PR #45324 Gemma4 resizing per layer inputs

gemma4

zucchini-nlp committed 74 days ago

mask out per-layer input correctly!

zucchini-nlp committed 74 days ago

revert gemma3n and add t5gemma

zucchini-nlp committed 73 days ago

fix repo

zucchini-nlp committed 73 days ago

the test

zucchini-nlp committed 73 days ago

fix some tests

zucchini-nlp committed 73 days ago

gemma3n also has a causal model

zucchini-nlp committed 73 days ago

fix repo

zucchini-nlp committed 73 days ago

Merge branch 'main' into gemma4-resizing

zucchini-nlp committed 73 days ago

Apply repo consistency fixes

github-actions[bot] committed 73 days ago

Merge remote-tracking branch 'upstream/main' into gemma4-resizing

zucchini-nlp committed 68 days ago

update attr even if weights are tied

zucchini-nlp committed 68 days ago

transformers Gemma4 resizing per layer inputs #45324 Merged