Commit 36fe2dee authored by Michael Yang's avatar Michael Yang
Browse files

only check system memory on macos

parent 4a8931f6
...@@ -56,38 +56,31 @@ func New(workDir, model string, adapters []string, opts api.Options) (LLM, error ...@@ -56,38 +56,31 @@ func New(workDir, model string, adapters []string, opts api.Options) (LLM, error
opts.NumGPU = 0 opts.NumGPU = 0
} }
} }
}
var requiredMemory int64
var f16Multiplier int64 = 2
switch ggml.ModelType() {
case "3B", "7B":
requiredMemory = 8 * format.GigaByte
case "13B":
requiredMemory = 16 * format.GigaByte
case "30B", "34B", "40B":
requiredMemory = 32 * format.GigaByte
case "65B", "70B":
requiredMemory = 64 * format.GigaByte
case "180B":
requiredMemory = 128 * format.GigaByte
f16Multiplier = 4
}
systemMemory := int64(memory.TotalMemory()) var requiredMemory int64
var f16Multiplier int64 = 2
videoMemory, err := CheckVRAM() switch ggml.ModelType() {
if err != nil{ case "3B", "7B":
videoMemory = 0 requiredMemory = 8 * format.GigaByte
} case "13B":
requiredMemory = 16 * format.GigaByte
case "30B", "34B", "40B":
requiredMemory = 32 * format.GigaByte
case "65B", "70B":
requiredMemory = 64 * format.GigaByte
case "180B":
requiredMemory = 128 * format.GigaByte
f16Multiplier = 4
}
totalMemory := systemMemory + videoMemory systemMemory := int64(memory.TotalMemory())
if ggml.FileType() == "F16" && requiredMemory*f16Multiplier > totalMemory { if ggml.FileType() == "F16" && requiredMemory*f16Multiplier > systemMemory {
return nil, fmt.Errorf("F16 model requires at least %s of total memory", format.HumanBytes(requiredMemory)) return nil, fmt.Errorf("F16 model requires at least %s of total memory", format.HumanBytes(requiredMemory))
} else if requiredMemory > totalMemory { } else if requiredMemory > systemMemory {
return nil, fmt.Errorf("model requires at least %s of total memory", format.HumanBytes(requiredMemory)) return nil, fmt.Errorf("model requires at least %s of total memory", format.HumanBytes(requiredMemory))
}
} }
switch ggml.Name() { switch ggml.Name() {
......
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