llama.cpp
sampling : refactor + optimize penalties sampler
#10803

Merged

sampling : refactor + optimize penalties sampler #10803

ggerganov merged 10 commits into master from gg/sampling-penalties

github-actions added testing

github-actions added examples

github-actions added devops

github-actions added server

ggerganov force pushed 1 year ago

ggerganov requested a review from

ngxson 1 year ago

ngxson approved these changes on 2024-12-12

slaren commented on 2024-12-12

ggerganov force pushed to a3125686 1 year ago

ggerganov commented on 2024-12-13

ggerganov requested a review from

slaren 1 year ago

ggerganov commented on 2024-12-14

sampling : refactor + optimize penalties sampler

0a1f7fb6

common : apply ignore_eos as logit bias

58a5c3bb

batched : remove penalties sampler

a04a5b52

params : allow penalty_last_n == -1 to be equal to context size

9847a375

common : by default, move the penalties at the end of the sampling chain

97261aa2

common : ignore all EOG tokens

1ff92962

common : move back the penalties at the front of the sampling chain

685c84c3

readme : restore hint about --ignore-eos flag [no ci]

60d26ded

llama : minor

e27c7119

webui : update

b58ebf30

ggerganov force pushed from 7415f3fd to b58ebf30 1 year ago

ggerganov merged 644fd71b into master 1 year ago

ggerganov deleted the gg/sampling-penalties branch 1 year ago

Reviewers

ngxson

slaren

Assignees

No one assigned

Labels

testing examples devops server

Milestone

No milestone

llama.cpp sampling : refactor + optimize penalties sampler #10803 Merged

sampling : refactor + optimize penalties sampler #10803

llama.cpp
sampling : refactor + optimize penalties sampler
#10803

Merged