Ensure lambda_factors and extension_ratio are not None in non_uniform…

…_interpolation

src/main.py

@@ -43,6 +43,9 @@ def non_uniform_interpolation(pos_embed, extension_ratio, lambda_factors, n_hat)
     if extension_ratio is None:
         raise ValueError("extension_ratio cannot be None")
 
+    if lambda_factors is None:
+        raise ValueError("lambda_factors cannot be None")
+
     d_model = pos_embed.shape[-1]
     interpolated_pos = pos_embed.clone()
 
@@ -53,9 +56,8 @@ def non_uniform_interpolation(pos_embed, extension_ratio, lambda_factors, n_hat)
             torch.ones_like(pos_embed[..., 0], device=pos_embed.device),
             1 / (lambda_factors[i] * extension_ratio),
         )
-        interpolated_pos[..., i * 2] *= scale
-        if i * 2 + 1 < d_model:  # Check if the index is within bounds
-            interpolated_pos[..., i * 2 + 1] *= scale
+        interpolated_pos[..., 2 * i] *= scale
+        interpolated_pos[..., 2 * i + 1] *= scale
 
     return interpolated_pos