【Fix】Fix ppl bug about grid sample (#325)

* fix ppl problems

* fix roialign

* fix grid_sampler bug

* fix grid sampler

* fix config

* fix test

configs/mmdet/_base_/base_instance-seg_static.py

@@ -1,3 +1,4 @@
 _base_ = ['./base_static.py']
 
 onnx_config = dict(output_names=['dets', 'labels', 'masks'])
+codebase_config = dict(post_processing=dict(export_postprocess_mask=False))

mmdeploy/pytorch/ops/grid_sampler.py

@@ -2,6 +2,7 @@
 from torch.onnx.symbolic_helper import parse_args
 
 from mmdeploy.core import SYMBOLIC_REWRITER
+from mmdeploy.utils import Backend, get_backend
 
 
 @parse_args('v', 'v', 'i', 'i', 'i')
@@ -26,10 +27,36 @@ def grid_sampler(g,
         align_corners_i=align_corners)
 
 
+@parse_args('v', 'v', 'i', 'i', 'i')
+def grid_sampler_ppl(g,
+                     input,
+                     grid,
+                     interpolation_mode,
+                     padding_mode,
+                     align_corners=False):
+    """Symbolic function for `grid_sampler`.
+
+    PyTorch does not support export grid_sampler to ONNX by default. We add the
+    support here. `grid_sampler` will be exported as ONNX node
+    'mmdeploy::grid_sampler'
+    """
+    return g.op(
+        'mmcv::grid_sampler',
+        input,
+        grid,
+        interpolation_mode_i=interpolation_mode,
+        padding_mode_i=padding_mode,
+        align_corners_i=align_corners)
+
+
 @SYMBOLIC_REWRITER.register_symbolic('grid_sampler', is_pytorch=True)
 def grid_sampler__default(ctx, *args):
     """Register default symbolic function for `grid_sampler`.
 
     Add support to grid_sample to ONNX.
     """
+    backend = get_backend(ctx.cfg)
+    if backend == Backend.PPLNN:
+        return grid_sampler_ppl(*args)
+    else:
         return grid_sampler(*args)

tests/test_pytorch/test_pytorch_ops.py

@@ -4,6 +4,7 @@ import tempfile
 import onnx
 import pytest
 import torch
+from mmcv import Config
 
 from mmdeploy.core import RewriterContext
 
@@ -12,7 +13,10 @@ onnx_file = tempfile.NamedTemporaryFile(suffix='onnx').name
 
 @pytest.fixture(autouse=True, scope='module')
 def prepare_symbolics():
-    context = RewriterContext({}, 'tensorrt', opset=11)
+    context = RewriterContext(
+        Config({'backend_config': {
+            'type': 'tensorrt'
+        }}), 'tensorrt', opset=11)
     context.enter()
 
     yield