pytorch-ignite 0.6.0.dev20250924__tar.gz → 0.6.0.dev20251106__tar.gz

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/.github/workflows/code-style-checks.yml

@@ -40,7 +40,7 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: ["3.9", "3.13"]
+        python-version: ["3.10", "3.13"]
 
     steps:
       - uses: actions/checkout@v5

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/.github/workflows/gpu-hvd-tests.yml

@@ -24,7 +24,7 @@ jobs:
         pytorch-channel: [pytorch]
       fail-fast: false
     env:
-      DOCKER_IMAGE: "pytorch/conda-builder:cuda12.1"
+      DOCKER_IMAGE: "pytorch/almalinux-builder:cuda12.8"
       REPOSITORY: ${{ github.repository }}
       PR_NUMBER: ${{ github.event.pull_request.number }}
     runs-on: linux.8xlarge.nvidia.gpu
@@ -113,6 +113,10 @@ jobs:
           pip install -r requirements-dev.txt
           pip install -e .
 
+          # Upgrade pyOpenSSL to avoid issue:
+          # AttributeError: module 'lib' has no attribute 'X509_V_FLAG_NOTIFY_POLICY'. Did you mean: 'X509_V_FLAG_EXPLICIT_POLICY'?
+          pip install -U pyOpenSSL
+
           EOF
           )
 

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/.github/workflows/gpu-tests.yml

@@ -24,7 +24,7 @@ jobs:
         pytorch-channel: [pytorch, pytorch-nightly]
       fail-fast: false
     env:
-      DOCKER_IMAGE: "pytorch/almalinux-builder:cuda12.4"
+      DOCKER_IMAGE: "pytorch/almalinux-builder:cuda12.8"
       REPOSITORY: ${{ github.repository }}
       PR_NUMBER: ${{ github.event.pull_request.number }}
     runs-on: linux.g4dn.12xlarge.nvidia.gpu
@@ -113,6 +113,10 @@ jobs:
           pip install -r requirements-dev.txt
           pip install -e .
 
+          # Upgrade pyOpenSSL to avoid issue:
+          # AttributeError: module 'lib' has no attribute 'X509_V_FLAG_NOTIFY_POLICY'. Did you mean: 'X509_V_FLAG_EXPLICIT_POLICY'?
+          pip install -U pyOpenSSL
+
           EOF
           )
 

pytorch_ignite-0.6.0.dev20251106/.github/workflows/pytorch-version-tests.yml

@@ -0,0 +1,83 @@
+name: PyTorch version tests
+
+on:
+  # https://docs.github.com/en/free-pro-team@latest/actions/reference/workflow-syntax-for-github-actions#onschedule
+  schedule:
+    # Run at 00:00 UTC Every Day
+    - cron: "0 0 * * *"
+  workflow_dispatch:
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    timeout-minutes: 85
+    strategy:
+      max-parallel: 5
+      fail-fast: false
+      matrix:
+        python-version: ["3.10", "3.11", "3.12"]
+        pytorch-version: [2.7.1, 2.5.1, 2.4.1, 2.3.1, 2.2.2]
+        # exclude:
+        #   - pytorch-version: 2.2.2
+        #     python-version: "3.12"
+
+    steps:
+      - uses: actions/checkout@v5
+
+      - name: Get year & week number
+        id: get-date
+        run: echo "date=$(/bin/date "+%Y-%U")" >> $GITHUB_OUTPUT
+        shell: bash -l {0}
+
+      - uses: astral-sh/setup-uv@v6
+        with:
+          version: "latest"
+          python-version: ${{ matrix.python-version }}
+          activate-environment: true
+          enable-cache: true
+          cache-suffix: "${{ steps.get-date.outputs.date }}-pytorch-${{ runner.os }}-${{ matrix.python-version }}-${{ matrix.pytorch-version }}"
+          cache-dependency-glob: |
+            **/requirements-dev.txt
+            **/pyproject.toml
+
+      - name: Install dependencies
+        shell: bash -l {0}
+        run: |
+          uv pip install "torch==${{ matrix.pytorch-version }}" torchvision --index-url https://download.pytorch.org/whl/cpu
+
+          # 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
+            uv pip install -U "numpy<2.0"
+          fi
+
+          uv pip install -r requirements-dev.txt
+          uv pip install .
+
+      - name: Download MNIST
+        uses: pytorch-ignite/download-mnist-github-action@master
+        with:
+          target_dir: /tmp
+
+      - name: Run Tests
+        uses: nick-fields/retry@v3.0.2
+        with:
+          max_attempts: 5
+          timeout_minutes: 15
+          shell: bash
+          command: bash -l tests/run_cpu_tests.sh "not test_time_profilers"
+          new_command_on_retry: USE_LAST_FAILED=1 bash -l tests/run_cpu_tests.sh "not test_time_profilers"
+
+  # create-issue:
+  #   runs-on: ubuntu-latest
+  #   # https://docs.github.com/en/actions/reference/context-and-expression-syntax-for-github-actions#needs-context
+  #   needs: build
+  #   if: always() && needs.build.result == 'failure'
+  #   steps:
+  #     - uses: actions/checkout@v5
+  #     - uses: JasonEtco/create-an-issue@v2
+  #       name: Create issue if pytorch version tests failed
+  #       with:
+  #         filename: .github/failed_schedule_issue_template.md
+  #       env:
+  #         GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/.github/workflows/stable-release-anaconda.yml

