Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-19 17:24:32 +02:00
Code Issues Packages Projects Releases Wiki Activity

fix global state tracking

Browse Source 
Signed-off-by: Vladimir Mandic <mandic00@live.com>
This commit is contained in:
Vladimir Mandic
2025-08-10 09:19:49 -04:00
parent e5637c3e8d
commit 41640773e5
7 changed files with 83 additions and 65 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/api/nvml.py
+14 -17
View File
@@ -25,9 +25,11 @@ def get_reason(val):
reason = ', '.join([throttle[i] for i in throttle if i & val])
return reason if len(reason) > 0 else 'ok'
def get_nvml():
global nvml_initialized # pylint: disable=global-statement
try:
from modules.memstats import ram_stats
if not nvml_initialized:
install('pynvml', quiet=True)
import pynvml # pylint: disable=redefined-outer-name
@@ -44,33 +46,28 @@ def get_nvml():
except Exception:
name = ''
load = pynvml.nvmlDeviceGetUtilizationRates(dev)
"""
'load': {
'gpu': round(load.gpu),
'memory': round(load.memory),
'temp': pynvml.nvmlDeviceGetTemperature(dev, 0),
'fan': pynvml.nvmlDeviceGetFanSpeed(dev),
},
'chart_val1': load.memory,
'chart_val2': load.gpu,
}
"""
mem = pynvml.nvmlDeviceGetMemoryInfo(dev)
ram = ram_stats()
data = {
"CUDA": f'version {pynvml.nvmlSystemGetCudaDriverVersion()} compute {pynvml.nvmlDeviceGetCudaComputeCapability(dev)}',
"CUDA": f'Version {pynvml.nvmlSystemGetCudaDriverVersion()} Compute {pynvml.nvmlDeviceGetCudaComputeCapability(dev)}',
"Driver": pynvml.nvmlSystemGetDriverVersion(),
"Hardware": f'VBIOS {pynvml.nvmlDeviceGetVbiosVersion(dev)} ROM {pynvml.nvmlDeviceGetInforomImageVersion(dev)}',
"PCI link": f'gen.{pynvml.nvmlDeviceGetCurrPcieLinkGeneration(dev)} x{pynvml.nvmlDeviceGetCurrPcieLinkWidth(dev)}',
"PCI link": f'Gen.{pynvml.nvmlDeviceGetCurrPcieLinkGeneration(dev)} x{pynvml.nvmlDeviceGetCurrPcieLinkWidth(dev)}',
"Power": f'{round(pynvml.nvmlDeviceGetPowerUsage(dev)/1000, 2)} W / {round(pynvml.nvmlDeviceGetEnforcedPowerLimit(dev)/1000, 2)} W',
"GPU clock": f'{pynvml.nvmlDeviceGetClockInfo(dev, 0)} Mhz / {pynvml.nvmlDeviceGetMaxClockInfo(dev, 0)} Mhz',
"SM clock": f'{pynvml.nvmlDeviceGetClockInfo(dev, 1)} Mhz / {pynvml.nvmlDeviceGetMaxClockInfo(dev, 1)} Mhz',
"Memory clock": f'{pynvml.nvmlDeviceGetClockInfo(dev, 2)} Mhz / {pynvml.nvmlDeviceGetMaxClockInfo(dev, 2)} Mhz',
"Memory usage": f'used {round(mem.used / 1024 / 1024)} MB | free {round(mem.free / 1024 / 1024)} MB | total {round(mem.total / 1024 / 1024)} MB',
"System load": f'GPU {load.gpu}% | Memory {load.memory}% | temp {pynvml.nvmlDeviceGetTemperature(dev, 0)}C | fan {pynvml.nvmlDeviceGetFanSpeed(dev)}%',
"VRAM clock": f'{pynvml.nvmlDeviceGetClockInfo(dev, 2)} Mhz / {pynvml.nvmlDeviceGetMaxClockInfo(dev, 2)} Mhz',
"VRAM usage": f'{round(100 * mem.used / mem.total)}% | {round(mem.used / 1024 / 1024)} MB used | {round(mem.free / 1024 / 1024)} MB free | {round(mem.total / 1024 / 1024)} MB total',
