Cloudmesh Command gpu
=====================
[](https://github.com/cloudmesh/cloudmesh-gpu)
[](https://pypi.org/project/cloudmesh-gpu)
[](https://pypi.org/project/cloudmesh-gpu/)
[](https://opensource.org/licenses/Apache-2.0)
[](https://shields.io/)
[](https://github.com/cloudmesh/cloudmesh-gpu/issues)
[](https://github.com/cloudmesh/cloudmesh-gpu/graphs/contributors)
[](https://github.com/cloudmesh/cloudmesh)
[](https://www.linux.org/)
[](https://www.apple.com/macos)
[](https://www.microsoft.com/windows)
Note: This file is automatically created. Please do not modify it.
Please change the code instead.
```
Usage:
gpu watch [--gpu=GPU] [--delay=SECONDS] [--logfile=LOGFILE] [--count=COUNT] [--dense]
gpu --json [--gpu=GPU] [--pretty] [FILE]
gpu --xml
gpu --yaml
gpu ps [--gpu=GPU] [--format=FORMAT] [--detail]
gpu system
gpu status
gpu count
gpu kill
gpu show --output=OUTPUT FILE [--plot=PLOT] [--frequency=FREQUENCY]
gpu probe
gpu
This command returns some information about NVIDIA GPUs if your
system has them. Please use the cms command in the commandline with
`cms gpu`
Options:
--json returns the information in json
--xml returns the information in xml
--yaml returns the information in xml
--logfile=LOGFILE the logfile
--count=COUNT how many times the watch is run [default: -1]
--dense do not print any spaces [default: False]
--detail short process names [default: False]
--format=FORMAT table, json, yaml [default: table]
--plot=PLOT timeseries, histogram [default: timeseries]
--frequency=FREQUENCY absolute, percent [default: percent]
--gpu=GPUS which graphics cards
Description:
Although some GPU information tools exist, we did not find all information
that we needed. Also, the output format was not convenient enough for later
analysis. The program we developed can obtain information at specified
intervals. Note that at this time the tool is restricted to NVIDIA GPUs.
cms gpu kill
In case you run the gpu command in the background you can kill it with the
`kill command`
cms gpu count
Returns the number of GPUS.
gpu watch [--gpu=GPU] [--delay=SECONDS] [--logfile=LOGFILE] [--count=COUNT] [--dense]
This command allows to print environmental variables in a continuous fashion.
If count is used the command is executed that number of times. If it is omitted
it runs continiously. The delay specifies how many seconds between invocations are
dealyed. The `logfile` specifies an output file in which the events are written.
By default spaces are included so the output has a table like format. The spaces
can be eliminated with dense so that less space is used in the log file.
In case multiple GPUs are present one can select specific GPUs for which the
monitoring entries are generated.
The watch command output is proceeded by a header that is adapted based on the
gpus specified. This makes it possible to read in the file as dataframe and apply
easily plotting tools.
An example output looks like (the newline in this example in the header was
introduced to increase readability of this documentation. In an execution it is
one line.
# ####################################################################################
# time, 0 id, 0 gpu_util %, 0 memory_util %, 0 encoder_util %, 0 decoder_util %,
# ... 0 gpu_temp C, 0 power_draw W
2022-03-18 11:26:40.877006, 0, 11, 16, 0, 0, 49, 47.47
2022-03-18 11:26:40.983229, 0, 17, 17, 0, 0, 49, 47.59
2022-03-18 11:26:41.093406, 0, 17, 17, 0, 0, 49, 47.88
gpu --json [--gpu=GPU] [--pretty] [FILE]
Prints out the information in json format. We have eliminated some of the attributes
that are not important to us at this time. with the `pretty` flag the json is printed
in indented pretty format.
The FILE parameter is used to read in a saved instance from nvidia-smi.
gpu --xml
Prints out the information in xml format. This is the format that is retrieved from
nvidia-smi. The gpu selection flag is not enabled for this format. If you want to
implement it, create a pull request.
gpu --yaml
Prints out the information in yaml format. The gpu selection flag is not enabled
for this format. If you want to implement it, create a pull request.
gpu ps [--gpu=GPU] [--format=FORMAT] [--detail]
Prints out the processes running on the GPU in the specified format for the selected GPUs.
The process name is shortened based on removing the path of the command. If the full path
is needed one can use the `detail` flag. Allowed formats are table, csv, json, and yaml.
