Initial benchmarks

tests/integration/test_image_inference.py

@@ -0,0 +1,18 @@
+from chandra.model import InferenceManager, BatchInputItem
+
+
+def test_inference_image(simple_text_image):
+    manager = InferenceManager(method="hf")
+    batch = [
+        BatchInputItem(
+            image=simple_text_image,
+            prompt_type="ocr_layout",
+        )
+    ]
+    outputs = manager.generate(batch, max_output_tokens=128)
+    assert len(outputs) == 1
+    output = outputs[0]
+    assert "Hello, World!" in output.markdown
+
+    chunks = output.chunks
+    assert len(chunks) == 1