Merge branch 'main' into prototype-datasets-inheritance

pmeier · pmeier · commit 1916bd772b6c · 2022-02-24T16:47:06.000+01:00
diff --git a/.circleci/config.yml b/.circleci/config.yml
diff --git a/.circleci/config.yml.in b/.circleci/config.yml.in
@@ -848,7 +848,7 @@ jobs:
     executor:
       name: windows-gpu
     environment:
-      CUDA_VERSION: "11.1"
+      CUDA_VERSION: "11.3"
       PYTHON_VERSION: << parameters.python_version >>
     steps:
       - checkout
diff --git a/.circleci/regenerate.py b/.circleci/regenerate.py
@@ -33,7 +33,7 @@ def build_workflows(prefix="", filter_branch=None, upload=False, indentation=6,
             python_versions = PYTHON_VERSIONS
             cu_versions_dict = {
                 "linux": ["cpu", "cu102", "cu111", "cu113", "cu115", "rocm4.3.1", "rocm4.5.2"],
-                "win": ["cpu", "cu111", "cu113", "cu115"],
+                "win": ["cpu", "cu113", "cu115"],
                 "macos": ["cpu"],
             }
             cu_versions = cu_versions_dict[os_type]
diff --git a/.circleci/unittest/windows/scripts/environment.yml b/.circleci/unittest/windows/scripts/environment.yml
@@ -9,10 +9,11 @@ dependencies:
   - libpng
   - jpeg
   - ca-certificates
-  - h5py
+  - hdf5
   - pip:
     - future
     - pillow >=5.3.0, !=8.3.*
     - scipy
     - av
     - dataclasses
+    - h5py
diff --git a/.circleci/unittest/windows/scripts/install.sh b/.circleci/unittest/windows/scripts/install.sh
@@ -28,7 +28,7 @@ else
 fi
 
 printf "Installing PyTorch with %s\n" "${cudatoolkit}"
-conda install -y -c "pytorch-${UPLOAD_CHANNEL}" -c conda-forge "pytorch-${UPLOAD_CHANNEL}"::pytorch[build="*${version}*"] "${cudatoolkit}" pytest
+conda install -y -c "pytorch-${UPLOAD_CHANNEL}" -c nvidia "pytorch-${UPLOAD_CHANNEL}"::pytorch[build="*${version}*"] "${cudatoolkit}" pytest
 
 torch_cuda=$(python -c "import torch; print(torch.cuda.is_available())")
 echo torch.cuda.is_available is $torch_cuda
diff --git a/scripts/release_notes/classify_prs.py b/scripts/release_notes/classify_prs.py
@@ -1,13 +1,18 @@
 # In[1]:
 
-
+# imports and set configuration
 import pandas as pd
+from retrieve_prs_data import run
 
+exclude_prototype = True
+data_filename = "10.0_to_11.0-rc2.json"
+previous_release = "v10.0"
+current_release = "v11.0-rc2"
 
 # In[2]:
 
 
-df = pd.read_json("10.0_to_11.0-rc2.json").T
+df = pd.read_json(data_filename).T
 df.tail()
 
 
@@ -76,6 +81,8 @@
 def format_prs(mod_df):
     out = []
     for idx, row in mod_df.iterrows():
+        if exclude_prototype and row["prototype"]:
+            continue
         modules = idx
         # Put "documentation" and "tests" first for sorting to be dece
         for last_module in ("documentation", "tests"):
@@ -119,3 +126,13 @@ def format_prs(mod_df):
 # Missing PRs are these ones... classify them manually
 missing_prs = pd.concat([mod_df, included_prs]).drop_duplicates(subset="pr_number", keep=False)
 print(format_prs(missing_prs))
+
+# In[12]:
+
+# Generate list of contributors
+print()
+print("## Contributors")
+
+command_to_run = f"{{ git shortlog -s {previous_release}..{current_release} | cut -f2- & git log -s {previous_release}..{current_release} | grep Co-authored | cut -f2- -d: | cut -f1 -d\\< | sed 's/^ *//;s/ *$//' ; }} | sort --ignore-case | uniq | tr '\\n' ';' | sed 's/;/, /g;s/, $//' | fold -s"
+rc, output, err = run(command_to_run)
+print(output)
diff --git a/test/test_transforms_tensor.py b/test/test_transforms_tensor.py
@@ -1,4 +1,5 @@
 import os
+import sys
 
 import numpy as np
 import pytest
@@ -958,6 +959,17 @@ def test_random_apply(device):
 )
 @pytest.mark.parametrize("channels", [1, 3])
 def test_gaussian_blur(device, channels, meth_kwargs):
+    if all(
+        [
+            device == "cuda",
+            channels == 1,
+            meth_kwargs["kernel_size"] in [23, [23]],
+            torch.version.cuda == "11.3",
+            sys.platform in ("win32", "cygwin"),
+        ]
+    ):
+        pytest.skip("Fails on Windows, see https://github.com/pytorch/vision/issues/5464")
+
     tol = 1.0 + 1e-10
     torch.manual_seed(12)
     _test_class_op(
diff --git a/torchvision/models/detection/_utils.py b/torchvision/models/detection/_utils.py
@@ -468,3 +468,27 @@ def retrieve_out_channels(model: nn.Module, size: Tuple[int, int]) -> List[int]:
         model.train()
 
