Harness #28
Merged
Harness #28
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Hide any extra arguments we may have to pass along
No need for our own stack trace. It's confusing.
Makes it more convenient for saving results
Instead of using the monolithic attelo evaluate command, break the evaluation process down into - extracting folds (attelo enfold) - looping over folds: - learning and saving model (attelo learn) - decoding with saved model (shared!) (attelo decode) - summarising the results (attelo report) This may also one day open the way to running folds concurrency
Only Pandoc supports definition lists, I suppose
We want to maximise code between various experiments
Pick up an evaluation where we left off
Kill: - any scratch directories - any incomplete eval dirs It's also tempting to get rid of feature directories without evaluations but it's not likely to be a common case in future development
I want it to be as easy to know where we are at a glance.
ENH add random forest and decision tree classifiers
Ooops! I must have done this only locally on the cluster
Ooops! I must have done this only locally on the cluster
Conflicts: irit_rst_dt/local.py
I just noticed that it seems possible to use joblib parallel in a sort of producer/consumer pattern: we don't need to have all our jobs ready in one go So if I understand correctly, this means we can have a generator expression that yields decoder jobs as soon as the parser for them has been fit. Outcome: 1. No more waiting for all the learners to complete before we decode; start decoding as soon as the relevant learner is done 2. But still apply parallelism across decoders for multiple configurations So if all learners are fast, that's great, we get to run lots of jobs in parallel. If some learners are slow, that's hopefully still OK because we can still work on decoding while they're crunching away. I hope this means less dead space. More cores humming away in parallel
…eptrons FIX update and fix calls to perceptrons
Conflicts: .gitignore README.md requirements.txt setup.py
You will need to define HarnessConfig which contains settings, file name conventions, and evaluations for the particular harness. This was taken from irit_rst_dt
Don't try to hash models we may not have built
kowey
added a commit
that referenced
this pull request
May 18, 2015
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Notes:
I may just merge this myself, as I understand Mathieu is busy