andreasjansson
Follow
Create your own GitHub profile
Sign up for your own profile on GitHub, the best place to host code, manage projects, and build software alongside 50 million developers.
Sign upPinned
1,124 contributions in the last year
Activity overview
Contributed to
replicate/replicate,
andreasjansson/coronahjalpen-habo,
andreasjansson/instrument-practice
and 5 other
repositories
Contribution activity
November 2020
Created 32 commits in 1 repository
Created 1 repository
- andreasjansson/replicate Python
Created a pull request in replicate/replicate that received 2 comments
Opened 15 other pull requests in 1 repository
replicate/replicate
15
merged
- Make targets to verify that Go and Python are installed correctly
- Delete unused cache package
- Add "Where to start?" section to the contributing guide
- Always ignore virtualenvs
- add long help to replicate analytics command
- Resolve todos
- Save repository version at project level
- add missing plot labels on checkpoints/experiments
- CheckpointList API with .metrics, .step, and .plot()
- constants for documentation URLs in python
- add python tests for tar get and put
- add new version file in release
- save replicate_version to experiment metadata
- require repository scheme
- cleanup dead code
Reviewed 1 pull request in 1 repository
replicate/replicate 1 pull request
Created an issue in replicate/replicate that received 2 comments
Filter name for creation time is called "started" instead of "created"
This is confusing and error prone. It should be called "created" to be consistent with the rest of the CLI and Python API.
2
comments
Opened 23 other issues in 1 repository
replicate/replicate
22
open
1
closed
- PyTorch Lightning integration
- Vendor pkg_resources?
- Add support for NO_COLOR
- Word wrapping in console.log
- Better handling of metadata load failures
- Add datetime type to param.Value
- Better validation for --filter string
- Generalize replicate diff to >2 experiments/checkpoints
- Add --json flag to replicate diff
- Make CheckpointList.metrics more discoverable
- More robust error checking in plotting functions
- Add smoothing argument to plotting functions
- Add status to experiments in Python API
- Speed up experiment delete
- Consolidate delete logic
- checkpoint.open() shouldn't read entire file into memory
- More rigorous config validation in Python library
- System metrics
- replicate.init() should fail hard, experiment.checkpoint() should just warn
- NaN values cannot be parsed in the CLI
- NaN values are listed as "best" in the Python API.
- replicate ls shows step 0 if step is not defined in checkpoints
- Creating new GCS bucket in Colab throws error 400
2
contributions
in private repositories
Nov 15 – Nov 17