Qwen3 hpu support (#656)

Signed-off-by: Liu, Kaixuan <[email protected]>

Author: kaixuanliu

backends/python/server/text_embeddings_server/models/__init__.py

@@ -13,6 +13,7 @@
 from text_embeddings_server.models.classification_model import ClassificationModel
 from text_embeddings_server.models.jinaBert_model import FlashJinaBert
 from text_embeddings_server.models.flash_mistral import FlashMistral
+from text_embeddings_server.models.flash_qwen3 import FlashQwen3
 from text_embeddings_server.utils.device import get_device, use_ipex
 
 __all__ = ["Model"]
@@ -121,6 +122,12 @@ def get_model(model_path: Path, dtype: Optional[str], pool: str):
         except FileNotFoundError:
             return create_model(DefaultModel, model_path, device, datatype, pool)
 
+    if config.model_type == "qwen3" and device.type == "hpu":
+        try:
+            return create_model(FlashQwen3, model_path, device, datatype, pool)
+        except FileNotFoundError:
+            return create_model(DefaultModel, model_path, device, datatype, pool)
+
     # Default case
     if config.architectures[0].endswith("Classification"):
         return create_model(ClassificationModel, model_path, device, datatype)