Move distillation config to runner default configs

Summary:
Pull Request resolved: https://github.com/facebookresearch/d2go/pull/557

To avoid circular dependencies, move the function `add_distillation_configs` that defines the default config for a `runner` making use of distillation from `mobile-vision/d2go/d2go/modeling/distillation.py` to `mobile-vision/d2go/d2go/runner/config_defaults.py`.

Reviewed By: wat3rBro

Differential Revision: D46096374

fbshipit-source-id: eb85d91b5239e7ab10809a9bf84c869d05d32401

d2go/modeling/distillation.py

@@ -37,25 +37,6 @@ logger = logging.getLogger(__name__)
 ModelOutput = Union[None, torch.Tensor, Iterable["ModelOutput"]]
 
 
-def add_distillation_configs(_C: CN) -> None:
-    """Add default parameters to config
-
-    The TEACHER.CONFIG field allows us to build a PyTorch model using an
-    existing config.  We can build any model that is normally supported by
-    D2Go (e.g., FBNet) because we just use the same config
-    """
-    _C.DISTILLATION = CN()
-    _C.DISTILLATION.ALGORITHM = "LabelDistillation"
-    _C.DISTILLATION.HELPER = "BaseDistillationHelper"
-    _C.DISTILLATION.TEACHER = CN()
-    _C.DISTILLATION.TEACHER.TORCHSCRIPT_FNAME = ""
-    _C.DISTILLATION.TEACHER.DEVICE = ""
-    _C.DISTILLATION.TEACHER.TYPE = "torchscript"
-    _C.DISTILLATION.TEACHER.CONFIG_FNAME = ""
-    _C.DISTILLATION.TEACHER.RUNNER_NAME = "d2go.runner.GeneralizedRCNNRunner"
-    _C.DISTILLATION.TEACHER.OVERWRITE_OPTS = []
-
-
 @dataclass
 class LayerLossMetadata:
     loss: nn.Module

d2go/runner/config_defaults.py

@@ -10,7 +10,6 @@ from d2go.data.build import (
 from d2go.data.config import add_d2go_data_default_configs
 from d2go.modeling import ema, kmeans_anchors
 from d2go.modeling.backbone.fbnet_cfg import add_fbnet_v2_default_configs
-from d2go.modeling.distillation import add_distillation_configs
 from d2go.modeling.meta_arch.fcos import add_fcos_configs
 from d2go.modeling.model_freezing_utils import add_model_freezing_configs
 from d2go.modeling.subclass import add_subclass_configs
@@ -39,6 +38,25 @@ def _add_base_runner_default_fb_cfg(_C: CN) -> None:
     pass
 
 
+def add_distillation_configs(_C: CN) -> None:
+    """Add default parameters to config
+
+    The TEACHER.CONFIG field allows us to build a PyTorch model using an
+    existing config.  We can build any model that is normally supported by
+    D2Go (e.g., FBNet) because we just use the same config
+    """
+    _C.DISTILLATION = CN()
+    _C.DISTILLATION.ALGORITHM = "LabelDistillation"
+    _C.DISTILLATION.HELPER = "BaseDistillationHelper"
+    _C.DISTILLATION.TEACHER = CN()
+    _C.DISTILLATION.TEACHER.TORCHSCRIPT_FNAME = ""
+    _C.DISTILLATION.TEACHER.DEVICE = ""
+    _C.DISTILLATION.TEACHER.TYPE = "torchscript"
+    _C.DISTILLATION.TEACHER.CONFIG_FNAME = ""
+    _C.DISTILLATION.TEACHER.RUNNER_NAME = "d2go.runner.GeneralizedRCNNRunner"
+    _C.DISTILLATION.TEACHER.OVERWRITE_OPTS = []
+
+
 def _add_detectron2go_runner_default_cfg(_C: CN) -> None:
     # _C.MODEL.FBNET_V2...
     add_fbnet_v2_default_configs(_C)

tests/modeling/test_modeling_distillation.py

@@ -13,7 +13,6 @@ from d2go.modeling import modeling_hook as mh
 from d2go.modeling.distillation import (
     _build_teacher,
     _set_device,
-    add_distillation_configs,
     BaseDistillationHelper,
     CachedLayer,
     compute_layer_losses,
@@ -38,6 +37,7 @@ from d2go.registry.builtin import (
     DISTILLATION_HELPER_REGISTRY,
     META_ARCH_REGISTRY,
 )
+from d2go.runner.config_defaults import add_distillation_configs
 from d2go.runner.default_runner import BaseRunner
 from d2go.utils.testing import helper
 from detectron2.checkpoint import DetectionCheckpointer