PR #22069 Fix position embeddings for GPT-J and CodeGen

Fix position embeddings for GPT-J and CodeGen #22069

gante merged 17 commits into huggingface:main from njhill:fix_pos_embeds

gante commented on 2023-03-10

njhill commented on 2023-03-10

njhill force pushed to e41a6f4c 2 years ago

gante approved these changes on 2023-03-11

gante requested a review from

sgugger 2 years ago

sgugger approved these changes on 2023-03-13

njhill force pushed 2 years ago

Revert "[GPT-J] add deprecation warning (#21869)"

eb9f614a

Fix position embeddings for GPT-J and CodeGen

7fdd8331

Address review comments from @gante

06a43e3e

Fix "Copied from" comment referencing wrong function

6c528598

Fix copy/paste mistake

b8c411c2

Fix training path

c64d9a31

Hopefully make torch.fx happy

129d0610

Move position_ids long cast

6955c28e

Revert "Hopefully make torch.fx happy"

ec274801

Changes to help with torch.fx tracing

a6c74708

Linter fix

67f36498

Correct position_ids tensor type hint

ca2e0a31

Work-around torch.fx tracing issue

95a75033

Get the changes to work with torch.fx

3c124f2a

njhill force pushed to 3c124f2a 2 years ago

njhill marked this pull request as ready for review 2 years ago

michaelbenayoun approved these changes on 2023-03-20

Address review comment from @michaelbenayoun

5f9cacd2

Another small adjustment

81599f1c

gante approved these changes on 2023-03-21

Add explanatory comment; small code tidyup

7888c100

gante merged 4e94c6c0 into main 2 years ago

njhill deleted the fix_pos_embeds branch 2 years ago

Reviewers

michaelbenayoun

gante

sgugger

Assignees

No one assigned

Labels

None yet

Milestone

No milestone

transformers Fix position embeddings for GPT-J and CodeGen #22069 Merged

Fix position embeddings for GPT-J and CodeGen #22069

transformers
Fix position embeddings for GPT-J and CodeGen
#22069

Merged