llama.cpp
Custom RoPE + bettter memory management for CUDA
#2295

Merged

Custom RoPE + bettter memory management for CUDA #2295

ikawrakow merged 2 commits into master from ik/context_extend_cuda

Custom RoPE + bettter memory management for CUDA

1cdbbbb3

ikawrakow requested a review from

JohannesGaessler 2 years ago

Adjusted look ahead in ggml_cuda_pool_malloc to 5%

b068f2f4

ggerganov approved these changes on 2023-07-21

ikawrakow merged d924522a into master 2 years ago

ikawrakow deleted the ik/context_extend_cuda branch 2 years ago

Reviewers

ggerganov

JohannesGaessler

Assignees

No one assigned

Labels

None yet

Milestone

No milestone