ExternalVendorCode/LocalAI
1
0
Fork 0
mirror of https://github.com/mudler/LocalAI.git synced 2024-12-18 20:27:57 +00:00
Code Issues Actions 12 Packages Projects Releases Wiki Activity

feat(diffusers): allow multiple lora adapters (#4081)

Browse Source 
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
This commit is contained in:
Ettore Di Giacinto 2024-11-05 15:14:33 +01:00 committed by GitHub
parent 20cd8814c1
commit 947224b952
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
4 changed files with 39 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
backend/backend.proto
View File

@ -235,6 +235,9 @@ message ModelOptions {
bool NoKVOffload = 57;
string ModelPath = 59;
repeated string LoraAdapters = 60;
repeated float LoraScales = 61;
}
message Result {

16
backend/python/diffusers/backend.py
View File

@ -311,10 +311,24 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
if request.LoraAdapter:
# Check if its a local file and not a directory ( we load lora differently for a safetensor file )
if os.path.exists(request.LoraAdapter) and not os.path.isdir(request.LoraAdapter):
# self.load_lora_weights(request.LoraAdapter, 1, device, torchType)
self.pipe.load_lora_weights(request.LoraAdapter)
else:
self.pipe.unet.load_attn_procs(request.LoraAdapter)
if len(request.LoraAdapters) > 0:
i = 0
adapters_name = []
adapters_weights = []
for adapter in request.LoraAdapters:
if not os.path.isabs(adapter):
adapter = os.path.join(request.ModelPath, adapter)
self.pipe.load_lora_weights(adapter, adapter_name=f"adapter_{i}")
adapters_name.append(f"adapter_{i}")
i += 1
for adapters_weight in request.LoraScales:
adapters_weights.append(adapters_weight)
self.pipe.set_adapters(adapters_name, adapter_weights=adapters_weights)
if request.CUDA:
self.pipe.to('cuda')

2
core/backend/options.go
View File

@ -125,6 +125,8 @@ func grpcModelOpts(c config.BackendConfig) *pb.ModelOptions {
CFGScale: c.Diffusers.CFGScale,
LoraAdapter: c.LoraAdapter,
LoraScale: c.LoraScale,
LoraAdapters: c.LoraAdapters,
LoraScales: c.LoraScales,
F16Memory: f16,
LoraBase: c.LoraBase,
IMG2IMG: c.Diffusers.IMG2IMG,

2
core/config/backend_config.go
View File

@ -138,6 +138,8 @@ type LLMConfig struct {
NUMA bool `yaml:"numa"`
LoraAdapter string `yaml:"lora_adapter"`
LoraBase string `yaml:"lora_base"`
LoraAdapters []string `yaml:"lora_adapters"`
LoraScales []float32 `yaml:"lora_scales"`
LoraScale float32 `yaml:"lora_scale"`
NoMulMatQ bool `yaml:"no_mulmatq"`
DraftModel string `yaml:"draft_model"`