update installation, support latest transformers version

.github/workflows/ci.yml

@@ -72,30 +72,30 @@ jobs:
         run: >
           pip install -r requirements.txt
 
-      - name: Install PyTorch(1.12.1) and TorchVision(0.13.1) on Linux and Windows
+      - name: Install PyTorch(1.13.1) and TorchVision(0.14.1) on Linux and Windows
         if: >
           matrix.operating-system == 'ubuntu-latest' ||
           matrix.operating-system == 'windows-latest'
         run: >
-          pip install torch==1.12.1+cpu torchvision==0.13.1+cpu
+          pip install torch==1.13.1+cpu torchvision==0.14.1+cpu
           -f https://download.pytorch.org/whl/torch_stable.html
 
       - name: Install PyTorch on MacOS
         if: matrix.operating-system == 'macos-latest'
-        run: pip install torch==1.12.1 torchvision==0.13.1
+        run: pip install torch==1.13.1 torchvision==0.14.1
 
       - name: Install MMDetection(2.26.0) with MMCV(1.7.0)
         run: >
-          pip install mmcv-full==1.7.0 -f https://download.openmmlab.com/mmcv/dist/cpu/torch1.12.0/index.html
+          pip install mmcv-full==1.7.0 -f https://download.openmmlab.com/mmcv/dist/cpu/torch1.13.0/index.html
           pip install mmdet==2.26.0
 
-      - name: Install YOLOv5(7.0.3)
+      - name: Install YOLOv5(7.0.4)
         run: >
-          pip install yolov5==7.0.3
+          pip install yolov5==7.0.4
 
-      - name: Install Transformers(4.24.0)
+      - name: Install Transformers(4.25.1)
         run: >
-          pip install transformers==4.24.0
+          pip install transformers==4.25.1
 
       - name: Install pycocotools(2.0.6)
         run: >

.github/workflows/ci_torch1.10.yml

@@ -90,19 +90,19 @@ jobs:
           pip install mmcv-full==1.7.0 -f https://download.openmmlab.com/mmcv/dist/cpu/torch1.10.0/index.html
           pip install mmdet==2.26.0
 
-      - name: Install YOLOv5(7.0.3)
+      - name: Install YOLOv5(7.0.4)
         run: >
-          pip install yolov5==7.0.3
+          pip install yolov5==7.0.4
 
       - name: Install Detectron2(0.6)
         run: >
           python -m pip install detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cpu/torch1.10/index.html
 
-      - name: Install Transformers(4.24.0)
+      - name: Install Transformers(4.25.1)
         if: >
           matrix.python-version != '3.6'
         run: >
-          pip install transformers==4.24.0
+          pip install transformers==4.25.1
 
       - name: Install pycocotools(2.0.6)
         run: >

.github/workflows/package_testing.yml

@@ -62,24 +62,22 @@ jobs:
         if: matrix.operating-system == 'macos-latest'
         run: pip install torch==1.10.1 torchvision==0.11.2
 
-      - name: Install MMDetection(2.25.3) with MMCV(1.7.0)
+      - name: Install MMDetection(2.26.9) with MMCV(1.7.0)
         run: >
           pip install mmcv-full==1.7.0 -f https://download.openmmlab.com/mmcv/dist/cpu/torch1.10.0/index.html
-          pip install mmdet==2.25.3
+          pip install mmdet==2.26.0
 
-      - name: Install YOLOv5(7.0.3)
+      - name: Install YOLOv5(7.0.4)
         run: >
-          pip install yolov5==7.0.3
+          pip install yolov5==7.0.4
 
       - name: Install Detectron2(0.6)
         run: >
           python -m pip install detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cpu/torch1.10/index.html
 
-      - name: Install Transformers(4.24.0)
-        if: >
-          matrix.python-version != '3.6'
+      - name: Install Transformers(4.25.1)
         run: >
-          pip install transformers==4.24.0
+          pip install transformers==4.25.1
 
       - name: Install pycocotools(2.0.6)
         run: >

README.md

@@ -109,22 +109,26 @@ pip install sahi
 conda install -c conda-forge shapely
 ```
 
-- Install your desired version of pytorch and torchvision:
+- Install your desired version of pytorch and torchvision (cuda 11.3 for detectron2, cuda 11.7 for rest):
 
 ```console
-conda install pytorch=1.12.1 torchvision=0.13.1 cudatoolkit=11.3 -c pytorch
+conda install pytorch=1.10.2 torchvision=0.11.3 cudatoolkit=11.3 -c pytorch
+```
+
+```console
+conda install pytorch=1.13.1 torchvision=0.14.1 pytorch-cuda=11.7 -c pytorch -c nvidia
 ```
 
 - Install your desired detection framework (yolov5):
 
 ```console
-pip install yolov5==6.2.3
+pip install yolov5==7.0.4
 ```
 
 - Install your desired detection framework (mmdet):
 
 ```console
-pip install mmcv-full==1.7.0 -f https://download.openmmlab.com/mmcv/dist/cu113/torch1.11.0/index.html
+pip install mmcv-full==1.7.0 -f https://download.openmmlab.com/mmcv/dist/cu117/torch1.13.0/index.html
 ```
 
