Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
OpenDAS
ollama
Commits
b1f71233
Unverified
Commit
b1f71233
authored
Oct 02, 2023
by
Bruce MacDonald
Committed by
GitHub
Oct 02, 2023
Browse files
clean up num_gpu calculation code (#673)
parent
1fbf3585
Changes
1
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
with
4 additions
and
4 deletions
+4
-4
llm/llama.go
llm/llama.go
+4
-4
No files found.
llm/llama.go
View file @
b1f71233
...
...
@@ -218,7 +218,6 @@ func NumGPU(numLayer, fileSizeBytes int64, opts api.Options) int {
if
opts
.
NumGPU
!=
-
1
{
return
opts
.
NumGPU
}
n
:=
1
// default to enable metal on macOS
if
runtime
.
GOOS
==
"linux"
{
vramMib
,
err
:=
CheckVRAM
()
if
err
!=
nil
{
...
...
@@ -235,10 +234,11 @@ func NumGPU(numLayer, fileSizeBytes int64, opts api.Options) int {
// TODO: this is a rough heuristic, better would be to calculate this based on number of layers and context size
bytesPerLayer
:=
fileSizeBytes
/
numLayer
// set n to the max number of layers we can fit in VRAM
return
int
(
totalVramBytes
/
bytesPerLayer
)
// max number of layers we can fit in VRAM
layers
:=
int
(
totalVramBytes
/
bytesPerLayer
)
log
.
Printf
(
"%d MiB VRAM available, loading up to %d GPU layers"
,
vramMib
,
layers
)
log
.
Printf
(
"%d MiB VRAM available, loading up to %d GPU layers"
,
vramMib
,
n
)
return
layers
}
// default to enable metal on macOS
return
1
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment