Comments (4)
In #79 Daniel mentioned that this should allow overriding Slurm parameters when starting an experiment, which is not addressed by this design.
However, I don't think it's ideal to have start
do side-effects like changing the database. I'd rather propose a new command (update
?) for updating config values. This would combine the filtering syntax of start
with the overriding syntax of this and #79:
seml <collection> update -b 7 -sb mem=25GB -o dataset=imagenet
I'd suggest to add the overriding options to both add
and update
then.
from seml.
Having an update command raises the potential issue that an overwrite parameter could cause two identical configurations. We would then have to check for duplicates again and potentially remove these. Still, having an override option at add time seems beneficial to me.
from seml.
I'd suggest breaking this up into two issues, as these are more or less orthogonal:
- Adding sbatch options to add (+ new syntax for jupyter)
- Adding an
update
command.
from seml.
Yes, good point! Let's keep the PRs small and manageable.
update could also break our assumption that every batch has consistent Slurm parameters. Creating batch arrays and sharing GPUs both rely on this. As a first step we could only allow changing Slurm parameters of full batches (and document the reasoning behind these restrictions).
from seml.
Related Issues (20)
- Feature request: restaging killed experiments without deleting info HOT 3
- Questions: MongoDB HOT 2
- Citing seml HOT 1
- Suggest to loosen the dependency on sacred HOT 1
- Unknown option directconnection HOT 1
- remove experiment-tracking topic label from project HOT 5
- Setting up mongodb on compute cluster HOT 3
- Current version of seml not available on conda-forge HOT 1
- Access to "seml jupyter" HOT 1
- Regular releases via pip and conda HOT 4
- Feature request: Automatic hyperparamter optimization HOT 2
- Source reload does not work correctly with file named configs
- Boolean configs get converted to integers HOT 2
- clean-db command fails due to the switch from tqdm to rich.progress HOT 1
- seml silently kills pending experiments, when observing them with the status command, while they are pending in slurm HOT 5
- create_neptune_observer does not use api_token argument
- TMPDIR is not configurable HOT 1
- Bug: Deleting a failed experiment does not delete all saved source files in the mongodb. HOT 7
- Bug: Step sizes for grid ranges are converted to integers HOT 1
- Feature Request: Add experiments from multiple config files at once
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from seml.