 ```console

sahi/models/huggingface.py

@@ -20,7 +20,7 @@ def __init__(
         self,
         model_path: Optional[str] = None,
         model: Optional[Any] = None,
-        feature_extractor: Optional[Any] = None,
+        processor: Optional[Any] = None,
         config_path: Optional[str] = None,
         device: Optional[str] = None,
         mask_threshold: float = 0.5,
@@ -31,7 +31,7 @@ def __init__(
         image_size: int = None,
     ):
 
-        self._feature_extractor = feature_extractor
+        self._processor = processor
         self._image_shapes = []
         super().__init__(
             model_path,
@@ -48,11 +48,11 @@ def __init__(
 
     def check_dependencies(self):
         check_requirements(["torch", "transformers"])
-        ensure_package_minimum_version("transformers", "4.24.0")
+        ensure_package_minimum_version("transformers", "4.25.1")
 
     @property
-    def feature_extractor(self):
-        return self._feature_extractor
+    def processor(self):
+        return self._processor
 
     @property
     def image_shapes(self):
@@ -67,31 +67,28 @@ def num_categories(self) -> int:
 
     def load_model(self):
 
-        from transformers import AutoFeatureExtractor, AutoModelForObjectDetection
+        from transformers import AutoModelForObjectDetection, AutoProcessor
 
         model = AutoModelForObjectDetection.from_pretrained(self.model_path)
         if self.image_size is not None:
-            feature_extractor = AutoFeatureExtractor.from_pretrained(
-                self.model_path, size=self.image_size, do_resize=True
+            processor = AutoProcessor.from_pretrained(
+                self.model_path, size={"shortest_edge": self.image_size, "longest_edge": None}, do_resize=True
             )
         else:
-            feature_extractor = AutoFeatureExtractor.from_pretrained(self.model_path)
-        self.set_model(model, feature_extractor)
-
-    def set_model(self, model: Any, feature_extractor: Any = None):
-        feature_extractor = feature_extractor or self.feature_extractor
-        if feature_extractor is None:
-            raise ValueError(f"'feature_extractor' is required to be set, got {feature_extractor}.")
-        elif (
-            "ObjectDetection" not in model.__class__.__name__
-            or "FeatureExtractor" not in feature_extractor.__class__.__name__
-        ):
+            processor = AutoProcessor.from_pretrained(self.model_path)
+        self.set_model(model, processor)
+
+    def set_model(self, model: Any, processor: Any = None):
+        processor = processor or self.processor
+        if processor is None:
+            raise ValueError(f"'processor' is required to be set, got {processor}.")
+        elif "ObjectDetection" not in model.__class__.__name__ or "ImageProcessor" not in processor.__class__.__name__:
             raise ValueError(
-                "Given 'model' is not an ObjectDetectionModel or 'feature_extractor' is not a valid FeatureExtractor."
+                "Given 'model' is not an ObjectDetectionModel or 'processor' is not a valid ImageProcessor."
             )
         self.model = model
         self.model.to(self.device)
-        self._feature_extractor = feature_extractor
+        self._processor = processor
         self.category_mapping = self.model.config.id2label
 
     def perform_inference(self, image: Union[List, np.ndarray]):
@@ -108,7 +105,7 @@ def perform_inference(self, image: Union[List, np.ndarray]):
             raise RuntimeError("Model is not loaded, load it by calling .load_model()")
 
         with torch.no_grad():
-            inputs = self.feature_extractor(images=image, return_tensors="pt")
+            inputs = self.processor(images=image, return_tensors="pt")
             inputs["pixel_values"] = inputs.pixel_values.to(self.device)
             if hasattr(inputs, "pixel_mask"):
                 inputs["pixel_mask"] = inputs.pixel_mask.to(self.device)

tests/test_huggingfacemodel.py

@@ -32,20 +32,18 @@ def test_load_model(self):
             self.assertNotEqual(huggingface_detection_model.model, None)
 
         def test_set_model(self):
-            from transformers import AutoFeatureExtractor, AutoModelForObjectDetection
+            from transformers import AutoModelForObjectDetection, AutoProcessor
 
             from sahi.models.huggingface import HuggingfaceDetectionModel
 
             huggingface_model = AutoModelForObjectDetection.from_pretrained(
                 HuggingfaceTestConstants.YOLOS_TINY_MODEL_PATH
             )
-            huggingface_feature_extractor = AutoFeatureExtractor.from_pretrained(
-                HuggingfaceTestConstants.YOLOS_TINY_MODEL_PATH
-            )
+            huggingface_processor = AutoProcessor.from_pretrained(HuggingfaceTestConstants.YOLOS_TINY_MODEL_PATH)
 
             huggingface_detection_model = HuggingfaceDetectionModel(
                 model=huggingface_model,
-                feature_extractor=huggingface_feature_extractor,
+                processor=huggingface_processor,
                 confidence_threshold=CONFIDENCE_THRESHOLD,
                 device=MODEL_DEVICE,
                 category_remapping=None,