chore: Move new ImageFeaturizer to onnx namespace (#1711)

deep-learning/src/main/python/synapse/ml/cntk/ImageFeaturizer.py → deep-learning/src/main/python/synapse/ml/onnx/ImageFeaturizer.py

@@ -6,7 +6,7 @@
 if sys.version >= "3":
     basestring = str
 
-from synapse.ml.cntk._ImageFeaturizer import _ImageFeaturizer
+from synapse.ml.onnx._ImageFeaturizer import _ImageFeaturizer
 from pyspark.ml.common import inherit_doc
 
 

deep-learning/src/main/scala/com/microsoft/azure/synapse/ml/cntk/ImageFeaturizer.scala → deep-learning/src/main/scala/com/microsoft/azure/synapse/ml/onnx/ImageFeaturizer.scala

@@ -1,14 +1,13 @@
 // Copyright (C) Microsoft Corporation. All rights reserved.
 // Licensed under the MIT License. See LICENSE in project root for information.
 
-package com.microsoft.azure.synapse.ml.cntk
+package com.microsoft.azure.synapse.ml.onnx
 
 import com.microsoft.azure.synapse.ml.codegen.Wrappable
 import com.microsoft.azure.synapse.ml.core.contracts.{HasInputCol, HasOutputCol}
 import com.microsoft.azure.synapse.ml.core.schema.DatasetExtensions
 import com.microsoft.azure.synapse.ml.core.schema.DatasetExtensions.findUnusedColumnName
 import com.microsoft.azure.synapse.ml.logging.BasicLogging
-import com.microsoft.azure.synapse.ml.onnx.{ONNXHub, ONNXModel, ONNXModelInfo}
 import com.microsoft.azure.synapse.ml.opencv.ImageTransformer
 import com.microsoft.azure.synapse.ml.param.TransformerParam
 import org.apache.spark.ml.linalg.DenseVector

deep-learning/src/main/scala/com/microsoft/azure/synapse/ml/onnx/ImageFeaturizer.txt

@@ -0,0 +1,16 @@
+The ``ImageFeaturizer`` relies on a ``ONNXModel`` to do the
+featurization of the image(s).  One can set this model directly using the
+onnxModel parameter, or use setModel to set a known model name that will
+be pulled from the ONNXHub.  See docs for other options to set the underlying
+model.
+
+The ``ImageFeaturizer`` takes an input column of images (the type
+returned by the ``ImageReader``), automatically resizes them to fit the
+``ONNXModel``'s inputs using an ``ImageTransformer``, and feeds them through
+a pre-trained ONNX model. Some of the ``ImageTransformer`` parameters are
+exposed on the featurizer so you can customize them.
+
+You may specify intermediate model nodes by either using the ``headless``
+parameter to automatically choose the full model output (headless = false),
+or the intermediate feature nodes as indicated by the OnnxHub manifest for that
+model (headless = true).

deep-learning/src/test/scala/com/microsoft/azure/synapse/ml/explainers/ImageExplainersSuite.scala

@@ -3,10 +3,9 @@
 
 package com.microsoft.azure.synapse.ml.explainers
 
-import com.microsoft.azure.synapse.ml.cntk.{ImageFeaturizer, TrainedONNXModelUtils}
 import com.microsoft.azure.synapse.ml.core.test.base.TestBase
 import com.microsoft.azure.synapse.ml.io.IOImplicits._
-import com.microsoft.azure.synapse.ml.onnx.ONNXModel
+import com.microsoft.azure.synapse.ml.onnx.{ImageFeaturizer, ONNXModel, TrainedONNXModelUtils}
 import com.microsoft.azure.synapse.ml.opencv.ImageTransformer
 import org.apache.commons.io.FileUtils
 import org.apache.spark.ml.{Pipeline, PipelineModel}

deep-learning/src/test/scala/com/microsoft/azure/synapse/ml/cntk/ImageFeaturizerSuite.scala → deep-learning/src/test/scala/com/microsoft/azure/synapse/ml/onnx/ImageFeaturizerSuite.scala

@@ -1,7 +1,7 @@
 // Copyright (C) Microsoft Corporation. All rights reserved.
 // Licensed under the MIT License. See LICENSE in project root for information.
 
-package com.microsoft.azure.synapse.ml.cntk
+package com.microsoft.azure.synapse.ml.onnx
 
 import breeze.linalg.argtopk
 import com.microsoft.azure.synapse.ml.Secrets

notebooks/features/other/DeepLearning - Flower Image Classification.ipynb

@@ -15,7 +15,6 @@
    "source": [
     "from pyspark.ml import Transformer, Estimator, Pipeline\n",
     "from pyspark.ml.classification import LogisticRegression\n",
-    "from synapse.ml.downloader import ModelDownloader\n",
     "import sys, time\n",
     "from pyspark.sql import SparkSession\n",
     "\n",
@@ -28,24 +27,6 @@
     "    from notebookutils.visualization import display"
    ]
   },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {
-    "mml-deploy": "local"
-   },
-   "outputs": [],
-   "source": [
-    "if running_on_synapse():\n",
-    "    modelDir = \"abfss://synapse@mmlsparkeuap.dfs.core.windows.net/models/\"\n",
-    "elif running_on_databricks():\n",
-    "    modelDir = \"dbfs:/models/\"\n",
-    "else:\n",
-    "    modelDir = \"/tmp/models/\"\n",
-    "\n",
-    "model = ModelDownloader(spark, modelDir).downloadByName(\"ResNet50\")"
-   ]
-  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -80,7 +61,7 @@
    "source": [
     "from synapse.ml.opencv import ImageTransformer\n",
     "from synapse.ml.image import UnrollImage\n",
-    "from synapse.ml.cntk import ImageFeaturizer\n",
+    "from synapse.ml.onnx import ImageFeaturizer\n",
     "from synapse.ml.stages import *\n",
     "\n",
     "# Make some featurizers\n",

notebooks/features/responsible_ai/Interpretability - Snow Leopard Detection.ipynb

@@ -223,7 +223,7 @@
     "from pyspark.ml.feature import StringIndexer\n",
     "from pyspark.ml.classification import LogisticRegression\n",
     "from pyspark.sql.functions import udf\n",
-    "from synapse.ml.cntk import ImageFeaturizer\n",
+    "from synapse.ml.onnx import ImageFeaturizer\n",
     "from synapse.ml.stages import UDFTransformer\n",
     "from pyspark.sql.types import *\n",
     "\n",