ELin2025 commented on code in PR #5568:
URL: https://github.com/apache/texera/pull/5568#discussion_r3445099673


##########
common/workflow-operator/src/main/scala/org/apache/texera/amber/operator/huggingFace/codegen/ImageTaskCodegen.scala:
##########
@@ -0,0 +1,166 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied.  See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+
+package org.apache.texera.amber.operator.huggingFace.codegen
+
+/**
+  * Codegen for the Hugging Face image-pipeline task family.
+  *
+  * Splits into two sub-families:
+  *  - "image-only" tasks send raw image bytes as the request body and don't
+  *    consume the prompt column: image-classification, object-detection,
+  *    image-segmentation, image-to-text.
+  *  - "image + prompt" tasks bundle a base64 image and a text prompt in a
+  *    JSON payload: visual-question-answering, document-question-answering,
+  *    zero-shot-image-classification, image-text-to-text, image-to-image.
+  *
+  * Per-row `current_image_bytes` is resolved upstream in
+  * [[PythonCodegenBase]]'s `process_table` (either from the operator's
+  * uploaded image or from `INPUT_IMAGE_COLUMN`). The image helpers
+  * (`_read_image_input`, `_compress_image_bytes`, `_image_input_as_base64`,
+  * `_read_binary_value`, `_looks_like_html`, `_html_to_image_bytes`,
+  * `_extract_json_arg`) live in PythonCodegenBase alongside the per-task
+  * tuples (`image_only_tasks`, `image_prompt_tasks`, `image_tasks`).
+  */
+object ImageTaskCodegen extends TaskCodegen {
+
+  /** Primary key for registration; the dispatcher maps every task in
+    * [[tasks]] to this codegen.
+    */
+  override val task: String = "image-classification"
+
+  /** All HF tasks routed through this codegen. */
+  override val tasks: Set[String] = Set(
+    // image-only
+    "image-classification",
+    "object-detection",
+    "image-segmentation",
+    "image-to-text",
+    // image + prompt
+    "visual-question-answering",
+    "document-question-answering",
+    "zero-shot-image-classification",
+    "image-text-to-text",
+    "image-to-image"
+  )
+
+  override def payloadPython(ctx: CodegenContext): String =
+    """            if task in image_only_tasks:
+      |                payload = current_image_bytes
+      |                use_raw_binary_body = True
+      |                raw_binary_headers = image_headers
+      |            elif task in ("visual-question-answering", 
"document-question-answering"):
+      |                payload = {
+      |                    "inputs": {
+      |                        "image": 
self._image_input_as_base64(current_image_bytes),
+      |                        "question": prompt_value,
+      |                    }
+      |                }
+      |            elif task == "image-text-to-text":
+      |                img_b64 = 
self._image_input_as_base64(current_image_bytes)
+      |                payload = {
+      |                    "model": self.MODEL_ID,
+      |                    "messages": [{
+      |                        "role": "user",
+      |                        "content": [
+      |                            {"type": "image_url", "image_url": {"url": 
f"data:image/png;base64,{img_b64}"}},
+      |                            {"type": "text", "text": prompt_value if 
prompt_value else "Describe this image."},
+      |                        ],
+      |                    }],
+      |                    "max_tokens": self.MAX_NEW_TOKENS,
+      |                }
+      |            elif task == "image-to-image":
+      |                payload = current_image_bytes
+      |                use_raw_binary_body = True
+      |                raw_binary_headers = image_headers
+      |            elif task == "zero-shot-image-classification":
+      |                # Zero-shot requires the caller to supply candidate 
labels.
+      |                # We reuse the prompt column as a comma-separated label 
list so
+      |                # the task is shippable without a dedicated operator 
field.
+      |                # TODO: replace with a first-class `candidateLabels` 
field once
+      |                # the property panel supports task-specific inputs.
+      |                #
+      |                # Fail fast if usable labels can't be derived. Both 
modes lead to
+      |                # a meaningless inference call:
+      |                #   1. Empty prompt column          -> labels = []
+      |                #      The HF API rejects candidate_labels: [] with an 
opaque 400.
+      |                #   2. Missing prompt column        -> upstream sets 
prompt_value
+      |                #      to the fallback "What is shown in this image?", 
which has
+      |                #      no comma, so labels collapses to a single 
nonsense entry.
+      |                # Zero-shot classification needs >= 2 candidate labels 
to be
+      |                # meaningful — surface a configuration error in both 
cases.
+      |                labels = [s.strip() for s in prompt_value.split(",") if 
s.strip()]
+      |                if len(labels) < 2:
+      |                    raise ValueError(
+      |                        "zero-shot-image-classification requires at 
least 2 candidate "
+      |                        "labels: provide a comma-separated list in the 
prompt column."
+      |                    )

Review Comment:
   Done



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to