mirror of
https://github.com/theroyallab/YALS.git
synced 2026-04-20 14:29:47 +00:00
Config: Clarify tensor split
Tensor split uses ratios rather than GBs of vram. This should solve size incosistencies. Signed-off-by: kingbri <8082010+kingbri1@users.noreply.github.com>
This commit is contained in:
@@ -79,7 +79,7 @@ model:
|
||||
# Set this to 999 to offload all layers to the GPU
|
||||
num_gpu_layers: 0
|
||||
|
||||
# An integer array of GBs of VRAM to split between GPUs (default: []).
|
||||
# An integer array defining the ratio of VRAM to split for each GPU (default: []).
|
||||
# Going over the max amount of GPUs will crash when loading the model
|
||||
gpu_split: []
|
||||
|
||||
|
||||
Reference in New Issue
Block a user