huggingface/lighteval

Pull Requests Commits

clefourrier committed 1 year ago

ce0015e2

we mock iep model creation

clefourrier committed 1 year ago

35b7e94e

added doc page to tree

clefourrier committed 1 year ago

730f226a

Merge branch 'main' into test_caching

clefourrier committed 1 year ago

Verified f4d5b145

we separate the interface function from the actual logic in models, and wrap up tests

clefourrier committed 1 year ago

d2ff97c0

Fix vLLM doc (#912)

lewtun committed 1 year ago

Verified bfa60767

clefourrier committed 1 year ago

348b7d23

cleanup, some things hanging when being removed from the refacto - no idea why the imports did not fail

clefourrier committed 1 year ago

19fa6d13

remove useless parameter

clefourrier committed 1 year ago

ba776ab8

removed loglikelihood single tokens since it was removed everywhere in theory - added docs on caching

clefourrier committed 1 year ago

3a9d1fbe

we now actively load cached samples after processing the needed items

clefourrier committed 1 year ago

94e31f18

cache management is working well with no DP for accelerate - need to 1)test with DP 2) add a system where we load cached samples in mem *after* processing the other items

clefourrier committed 1 year ago

03eb193c

Merge branch 'main' into test_caching

clefourrier committed 1 year ago

187aaa72

fixed caching system for predictions, and better logging

clefourrier committed 1 year ago

205ab1e5

added caching to transformers

clefourrier committed 1 year ago

39c03e7c

Merge branch 'main' into test_caching

clefourrier committed 1 year ago

76c8b6e8

Fix log system prompt (#907)

clefourrier committed 1 year ago

Verified 7693a0fd

Moved some files (#905)

clefourrier committed 1 year ago

Verified d0cd4c91

Debug log likelihood evals which are broken for accelerate (#901)

clefourrier committed 1 year ago

Verified ea1dd184

Number of fixes to run accelerate evaluations (#898)

clefourrier committed 1 year ago

Verified 64f93b0e

Update transformers and vllm dependencies (#899)

clefourrier committed 1 year ago

Verified 865335e4

Added post processing (for reasoning tokens) to pipeline (#882)

clefourrier committed 1 year ago

Verified d7beacbc

fix path concatenation (#891)

clefourrier committed 1 year ago

Verified 994e9e07

Fix DP>1 & TP>1 evals with vllm (#841)

nmntz committed 1 year ago

Verified 24895519

Adds continuous batching (#850)

NathanHB committed 1 year ago

Verified 99bfd9f2

Fix typo: CONCURENT_CALLS -> CONCURRENT_CALLS (#884)

muupan committed 1 year ago

Verified 1404ba1d

Fixed typo in Python API documentation (#862)

dtung8068 committed 1 year ago

Verified 3b1126ad

Automatically infer whether to use a chat template or not instead of using kwargs (#885)

clefourrier committed 1 year ago

Verified e7d70902

clefourrier committed 1 year ago

aaae0ea2

clefourrier committed 1 year ago

e0cd1330

Older