dearwolf/LocalAI
1
0
Fork 0
mirror of https://github.com/mudler/LocalAI.git synced 2025-05-20 02:24:59 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 21

feat: add llama-master backend (#752)

Browse source 
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
This commit is contained in:
Ettore Di Giacinto 2023-07-17 23:58:15 +02:00 • committed by GitHub
parent fb6cce487f
commit 6352448b72
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
7 changed files with 223 additions and 12 deletions
Download patch file Download diff file Expand all files Collapse all files

17
pkg/model/initializers.go
View file

@ -37,6 +37,7 @@ const (
Gpt4All = "gpt4all"
FalconBackend = "falcon"
FalconGGMLBackend = "falcon-ggml"
LlamaMasterBackend = "llama-master"
BertEmbeddingsBackend = "bert-embeddings"
RwkvBackend = "rwkv"
@ -47,14 +48,13 @@ const (
//GGLLMFalconBackend = "falcon"
)
var autoLoadBackends []string = []string{
var AutoLoadBackends []string = []string{
LlamaBackend,
Gpt4All,
RwkvBackend,
FalconBackend,
WhisperBackend,
GPTNeoXBackend,
BertEmbeddingsBackend,
LlamaMasterBackend,
FalconGGMLBackend,
GPTJBackend,
Gpt2Backend,
@ -62,7 +62,6 @@ var autoLoadBackends []string = []string{
MPTBackend,
ReplitBackend,
StarcoderBackend,
BloomzBackend,
}
func (ml *ModelLoader) StopGRPC() {
@ -186,7 +185,7 @@ func (ml *ModelLoader) BackendLoader(opts ...Option) (model *grpc.Client, err er
backend := strings.ToLower(o.backendString)
switch backend {
case LlamaBackend, GPTJBackend, DollyBackend,
case LlamaBackend, LlamaMasterBackend, GPTJBackend, DollyBackend,
MPTBackend, Gpt2Backend, FalconBackend,
GPTNeoXBackend, ReplitBackend, StarcoderBackend, BloomzBackend,
RwkvBackend, LCHuggingFaceBackend, BertEmbeddingsBackend, FalconGGMLBackend, StableDiffusionBackend, WhisperBackend:
@ -217,10 +216,7 @@ func (ml *ModelLoader) GreedyLoader(opts ...Option) (*grpc.Client, error) {
ml.mu.Unlock()
var err error
for _, b := range autoLoadBackends {
if b == BloomzBackend || b == WhisperBackend || b == RwkvBackend { // do not autoload bloomz/whisper/rwkv
continue
}
for _, b := range AutoLoadBackends {
log.Debug().Msgf("[%s] Attempting to load", b)
model, modelerr := ml.BackendLoader(
@ -236,6 +232,9 @@ func (ml *ModelLoader) GreedyLoader(opts ...Option) (*grpc.Client, error) {
} else if modelerr != nil {
err = multierror.Append(err, modelerr)
log.Debug().Msgf("[%s] Fails: %s", b, modelerr.Error())
} else if model == nil {
err = multierror.Append(err, modelerr)
log.Debug().Msgf("[%s] Fails: %s", b, "backend returned no usable model")
}
}