     return out_channels
+
+
+def _topk_min(input: Tensor, orig_kval: int, axis: int) -> Tensor:
+    """
+    ONNX spec requires the k-value to be less than or equal to the number of inputs along
+    provided dim. Certain models use the number of elements along a particular axis instead of K
+    if K exceeds the number of elements along that axis. Previously, python's min() function was
+    used to determine whether to use the provided k-value or the specified dim axis value.
+
+    However in cases where the model is being exported in tracing mode, python min() is
+    static causing the model to be traced incorrectly and eventually fail at the topk node.
+    In order to avoid this situation, in tracing mode, torch.min() is used instead.
+
+    Args:
+        input (Tensor): The orignal input tensor.
+        orig_kval (int): The provided k-value.
+        axis(int): Axis along which we retreive the input size.
+
+    Returns:
+        min_kval (Tensor): Appropriately selected k-value.
+    """
+    axis_dim_val = torch._shape_as_tensor(input)[axis].unsqueeze(0)
+    min_kval = torch.min(torch.cat((torch.tensor([orig_kval], dtype=axis_dim_val.dtype), axis_dim_val), 0))
+    return min_kval  # type: ignore[arg-type]
diff --git a/torchvision/models/detection/fcos.py b/torchvision/models/detection/fcos.py
@@ -501,7 +501,7 @@ def postprocess_detections(
                 topk_idxs = torch.where(keep_idxs)[0]
 
                 # keep only topk scoring predictions
-                num_topk = min(self.topk_candidates, topk_idxs.size(0))
+                num_topk = det_utils._topk_min(topk_idxs, self.topk_candidates, 0)
                 scores_per_level, idxs = scores_per_level.topk(num_topk)
                 topk_idxs = topk_idxs[idxs]
 
diff --git a/torchvision/models/detection/retinanet.py b/torchvision/models/detection/retinanet.py
@@ -436,7 +436,7 @@ def postprocess_detections(self, head_outputs, anchors, image_shapes):
                 topk_idxs = torch.where(keep_idxs)[0]
 
                 # keep only topk scoring predictions
-                num_topk = min(self.topk_candidates, topk_idxs.size(0))
+                num_topk = det_utils._topk_min(topk_idxs, self.topk_candidates, 0)
                 scores_per_level, idxs = scores_per_level.topk(num_topk)
                 topk_idxs = topk_idxs[idxs]
 
diff --git a/torchvision/models/detection/rpn.py b/torchvision/models/detection/rpn.py
@@ -1,7 +1,6 @@
-from typing import List, Optional, Dict, Tuple, cast
+from typing import List, Optional, Dict, Tuple
 
 import torch
-import torchvision
 from torch import nn, Tensor
 from torch.nn import functional as F
 from torchvision.ops import boxes as box_ops
@@ -13,17 +12,6 @@
 from .image_list import ImageList
 
 
-@torch.jit.unused
-def _onnx_get_num_anchors_and_pre_nms_top_n(ob: Tensor, orig_pre_nms_top_n: int) -> Tuple[int, int]:
-    from torch.onnx import operators
-
-    num_anchors = operators.shape_as_tensor(ob)[1].unsqueeze(0)
-    pre_nms_top_n = torch.min(torch.cat((torch.tensor([orig_pre_nms_top_n], dtype=num_anchors.dtype), num_anchors), 0))
-
-    # for mypy we cast at runtime
-    return cast(int, num_anchors), cast(int, pre_nms_top_n)
-
-
 class RPNHead(nn.Module):
     """
     Adds a simple RPN Head with classification and regression heads
@@ -206,11 +194,8 @@ def _get_top_n_idx(self, objectness: Tensor, num_anchors_per_level: List[int]) -
         r = []
         offset = 0
         for ob in objectness.split(num_anchors_per_level, 1):
-            if torchvision._is_tracing():
-                num_anchors, pre_nms_top_n = _onnx_get_num_anchors_and_pre_nms_top_n(ob, self.pre_nms_top_n())
-            else:
-                num_anchors = ob.shape[1]
-                pre_nms_top_n = min(self.pre_nms_top_n(), num_anchors)
+            num_anchors = ob.shape[1]
+            pre_nms_top_n = det_utils._topk_min(ob, self.pre_nms_top_n(), 1)
             _, top_n_idx = ob.topk(pre_nms_top_n, dim=1)
             r.append(top_n_idx + offset)
             offset += num_anchors
diff --git a/torchvision/models/detection/ssd.py b/torchvision/models/detection/ssd.py
@@ -407,7 +407,7 @@ def postprocess_detections(
                 box = boxes[keep_idxs]
 