A sample output for a table looks similar to
+-----+-----+-----+------+-------------+---------------------+-----------------+--------------+
| job | gpu | pid | type | used_memory | compute_instance_id | gpu_instance_id | process_name |
+-----+-----+-----+------+-------------+---------------------+-----------------+--------------+
| 1 | 0 | 173 | G | 198 MiB | N/A | N/A | Xorg |
| 2 | 0 | 260 | G | 610 MiB | N/A | N/A | Xorg |
| 3 | 0 | 274 | G | 49 MiB | N/A | N/A | gnome-shell |
| 4 | 0 | 436 | G | 27 MiB | N/A | N/A | zoom |
| 5 | 0 | 321 | G | 100 MiB | N/A | N/A | slack |
| 6 | 0 | 591 | G | 11 MiB | N/A | N/A | firefox |
+-----+-----+-----+------+-------------+---------------------+-----------------+--------------+
gpu system
Returns information about the GPU card
[
{
"product_name": "NVIDIA GeForce RTX 3090",
"product_brand": "GeForce",
"product_architecture": "Ampere",
"vbios_version": "94.02.42.00.A9",
"inforom_version": {
"img_version": "G001.0000.03.03",
"oem_object": "2.0",
"ecc_object": "N/A",
"pwr_object": "N/A"
},
"accounted_processes": null,
"vendor": "ASUSTeK Computer Inc. Device [abcd:1234]"
}
]
gpu status
Returns elementary status information such as environmental sensors
[
{
"fan_speed": "0 %",
"utilization": {
"gpu_util": "18 %",
"memory_util": "18 %",
"encoder_util": "0 %",
"decoder_util": "0 %"
},
"temperature": {
"gpu_temp": "48 C",
"gpu_temp_max_threshold": "98 C",
"gpu_temp_slow_threshold": "95 C",
"gpu_temp_max_gpu_threshold": "93 C",
"gpu_target_temperature": "83 C",
"memory_temp": "N/A",
"gpu_temp_max_mem_threshold": "N/A"
},
"supported_gpu_target_temp": {
"gpu_target_temp_min": "65 C",
"gpu_target_temp_max": "91 C"
},
"power_readings": {
"power_state": "P8",
"power_management": "Supported",
"power_draw": "47.28 W",
"power_limit": "390.00 W",
"default_power_limit": "390.00 W",
"enforced_power_limit": "390.00 W",
"min_power_limit": "100.00 W",
"max_power_limit": "480.00 W"
},
"clocks": {
"graphics_clock": "210 MHz",
"sm_clock": "210 MHz",
"mem_clock": "405 MHz",
"video_clock": "555 MHz"
},
"voltage": {
"graphics_volt": "737.500 mV"
},
"accounted_processes": null
}
]
```
Raw data
{
"_id": null,
"home_page": "",
"name": "cloudmesh-gpu",
"maintainer": "",
"docs_url": null,
"requires_python": ">=3.8",
"maintainer_email": "Gregor von Laszewski <laszewski@gmail.com>",
"keywords": "helper library,cloudmesh",
"author": "",
"author_email": "Gregor von Laszewski <laszewski@gmail.com>",
"download_url": "https://files.pythonhosted.org/packages/b1/41/a0a0255dea327e227c2f96726301998da4a580c57136ae8cb8bf9fc9e1bf/cloudmesh-gpu-5.0.3.tar.gz",
"platform": null,
"description": "Cloudmesh Command gpu\n=====================\n\n[](https://github.com/cloudmesh/cloudmesh-gpu)\n[](https://pypi.org/project/cloudmesh-gpu)\n[](https://pypi.org/project/cloudmesh-gpu/)\n[](https://opensource.org/licenses/Apache-2.0)\n\n[](https://shields.io/)\n[](https://github.com/cloudmesh/cloudmesh-gpu/issues)\n[](https://github.com/cloudmesh/cloudmesh-gpu/graphs/contributors)\n[](https://github.com/cloudmesh/cloudmesh)\n\n\n[](https://www.linux.org/)\n[](https://www.apple.com/macos)\n[](https://www.microsoft.com/windows)\n\n\nNote: This file is automatically created. Please do not modify it.\n Please change the code instead.\n\n```\n\n Usage:\n gpu watch [--gpu=GPU] [--delay=SECONDS] [--logfile=LOGFILE] [--count=COUNT] [--dense]\n gpu --json [--gpu=GPU] [--pretty] [FILE]\n gpu --xml\n gpu --yaml\n gpu ps [--gpu=GPU] [--format=FORMAT] [--detail]\n gpu system\n gpu status\n gpu count\n gpu kill\n gpu show --output=OUTPUT FILE [--plot=PLOT] [--frequency=FREQUENCY]\n gpu probe\n gpu\n\n This command returns some information about NVIDIA GPUs if your\n system has them. Please use the cms command in the commandline with\n `cms gpu`\n\n Options:\n --json returns the information in json\n --xml returns the information in xml\n --yaml returns the information in xml\n --logfile=LOGFILE the logfile\n --count=COUNT how many times the watch is run [default: -1]\n --dense do not print any spaces [default: False]\n --detail short process names [default: False]\n --format=FORMAT table, json, yaml [default: table]\n --plot=PLOT timeseries, histogram [default: timeseries]\n --frequency=FREQUENCY absolute, percent [default: percent]\n --gpu=GPUS which graphics cards\n\n Description:\n\n Although some GPU information tools exist, we did not find all information\n that we needed. Also, the output format was not convenient enough for later\n analysis. The program we developed can obtain information at specified\n intervals. Note that at this time the tool is restricted to NVIDIA GPUs.