@@ -14,13 +14,7 @@ jobs:
         uses: conda-incubator/setup-miniconda@v3
         with:
           miniconda-version: "latest"
-          python-version: "3.10"
-
-      - name: Install dependencies
-        shell: bash -l {0}
-        run: |
-          conda install -y pytorch torchvision cpuonly -c pytorch
-          pip install .
+          python-version: "3.13"
 
       - name: Build and Publish Conda binaries
         shell: bash -l {0}

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/.github/workflows/typing-checks.yml

@@ -34,7 +34,7 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: ["3.9", "3.13"]
+        python-version: ["3.10", "3.13"]
         pytorch-channel: [pytorch]
 
     steps:

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/.github/workflows/unit-tests.yml

@@ -43,7 +43,7 @@ jobs:
       fail-fast: false
       matrix:
         os: [ubuntu-latest]
-        python-version: ["3.9", "3.10", "3.11", "3.12", "3.13"]
+        python-version: ["3.10", "3.11", "3.12", "3.13"]
         pytorch-channel: [pytorch, pytorch-nightly]
         include:
           # includes a single build on windows
@@ -76,9 +76,6 @@ jobs:
             **/requirements-dev.txt
             **/pyproject.toml
 
-      - name: Install dev tools
-        run: uv pip install wheel setuptools -Uqq
-
       - name: Install PyTorch
         if: ${{ matrix.pytorch-channel == 'pytorch' }}
         run: uv pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pytorch-ignite
-Version: 0.6.0.dev20250924
+Version: 0.6.0.dev20251106
 Summary: A lightweight library to help with training neural networks in PyTorch.
 Project-URL: Homepage, https://pytorch-ignite.ai
 Project-URL: Repository, https://github.com/pytorch/ignite
@@ -412,7 +412,7 @@ Few pointers to get you started:
 - [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/pytorch/ignite/blob/master/examples/notebooks/FastaiLRFinder_MNIST.ipynb) [Basic example of LR finder on
   MNIST](https://github.com/pytorch/ignite/blob/master/examples/notebooks/FastaiLRFinder_MNIST.ipynb)
 - [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/pytorch/ignite/blob/master/examples/notebooks/Cifar100_bench_amp.ipynb) [Benchmark mixed precision training on Cifar100:
-  torch.cuda.amp vs nvidia/apex](https://github.com/pytorch/ignite/blob/master/examples/notebooks/Cifar100_bench_amp.ipynb)
+  torch.amp vs nvidia/apex](https://github.com/pytorch/ignite/blob/master/examples/notebooks/Cifar100_bench_amp.ipynb)
 - [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/pytorch/ignite/blob/master/examples/notebooks/MNIST_on_TPU.ipynb) [MNIST training on a single
   TPU](https://github.com/pytorch/ignite/blob/master/examples/notebooks/MNIST_on_TPU.ipynb)
 - [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1E9zJrptnLJ_PKhmaP5Vhb6DTVRvyrKHx) [CIFAR10 Training on multiple TPUs](https://github.com/pytorch/ignite/tree/master/examples/cifar10)

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/README.md

@@ -397,7 +397,7 @@ Few pointers to get you started:
 - [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/pytorch/ignite/blob/master/examples/notebooks/FastaiLRFinder_MNIST.ipynb) [Basic example of LR finder on
   MNIST](https://github.com/pytorch/ignite/blob/master/examples/notebooks/FastaiLRFinder_MNIST.ipynb)
 - [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/pytorch/ignite/blob/master/examples/notebooks/Cifar100_bench_amp.ipynb) [Benchmark mixed precision training on Cifar100:
-  torch.cuda.amp vs nvidia/apex](https://github.com/pytorch/ignite/blob/master/examples/notebooks/Cifar100_bench_amp.ipynb)
+  torch.amp vs nvidia/apex](https://github.com/pytorch/ignite/blob/master/examples/notebooks/Cifar100_bench_amp.ipynb)
 - [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/pytorch/ignite/blob/master/examples/notebooks/MNIST_on_TPU.ipynb) [MNIST training on a single
   TPU](https://github.com/pytorch/ignite/blob/master/examples/notebooks/MNIST_on_TPU.ipynb)
 - [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1E9zJrptnLJ_PKhmaP5Vhb6DTVRvyrKHx) [CIFAR10 Training on multiple TPUs](https://github.com/pytorch/ignite/tree/master/examples/cifar10)

