]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
llama : dynamic temperature sampling (#4972)
authorl3utterfly <redacted>
Thu, 25 Jan 2024 20:06:22 +0000 (05:06 +0900)
committerGitHub <redacted>
Thu, 25 Jan 2024 20:06:22 +0000 (22:06 +0200)
commit5eaf9964fc797d4585c214db32a463d557f3ed33
treee41be4957a366e1210769298156de9f09b9593c8
parentd292f4f2047963f558dd516f1baaa71793e9acf2
llama : dynamic temperature sampling (#4972)

* implemented dynamic temperature sampling from koboldcpp

* removed trailing whitespace

* removed unused temp parameter in llama_sample_entropy

* exposed exponent_val in dynamic temp sampler

* added debug check for printf statements

* use nullptr in llama_sample_softmax call during llama_sample_entropy

this avoids counting the time taken stats twice

Co-authored-by: Georgi Gerganov <redacted>
* return earlier if there is only 1 candiate (i.e. max_entropy == 0)

* reformat 't' case in llama_sample_queue

Co-authored-by: Jared Van Bortel <redacted>
* check for one or zero candidates case in llama_sample_entropy

---------

Co-authored-by: Georgi Gerganov <redacted>
Co-authored-by: Jared Van Bortel <redacted>
common/sampling.cpp
common/sampling.h
llama.cpp
llama.h