\n\n cms gpu kill\n In case you run the gpu command in the background you can kill it with the\n `kill command`\n\n cms gpu count\n Returns the number of GPUS.\n\n gpu watch [--gpu=GPU] [--delay=SECONDS] [--logfile=LOGFILE] [--count=COUNT] [--dense]\n This command allows to print environmental variables in a continuous fashion.\n If count is used the command is executed that number of times. If it is omitted\n it runs continiously. The delay specifies how many seconds between invocations are\n dealyed. The `logfile` specifies an output file in which the events are written.\n By default spaces are included so the output has a table like format. The spaces\n can be eliminated with dense so that less space is used in the log file.\n In case multiple GPUs are present one can select specific GPUs for which the\n monitoring entries are generated.\n The watch command output is proceeded by a header that is adapted based on the\n gpus specified. This makes it possible to read in the file as dataframe and apply\n easily plotting tools.\n\n An example output looks like (the newline in this example in the header was\n introduced to increase readability of this documentation. In an execution it is\n one line.\n\n # ####################################################################################\n # time, 0 id, 0 gpu_util %, 0 memory_util %, 0 encoder_util %, 0 decoder_util %,\n # ... 0 gpu_temp C, 0 power_draw W\n 2022-03-18 11:26:40.877006, 0, 11, 16, 0, 0, 49, 47.47\n 2022-03-18 11:26:40.983229, 0, 17, 17, 0, 0, 49, 47.59\n 2022-03-18 11:26:41.093406, 0, 17, 17, 0, 0, 49, 47.88\n\n gpu --json [--gpu=GPU] [--pretty] [FILE]\n Prints out the information in json format. We have eliminated some of the attributes\n that are not important to us at this time. with the `pretty` flag the json is printed\n in indented pretty format.\n The FILE parameter is used to read in a saved instance from nvidia-smi.\n\n gpu --xml\n Prints out the information in xml format. This is the format that is retrieved from\n nvidia-smi. The gpu selection flag is not enabled for this format. If you want to\n implement it, create a pull request.\n\n gpu --yaml\n Prints out the information in yaml format. The gpu selection flag is not enabled\n for this format. If you want to implement it, create a pull request.\n\n gpu ps [--gpu=GPU] [--format=FORMAT] [--detail]\n Prints out the processes running on the GPU in the specified format for the selected GPUs.\n The process name is shortened based on removing the path of the command. If the full path\n is needed one can use the `detail` flag. Allowed formats are table, csv, json, and yaml.\n\n A sample output for a table looks similar to\n\n +-----+-----+-----+------+-------------+---------------------+-----------------+--------------+\n | job | gpu | pid | type | used_memory | compute_instance_id | gpu_instance_id | process_name |\n +-----+-----+-----+------+-------------+---------------------+-----------------+--------------+\n | 1 | 0 | 173 | G | 198 MiB | N/A | N/A | Xorg |\n | 2 | 0 | 260 | G | 610 MiB | N/A | N/A | Xorg |\n | 3 | 0 | 274 | G | 49 MiB | N/A | N/A | gnome-shell |\n | 4 | 0 | 436 | G | 27 MiB | N/A | N/A | zoom |\n | 5 | 0 | 321 | G | 100 MiB | N/A | N/A | slack |\n | 6 | 0 | 591 | G | 11 MiB | N/A | N/A | firefox |\n +-----+-----+-----+------+-------------+---------------------+-----------------+--------------+\n\n\n gpu system\n Returns information about the GPU card\n\n [\n {\n \"product_name\": \"NVIDIA GeForce RTX 3090\",\n \"product_brand\": \"GeForce\",\n \"product_architecture\": \"Ampere\",\n \"vbios_version\": \"94.02.42.00.A9\",\n \"inforom_version\": {\n \"img_version\": \"G001.0000.03.03\",\n \"oem_object\": \"2.0\",\n \"ecc_object\": \"N/A\",\n \"pwr_object\": \"N/A\"\n },\n \"accounted_processes\": null,\n \"vendor\": \"ASUSTeK Computer Inc. Device [abcd:1234]\"\n }\n ]\n\n\n gpu status\n Returns elementary status information such as environmental sensors\n\n [\n {\n \"fan_speed\": \"0 %\",\n \"utilization\": {\n \"gpu_util\": \"18 %\",\n \"memory_util\": \"18 %\",\n \"encoder_util\": \"0 %\",\n \"decoder_util\": \"0 %\"\n },\n \"temperature\": {\n \"gpu_temp\": \"48 C\",\n \"gpu_temp_max_threshold\": \"98 C\",\n \"gpu_temp_slow_threshold\": \"95 C\",\n \"gpu_temp_max_gpu_threshold\": \"93 C\",\n \"gpu_target_temperature\": \"83 C\",\n \"memory_temp\": \"N/A\",\n \"gpu_temp_max_mem_threshold\": \"N/A\"\n },\n \"supported_gpu_target_temp\": {\n \"gpu_target_temp_min\": \"65 C\",\n \"gpu_target_temp_max\": \"91 C\"\n },\n \"power_readings\": {\n \"power_state\": \"P8\",\n \"power_management\": \"Supported\",\n \"power_draw\": \"47.28 W\",\n \"power_limit\": \"390.00 W\",\n \"default_power_limit\": \"390.00 W\",\n \"enforced_power_limit\": \"390.00 W\",\n \"min_power_limit\": \"100.00 W\",\n \"max_power_limit\": \"480.00 W\"\n },\n \"clocks\": {\n \"graphics_clock\": \"210 MHz\",\n \"sm_clock\": \"210 MHz\",\n \"mem_clock\": \"405 MHz\",\n \"video_clock\": \"555 MHz\"\n },\n \"voltage\": {\n \"graphics_volt\": \"737.500 mV\"\n },\n \"accounted_processes\": null\n }\n ]\n\n\n\n\n```\n",
"bugtrack_url": null,
"license": "Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ Copyright 2017-2021 Gregor von Laszewski, Indiana University Copyright 2021,2022 Gregor von Laszewski, University of Virginia Licensed under the Apache License, Version 2.0 (the \"License\"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. ",
"summary": "A cloudmesh command called gpu that can be used to monitor the GPU during runtime",
"version": "5.0.3",
"project_urls": {
"Changelog": "https://github.com/cloudmesh/cloudmesh-bar/blob/main/CHANGELOG.md",
"Documentation": "https://github.com/cloudmesh/cloudmesh-bar/blob/main/README.md",
"Homepage": "https://github.com/cloudmesh/cloudmesh-bar",
"Issues": "https://github.com/cloudmesh/cloudmesh-bar/issues",
"Repository": "https://github.com/cloudmesh/cloudmesh-bar.git"
},
"split_keywords": [
"helper library",
"cloudmesh"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "410e404cdc2a3d0fa32f7a84a2377e04c046fcb6e38892b17e8e70970fc527b3",
"md5": "2781debe17367027b9fca39a75ed125d",
"sha256": "afd81105d9e811f5156881a06940dfc79baa923f755ac85a06679d23d7f144e2"
},
"downloads": -1,
"filename": "cloudmesh_gpu-5.0.3-py2.py3-none-any.whl",
"has_sig": false,
"md5_digest": "2781debe17367027b9fca39a75ed125d",
"packagetype": "bdist_wheel",
"python_version": "py2.py3",
"requires_python": ">=3.8",
"size": 15542,
"upload_time": "2024-01-01T16:35:43",
"upload_time_iso_8601": "2024-01-01T16:35:43.631492Z",
"url": "https://files.pythonhosted.org/packages/41/0e/404cdc2a3d0fa32f7a84a2377e04c046fcb6e38892b17e8e70970fc527b3/cloudmesh_gpu-5.0.3-py2.py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "b141a0a0255dea327e227c2f96726301998da4a580c57136ae8cb8bf9fc9e1bf",
"md5": "e8f192bbe2013a437b090d46cb6e9392",
"sha256": "c518fea32f50d301cecd3f53cd2c75d14dbd2dadb0c56532c388c977949a753c"
},
"downloads": -1,
"filename": "cloudmesh-gpu-5.0.3.tar.gz",
"has_sig": false,
"md5_digest": "e8f192bbe2013a437b090d46cb6e9392",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.8",
"size": 18759,
"upload_time": "2024-01-01T16:35:45",
"upload_time_iso_8601": "2024-01-01T16:35:45.804380Z",
"url": "https://files.pythonhosted.org/packages/b1/41/a0a0255dea327e227c2f96726301998da4a580c57136ae8cb8bf9fc9e1bf/cloudmesh-gpu-5.0.3.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-01-01 16:35:45",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "cloudmesh",
"github_project": "cloudmesh-bar",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"requirements": [],
"lcname": "cloudmesh-gpu"
}
