I suppose the flow would be something like:
select all experiments from project x with iterations greater than y, pull runtime for each one add them all up. I just don't know what API calls to make for 1 and 2
Regarding 1 & 2 - I suggest always keeping the API docs handy - https://clear.ml/docs/latest/docs/references/api/definitions
I love using the API since it's so convenient. So to get to business -
To select all experiments from a certain project you can use tasks.get_all with filtering according to the API docs (I suggest you also use the web UI as reference - if you hit F12 you can see all the API calls and their responses. This can really help to get an understanding of it's capabilities I'm not sure the runtime sits in the database as an exact number, but you can easily calculate it as 'completed' minus 'started' time of experiments.
Or examples of, like, "select all experiments in project with iterations > 0"?
If you get GPU-hours per project stats it would be really cool if you added this as a pull request
I might not be able to get to that but if you create an issue I'd be happy to link or post what I came up with, wdyt?
CostlyOstrich36 I made a code snippet for you:
` from clearml import Task
figuring out the project ID
project_list = Task.get_projects() # get all the projects
project_id = Task.get_project_id("your project name here")
getting all the tasks for a project
tasks = Task.get_all(project=[project_id]).response.tasks
loop through and get approximate maximum gpu-seconds by type.
import random
from collections import defaultdict
task = random.choice(tasks)
print(dir(task))
print(task.runtime)
gpu_seconds = defaultdict(int)
for task in tasks:
if task.runtime.get("gpu_count", 0) >0:
gpu_type = task.runtime.get("gpu_type")
active_duration = task.active_duration # in seconds, according to
print(f"active_duration of {gpu_type} is {active_duration} seconds")
gpu_seconds[gpu_type] = gpu_seconds[gpu_type] + active_duration Result of printing gpu_hours now will look like defaultdict(int,
{'A100-PCIE-40GB': 3563829,
'GeForce RTX 2080 Ti': 119211,
'Quadro P5200': 80997,
'TITAN RTX': 484239,
'Tesla P100-PCIE-16GB': 99454,
'Tesla T4': 193,
'Tesla V100-SXM2-16GB': 1278}) `
Hang on,
I just noticed that there's a "project compute time" on the dashboard? Do you know how that is calculated/what that is?
Are you referring to the to the example in services?
I might not be able to get to that but if you create an issue I'd be happy to link or post what I came up with, wdyt?
Taking a look at your snippet, I wouldn't mind submitting a PR for such a cool feature 🙂
One has an active duration of 185502. dividing that by 60 gives you minutes, oh I did the math wrong. Need to divide by 60 again to get hours,
CostlyOstrich36 at the bottom of the screenshot it says "Compute Time: 440 days"
Ok, will do once I get back to the office, thanks for the heads up! 🙂
Ah I see. I'm guessing UI is summing up runtimes of experiments in project.
Here's the hours/days version, corrected now lol:gpu_hours = {} gpu_days = {} for gpu_type, gpu_time_seconds in gpu_seconds.items(): gpu_time_hours = gpu_time_seconds/3600 gpu_hours[gpu_type] = gpu_time_hours gpu_days[gpu_type] = gpu_time_hours/24
You can do this quite easily with some code and the API 🙂
CostlyOstrich36 nice, thanks for the link. I know that in "info" on the experiments dashboard it includes gpu_type and started/completed times, I'll give it a go based on that
Hang on, CostlyOstrich36 I just noticed that there's a "project compute time" on the dashboard? Do you know how that is calculated/what that is?
The UI uses also the API so any data you see the the UI you can directly extract from the API that's why I personally love using it so much for similar tasks
SmallDeer34 Hi 🙂
I don't think there is a way out of the box to see GPU hours per project, but it can be a pretty cool feature! Maybe open a github feature request for this.
Regarding on how to calculate this, I think an easier solution for you would be to sum up the runtime of all experiments in a certain project rather than looking by GPU utilization graphs
Good point! Any pointers to API docs to start looking?
CostlyOstrich36 I get some weird results, for "active duration".
For example, several of the experiments show that their active duration is more than 90 days, but I definitely didn't run them that long.