"RAM usage": f'{round(100 * ram["used"] / ram["total"])}% | {round(1024 * ram["used"])} MB used | {round(1024 * ram["free"])} MB free | {round(1024 * ram["total"])} MB total',
"System load": f'GPU {load.gpu}% | VRAM {load.memory}% | Temp {pynvml.nvmlDeviceGetTemperature(dev, 0)}C | Fan {pynvml.nvmlDeviceGetFanSpeed(dev)}%',
'State': get_reason(pynvml.nvmlDeviceGetCurrentClocksThrottleReasons(dev)),
}
chart = [load.memory, load.gpu]
devices.append({ 'name': name, 'data': data, 'chart': chart })
devices.append({
'name': name,
'data': data,
'chart': chart,
})
# log.debug(f'nmvl: {devices}')
return devices
except Exception as e:
modules/api/rocm_smi.py
+12 -5
View File
@@ -3,12 +3,14 @@ import json
import subprocess as sp
from enum import IntFlag
try:
from installer import log
except Exception:
import logging
log = logging.getLogger(__name__)
try:
from modules.rocm import version as rocm_version
except Exception:
@@ -87,14 +89,19 @@ def get_rocm_smi():
"PCI link": f'Gen.{int(math.log2(float(rocm_smi_data[key].get("pcie_link_speed (0.1 GT/s)", 10)) / 10))} x{rocm_smi_data[key].get("pcie_link_width (Lanes)", "unknown")}',
"Power": f'{round(float(rocm_smi_data[key].get("Average Graphics Package Power (W)", 0)), 2)} W / {round(float(rocm_smi_data[key].get("Max Graphics Package Power (W)", 0)), 2)} W',
"GPU clock": f'{rocm_smi_data[key].get("average_gfxclk_frequency (MHz)", 0)} Mhz / {rocm_smi_data[key].get("Valid sclk range", "0").split(" - ")[-1].removesuffix("Mhz")} Mhz',
"Memory clock": f'{rocm_smi_data[key].get("current_uclk (MHz)", 0)} Mhz / {rocm_smi_data[key].get("Valid mclk range", "0").split(" - ")[-1].removesuffix("Mhz")} Mhz',
"Memory usage": f'used {load["memory"]}% | activity {rocm_smi_data[key].get("GPU Memory Read/Write Activity (%)", "unknown")}%',
"GPU usage": f'GPU {load["gpu"]}% | fan {load["fan"]}%',
"GPU temp": f'edge {load["temp"]}C | junction {load["temp_junction"]}C | memory {load["temp_memory"]}C',
"VRAM clock": f'{rocm_smi_data[key].get("current_uclk (MHz)", 0)} Mhz / {rocm_smi_data[key].get("Valid mclk range", "0").split(" - ")[-1].removesuffix("Mhz")} Mhz',
"VRAM usage": f'{load["memory"]}% Used | {rocm_smi_data[key].get("GPU Memory Read/Write Activity (%)", "unknown")}% Activity',
"GPU usage": f'GPU {load["gpu"]}% | Fan {load["fan"]}%',
"GPU temp": f'Edge {load["temp"]}C | Junction {load["temp_junction"]}C | Memory {load["temp_memory"]}C',
'Throttle reason': str(ThrottleStatus(int(rocm_smi_data[key].get("throttle_status", 0)))),
}
name = rocm_smi_data[key].get('Device Name', 'unknown')
chart = [load["memory"], load["gpu"]]
devices.append({ 'name': rocm_smi_data[key].get('Device Name', 'unknown'), 'data': data, 'chart': chart })
devices.append({
'name': name,
'data': data,
'chart': chart,
})
return devices
except Exception as e:
log.error(f'ROCm SMI: {e}')