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/conda.recipe/build_and_upload.sh

@@ -20,6 +20,11 @@ fi
 
 set -xeu
 
+# Set version in meta.yaml
+version=$(sed -nE 's/__version__ = "(.*)"/\1/p' ignite/__init__.py)
+sed -i "s/__version__ = \"\(.*\)\"/__version__ = \"$version\"/g" conda.recipe/meta.yaml
+cat conda.recipe/meta.yaml | grep version
+
 conda install -y conda-build conda-verify anaconda-client conda-package-handling
 conda config --set anaconda_upload no
 

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/conda.recipe/meta.yaml

@@ -1,4 +1,4 @@
-{% set __version__ = "0.6.0.dev20250924" %}
+{% set __version__ = "0.6.0.dev20251106" %}
 
 package:
   name: ignite

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/docker/docker.cfg

@@ -1,4 +1,4 @@
 [DEFAULT]
-build_docker_image_pytorch_version = 2.7.1-cuda12.6-cudnn9
+build_docker_image_pytorch_version = 2.9.0-cuda12.8-cudnn9
 build_docker_image_hvd_version = v0.28.1
 build_docker_image_msdp_version = v0.14.0

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/docs/source/conf.py

@@ -354,7 +354,7 @@ nitpick_ignore = [
     ("py:class", "torch.optim.optimizer.Optimizer"),
     ("py:class", "torch.utils.data.dataset.Dataset"),
     ("py:class", "torch.utils.data.sampler.BatchSampler"),
-    ("py:class", "torch.cuda.amp.grad_scaler.GradScaler"),
+    ("py:class", "torch.amp.grad_scaler.GradScaler"),
     ("py:class", "torch.optim.lr_scheduler._LRScheduler"),
     ("py:class", "torch.optim.lr_scheduler.LRScheduler"),
     ("py:class", "torch.utils.data.dataloader.DataLoader"),

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/docs/source/handlers.rst

@@ -11,6 +11,7 @@ Complete list of generic handlers
     :toctree: generated
 
     checkpoint.Checkpoint
+    checkpoint.CheckpointEvents
     DiskSaver
     checkpoint.ModelCheckpoint
     ema_handler.EMAHandler

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/examples/cifar10/main.py

@@ -7,8 +7,7 @@ import torch
 import torch.nn as nn
 import torch.optim as optim
 import utils
-from torch.amp import autocast
-from torch.cuda.amp import GradScaler
+from torch.amp import autocast, GradScaler
 
 import ignite
 import ignite.distributed as idist

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/examples/cifar100_amp_benchmark/benchmark_torch_cuda_amp.py

@@ -1,7 +1,6 @@
 import fire
 import torch
-from torch.amp import autocast
-from torch.cuda.amp import GradScaler
+from torch.amp import autocast, GradScaler
 from torch.nn import CrossEntropyLoss
 from torch.optim import SGD
 from torchvision.models import wide_resnet50_2

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/examples/cifar10_qat/main.py

@@ -6,8 +6,7 @@ import torch
 import torch.nn as nn
 import torch.optim as optim
 import utils
-from torch.amp import autocast
-from torch.cuda.amp import GradScaler
+from torch.amp import autocast, GradScaler
 
 import ignite
 import ignite.distributed as idist

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/examples/notebooks/CycleGAN_with_torch_cuda_amp.ipynb

