add mse (#1110)

* add mse

* fix mse

* fix compare_util

* fix compare_util

---------

Co-authored-by: hejunchao <hejunchao@canaan-creative.com>
Co-authored-by: HeJunchao100813 <HeJunchao100813@users.noreply.github.com>

tests/compare_util.py

@@ -51,7 +51,6 @@ def cosine(gt: np.ndarray, pred: np.ndarray, *args):
 
     result = (gt @ pred) / (np.linalg.norm(gt, 2) * np.linalg.norm(pred, 2))
 
-    # When tensor gt is a multiple of tensor pred, their similarity is also 1.
     return -1 if math.isnan(result) else result
 
 
@@ -63,6 +62,29 @@ def euclidean(gt: np.ndarray, pred: np.ndarray, *args):
     return np.linalg.norm(gt.reshape(-1) - pred.reshape(-1))
 
 
+# def mse(gt: np.ndarray, pred: np.ndarray, *args):
+#     return np.mean((gt - pred) ** 2)
+
+def divide(gt: np.ndarray, pred: np.ndarray):
+
+    # remove the zero values in the same location.
+    gt_mask = np.equal(gt, 0)
+    pred_mask = np.equal(pred, 0)
+    mask = gt_mask & pred_mask
+    gt = gt[~mask]
+    pred = pred[~mask]
+
+    # to avoid divide zero.
+    pred = np.where(np.equal(pred, 0), 1e-7, pred)
+
+    result = np.divide(gt, pred)
+    return result
+
+
+def mean(gt: np.ndarray):
+    return np.mean(gt)
+
+
 def allclose(gt: np.ndarray, pred: np.ndarray, thresh: float):
     return np.allclose(gt, pred, atol=thresh)
 
@@ -122,6 +144,8 @@ def compare_binfile(result_path: Tuple[str, str],
         compare_op = gt
     if compare_op(similarity, threshold):
         return False, similarity_info
+    if (mean(divide(gt_arr, pred_arr)) > 1.5 or mean(divide(gt_arr, pred_arr)) < 0.6):
+        return False, similarity_info, f"\nmaybe a case of multiples"
     return True, similarity_info