1132-fix-the-bug-in-loading-tool-segmentation-model (#1135)

Signed-off-by: binliu <binliu@nvidia.com>

Fixes #1132 .

### Description
The pretrained `endoscopic_tool_segmentation` model in MONAI model zoo
cannot be loaded to `video_seg` tutorial since the different output.
Therefore, the whole model except segmentation head should be loaded to
perform pretrain.

### Checks
<!--- Put an `x` in all the boxes that apply, and remove the not
applicable items -->
- [ ] Notebook runs automatically `./runner [-p <regex_pattern>]`

Signed-off-by: binliu <binliu@nvidia.com>

Author: binliunls

computer_assisted_intervention/video_seg.ipynb

@@ -372,12 +372,11 @@
     "## Normal training loop\n",
     "\n",
     "A normal training loop will be excuted. Meanwhile, the iou value of each epoch will be recorded as the accuracy. \n",
-    "There is a pretrained bundle for surgical tool segmentation task in the MONAI model-zoo, the model can be automatically downloaded and loaded to the network by adding codes to the block below:\n",
+    "There is a pretrained bundle for surgical tool segmentation task in the MONAI model-zoo, the model can be automatically downloaded and loaded to the network by adding codes to the block below. Please add `from monai.networks.utils import copy_model_state` in the `Imports` part of this tutorial, when using it.\n",
     "```\n",
-    "pretrained_weights = monai.bundle.load(\n",
-    "    name=\"endoscopic_tool_segmentation\", bundle_dir=\"./\", version=\"0.2.0\"\n",
-    ")\n",
-    "model.load_state_dict(pretrained_weights)\n",
+    "pretrained_weights = monai.bundle.load(name=\"endoscopic_tool_segmentation\", bundle_dir=\"./\", version=\"0.2.0\")\n",
+    "new_model_dict, _, _ = copy_model_state(model, pretrained_weights, exclude_vars=\"segmentation_head\")\n",
+    "model.load_state_dict(new_model_dict)\n",
     "```"
    ]
   },
@@ -691,9 +690,9 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "conda_base",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
-   "name": "base"
+   "name": "python3"
   },
   "language_info": {
    "codemirror_mode": {