Upload folder using huggingface_hub

config.json

@@ -0,0 +1,42 @@
+{
+  "model_name": "Physics-Informed UNet++",
+  "task": "Disaster Risk Segmentation",
+  "architecture": {
+    "backbone": "UNet++",
+    "encoder": "resnet34",
+    "in_channels": 32,
+    "out_channels": 1,
+    "film_dim": 32,
+    "dropout_p": 0.016687189127500446
+  },
+  "training": {
+    "best_val_iou": 1.0,
+    "optimizer": "AdamW",
+    "learning_rate": 4.7031995111855064e-05,
+    "weight_decay": 0.00025303130757493825,
+    "batch_size": 8,
+    "epochs": 50,
+    "mixed_precision": "FP16"
+  },
+  "inference": {
+    "input_shape": [
+      1,
+      32,
+      512,
+      512
+    ],
+    "output": "probability_map",
+    "recommended_threshold": 0.000175
+  },
+  "framework": {
+    "library": "PyTorch",
+    "torch_version": "2.x",
+    "export_formats": [
+      "TorchScript",
+      "ONNX"
+    ]
+  },
+  "license": "Apache-2.0",
+  "author": "Lokesh Reddy Poreddy",
+  "repository": "https://huggingface.co/loki200519/urop"
+}

inference_example.py

@@ -0,0 +1,15 @@
+import torch
+
+# Load TorchScript model
+model = torch.jit.load("model_logits_traced.pt")
+model.eval()
+
+# Dummy input (B, C, H, W)
+x = torch.randn(1, 32, 512, 512)
+
+with torch.no_grad():
+    y = model(x)
+
+print("Inference successful!")
+print("Output shape:", y.shape)
+print("Min/Max:", y.min().item(), y.max().item())