llama.cpp
llama : minimize size used for state save/load
#4820
Merged

llama : minimize size used for state save/load #4820

ggerganov merged 7 commits into ggml-org:master from dfriehs:min-state-size
dfriehs
dfriehs examples : save-load-state: save only required state
aee95df8
dfriehs llama : only reserve n_vocab * n_batch at most for logits
0093dea9
dfriehs llama : always reserve n_vocab * n_batch for logits
b9c60dec
dfriehs llama : only save and restore used logits
5ee58147
dfriehs llama : use ostringstream and istringstream for save and load
e872af8d
dfriehs llama : serialize rng into minimum amount of space required
69d44e3e
dfriehs llama : break session version due to serialization changes
06e3b4f5
ggerganov
dfriehs
slaren
ggerganov ggerganov merged df845cc9 into master 1 year ago
dfriehs dfriehs deleted the min-state-size branch 1 year ago

Login to write a write a comment.

Login via GitHub

Reviewers
No reviews
Assignees
No one assigned
Labels
Milestone