rewrote the comment

Author: cehongwang

py/torch_tensorrt/dynamo/conversion/aten_ops_converters.py

@@ -2726,12 +2726,14 @@ def topk_sort_validator(k: int) -> bool:
     # topk layer supports dynamic k value but we cannot determine supported dynamic topk value at
     # compile time.
     if k == DYNAMIC_DIM:
-        _LOGGER.debug("k value cannot be dynamic!")
+        _LOGGER.debug(
+            "[top_k validator] Converter does not support k being a dynamic value. Therefore, aten::topk will run in PyTorch"
+        )
         return False
 
     if k > 3840:
         _LOGGER.debug(
-            f"Currently only topk values up to 3840 are supported, got k={k}."
+            f"[top_k validator] Currently only topk values up to 3840 are supported, got k={k}. Therefore, aten::topk will run in PyTorch"
         )
         return False
     return True

py/torch_tensorrt/dynamo/conversion/impl/topk.py

@@ -209,10 +209,6 @@ def topk(
             get_axes_for_reduce_op(get_positive_dim(dim, len(input.shape))),
         )
 
-    # topk layer supports dynamic k value but we cannot dertermin supported dynamic topk value at
-    # compile time.
-    assert k != DYNAMIC_DIM, "k value cannot be dynamic!"
-
     # TensorRT ITopKLayer does not have a sorted flag, it is always returning the sorted topk elements
     # so here no matter sorted is True or False the returned the topk Tensor object is always sorted
     set_layer_name(topk_layer, target, f"{name}_topk", source_ir)