Merge branch 'hugging-face' of https://github.com/binliunls/tutorials into hugging-face

Author: binliunls

hugging_face/hugging_face_pipeline_for_monai.ipynb

@@ -170,7 +170,7 @@
     "        strides=(2, 2, 2, 2),\n",
     "        num_res_units=2,\n",
     "        norm=\"batch\",\n",
-    "        **kwargs\n",
+    "        **kwargs,\n",
     "    ):\n",
     "        super().__init__(**kwargs)\n",
     "        self.spatial_dims = spatial_dims\n",
@@ -179,7 +179,7 @@
     "        self.channels = channels\n",
     "        self.strides = strides\n",
     "        self.num_res_units = num_res_units\n",
-    "        self.norm=norm\n"
+    "        self.norm = norm"
    ]
   },
   {
@@ -211,7 +211,7 @@
     "            channels=config.channels,\n",
     "            strides=config.strides,\n",
     "            num_res_units=config.num_res_units,\n",
-    "            norm=config.norm\n",
+    "            norm=config.norm,\n",
     "        )\n",
     "\n",
     "    def forward(self, x):\n",
@@ -268,20 +268,15 @@
     "\n",
     "    def _init_preprocessing_transforms(self, image_key=\"image\", load_image=True):\n",
     "        transform_list = [LoadImaged(keys=image_key)] if load_image else []\n",
-    "        transform_list = transform_list.extend([\n",
-    "            EnsureChannelFirstd(keys=image_key),\n",
-    "            Orientationd(keys=image_key, axcodes=\"RAS\"),\n",
-    "            Spacingd(keys=image_key, pixdim=(1.5, 1.5, 2.0), mode=\"bilinear\"),\n",
-    "            ScaleIntensityRanged(\n",
-    "                keys=image_key,\n",
-    "                a_min=-57,\n",
-    "                a_max=164,\n",
-    "                b_min=0,\n",
-    "                b_max=1,\n",
-    "                clip=True\n",
-    "            ),\n",
-    "            EnsureTyped(keys=image_key)\n",
-    "        ])\n",
+    "        transform_list = transform_list.extend(\n",
+    "            [\n",
+    "                EnsureChannelFirstd(keys=image_key),\n",
+    "                Orientationd(keys=image_key, axcodes=\"RAS\"),\n",
+    "                Spacingd(keys=image_key, pixdim=(1.5, 1.5, 2.0), mode=\"bilinear\"),\n",
+    "                ScaleIntensityRanged(keys=image_key, a_min=-57, a_max=164, b_min=0, b_max=1, clip=True),\n",
+    "                EnsureTyped(keys=image_key),\n",
+    "            ]\n",
+    "        )\n",
     "        preprocessing_transforms = Compose(transform_list)\n",
     "        return preprocessing_transforms\n",
     "\n",
@@ -303,33 +298,35 @@
     "                transform=copy.deepcopy(self.preprocessing_transforms),\n",
     "                orig_keys=image_key,\n",
     "                nearest_interp=False,\n",
-    "                to_tensor=True\n",
+    "                to_tensor=True,\n",
     "            ),\n",
     "            AsDiscreted(keys=pred_key, argmax=True),\n",
     "        ]\n",
-    "        transform_list = transform_list.append(SaveImaged(\n",
-    "                keys=pred_key,\n",
-    "                output_dir=output_dir,\n",
-    "                output_ext=output_ext,\n",
-    "                output_dtype=output_dtype,\n",
-    "                output_postfix=output_postfix,\n",
-    "                separate_folder=separate_folder\n",
-    "            )) if save_output else transform_list\n",
+    "        transform_list = (\n",
+    "            transform_list.append(\n",
+    "                SaveImaged(\n",
+    "                    keys=pred_key,\n",
+    "                    output_dir=output_dir,\n",
+    "                    output_ext=output_ext,\n",
+    "                    output_dtype=output_dtype,\n",
+    "                    output_postfix=output_postfix,\n",
+    "                    separate_folder=separate_folder,\n",
+    "                )\n",
+    "            )\n",
+    "            if save_output\n",
+    "            else transform_list\n",
+    "        )\n",
     "\n",
     "        postprocessing_transforms = Compose(transform_list)\n",
     "        return postprocessing_transforms\n",
-    "    \n",
+    "\n",
     "    def _init_inferer(\n",
     "        self,\n",
     "        roi_size=(96, 96, 96),\n",
     "        overlap=0.5,\n",
     "        sw_batch_size=4,\n",
     "    ):\n",
-    "        return SlidingWindowInferer(\n",
-    "            roi_size=roi_size,\n",
-    "            sw_batch_size=sw_batch_size,\n",
-    "            overlap=overlap\n",
-    "        )\n",
+    "        return SlidingWindowInferer(roi_size=roi_size, sw_batch_size=sw_batch_size, overlap=overlap)\n",
     "\n",
     "    def _sanitize_parameters(self, **kwargs):\n",
     "        preprocessing_kwargs = {}\n",
@@ -356,9 +353,7 @@
     "    ):\n",
     "        for key, value in kwargs.items():\n",
     "            if key in self._preprocess_params and value != self._preprocess_params[key]:\n",
-    "                logging.warning(\n",
-    "                    f\"Please set the parameter {key} during initialization.\"\n",
-    "                )\n",
+    "                logging.warning(f\"Please set the parameter {key} during initialization.\")\n",
     "\n",
     "            if key not in self.PREPROCESSING_EXTRA_ARGS:\n",
     "                logging.warning(f\"Cannot set parameter {key} for preprocessing.\")\n",
@@ -372,7 +367,7 @@
     "    ):\n",
     "        inputs.to(self.device)\n",
     "        self.model.to(self.device)\n",
-    "        mode=eval_mode,\n",
+    "        mode = (eval_mode,)\n",
     "        outputs = {Keys.IMAGE: inputs, Keys.LABEL: None}\n",
     "        with mode(self.model):\n",
     "            if amp:\n",
@@ -388,7 +383,7 @@
     "                logging.warning(f\"Cannot set parameter {key} for postprocessing.\")\n",
     "\n",
     "        outputs = self.postprocessing_transforms(decollate_batch(outputs))\n",
-    "        return outputs\n"
+    "        return outputs"
    ]
   },
   {