Refactor ContrastiveDataset and ContrastiveDistillationDataset #579

Open

DemirTonchev wants to merge 24 commits intohuggingface:mainfrom

DemirTonchev:refactor-contrds

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Refactor ContrastiveDataset and ContrastiveDistillationDataset #579

Refactor ContrastiveDataset and ContrastiveDistillationDataset #579
DemirTonchev wants to merge 24 commits intohuggingface:mainfrom
DemirTonchev:refactor-contrds

Commits on Dec 20, 2024

fixed to work with processing_class instead tokenizer after transformers 4.45.2

Commits on Dec 23, 2024

refactor attempt for ContrastiveDataset so that it does not blow RAM with bigger dataset

Commits on Dec 24, 2024

added Samplit strategy enum

improved logic and fixed iterator pattern

fix for negative samples formula

added multilalbel support as in the original implementation

Commits on Dec 25, 2024

ContrastiveDataset iterator refactor

trainer fixed to work with ContrastiveDataset iter method

ContrastiveDistillationDataset iter refactor

Commits on Dec 26, 2024

typing fix

Commits on Jan 10, 2025

Merge branch 'main' into pr-579

TypeAlias will be deprecated in 3.12 again, so let's avoid it

Remove args.sampling_strategy from ContrastiveDistillationDataset init

Run formatting

Add 'save_strategy="no"' in tests to counteract transformers v4.48.0 bug

Commits on Jan 11, 2025

fix some autocomplete mistake

Merge branch 'refactor-contrds' of github.com:DemirTonchev/setfit into refactor-contrds

Commits on Jan 12, 2025

cleanup leftovers attributes from old logic

Commits on Jan 13, 2025

docs

fix to correctly specify positives and negatives when passing num_iterations

fix for negative or positive examples only(not really possible)

removed unnecessary np uses and casting

safeguard for oversampling strategy, when DS is negatives only or single pos pair

Commits on Apr 9, 2025

Merge branch 'huggingface:main' into refactor-contrds

Refactor ContrastiveDataset and ContrastiveDistillationDataset #579DemirTonchev wants to merge 24 commits intohuggingface:mainhuggingface/setfit:mainfrom DemirTonchev:refactor-contrdsDemirTonchev/setfit:refactor-contrdsCopy head branch name to clipboard

Commits

Commits on Dec 20, 2024

Commits on Dec 23, 2024

Commits on Dec 24, 2024

Commits on Dec 25, 2024

Commits on Dec 26, 2024

Commits on Jan 10, 2025

Commits on Jan 11, 2025

Commits on Jan 12, 2025

Commits on Jan 13, 2025

Commits on Apr 9, 2025

Refactor ContrastiveDataset and ContrastiveDistillationDataset #579
DemirTonchev wants to merge 24 commits intohuggingface:mainfrom
DemirTonchev:refactor-contrds