llama.cpp
sampling : refactor init to use llama_sampling_params
#3696

Merged

Commits

sampling : refactor init to use llama_sampling_params

ggerganov committed 2 years ago
llama : combine repetition, frequency and presence penalties in 1 call

ggerganov committed 2 years ago
examples : remove embd-input and gptneox-wip

ggerganov committed 2 years ago
sampling : rename penalty params + reduce size of "prev" vector

ggerganov committed 2 years ago
sampling : add llama_sampling_print helper

ggerganov committed 2 years ago
sampling : hide prev behind API and apply #3661

ggerganov committed 2 years ago