fix: support sagemaker batch transform for clip (#6171)

jina/serve/runtimes/worker/http_csp_app.py

@@ -189,9 +189,14 @@ def construct_model_from_line(
                                 )
                             else:
                                 parsed_fields[field_name] = parsed_list
-                        # Handle direct assignment for basic types
+                        # General parsing attempt for other types
                         else:
-                            parsed_fields[field_name] = field_info.type_(field_str)
+                            if field_str:
+                                try:
+                                    parsed_fields[field_name] = field_info.type_(field_str)
+                                except (ValueError, TypeError):
+                                    # Fallback to parse_obj_as when type is more complex, e., AnyUrl or ImageBytes
+                                    parsed_fields[field_name] = parse_obj_as(field_info.type_, field_str)
 
                     return model(**parsed_fields)
 

tests/integration/docarray_v2/csp/SampleClipExecutor/README.md

@@ -0,0 +1,2 @@
+# SampleClipExecutor
+

tests/integration/docarray_v2/csp/SampleClipExecutor/config.yml

@@ -0,0 +1,8 @@
+jtype: SampleClipExecutor
+py_modules:
+  - executor.py
+metas:
+  name: SampleClipExecutor
+  description: 
+  url: 
+  keywords: []

tests/integration/docarray_v2/csp/SampleClipExecutor/executor.py

@@ -0,0 +1,43 @@
+from typing import Optional
+
+import numpy as np
+from docarray import BaseDoc, DocList
+from docarray.typing import NdArray
+from docarray.typing.bytes import ImageBytes
+from docarray.typing.url import AnyUrl
+from jina import Executor, requests
+from pydantic import Field
+
+
+class TextAndImageDoc(BaseDoc):
+    text: Optional[str] = None
+    url: Optional[AnyUrl] = None
+    bytes: Optional[ImageBytes] = None
+
+
+class EmbeddingResponseModel(TextAndImageDoc):
+    embeddings: NdArray = Field(description="The embedding of the texts", default=[])
+
+    class Config(BaseDoc.Config):
+        allow_population_by_field_name = True
+        arbitrary_types_allowed = True
+        json_encoders = {NdArray: lambda v: v.tolist()}
+
+
+class SampleClipExecutor(Executor):
+    @requests(on="/encode")
+    def foo(
+        self, docs: DocList[TextAndImageDoc], **kwargs
+    ) -> DocList[EmbeddingResponseModel]:
+        ret = []
+        for doc in docs:
+            ret.append(
+                EmbeddingResponseModel(
+                    id=doc.id,
+                    text=doc.text,
+                    url=doc.url,
+                    bytes=doc.bytes,
+                    embeddings=np.random.random((1, 64)),
+                )
+            )
+        return DocList[EmbeddingResponseModel](ret)

tests/integration/docarray_v2/csp/test_sagemaker_clip.py

@@ -0,0 +1,89 @@
+import csv
+import io
+import os
+
+import requests
+from jina.orchestrate.pods import Pod
+from jina.parsers import set_pod_parser
+
+sagemaker_port = 8080
+
+
+def test_provider_sagemaker_pod_rank():
+    args, _ = set_pod_parser().parse_known_args(
+        [
+            "--uses",
+            os.path.join(os.path.dirname(__file__), "SampleClipExecutor", "config.yml"),
+            "--provider",
+            "sagemaker",
+            "--provider-endpoint",
+            "encode",
+            "serve",  # This is added by sagemaker
+        ]
+    )
+    with Pod(args):
+        # Test the `GET /ping` endpoint (added by jina for sagemaker)
+        resp = requests.get(f"http://localhost:{sagemaker_port}/ping")
+        assert resp.status_code == 200
+        assert resp.json() == {}
+
+        # Test the `POST /invocations` endpoint for inference
+        # Note: this endpoint is not implemented in the sample executor
+        resp = requests.post(
+            f"http://localhost:{sagemaker_port}/invocations",
+            json={
+                "data": [
+                    {"url": "http://google.com"},
+                ]
+            },
+        )
+        assert resp.status_code == 200
+        resp_json = resp.json()
+        assert len(resp_json["data"]) == 1
+        assert len(resp_json["data"][0]["embeddings"][0]) == 64
+        assert resp_json["data"][0]["url"] == "http://google.com"
+
+
+def test_provider_sagemaker_pod_batch_transform_valid():
+    args, _ = set_pod_parser().parse_known_args(
+        [
+            "--uses",
+            os.path.join(os.path.dirname(__file__), "SampleClipExecutor", "config.yml"),
+            "--provider",
+            "sagemaker",
+            "serve",  # This is added by sagemaker
+        ]
+    )
+    with Pod(args):
+        # Test `POST /invocations` endpoint for batch-transform with valid input
+        with open(
+            os.path.join(os.path.dirname(__file__), "valid_clip_input.csv"), "r"
+        ) as f:
+            csv_data = f.read()
+
+        text = []
+        for line in csv.reader(
+            io.StringIO(csv_data),
+            delimiter=",",
+            quoting=csv.QUOTE_NONE,
+            escapechar="\\",
+        ):
+            text.append(line)
+
+        resp = requests.post(
+            f"http://localhost:{sagemaker_port}/invocations",
+            headers={
+                "accept": "application/json",
+                "content-type": "text/csv",
+            },
+            data=csv_data,
+        )
+        assert resp.status_code == 200
+        resp_json = resp.json()
+        assert len(resp_json["data"]) == 3
+        assert resp_json["data"][0]["text"] == "the cat is in my house"
+        assert (
+            resp_json["data"][1]["url"]
+            == "https://dummyimage3.com/333/000/fff.jpg&text=embed+this"
+        )
+        assert "hWjj1RNtNftP" in resp_json["data"][2]["bytes"]