Custom qconfig for eager-mode quantization

Summary:
Pull Request resolved: https://github.com/facebookresearch/d2go/pull/664

Generate qconfig based on the config

Reviewed By: wat3rBro

Differential Revision: D58210321

fbshipit-source-id: 7a86f8b6e9d112302c978080c2bd5721e3c7dbff

d2go/quantization/qconfig.py

@@ -26,9 +26,11 @@ def set_backend_and_create_qconfig(cfg, *, is_train):
     return QCONFIG_CREATOR_REGISTRY.get("smart")(cfg, is_train=is_train)
 
 
-def holistic_get_qconfig(backend, is_qat, use_symmetric=False):
+@fb_overwritable()
+def holistic_get_qconfig(backend, is_qat, use_symmetric=False, cfg=None):
     """
-    Config-less vanilla way to create the QConfig, suitable for explicitly creating qconfig.
+    Utility to create the QConfig based on backend, is_qat, and use_symmetric.
+    cfg (unused) is to customize QConfig based on the cfg.
     """
 
     if use_symmetric:
@@ -66,7 +68,7 @@ def _smart_set_backend_and_create_qconfig(cfg, *, is_train):
     assert qat_method in ["default", "learnable"]
 
     qconfig = holistic_get_qconfig(
-        backend=backend, is_qat=is_train, use_symmetric=is_symmetric
+        backend=backend, is_qat=is_train, use_symmetric=is_symmetric, cfg=cfg
     )
     if is_train and qat_method == "learnable":
         qconfig = convert_to_learnable_qconfig(qconfig)