Add enable_tvm_constant_prop compiler flag

python/tvm/contrib/forge_utils.py

@@ -186,23 +186,21 @@ def construct_tvm_ir(framework: str, model, tvm_mod, params, compiler_cfg: Compi
     if framework == "pytorch":
         param_name_lookup = {}
 
-        if not compiler_cfg.enable_tvm_constant_prop:
-            tvm_mod = tvm.IRModule.from_expr(
-                tvm.relay.build_module.bind_params_by_name(tvm_mod["main"], {})
-            )
-        else:
-            if len(compiler_cfg.tvm_constnat_prop_mask):
-                propped_params = {
+        if params is not None:
+            propped_params = {
                     k: (v, True)
                     for k, v, in params.items()
                     if any([mask in k for mask in compiler_cfg.tvm_constnat_prop_mask])
                 }
-            else:
-                propped_params = {k: (v, True) for k, v, in params.items()}
             tvm_mod = tvm.IRModule.from_expr(
                 tvm.relay.build_module.bind_params_by_name(tvm_mod["main"], propped_params)
             )
-
+        else:
+            tvm_mod = tvm.IRModule.from_expr(
+                tvm.relay.build_module.bind_params_by_name(tvm_mod["main"], {})
+            )
+
+
     elif framework == "tensorflow":
         # TODO: Destupidify this! (Maybe we can sort by a substring of the weight names to make this more efficient)
         found_weights = []