unknown GPU (decoding error)'
                print(f"UnicodeDecodeError: {e}")

            return gpuName
        # elif self.pyamdLoaded:
        #     return rocml.smi_get_device_name(deviceIndex)
        else:
            return ''

    def systemGetDriverVersion(self):
        if self.pynvmlLoaded:
            return f'NVIDIA Driver: {pynvml.nvmlSystemGetDriverVersion()}'
        # elif self.pyamdLoaded:
        #     ver_str = create_string_buffer(256)
        #     rocml.rocm_lib.rsmi_version_str_get(0, ver_str, 256)
        #     return f'AMD Driver: {ver_str.value.decode()}'
        else:
            return 'Driver unknown'

    def deviceGetUtilizationRates(self, deviceHandle):
        if self.pynvmlLoaded:
            return pynvml.nvmlDeviceGetUtilizationRates(deviceHandle).gpu
        # elif self.pyamdLoaded:
        #     return rocml.smi_get_device_utilization(deviceHandle)
        else:
            return 0

    def deviceGetMemoryInfo(self, deviceHandle):
        if self.pynvmlLoaded:
            mem = pynvml.nvmlDeviceGetMemoryInfo(deviceHandle)
            return {'total': mem.total, 'used': mem.used}
        # elif self.pyamdLoaded:
        #     mem_used = rocml.smi_get_device_memory_used(deviceHandle)
        #     mem_total = rocml.smi_get_device_memory_total(deviceHandle)
        #     return {'total': mem_total, 'used': mem_used}
        else:
            return {'total': 1, 'used': 1}

    def deviceGetTemperature(self, deviceHandle):
        if self.pynvmlLoaded:
            return pynvml.nvmlDeviceGetTemperature(deviceHandle, pynvml.NVML_TEMPERATURE_GPU)
        # elif self.pyamdLoaded:
        #     temp = c_int64(0)
        #     rocml.rocm_lib.rsmi_dev_temp_metric_get(deviceHandle, 1, 0, byref(temp))
        #     return temp.value / 1000
        else:
            return 0