Files
LocalAI/pkg/vram/gguf_reader.go
2026-04-04 15:14:35 +02:00

48 lines
1.1 KiB
Go

package vram
import (
"context"
"strings"
gguf "github.com/gpustack/gguf-parser-go"
"github.com/mudler/LocalAI/pkg/downloader"
)
type defaultGGUFReader struct{}
func (defaultGGUFReader) ReadMetadata(ctx context.Context, uri string) (*GGUFMeta, error) {
u := downloader.URI(uri)
urlStr := u.ResolveURL()
if strings.HasPrefix(uri, downloader.LocalPrefix) {
f, err := gguf.ParseGGUFFile(urlStr)
if err != nil {
return nil, err
}
return ggufFileToMeta(f), nil
}
if !u.LooksLikeHTTPURL() {
return nil, nil
}
f, err := gguf.ParseGGUFFileRemote(ctx, urlStr)
if err != nil {
return nil, err
}
return ggufFileToMeta(f), nil
}
func ggufFileToMeta(f *gguf.GGUFFile) *GGUFMeta {
arch := f.Architecture()
meta := &GGUFMeta{
BlockCount: uint32(arch.BlockCount),
EmbeddingLength: uint32(arch.EmbeddingLength),
HeadCount: uint32(arch.AttentionHeadCount),
HeadCountKV: uint32(arch.AttentionHeadCountKV),
MaximumContextLength: arch.MaximumContextLength,
}
if meta.HeadCountKV == 0 {
meta.HeadCountKV = meta.HeadCount
}
return meta
}