                 # keep only topk scoring predictions
-                num_topk = min(self.topk_candidates, score.size(0))
+                num_topk = det_utils._topk_min(score, self.topk_candidates, 0)
                 score, idxs = score.topk(num_topk)
                 box = box[idxs]
 
diff --git a/torchvision/transforms/functional.py b/torchvision/transforms/functional.py
@@ -338,30 +338,9 @@ def normalize(tensor: Tensor, mean: List[float], std: List[float], inplace: bool
     if not torch.jit.is_scripting() and not torch.jit.is_tracing():
         _log_api_usage_once(normalize)
     if not isinstance(tensor, torch.Tensor):
-        raise TypeError(f"Input tensor should be a torch tensor. Got {type(tensor)}.")
+        raise TypeError(f"img should be Tensor Image. Got {type(tensor)}")
 
-    if not tensor.is_floating_point():
-        raise TypeError(f"Input tensor should be a float tensor. Got {tensor.dtype}.")
-
-    if tensor.ndim < 3:
-        raise ValueError(
-            f"Expected tensor to be a tensor image of size (..., C, H, W). Got tensor.size() = {tensor.size()}"
-        )
-
-    if not inplace:
-        tensor = tensor.clone()
-
-    dtype = tensor.dtype
-    mean = torch.as_tensor(mean, dtype=dtype, device=tensor.device)
-    std = torch.as_tensor(std, dtype=dtype, device=tensor.device)
-    if (std == 0).any():
-        raise ValueError(f"std evaluated to zero after conversion to {dtype}, leading to division by zero.")
-    if mean.ndim == 1:
-        mean = mean.view(-1, 1, 1)
-    if std.ndim == 1:
-        std = std.view(-1, 1, 1)
-    tensor.sub_(mean).div_(std)
-    return tensor
+    return F_t.normalize(tensor, mean=mean, std=std, inplace=inplace)
 
 
 def resize(
@@ -1281,11 +1260,7 @@ def erase(img: Tensor, i: int, j: int, h: int, w: int, v: Tensor, inplace: bool
     if not isinstance(img, torch.Tensor):
         raise TypeError(f"img should be Tensor Image. Got {type(img)}")
 
-    if not inplace:
-        img = img.clone()
-
-    img[..., i : i + h, j : j + w] = v
-    return img
+    return F_t.erase(img, i, j, h, w, v, inplace=inplace)
 
 
 def gaussian_blur(img: Tensor, kernel_size: List[int], sigma: Optional[List[float]] = None) -> Tensor:
diff --git a/torchvision/transforms/functional_tensor.py b/torchvision/transforms/functional_tensor.py
@@ -918,3 +918,40 @@ def equalize(img: Tensor) -> Tensor:
         return _equalize_single_image(img)
 
     return torch.stack([_equalize_single_image(x) for x in img])
+
+
+def normalize(tensor: Tensor, mean: List[float], std: List[float], inplace: bool = False) -> Tensor:
+    _assert_image_tensor(tensor)
+
+    if not tensor.is_floating_point():
+        raise TypeError(f"Input tensor should be a float tensor. Got {tensor.dtype}.")
+
+    if tensor.ndim < 3:
+        raise ValueError(
+            f"Expected tensor to be a tensor image of size (..., C, H, W). Got tensor.size() = {tensor.size()}"
+        )
+
+    if not inplace:
+        tensor = tensor.clone()
+
+    dtype = tensor.dtype
+    mean = torch.as_tensor(mean, dtype=dtype, device=tensor.device)
+    std = torch.as_tensor(std, dtype=dtype, device=tensor.device)
+    if (std == 0).any():
+        raise ValueError(f"std evaluated to zero after conversion to {dtype}, leading to division by zero.")
+    if mean.ndim == 1:
+        mean = mean.view(-1, 1, 1)
+    if std.ndim == 1:
+        std = std.view(-1, 1, 1)
+    tensor.sub_(mean).div_(std)
+    return tensor
+
+
+def erase(img: Tensor, i: int, j: int, h: int, w: int, v: Tensor, inplace: bool = False) -> Tensor:
+    _assert_image_tensor(img)
+
+    if not inplace:
+        img = img.clone()
+
+    img[..., i : i + h, j : j + w] = v
+    return img

Original file line number	Diff line number	Diff line change
`@@ -33,7 +33,7 @@ def build_workflows(prefix="", filter_branch=None, upload=False, indentation=6,`
`33`	`33`	`python_versions = PYTHON_VERSIONS`
`34`	`34`	`cu_versions_dict = {`
`35`	`35`	`"linux": ["cpu", "cu102", "cu111", "cu113", "cu115", "rocm4.3.1", "rocm4.5.2"],`
`36`		`- "win": ["cpu", "cu111", "cu113", "cu115"],`
	`36`	`+ "win": ["cpu", "cu113", "cu115"],`
`37`	`37`	`"macos": ["cpu"],`
`38`	`38`	`}`
`39`	`39`	`cu_versions = cu_versions_dict[os_type]`