Merge pull request #1218 from vespa-engine/thomasht86/no-model-url-validation-default

Thomasht86/no model url validation default

Author: thomasht86

tests/unit/test_models.py

@@ -2067,14 +2067,20 @@ def test_hf_model_urls_are_valid(self, model_name):
         HF_MODELS entry return HTTP 200 status codes.
         """
         config = HF_MODELS[model_name]
-        # URL validation happens in __post_init__, so if we get here without
-        # an exception, the URLs are valid. But let's also verify the config
-        # has the expected URLs set.
-        print(f"Validating URLs for model: {model_name}")
+
+        # Verify config has URLs set
         assert config.model_url is not None, f"{model_name} should have model_url"
         assert (
             config.tokenizer_url is not None
         ), f"{model_name} should have tokenizer_url"
-        assert (
-            config.validate_urls is True
-        ), f"{model_name} should have validate_urls=True"
+
+        # Explicitly validate URLs by creating a new config with validation enabled
+        # (HF_MODELS uses validate_urls=False by default for fast import)
+        print(f"Validating URLs for model: {model_name}")
+        ModelConfig(
+            model_id=config.model_id,
+            embedding_dim=config.embedding_dim,
+            model_url=config.model_url,
+            tokenizer_url=config.tokenizer_url,
+            validate_urls=True,  # Explicitly enable validation
+        )

vespa/models.py

@@ -131,7 +131,7 @@ class ModelConfig:
             Optional, if not set the Vespa embedder uses its internal default (False).
         query_prepend: Optional instruction to prepend to query text
         document_prepend: Optional instruction to prepend to document text
-        validate_urls: Whether to validate URLs by checking they return HTTP 200 (default: True)
+        validate_urls: Whether to validate URLs by checking they return HTTP 200 (default: False)
     """
 
     model_id: str
@@ -154,7 +154,7 @@ class ModelConfig:
     normalize: Optional[bool] = None
     query_prepend: Optional[str] = None
     document_prepend: Optional[str] = None
-    validate_urls: bool = True
+    validate_urls: bool = False
     config_hash: str = field(default="", init=False)
 
     def __post_init__(self):