Remove zoedepth non-automated zoedepth changes and zoedepth test

Signed-off-by: Phillip Kuznetsov <[email protected]>

Author: philkuz

src/transformers/models/zoedepth/modeling_zoedepth.py

@@ -367,32 +367,29 @@ def __init__(self, config):
         self.projection = None
         if config.add_projection:
             self.projection = nn.Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
-            self.projection_act = nn.ReLU()
 
         features = config.fusion_hidden_size
         self.conv1 = nn.Conv2d(features, features // 2, kernel_size=3, stride=1, padding=1)
         self.upsample = nn.Upsample(scale_factor=2, mode="bilinear", align_corners=True)
         self.conv2 = nn.Conv2d(features // 2, config.num_relative_features, kernel_size=3, stride=1, padding=1)
-        self.conv2_act = nn.ReLU()
         self.conv3 = nn.Conv2d(config.num_relative_features, 1, kernel_size=1, stride=1, padding=0)
-        self.conv3_act = nn.ReLU()
 
     def forward(self, hidden_states: List[torch.Tensor]) -> torch.Tensor:
         # use last features
         hidden_states = hidden_states[self.head_in_index]
 
         if self.projection is not None:
             hidden_states = self.projection(hidden_states)
-            hidden_states = self.projection_act(hidden_states)
+            hidden_states = nn.ReLU()(hidden_states)
 
         hidden_states = self.conv1(hidden_states)
         hidden_states = self.upsample(hidden_states)
         hidden_states = self.conv2(hidden_states)
-        hidden_states = self.conv2_act(hidden_states)
+        hidden_states = nn.ReLU()(hidden_states)
         # we need the features here (after second conv + ReLu)
         features = hidden_states
         hidden_states = self.conv3(hidden_states)
-        hidden_states = self.conv3_act(hidden_states)
+        hidden_states = nn.ReLU()(hidden_states)
 
         predicted_depth = hidden_states.squeeze(dim=1)
 

tests/models/zoedepth/test_modeling_zoedepth.py

@@ -20,7 +20,6 @@
 
 from transformers import Dinov2Config, ZoeDepthConfig
 from transformers.file_utils import is_torch_available, is_vision_available
-from transformers.pytorch_utils import is_torch_greater_or_equal_than_2_4
 from transformers.testing_utils import require_torch, require_vision, slow, torch_device
 
 from ...test_configuration_common import ConfigTester
@@ -355,29 +354,3 @@ def test_inference_depth_estimation_post_processing_pad_flip(self):
         model = ZoeDepthForDepthEstimation.from_pretrained("Intel/zoedepth-nyu-kitti").to(torch_device)
 
         self.check_post_processing_test(image_processor, images, model, pad_input=True, flip_aug=True)
-
-    def test_export(self):
-        self.skipTest(
-            reason="This test fails because the beit backbone of ZoeDepth is not compatible with torch.export"
-        )
-        for strict in [True, False]:
-            with self.subTest(strict=True):
-                if not is_torch_greater_or_equal_than_2_4:
-                    self.skipTest(reason="This test requires torch >= 2.4 to run.")
-                model = ZoeDepthForDepthEstimation.from_pretrained("Intel/zoedepth-nyu").to(torch_device).eval()
-                image_processor = ZoeDepthImageProcessor.from_pretrained("Intel/zoedepth-nyu")
-                image = prepare_img()
-                inputs = image_processor(images=image, return_tensors="pt").to(torch_device)
-
-                exported_program = torch.export.export(
-                    model,
-                    args=(inputs["pixel_values"],),
-                    strict=strict,
-                )
-                with torch.no_grad():
-                    eager_outputs = model(**inputs)
-                    exported_outputs = exported_program.module().forward(inputs["pixel_values"])
-                self.assertEqual(eager_outputs.predicted_depth.shape, exported_outputs.predicted_depth.shape)
-                self.assertTrue(
-                    torch.allclose(eager_outputs.predicted_depth, exported_outputs.predicted_depth, atol=1e-4)
-                )