Merge branch 'master' into more-gpu-test-fixes

.github/workflows/pytorch-version-tests.yml

@@ -17,7 +17,7 @@ jobs:
       matrix:
         python-version: [3.9, "3.10", "3.11"]
         pytorch-version:
-          [2.4.1, 2.3.1, 2.2.2, 2.0.1, 1.13.1, 1.12.1, 1.10.0]
+          [2.5.1, 2.4.1, 2.3.1, 2.2.2, 1.13.1, 1.12.1, 1.10.0]
         exclude:
           - pytorch-version: 1.10.0
             python-version: "3.10"
@@ -70,7 +70,7 @@ jobs:
         shell: bash -l {0}
         run: |
           conda install pytorch=${{ matrix.pytorch-version }} torchvision cpuonly python=${{ matrix.python-version }} -c pytorch
-          
+
           # We should install numpy<2.0 for pytorch<2.3
           numpy_one_pth_version=$(python -c "import torch; print(float('.'.join(torch.__version__.split('.')[:2])) < 2.3)")
           if [ "${numpy_one_pth_version}" == "True" ]; then

docker/docker.cfg

@@ -1,4 +1,4 @@
 [DEFAULT]
-build_docker_image_pytorch_version = 2.5.1-cuda12.4-cudnn9
+build_docker_image_pytorch_version = 2.6.0-cuda12.4-cudnn9
 build_docker_image_hvd_version = v0.28.1
 build_docker_image_msdp_version = v0.14.0

docs/source/conf.py

@@ -359,6 +359,8 @@ def run(self):
     "https://github.com/pytorch/ignite/tree/master/examples/cifar10#check-resume-training",
     "https://github.com/pytorch/ignite/tree/master/examples/mnist#training-save--resume",
     "https://machinelearningmastery.com/gentle-introduction-backpropagation-time/",
+    "https://github.com/pytorch/pytorch/issues/7844#issuecomment-503713840",
+    "https://github.com/pytorch/pytorch/issues/23430#issuecomment-562350407",
 ]
 
 

examples/notebooks/FashionMNIST.ipynb

@@ -131,7 +131,7 @@
       "source": [
         "The code below first sets up transform using `torhvision transfroms` for converting images to pytorch tensors and normalizing the images.\n",
         "\n",
-        "Next, We use `torchvision datasets` for dowloading the fashion mnist dataset and applying transforms which we defined above.\n",
+        "Next, We use `torchvision datasets` for downloading the fashion mnist dataset and applying transforms which we defined above.\n",
         "\n",
         "* `trainset` contains the training data.\n",
         "* `validationset` contains the validation data\n",
@@ -342,7 +342,7 @@
       "source": [
         "### EarlyStopping - Tracking Validation Loss\n",
         "\n",
-        "Now we will setup a `EarlyStopping` handler for this training process. EarlyStopping requires a score_function that allows the user to define whatever criteria to stop trainig. In this case, if the loss of the validation set does not decrease in 10 epochs, the training process will stop early. Since the `EarlyStopping` handler relies on the validation loss, it's attached to the `val_evaluator`. "
+        "Now we will setup a `EarlyStopping` handler for this training process. EarlyStopping requires a score_function that allows the user to define whatever criteria to stop training. In this case, if the loss of the validation set does not decrease in 10 epochs, the training process will stop early. Since the `EarlyStopping` handler relies on the validation loss, it's attached to the `val_evaluator`. "
       ]
     },
     {
@@ -605,7 +605,7 @@
         "classes = ['T-shirt/top','Trouser','Pullover','Dress','Coat','Sandal','Shirt','Sneaker','Bag','Ankle Boot']\n",
         "# creating iterator for iterating the dataset\n",
         "dataiter = iter(val_loader)\n",
-        "images, labels = dataiter.next()\n",
+        "images, labels = next(dataiter)\n",
         "images_arr = []\n",
         "labels_arr = []\n",
         "pred_arr = []\n",
@@ -621,7 +621,7 @@
         "# plotting the results\n",
         "fig = plt.figure(figsize=(25,4))\n",
         "for i in range(10):\n",
-        "    ax = fig.add_subplot(2, 20/2, i+1, xticks=[], yticks=[])\n",
+        "    ax = fig.add_subplot(2, 20//2, i+1, xticks=[], yticks=[])\n",
         "    ax.imshow(images_arr[i].resize_(1, 28, 28).numpy().squeeze())\n",
         "    ax.set_title(\"{} ({})\".format(classes[pred_arr[i]], classes[labels_arr[i]]),\n",
         "                 color=(\"green\" if pred_arr[i]==labels_arr[i] else \"red\"))"
@@ -633,7 +633,7 @@
         "id": "cWXOEpQ4abyv"
       },
       "source": [
-        "### Refrences \n",
+        "### References  \n",
         "* [Pytorch Ignite Text CNN example notebook](https://github.com/pytorch/ignite/blob/master/examples/notebooks/TextCNN.ipynb)\n",
         "* [Pytorch Ignite MNIST example](https://github.com/pytorch/ignite/blob/master/examples/mnist/mnist.py)"
       ]

examples/notebooks/TextCNN.ipynb

@@ -426,7 +426,7 @@
     "batch_size = 8  # A batch size of 8\n",
     "\n",
     "def create_iterators(batch_size=8):\n",
-    "    \"\"\"Heler function to create the iterators\"\"\"\n",
+    "    \"\"\"Helper function to create the iterators\"\"\"\n",
     "    dataloaders = []\n",
     "    for split in [train_list, validation_list, test_list]:\n",
     "        dataloader = DataLoader(\n",
@@ -695,7 +695,7 @@
     "Similar to the training process function, we set up a function to evaluate a single batch. Here is what the eval_function does:\n",
     "\n",
     "* Sets model in eval mode.\n",
-    "* With torch.no_grad(), no gradients are calculated for any succeding steps.\n",
+    "* With torch.no_grad(), no gradients are calculated for any succeeding steps.\n",
     "* Generates x and y from batch.\n",
     "* Performs a forward pass on the model to calculate y_pred based on model and x.\n",
     "* Returns y_pred and y.\n",
@@ -1002,4 +1002,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 4
-}
+}

ignite/metrics/clustering/calinski_harabasz_score.py

@@ -86,7 +86,7 @@ class CalinskiHarabaszScore(_ClusteringMetricBase):
 
         .. testoutput::
 
-            5.733935121807529
+            5.733936
 
     .. versionadded:: 0.5.2
     """

tests/ignite/metrics/test_confusion_matrix.py

@@ -261,7 +261,7 @@ def test_miou():
 
     res = iou_metric.compute().numpy()
 
-    assert res == true_res_
+    assert pytest.approx(res) == true_res_
 
     for ignore_index in range(3):
         cm = ConfusionMatrix(num_classes=3)
@@ -271,7 +271,7 @@ def test_miou():
         cm.update(output)
         res = iou_metric.compute().numpy()
         true_res_ = np.mean(true_res[:ignore_index] + true_res[ignore_index + 1 :])
-        assert res == true_res_, f"{ignore_index}: {res} vs {true_res_}"
+        assert pytest.approx(res) == true_res_, f"{ignore_index}: {res} vs {true_res_}"
 
 
 def test_cm_accuracy():