add dacon metric

rahulscripts/metrics/__init__.py

@@ -0,0 +1 @@
+import * from custom_metrics

rahulscripts/metrics/custom_metrics.py

@@ -0,0 +1,45 @@
+import torch
+from torchmetrics import Metric
+
+
+class DaconScore(Metric):
+    def __init__(self, dist_sync_on_step=False, **kwargs):
+        super().__init__(dist_sync_on_step=dist_sync_on_step, **kwargs)
+        self.add_state("score", default=torch.tensor(0.0), dist_reduce_fx="sum")
+        self.add_state("count", default=torch.tensor(0), dist_reduce_fx="sum")
+        self.unit = kwargs.get("unit", 9)
+
+    def update(self, pred: torch.Tensor, target: torch.Tensor):
+        score = self.score_compute(pred, target)
+        self.score += score
+        self.count += 1
+
+    def compute(self) -> torch.Tensor:
+        return self.score / self.count
+
+    def rmse_compute(self, pred, target):
+        return torch.sqrt(((pred - target) ** 2).mean())
+
+    def normalized_rmse_compute(self, pred, target):
+        return self.rmse_compute(pred, target) / (target.max() - target.min())
+
+    def correct_ratio(self, pred, target):
+        pIC50_pred = self.ic50_to_pic50(pred, self.unit)
+        pIC50_target = self.ic50_to_pic50(target, self.unit)
+        diff = pIC50_pred - pIC50_target
+        correct = torch.sum(diff <= 0.5)
+        return correct / len(pred)
+
+    def score_compute(self, pred, target):
+        if not isinstance(pred, torch.Tensor):
+            pred = torch.tensor(pred)
+        if not isinstance(target, torch.Tensor):
+            target = torch.tensor(target)
+        score = 0.5 * (1 - min(self.normalized_rmse_compute(pred, target), 1)) + (
+            0.5 * self.correct_ratio(pred, target)
+        )
+        return score
+
+    def ic50_to_pic50(self, ic50_value: float, unit: int = 9) -> float:
+        pic50 = unit - torch.log10(ic50_value)
+        return pic50