Speeds up datatype constraints tests

Makes datatype constraints tests significantly faster by not evaluating
inputs.

Author: pranavm-nvidia

tripy/tests/constraints/object_builders.py

@@ -26,15 +26,15 @@
 def tensor_builder(init, dtype, namespace):
     if init is None:
         out = tp.ones(dtype=namespace[dtype], shape=(3, 2))
-        out.eval()
         return out
     elif not isinstance(init, tp.Tensor):
         return init
 
     out = init
     if dtype is not None:
         out = tp.cast(out, dtype=namespace[dtype])
-    out.eval()
+        # Need to evaluate when casting because we run into MLIR-TRT bugs while deriving upper bounds.
+        out.eval()
     return out
 
 
@@ -47,8 +47,6 @@ def tensor_list_builder(init, dtype, namespace):
         out = [tp.ones(shape=(3, 2), dtype=namespace[dtype]) for _ in range(2)]
     else:
         out = [tp.cast(tens, dtype=namespace[dtype]) for tens in init]
-    for t in out:
-        t.eval()
     return out
 
 
@@ -132,7 +130,7 @@ def default_builder(init, dtype, namespace):
     "pad": {"pad": [(0, 1), (1, 0)]},
     "permute": {"perm": [1, 0]},
     "prod": {"dim": 0},
-    "quantize": {"scale": tp.Tensor([1, 1, 1]), "dim": 0},
+    "quantize": {"input": tp.ones((3, 2)), "scale": tp.Tensor([1, 1, 1]), "dim": 0},
     "repeat": {"repeats": 2, "dim": 0},
     "reshape": {"shape": [6]},
     "resize": {