Add a custom handler for Inference Endpoints (#1)

- Add a custom handler for Inference Endpoints (70197b67f05f194a4e06072c22668cc6313922be)
- extend transforms and change requirements.txt encoding (7c184f56abb85493b97e2c5ccf035a4918b23335)


Co-authored-by: Jacek Nowak <[email protected]>

.gitignore

@@ -0,0 +1,2 @@
+.venv
+.idea

handler.py

@@ -0,0 +1,58 @@
+import torch
+from torchvision import transforms
+from huggingface_hub import hf_hub_download
+import json
+import io
+import base64
+from PIL import Image
+from omegaconf import OmegaConf
+
+from model import Generator
+
+
+class EndpointHandler:
+
+    def __init__(self, path=''):
+        self.transform = transforms.Compose(
+            [
+                transforms.ToImage(),
+                transforms.ToDtype(torch.float32, scale=True),
+                transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5]),
+            ]
+        )
+
+        repo_id = "Kiwinicki/sat2map-generator"
+        generator_path = hf_hub_download(repo_id=repo_id, filename="generator.pth")
+        config_path = hf_hub_download(repo_id=repo_id, filename="config.json")
+        model_path = hf_hub_download(repo_id=repo_id, filename="model.py")
+
+        with open(config_path, "r") as f:
+            config_dict = json.load(f)
+        cfg = OmegaConf.create(config_dict)
+
+        self.generator = Generator(cfg)
+        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        self.generator.load_state_dict(torch.load(generator_path, map_location=self.device))
+        self.generator.eval()
+
+
+    def __call__(self, data: dict[str, any]) -> dict[str, str]:
+        base64_image = data.get('inputs')
+        input_tensor = self._decode_base64_image(base64_image)
+        # print('Input tensor shape: ' + str(input_tensor.shape))
+        output_tensor = self.generator(input_tensor.to(self.device))
+        output_tensor = output_tensor.squeeze(0)
+        output_image = transforms.ToPILImage()(output_tensor)
+        output_image = output_image.convert('RGB')
+        output_buffer = io.BytesIO()
+        output_image.save(output_buffer, format="png")
+        base64_output = base64.b64encode(output_buffer.getvalue()).decode('utf-8')
+        return {"output": base64_output}
+
+
+    def _decode_base64_image(self, base64_image: str) -> torch.Tensor:
+        image_decoded = base64.b64decode(base64_image)
+        image = Image.open(io.BytesIO(image_decoded)).convert('RGB')
+        image_tensor: torch.Tensor = self.transform(image)
+        image_tensor = image_tensor.unsqueeze(0)
+        return image_tensor

requirements.txt

@@ -0,0 +1,25 @@
+antlr4-python3-runtime==4.9.3
+certifi==2024.12.14
+charset-normalizer==3.4.1
+colorama==0.4.6
+filelock==3.17.0
+fsspec==2024.12.0
+huggingface-hub==0.28.0
+idna==3.10
+Jinja2==3.1.5
+MarkupSafe==3.0.2
+mpmath==1.3.0
+networkx==3.4.2
+numpy==2.2.2
+omegaconf==2.3.0
+packaging==24.2
+pillow==11.1.0
+PyYAML==6.0.2
+requests==2.32.3
+setuptools==75.8.0
+sympy==1.13.1
+torch==2.5.1
+torchvision==0.20.1
+tqdm==4.67.1
+typing_extensions==4.12.2
+urllib3==2.3.0