[inference] Necessary breaking change: nest task-specific route inside of model route (#3044)

* nest task-specific route inside of model route

* add unit tests

---------

Co-authored-by: Celina Hanouti <[email protected]>

Author: julien-c

src/huggingface_hub/inference/_providers/hf_inference.py

@@ -42,7 +42,7 @@ def _prepare_url(self, api_key: str, mapped_model: str) -> str:
             return mapped_model
         return (
             # Feature-extraction and sentence-similarity are the only cases where we handle models with several tasks.
-            f"{self.base_url}/pipeline/{self.task}/{mapped_model}"
+            f"{self.base_url}/models/{mapped_model}/pipeline/{self.task}"
             if self.task in ("feature-extraction", "sentence-similarity")
             # Otherwise, we use the default endpoint
             else f"{self.base_url}/models/{mapped_model}"

tests/test_inference_providers.py

@@ -435,6 +435,20 @@ def test_prepare_url(self):
 
         assert helper._prepare_url("hf_test_token", "https://any-url.com") == "https://any-url.com"
 
+    def test_prepare_url_feature_extraction(self):
+        helper = HFInferenceTask("feature-extraction")
+        assert (
+            helper._prepare_url("hf_test_token", "username/repo_name")
+            == "https://router.huggingface.co/hf-inference/models/username/repo_name/pipeline/feature-extraction"
+        )
+
+    def test_prepare_url_sentence_similarity(self):
+        helper = HFInferenceTask("sentence-similarity")
+        assert (
+            helper._prepare_url("hf_test_token", "username/repo_name")
+            == "https://router.huggingface.co/hf-inference/models/username/repo_name/pipeline/sentence-similarity"
+        )
+
     def test_prepare_payload_as_dict(self):
         helper = HFInferenceTask("text-classification")
         mapping_info = InferenceProviderMapping(