Skip to content
This repository has been archived by the owner on Dec 16, 2022. It is now read-only.

Using only a few instances for guessing sorting keys #3812

Closed
wants to merge 2 commits into from

Conversation

matt-gardner
Copy link
Contributor

Partial fix for #3664 (and maybe enough to close that issue entirely as good enough). This just grabs the first 10 instances and uses those to guess a sorting key.

@matt-gardner
Copy link
Contributor Author

I know this is in a class that you're looking to remove @DeNeutoy, but it's a small change, and the code here will likely still be needed in whatever you're replacing this file with.

@matt-gardner
Copy link
Contributor Author

Also, I accidentally requested a review before checking if the tests were good enough; I was going to see in the coverage report if this needed any more tests. I'm somewhat inclined to leave it, though, as you'll be changing things here, unless there's a particular test you'd find helpful that you'd like me to add.

DeNeutoy added a commit to DeNeutoy/allennlp that referenced this pull request Feb 22, 2020
@DeNeutoy
Copy link
Contributor

Closing, as I incorporated this in #3700

@DeNeutoy DeNeutoy closed this Feb 22, 2020
@matt-gardner matt-gardner deleted the sorting_key branch February 22, 2020 15:06
DeNeutoy added a commit to DeNeutoy/allennlp that referenced this pull request Feb 23, 2020
This reverts commit 8a08899.
DeNeutoy added a commit that referenced this pull request Feb 26, 2020
* example for feedback

* remove all existing multiprocessing

* sneak torch datasets inside DatasetReader

* lint

* trainer_v2, We Love To See It

* datasets have index_with now, not iterators

* use iter, custom collate function in allennlp wrapper

* we don't even need the data in the trainer anymore

* all trainer tests passing

* black

* make find learning rate work

* update test fixtures to new config

* get train command tests mostly working

* lazily construct samplers, index lazy datasets

* update some fixtures

* evaluate tests passing

* all command tests passing

* lint

* update model test case, common and module tests passing

* fix test interdependence introduced by #3762

* more test interdependence

* tests tests tests

* remove unnecessary brackets

Co-Authored-By: Santiago Castro <[email protected]>

* update a chunk of the configs

* fix archival test, couple more configs

* rm pointless gan test

* more tests passing

* add current state of from params changes

* Revert "add current state of from params changes"

This reverts commit ad45659.

* updated understanding of Lazy

* add discussion of None comparison to Lazy

* lint

* it's a hard doc life

* pull samplers into separate file

* more docs updates

* fold in #3812

* remove torch dataset

* add example to lazy

* rename to collate

* no kwargs

* Revert "fold in #3812"

This reverts commit 8a08899.

* don't break up dataset

* add comment to iterable dataset len

* improve docstrings, build dataloader using partial_objects

* flake

* give dataloader a default implementation

* safer default for DataLoader init

* more coherent dir structure

* update imports

* add a test for the BucketBatchSampler

* split bucket sampler into own file, tests

* PR comments

Co-authored-by: Santiago Castro <[email protected]>
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants