v0.2.71

LICENSE

@@ -0,0 +1,11 @@
+Copyright 2024 Qualcomm® Technologies, Inc.
+
+Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+
+1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+
+3. Neither the name of the copyright holder nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

mypy.ini

@@ -0,0 +1,78 @@
+[mypy]
+# https://mypy.readthedocs.io/en/stable/config_file.html#import-discovery
+exclude = "qai_hub_models/models"
+
+[mypy-huggingface_hub.*]
+ignore_missing_imports = True
+
+[mypy-onnx.*]
+ignore_missing_imports = True
+
+[mypy-onnxsim.*]
+ignore_missing_imports = True
+
+[mypy-onnxruntime.*]
+ignore_missing_imports = True
+
+[mypy-pytimeparse.*]
+ignore_missing_imports = True
+
+[mypy-skimage.*]
+ignore_missing_imports = True
+
+[mypy-setuptools.*]
+ignore_missing_imports = True
+
+[mypy-tensorflow.*]
+ignore_missing_imports = True
+
+[mypy-torchvision.*]
+ignore_missing_imports = True
+
+[mypy-transformers.*]
+ignore_missing_imports = True
+
+[mypy-tqdm.*]
+ignore_missing_imports = True
+
+[mypy-tap.*]
+ignore_missing_imports = True
+
+[mypy-h5py.*]
+ignore_missing_imports = True
+
+[mypy-flatbuffers.*]
+ignore_missing_imports = True
+
+[mypy-soundfile.*]
+ignore_missing_imports = True
+
+[mypy-datasets.*]
+ignore_missing_imports = True
+
+[mypy-keras.*]
+ignore_missing_imports = True
+
+[mypy-rangefilter.filters.*]
+ignore_missing_imports = True
+
+[mypy-schema.*]
+ignore_missing_imports = True
+
+[mypy-gdown.*]
+ignore_missing_imports = True
+
+[mypy-aimet_torch.*]
+ignore_missing_imports = True
+
+[mypy-boto3.*]
+ignore_missing_imports = True
+
+[mypy-botocore.*]
+ignore_missing_imports = True
+
+[mypy-ruamel.*]
+ignore_missing_imports = True
+
+[mypy-qai_hub_models.models.*]
+ignore_errors = true

pyrightconfig.json

@@ -0,0 +1,26 @@
+{
+    "exclude": [
+        "**/.mypy_cache",
+        "**/.pytest_cache",
+        "**/__pycache__",
+        "**/node_modules",
+        "build/tungsten",
+        "src/public/staging_python/qai_hub_staging",
+        "src/tungsten",
+        "src/www/onnx-optimizer/third_party",
+        "src/www/onnx-simplifier/third_party",
+    ],
+
+    "extraPaths": [
+        "./build/proto"
+    ],
+
+    "reportMissingModuleSource": "none",
+    "reportMissingImports": true,
+    "reportMissingTypeStubs": false,
+    "reportShadowedImports": false,
+    "verboseOutput": false,
+
+    "venvPath": ".",
+    "venv": "qaism-dev"
+}

pytest.ini

@@ -0,0 +1,16 @@
+[pytest]
+testpaths = qai_hub_models
+norecursedirs = build
+python_files = tests.py test_*.py test.py
+filterwarnings =
+    ignore::DeprecationWarning:coremltools.*:
+    ignore::DeprecationWarning:torch.*:
+    ignore::DeprecationWarning:torchvision.*:
+    ignore::DeprecationWarning:tensorflow.*:
+    ignore::DeprecationWarning:tensorflow-macos.*:
+    ignore::DeprecationWarning:tensorflow-metal.*:
+    ignore::DeprecationWarning:tensorflow-probability.*:
+markers =
+    serial: test must not be run in parallel
+    slow: marks tests as slow
+    slow_cloud: marks test as slow and cloud-dependent

qai_hub_models/_version.py

@@ -0,0 +1 @@
+__version__ = "0.2.71"

qai_hub_models/asset_bases.yaml

@@ -0,0 +1,12 @@
+store_url: https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models
+web_asset_folder: models/{model_id}/web-assets
+static_web_banner_filename: banner.png
+animated_web_banner_filename: banner.mp4
+model_asset_folder: models/{model_id}/v{version}
+dataset_asset_folder: datasets/{dataset_id}/v{version}
+repo_url: https://github.com/quic/ai-hub-models/blob/main
+qaihm_repo: qai_hub_models/models/{model_id}
+example_use: qai_hub_models/models/{model_id}#example--usage
+huggingface_path: qualcomm/{model_name}
+models_website_url: https://aihub.qualcomm.com
+models_website_relative_path: models/{model_id}

qai_hub_models/conftest.py

@@ -0,0 +1,4 @@
+def pytest_configure(config):
+    config.addinivalue_line("markers", "compile: Run compile tests.")
+    config.addinivalue_line("markers", "profile: Run profile tests.")
+    config.addinivalue_line("markers", "inference: Run inference tests.")

qai_hub_models/datasets/bsd300.py

@@ -0,0 +1,124 @@
+from __future__ import annotations
+
+import os
+from typing import Tuple
+
+import numpy as np
+import torch
+from PIL import Image
+
+from qai_hub_models.datasets.common import BaseDataset
+from qai_hub_models.utils.asset_loaders import CachedWebDatasetAsset
+
+BSD300_URL = (
+    "https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/bsds/BSDS300-images.tgz"
+)
+BSD300_FOLDER_NAME = "BSDS300"
+BSD300_VERSION = 1
+BSD300_ASSET = CachedWebDatasetAsset(
+    BSD300_URL, BSD300_FOLDER_NAME, BSD300_VERSION, "BSDS300.tgz"
+)
+DATASET_LENGTH = 200
+
+
+class BSD300Dataset(BaseDataset):
+    """
+    BSD300 published here: https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/bsds/
+    """
+
+    def __init__(self, scaling_factor=4):
+        self.bsd_path = BSD300_ASSET.path(extracted=True)
+        self.images_path = os.path.join(self.bsd_path, "images/train")
+        BaseDataset.__init__(self, self.bsd_path)
+        self.scaling_factor = scaling_factor
+
+    def _validate_data(self) -> bool:
+        images_path = os.path.join(self.dataset_path, "images/train")
+
+        # Check image path exists
+        if not os.path.exists(images_path):
+            return False
+
+        # Ensure the correct number of images are there
+        files = os.listdir(images_path)
+        images = [f for f in files if ".jpg" in f]
+        if len(images) != DATASET_LENGTH:
+            return False
+
+        return True
+
+    def _prepare_data(self):
+        # Rename images to be more friendly to enumeration
+        directory = os.path.join(self.dataset_path, "images/train")
+        files = os.listdir(directory)
+        for i, filename in enumerate(files):
+            if filename.endswith(".jpg"):
+                # Open the image and convert it to png
+                try:
+                    with Image.open(os.path.join(directory, filename)) as img:
+                        img.save(os.path.join(directory, f"img_{i + 1:03d}_HR.jpg"))
+                    # delete the old image
+                    os.remove(os.path.join(directory, filename))
+                except ValueError:
+                    print(f"File {filename} does not exist!")
+
+    def __len__(self):
+        return DATASET_LENGTH
+
+    def __getitem__(self, item) -> Tuple[torch.Tensor, torch.Tensor]:
+        # We use the super resolution GT-and-test image preparation from AIMET zoo:
+        # https://github.com/quic/aimet-model-zoo/blob/d09d2b0404d10f71a7640a87e9d5e5257b028802/aimet_zoo_torch/quicksrnet/dataloader/utils.py#L51
+
+        img = np.asarray(
+            Image.open(os.path.join(self.images_path, f"img_{item + 1:03d}_HR.jpg"))
+        )
+        height, width = img.shape[0:2]
+
+        # If portrait, transpose to landscape so that all tensors are equal size
+        if height > width:
+            img = np.transpose(img, (1, 0, 2))
+            height, width = img.shape[0:2]
+
+        # Take the largest possible center-crop of it such that its dimensions are perfectly divisible by the scaling factor
+        x_remainder = width % (
+            2 * self.scaling_factor
+            if self.scaling_factor == 1.5
+            else self.scaling_factor
+        )
+        y_remainder = height % (
+            2 * self.scaling_factor
+            if self.scaling_factor == 1.5
+            else self.scaling_factor
+        )
+        left = int(x_remainder // 2)
+        top = int(y_remainder // 2)
+        right = int(left + (width - x_remainder))
+        bottom = int(top + (height - y_remainder))
+        hr_img = img[top:bottom, left:right]
+
+        hr_height, hr_width = hr_img.shape[0:2]
+
+        hr_img = np.array(hr_img, dtype="uint8")
+        new_size = (int(width / self.scaling_factor), int(height / self.scaling_factor))
+        lr_img = np.asarray(Image.fromarray(hr_img).resize(new_size))
+        lr_img = np.clip(lr_img, 0.0, 255.0).astype(np.uint8)
+
+        lr_height, lr_width = lr_img.shape[0:2]
+
+        # Sanity check
+        assert (
+            hr_width == lr_width * self.scaling_factor
+            and hr_height == lr_height * self.scaling_factor
+        )
+
+        lr_img_tensor = torch.from_numpy(lr_img.transpose((2, 0, 1))).contiguous()
+        lr_img_tensor = lr_img_tensor.to(dtype=torch.float32).div(255)
+
+        hr_img_tensor = torch.from_numpy(hr_img.transpose((2, 0, 1))).contiguous()
+        hr_img_tensor = hr_img_tensor.to(dtype=torch.float32).div(255)
+
+        return lr_img_tensor, hr_img_tensor
+
+    def _download_data(self) -> None:
+        BSD300_ASSET.fetch(extract=True)
+        self._prepare_data()

qai_hub_models/datasets/coco.py

@@ -0,0 +1,117 @@
+import os
+from typing import Tuple, Union
+
+import torch
+from torch.utils.data.dataloader import default_collate
+from torchvision.datasets.coco import CocoDetection
+
+from qai_hub_models.datasets.common import BaseDataset
+from qai_hub_models.utils.asset_loaders import CachedWebDatasetAsset
+from qai_hub_models.utils.image_processing import app_to_net_image_inputs
+
+DATASET_ID = "coco"
+DATASET_ASSET_VERSION = 1
+COCO_DATASET = CachedWebDatasetAsset(
+    "http://images.cocodataset.org/zips/val2017.zip",
+    DATASET_ID,
+    DATASET_ASSET_VERSION,
+    "val2017.zip",
+)
+COCO_ANNOTATIONS = CachedWebDatasetAsset(
+    "http://images.cocodataset.org/annotations/annotations_trainval2017.zip",
+    DATASET_ID,
+    DATASET_ASSET_VERSION,
+    "annotations_trainval2017.zip",
+)
+
+
+def collate_fn(batch):
+    try:
+        image, gt = batch[0][0], batch[0][1]
+        image_id, height, width, boxes, labels = gt
+        new_list = []
+        new_list.append(default_collate([i for i in image if torch.is_tensor(i)]))
+        target = (
+            torch.tensor(image_id),
+            torch.tensor(height),
+            torch.tensor(width),
+            default_collate([i for i in boxes if torch.is_tensor(i)]),
+            default_collate([i for i in labels if torch.is_tensor(i)]),
+        )
+        new_list.append(target)
+        return new_list
+    except Exception:
+        return [], ([], [], [], [], [])
+
+
+class CocoDataset(BaseDataset, CocoDetection):
+    """
+    Class for using the COCODetection dataset published here:
+
+
+    Contains ~5k images spanning 80 classes.
+    """
+
+    def __init__(self, target_image_size: Union[int, Tuple[int, int]] = 640):
+        BaseDataset.__init__(self, str(COCO_DATASET.path(extracted=True)))
+        CocoDetection.__init__(
+            self,
+            root=COCO_DATASET.path() / "val2017",
+            annFile=COCO_ANNOTATIONS.path() / "annotations" / "instances_val2017.json",
+        )
+
+        categories = self.coco.loadCats(self.coco.getCatIds())
+        categories.sort(key=lambda x: x["id"])
+        self.label_map = {}
+        counter = 0
+        for c in categories:
+            self.label_map[c["id"]] = counter
+            counter += 1
+        self.target_image_size = (
+            target_image_size
+            if isinstance(target_image_size, tuple)
+            else (target_image_size, target_image_size)
+        )
+
+    def __getitem__(self, item):
+        image, target = super(CocoDataset, self).__getitem__(item)
+        width, height = image.size
+        boxes = []
+        labels = []
+        if len(target) == 0:
+            return None, (None, None)
+        for annotation in target:
+            bbox = annotation.get("bbox")
+            boxes.append(
+                [
+                    bbox[0] / width,
+                    bbox[1] / height,
+                    (bbox[0] + bbox[2]) / width,
+                    (bbox[1] + bbox[3]) / height,
+                ]
+            )
+            labels.append(self.label_map[annotation.get("category_id")])
+        boxes = torch.tensor(boxes)
+        labels = torch.tensor(labels)
+        image = image.resize(self.target_image_size)
+        image = app_to_net_image_inputs(image)[1]
+        return image, (target[0]["image_id"], height, width, boxes, labels)
+
+    def _validate_data(self) -> bool:
+        # Check validation data exists
+        if not COCO_DATASET.path().exists():
+            return False
+
+        # Check annotations exist
+        if not COCO_ANNOTATIONS.path().exists():
+            return False
+
+        # Ensure there are 5000 samples
+        if len(os.listdir(COCO_DATASET.path() / "val2017")) < 5000:
+            return False
+
+        return True
+
+    def _download_data(self) -> None:
+        COCO_DATASET.fetch(extract=True)
+        COCO_ANNOTATIONS.fetch(extract=True)