huggingface/transformers

Pull Requests Commits

run feat/kv_cache_class

ydshieh committed 2 years ago

6a9f6d7a

gante committed 2 years ago

6ac272f6

add Cache support to Phi+FA2

gante committed 2 years ago

d08ac80f

gante committed 2 years ago

58618a78

now also compatible with input_embeds input

gante committed 2 years ago

220dc8f4

fix sink cache and add hard integration test

gante committed 2 years ago

5b701e6a

all but sinkcache integration tests

gante committed 2 years ago

0f20709e

fix group beam search

gante committed 2 years ago

fa813044

cache can be a list of tuples in some models

gante committed 2 years ago

bb4f5bda

fix beam search bug

gante committed 2 years ago

a224b992

gante committed 2 years ago

d8540452

gante committed 2 years ago

3c4be9b9

gante committed 2 years ago

eefb64a0

gante committed 2 years ago

86d5db91

more tests, more docstrings, add to docs

gante committed 2 years ago

5eefc2c7

gante committed 2 years ago

6e50fc25

gante committed 2 years ago

e3ca76be

gante committed 2 years ago

13a235a3

Apply suggestions from code review

gante committed 2 years ago

ceb28268

fix failing test

gante committed 2 years ago

26f78885

apply PR suggestions

gante committed 2 years ago

7033f118

Apply suggestions from code review

gante committed 2 years ago

0c43eba1

correct failing test

gante committed 2 years ago

93dbd7d0

add default to model_kwargs.get('use_legacy_cache')

gante committed 2 years ago

53959cc9

gante committed 2 years ago

631427f5

Cache class working with generate (#1)

gante committed 2 years ago

558541a1

manual rebase with main

gante committed 2 years ago

46da03d9

Match import style

tomaarsen committed 2 years ago

d002c5e1

Remove copy utility from deprecated OpenLlama

tomaarsen committed 2 years ago

0ce084e8

Undo unnecessary newline change

tomaarsen committed 2 years ago

89e9e609

Older