deprecate some more caffe2 stuff

Summary:
1. make 'import caffe2 ' optional because pytorch is removing it from OSS. this should help with
	 https://fb.workplace.com/groups/277527419809135/posts/955879138640623/

2. remove two functions `export_{caffe2,onnx}_model` that are marked deprecated a long time ago

3. make `add_export_config` a no-op and remove all callsites (there is no callsite of it in d2go).

Reviewed By: wat3rBro

Differential Revision: D32556033

fbshipit-source-id: 6254562ee892c0b57c8455003db17bb00998f5ed

detectron2/export/__init__.py

@@ -1,6 +1,14 @@
 # -*- coding: utf-8 -*-
 
-from .api import *
+try:
+    from caffe2.proto import caffe2_pb2 as _tmp
+
+    # caffe2 is optional
+except ImportError:
+    pass
+else:
+    from .api import *
+
 from .flatten import TracingAdapter
 from .torchscript import scripting_with_instances, dump_torchscript_IR
 

detectron2/export/api.py

@@ -15,30 +15,12 @@
 
 __all__ = [
     "add_export_config",
-    "export_caffe2_model",
     "Caffe2Model",
-    "export_onnx_model",
     "Caffe2Tracer",
 ]
 
 
 def add_export_config(cfg):
-    """
-    Add options needed by caffe2 export.
-
-    Args:
-        cfg (CfgNode): a detectron2 config
-
-    Returns:
-        CfgNode:
-            an updated config with new options that will be used by :class:`Caffe2Tracer`.
-    """
-    is_frozen = cfg.is_frozen()
-    cfg.defrost()
-    cfg.EXPORT_CAFFE2 = CfgNode()
-    cfg.EXPORT_CAFFE2.USE_HEATMAP_MAX_KEYPOINT = False
-    if is_frozen:
-        cfg.freeze()
     return cfg
 
 
@@ -68,9 +50,7 @@ class Caffe2Tracer:
     def __init__(self, cfg: CfgNode, model: nn.Module, inputs):
         """
         Args:
-            cfg (CfgNode): a detectron2 config, with extra export-related options
-                added by :func:`add_export_config`. It's used to construct
-                caffe2-compatible model.
+            cfg (CfgNode): a detectron2 config used to construct caffe2-compatible model.
             model (nn.Module): An original pytorch model. Must be among a few official models
                 in detectron2 that can be converted to become caffe2-compatible automatically.
                 Weights have to be already loaded to this model.
@@ -81,8 +61,6 @@ def __init__(self, cfg: CfgNode, model: nn.Module, inputs):
         assert isinstance(cfg, CfgNode), cfg
         assert isinstance(model, torch.nn.Module), type(model)
 
-        if "EXPORT_CAFFE2" not in cfg:
-            cfg = add_export_config(cfg)  # will just the defaults
         # TODO make it support custom models, by passing in c2 model directly
         C2MetaArch = META_ARCH_CAFFE2_EXPORT_TYPE_MAP[cfg.MODEL.META_ARCHITECTURE]
         self.traceable_model = C2MetaArch(cfg, copy.deepcopy(model))
@@ -255,19 +233,3 @@ def __call__(self, inputs):
         if self._predictor is None:
             self._predictor = ProtobufDetectionModel(self._predict_net, self._init_net)
         return self._predictor(inputs)
-
-
-def export_caffe2_model(cfg, model, inputs):
-    logger = logging.getLogger(__name__)
-    logger.warning(
-        "export_caffe2_model() is deprecated. Please use `Caffe2Tracer().export_caffe2() instead."
-    )
-    return Caffe2Tracer(cfg, model, inputs).export_caffe2()
-
-
-def export_onnx_model(cfg, model, inputs):
-    logger = logging.getLogger(__name__)
-    logger.warning(
-        "export_caffe2_model() is deprecated. Please use `Caffe2Tracer().export_onnx() instead."
-    )
-    return Caffe2Tracer(cfg, model, inputs).export_onnx()

detectron2/export/caffe2_modeling.py

@@ -248,8 +248,12 @@ def __init__(self, cfg, torch_model):
         torch_model = patch_generalized_rcnn(torch_model)
         super().__init__(cfg, torch_model)
 
+        try:
+            use_heatmap_max_keypoint = cfg.EXPORT_CAFFE2.USE_HEATMAP_MAX_KEYPOINT
+        except AttributeError:
+            use_heatmap_max_keypoint = False
         self.roi_heads_patcher = ROIHeadsPatcher(
-            self._wrapped_model.roi_heads, cfg.EXPORT_CAFFE2.USE_HEATMAP_MAX_KEYPOINT
+            self._wrapped_model.roi_heads, use_heatmap_max_keypoint
         )
 
     def encode_additional_info(self, predict_net, init_net):

docs/conf.py

@@ -282,8 +282,6 @@ def autodoc_skip_member(app, what, name, obj, skip, options):
         "ResNetBlockBase",
         "GroupedBatchSampler",
         "build_transform_gen",
-        "export_caffe2_model",
-        "export_onnx_model",
         "apply_transform_gens",
         "TransformGen",
         "apply_augmentations",
@@ -294,6 +292,7 @@ def autodoc_skip_member(app, what, name, obj, skip, options):
         "WarmupMultiStepLR",
         "downgrade_config",
         "upgrade_config",
+        "add_export_config",
     }
     try:
         if name in HIDDEN or (

tests/test_export_caffe2.py

@@ -8,7 +8,7 @@
 import torch
 
 from detectron2 import model_zoo
-from detectron2.export import Caffe2Model, Caffe2Tracer, add_export_config
+from detectron2.export import Caffe2Model, Caffe2Tracer
 from detectron2.utils.logger import setup_logger
 from detectron2.utils.testing import get_sample_coco_image
 
@@ -22,7 +22,6 @@ def setUp(self):
 
     def _test_model(self, config_path, device="cpu"):
         cfg = model_zoo.get_config(config_path)
-        add_export_config(cfg)
         cfg.MODEL.DEVICE = device
         model = model_zoo.get(config_path, trained=True, device=device)
 

tools/deploy/export_model.py

@@ -11,13 +11,7 @@
 from detectron2.config import get_cfg
 from detectron2.data import build_detection_test_loader, detection_utils
 from detectron2.evaluation import COCOEvaluator, inference_on_dataset, print_csv_format
-from detectron2.export import (
-    Caffe2Tracer,
-    TracingAdapter,
-    add_export_config,
-    dump_torchscript_IR,
-    scripting_with_instances,
-)
+from detectron2.export import TracingAdapter, dump_torchscript_IR, scripting_with_instances
 from detectron2.modeling import GeneralizedRCNN, RetinaNet, build_model
 from detectron2.modeling.postprocessing import detector_postprocess
 from detectron2.projects.point_rend import add_pointrend_config
@@ -31,7 +25,6 @@ def setup_cfg(args):
     cfg = get_cfg()
     # cuda context is initialized before creating dataloader, so we don't fork anymore
     cfg.DATALOADER.NUM_WORKERS = 0
-    cfg = add_export_config(cfg)
     add_pointrend_config(cfg)
     cfg.merge_from_file(args.config_file)
     cfg.merge_from_list(args.opts)
@@ -40,6 +33,8 @@ def setup_cfg(args):
 
 
 def export_caffe2_tracing(cfg, torch_model, inputs):
+    from detectron2.export import Caffe2Tracer
+
     tracer = Caffe2Tracer(cfg, torch_model, inputs)
     if args.format == "caffe2":
         caffe2_model = tracer.export_caffe2()