Reputation
Badges 1
25 × Eureka!ElegantCoyote26 point me to where Keras stores the data 🙂
If in the process of integration you had to add a logger/callback to your Keras code, that is the equivalent of using the TB.
I'll make sure we fix the example, because as you pointed, it is broken :(
Thanks ElegantCoyote26 I'll look into it. Seems like someone liked our automagical approach 🙂
A few epochs is just fine
`
Example use case:
an_optimizer = HyperParameterOptimizer(
# This is the experiment we want to optimize
base_task_id=args['template_task_id'],
# here we define the hyper-parameters to optimize
hyper_parameters=[
UniformIntegerParameterRange('General/layer_1', min_value=128, max_value=512, step_size=128),
UniformIntegerParameterRange('General/layer_2', min_value=128, max_value=512, step_size=128),
DiscreteParameterRange('General/batch_size', values=[...
ElegantCoyote26 It means we need to have a keras logger that logs everything to trains, then we need to hook it automatically.
Do you feel like PR-ing the logger (the hooking I can take care of 🙂 )?
Hi ElegantCoyote26 , yes I did 🙂
It seems cometml puts their default callback logger for you, that's it.
PungentLouse55 from the screenshot I assume the experiment template you are trying to optimize is not the one from the trains/examples 🙂
In that case, and based on the screenshots, the prefix is "Args/" as this is the section name.
Regrading objective metric, again based on your screenshots:objective_metric_title="Accuracy" objective_metric_series="Validation"
Make sense ?
Hi PungentLouse55
Hope you are not tired of me
Lol 🙂 No worries
I am using trains 0.16.1
Are you referring to the trains-server version or the python package ? (they are not the same and can be of totally different versions)
PungentLouse55 I'm checking something here, you might stumbled on a bug in parameter overriding. Updating here soon ...
The latest RC (0.17.5rc6) moved all logs into separate subprocess to improve speed with pytorch dataloaders
I want to optimizer hyperparameters with trains.automation but: ...
Yes you are correct, in case of the example code, it should be "General/..." if you have ArgParser, it should be "Args/..." Yes it looks like the metric is wrong, it should be "epoch_accuracy" & "epoch_accuracy"
Hi PungentLouse55
Are you referring to the example code ?
Hi @<1545216070686609408:profile|EnthusiasticCow4>
will ClearML remove the corresponding folders and files on S3?
Yes and it will ask you for credentials as well. I think there is a way to configure it so that the backend has access to it (somehow) but this breaks the "federated" approach
It's just another flag when running the trains-agent
You can have multiple service-mode instances, there is no actual limit 🙂
GreasyLeopard35
I can update that the fix to UniformIntegerParameterRange should be pushed with tomorrows release 🙂
(which would fix in turn LogUniformParameterRange)
Hi PompousBeetle71 , what exactly is the scenario / problem we are trying to solve ?
GrittyKangaroo27 any chance you can open a GitHub issue so this is not forgotten ?
(btw: we I think 1.1.6 is going to be released later today, then we will have a few RC with improvements on the pipeline, I will make sure we add that as well)
Hi GrittyKangaroo27
How could I turn off model logging when running this training step?
This is a good point! I think we cannot pass these arguments.
Would this make sense to you?PipelineDecorator.component(...,
auto_connect_frameworks)
wdyt?
Hi GrievingTurkey78
Can you test with the latest clearml-agent RC (I remember a fix just for that)pip install clearml-agent==1.2.0rc0
Great if this is what you do how come you need to change the entry script in the ui?
I did nothing to generate a command-line. Just cloned the experiment and enqueued it. Used the server GUI.
Who/What created the initial experiment ?
I noticed that if I run the initial experiment by "python -m folder_name.script_name"
"-m module" as script entry is used to launch entry points like python modules (which is translated to "python -m script")
Why isn't the entry point just the python script?
The command line arguments are passed as arguments on the Args section of t...
Oh I see, what you need is to pass '--script script.py' as entry-point and ' --cwd folder' as working dir
Actually it is better to leave it as is, it will just automatically mount the .ssh folder into the container, i will make sure the docs point to this option first