Skip to content

GitLab

Unverified Commit 0b7ade0d authored by Michael Yang's avatar Michael Yang Committed by GitHub
Browse files

Merge pull request #1212 from jmorganca/mxyng/metal 

enable metal for fp32, q5_0, q5_1
parents 31ab453d 19b7a4d7
Hide whitespace changes
Inline Side-by-side
Showing with 1 addition and 8 deletions
llm/llm.go
View file @ 0b7ade0d
...@@ -41,20 +41,13 @@ func New(workDir, model string, adapters []string, opts api.Options) (LLM, error ...@@ -41,20 +41,13 @@ func New(workDir, model string, adapters []string, opts api.Options) (LLM, error
if runtime.GOOS == "darwin" { if runtime.GOOS == "darwin" {
switch ggml.FileType() { switch ggml.FileType() {
case "Q8_0": case "F32", "Q5_0", "Q5_1", "Q8_0":
if ggml.Name() != "gguf" && opts.NumGPU != 0 { if ggml.Name() != "gguf" && opts.NumGPU != 0 {
// GGML Q8_0 do not support Metal API and will // GGML Q8_0 do not support Metal API and will
// cause the runner to segmentation fault so disable GPU // cause the runner to segmentation fault so disable GPU
log.Printf("WARNING: GPU disabled for F32, Q5_0, Q5_1, and Q8_0") log.Printf("WARNING: GPU disabled for F32, Q5_0, Q5_1, and Q8_0")
opts.NumGPU = 0 opts.NumGPU = 0
} }
case "F32", "Q5_0", "Q5_1":
if opts.NumGPU != 0 {
// F32, Q5_0, Q5_1, and Q8_0 do not support Metal API and will
// cause the runner to segmentation fault so disable GPU
log.Printf("WARNING: GPU disabled for F32, Q5_0, Q5_1, and Q8_0")
opts.NumGPU = 0
}
} }
var requiredMemory int64 var requiredMemory int64
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment