functional handler with gcs support

Dockerfile

@@ -0,0 +1,30 @@
+FROM pytorch/torchserve:latest-gpu
+
+# set user root
+USER root
+
+RUN pip install --upgrade pip
+
+# Install dependencies
+RUN pip install diffusers transformers accelerate invisible-watermark nvgpu google-cloud-storage tensorrt
+
+# Copying model files
+COPY ./config.properties /home/model-server/config.properties
+COPY ./sketch-model-3.mar /home/model-server/sketch-model-3.mar
+
+USER model-server
+
+# Expose health and prediction listener ports from the image
+EXPOSE 7080
+EXPOSE 7081
+
+# # Generate MAR file
+
+CMD ["torchserve", \
+     "--start", \
+     "--ts-config=/home/model-server/config.properties", \
+     "--models", \
+     "sketch-model-3.mar", \
+     "--model-store", \
+     "/home/model-server"]
+

replace_handler.ipynb

@@ -0,0 +1,143 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "db06489e-2e2e-4d7f-bfab-13cf42261688",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "im here\n"
+     ]
+    }
+   ],
+   "source": [
+    "print('im here')\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "9217fed7-ed85-4592-a480-ef15f0632501",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "/home/jupyter\n"
+     ]
+    }
+   ],
+   "source": [
+    "cd .."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "502b64af-59f1-4ae9-89f3-378458ba52be",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "replacing old file\n",
+      "Untitled.ipynb\n",
+      "README.md\n",
+      "config.properties\n",
+      ".gitattributes\n",
+      "model_index.json\n",
+      "Untitled1.ipynb\n",
+      "Dockerfile\n",
+      "stable_diffusion_handler.py\n",
+      "replacing  stable_diffusion_handler.py\n",
+      "tokenizer_2/tokenizer_config.json\n",
+      "tokenizer_2/special_tokens_map.json\n",
+      "tokenizer_2/vocab.json\n",
+      "tokenizer_2/merges.txt\n",
+      "text_encoder/config.json\n",
+      "text_encoder/pytorch_model.bin\n",
+      "text_encoder_2/config.json\n",
+      "text_encoder_2/pytorch_model.bin\n",
+      "unet/config.json\n",
+      "unet/diffusion_pytorch_model.bin\n",
+      "tokenizer/tokenizer_config.json\n",
+      "tokenizer/special_tokens_map.json\n",
+      "tokenizer/vocab.json\n",
+      "tokenizer/merges.txt\n",
+      "scheduler/scheduler_config.json\n",
+      "vae/config.json\n",
+      "vae/diffusion_pytorch_model.bin\n",
+      "MAR-INF/MANIFEST.json\n"
+     ]
+    }
+   ],
+   "source": [
+    "import zipfile\n",
+    "\n",
+    "# Path to the existing .mar file and the file to be replaced\n",
+    "mar_path = \"export/sketch-model-3.mar\"\n",
+    "file_to_replace = \"stable_diffusion_handler.py\"\n",
+    "new_file_path = \"sketch-model-3/stable_diffusion_handler.py\"\n",
+    "\n",
+    "# # Create a temporary .mar file\n",
+    "temp_mar_path = \"sketch-model-3-updated.mar\"\n",
+    "\n",
+    "print(\"replacing old file\")\n",
+    "# Open the existing .mar and the temporary .mar\n",
+    "with zipfile.ZipFile(mar_path, 'r') as zip_ref, zipfile.ZipFile(temp_mar_path, 'w', zipfile.ZIP_STORED) as new_zip:\n",
+    "    # Loop through existing files\n",
+    "    for item in zip_ref.infolist():\n",
+    "        print(item.filename)\n",
+    "        buffer = zip_ref.read(item.filename)\n",
+    "        \n",
+    "        # Replace the file if it matches the target file name\n",
+    "        if item.filename == file_to_replace:\n",
+    "            print('replacing ', item.filename)\n",
+    "            with open(new_file_path, \"rb\") as f:\n",
+    "                new_buffer = f.read()\n",
+    "            new_zip.writestr(item, new_buffer)\n",
+    "        else:\n",
+    "            new_zip.writestr(item, buffer)\n",
+    "\n",
+    "print('done')\n",
+    "# Remove the original .mar file and replace it with the new one\n",
+    "import os\n",
+    "os.remove(mar_path)\n",
+    "os.rename(temp_mar_path, mar_path)"
+   ]
+  }
+ ],
+ "metadata": {
+  "environment": {
+   "kernel": "conda-root-py",
+   "name": "workbench-notebooks.m110",
+   "type": "gcloud",
+   "uri": "gcr.io/deeplearning-platform-release/workbench-notebooks:m110"
+  },
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "conda-root-py"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.10.12"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

stable_diffusion_handler.py

@@ -36,6 +36,7 @@ class DiffusersHandler(BaseHandler, ABC):
         """
         
         logger.info("Loading diffusion model")
+        logger.info("I'm totally new and updated")
 
         self.manifest = ctx.manifest
         properties = ctx.system_properties
@@ -97,10 +98,10 @@ class DiffusersHandler(BaseHandler, ABC):
         logger.info("Generated image: '%s'", inferences)
         return inferences
 
-    def postprocess(self, inference_output):
+    def postprocess(self, inference_outputs):
         """Post Process Function converts the generated image into Torchserve readable format.
         Args:
-            inference_output (list): It contains the generated image of the input text.
+            inference_outputs (list): It contains the generated image of the input text.
         Returns:
             (list): Returns a list of the images.
         """
@@ -108,7 +109,7 @@ class DiffusersHandler(BaseHandler, ABC):
         client = storage.Client()
         bucket = client.get_bucket(bucket_name)
         outputs = []
-        for image in inference_output.images:
+        for image in inference_outputs:
             image_name = str(uuid.uuid4())
 
             blob = bucket.blob(image_name + '.png')
@@ -119,8 +120,8 @@ class DiffusersHandler(BaseHandler, ABC):
                 blob.upload_from_file(tmp, content_type='image/png')
 
             # generate txt file with the image name and the prompt inside
-            blob = bucket.blob(image_name + '.txt')
-            blob.upload_from_string(self.prompt)
+            # blob = bucket.blob(image_name + '.txt')
+            # blob.upload_from_string(self.prompt)
 
             outputs.append('https://storage.googleapis.com/' + bucket_name + '/' + image_name + '.png')
         return outputs