Upload inference.py

inference.py

@@ -0,0 +1,11 @@
+# inference.py
+
+from text_embeddings_runner import EmbeddingRunner
+
+def create_runner():
+    # Force PyTorch runner on GPU
+    return EmbeddingRunner.from_pretrained(
+        repo_id="magnifi/optimized-semcache-embeds-en-final",
+        backend="pytorch",
+        device="cuda"
+    )