PR #306 Fix causal attention mask

Fix causal attention mask #306

thomasw21 merged 8 commits into main from thomas/fix_causal_attention_mask

Raname attn_mask as it's bad IMO

dc9709b7

WIP: make sure that custom kernels have the same API as their torch v…

fbabf8ca

Add a test that we are in fact using a custom kernel

3a857ddb

Woops wrong import

80613763

Woops

946ded09

Fix device issue + relax test to assert close instead of equal

038bddcb

Woops fix the causal mask

96823a83

thomasw21 marked this pull request as ready for review 3 years ago

thomasw21 requested a review from

stas00 3 years ago

thomasw21 requested a review from

Muennighoff 3 years ago

thomasw21 commented on 2022-07-07

stas00 commented on 2022-07-07

Update test + use getattr

c20dd196

thomasw21 requested a review from

stas00 3 years ago

Muennighoff approved these changes on 2022-07-07

stas00 approved these changes on 2022-07-07

thomasw21 merged 38607ae9 into main 3 years ago

thomasw21 deleted the thomas/fix_causal_attention_mask branch 3 years ago

Reviewers

Muennighoff

stas00

Assignees

No one assigned

Labels

None yet

Milestone

No milestone