feat: 实现模型动态注册机制

This commit is contained in:
2026-04-05 19:38:12 +08:00
parent ff43a2fab8
commit fc278d17ab
25 changed files with 686 additions and 651 deletions
+6 -2
View File
@@ -90,8 +90,12 @@ from astrai.inference import InferenceEngine, GenerationRequest
param = ModelParameter.load("your_model_dir")
param.to(device="cuda", dtype=torch.bfloat16)
# Create engine
engine = InferenceEngine(param)
# Create engine with separate model and tokenizer
engine = InferenceEngine(
model=param.model,
tokenizer=param.tokenizer,
config=param.config,
)
# Build request
request = GenerationRequest(