@@ -875,10 +875,10 @@
     "As suggested, we divide the objective by 2 while optimizing D, which slows down the rate at which D learns, relative to the rate of G. \n",
     "\n",
     "According to the paper:\n",
-    "- generator A is trained  minimize $\\text{mean}_{x \\in A}[(D_B(G(x)) − 1)^2]$ and cycle loss $\\text{mean}_{x \\in A}\\left[ |F(G(x)) - x|_1 \\right]$\n",
-    "- generator B is trained  minimize $\\text{mean}_{y \\in B}[(D_A(F(y)) − 1)^2]$ and cycle loss $\\text{mean}_{y \\in B}\\left[ |G(F(y)) - y|_1 \\right]$\n",
-    "- discriminators A is trained to minimize $\\text{mean}_{x \\in A}[(D_A(x) − 1)^2] + \\text{mean}_{y \\in B}[D_A(F(y))^2]$.\n",
-    "- discriminator B is trained to minimize $\\text{mean}_{y \\in B}[(D_B(y) − 1)^2] + \\text{mean}_{x \\in A}[D_B(G(x))^2]$."
+    "- generator A is trained  minimize $\\text{mean}_{x \\in A}[(D_B(G(x)) \u2212 1)^2]$ and cycle loss $\\text{mean}_{x \\in A}\\left[ |F(G(x)) - x|_1 \\right]$\n",
+    "- generator B is trained  minimize $\\text{mean}_{y \\in B}[(D_A(F(y)) \u2212 1)^2]$ and cycle loss $\\text{mean}_{y \\in B}\\left[ |G(F(y)) - y|_1 \\right]$\n",
+    "- discriminators A is trained to minimize $\\text{mean}_{x \\in A}[(D_A(x) \u2212 1)^2] + \\text{mean}_{y \\in B}[D_A(F(y))^2]$.\n",
+    "- discriminator B is trained to minimize $\\text{mean}_{y \\in B}[(D_B(y) \u2212 1)^2] + \\text{mean}_{x \\in A}[D_B(G(x))^2]$."
    ]
   },
   {
@@ -887,7 +887,7 @@
     "id": "JE8dLeEfIl_Z"
    },
    "source": [
-    "We will use [`torch.amp.autocast`](https://pytorch.org/docs/master/amp.html#torch.amp.autocast) and [`torch.cuda.amp.GradScaler`](https://pytorch.org/docs/master/amp.html#torch.cuda.amp.GradScaler) to perform automatic mixed precision training. Our code follows a [typical mixed precision training example](https://pytorch.org/docs/master/notes/amp_examples.html#typical-mixed-precision-training)."
+    "We will use [`torch.amp.autocast`](https://pytorch.org/docs/master/amp.html#torch.amp.autocast) and [`torch.amp.GradScaler`](https://pytorch.org/docs/master/amp.html#torch.amp.GradScaler) to perform automatic mixed precision training. Our code follows a [typical mixed precision training example](https://pytorch.org/docs/master/notes/amp_examples.html#typical-mixed-precision-training)."
    ]
   },
   {
@@ -896,8 +896,7 @@
     "id": "vrJls4p-FRcA"
    },
    "source": [
-    "from torch.cuda.amp import GradScaler\n",
-    "from torch.amp import autocast\n",
+    "from torch.amp import autocast, GradScaler\n",
     "\n",
     "from ignite.utils import convert_tensor\n",
     "import torch.nn.functional as F\n",
@@ -924,7 +923,7 @@
     "\n",
     "\n",
     "def compute_loss_discriminator(decision_real, decision_fake):\n",
-    "    # loss = mean (D_b(y) − 1)^2 + mean D_b(G(x))^2    \n",
+    "    # loss = mean (D_b(y) \u2212 1)^2 + mean D_b(G(x))^2    \n",
     "    loss = F.mse_loss(decision_fake, torch.zeros_like(decision_fake))\n",
     "    loss += F.mse_loss(decision_real, torch.ones_like(decision_real))\n",
     "    return loss\n",
@@ -954,10 +953,10 @@
     "        decision_fake_b = discriminator_B(fake_b)\n",
     "\n",
     "        # Compute loss for generators and update generators\n",
-    "        # loss_a2b = GAN loss: mean (D_b(G(x)) − 1)^2 + Forward cycle loss: || F(G(x)) - x ||_1    \n",
+    "        # loss_a2b = GAN loss: mean (D_b(G(x)) \u2212 1)^2 + Forward cycle loss: || F(G(x)) - x ||_1    \n",
     "        loss_a2b = compute_loss_generator(decision_fake_b, real_a, rec_a, lambda_value)    \n",
     "\n",
-    "        # loss_b2a = GAN loss: mean (D_a(F(x)) − 1)^2 + Backward cycle loss: || G(F(y)) - y ||_1\n",
+    "        # loss_b2a = GAN loss: mean (D_a(F(x)) \u2212 1)^2 + Backward cycle loss: || G(F(y)) - y ||_1\n",
     "        loss_b2a = compute_loss_generator(decision_fake_a, real_b, rec_b, lambda_value)\n",
     "\n",
     "        # total generators loss:\n",
@@ -977,10 +976,10 @@
     "        decision_real_a, decision_fake_a = discriminator_forward_pass(discriminator_A, real_a, fake_a.detach(), fake_a_buffer)    \n",
     "        decision_real_b, decision_fake_b = discriminator_forward_pass(discriminator_B, real_b, fake_b.detach(), fake_b_buffer)    \n",
     "        # Compute loss for discriminators and update discriminators\n",
-    "        # loss_a = mean (D_a(y) − 1)^2 + mean D_a(F(x))^2\n",
+    "        # loss_a = mean (D_a(y) \u2212 1)^2 + mean D_a(F(x))^2\n",
     "        loss_a = compute_loss_discriminator(decision_real_a, decision_fake_a)\n",
     "\n",
-    "        # loss_b = mean (D_b(y) − 1)^2 + mean D_b(G(x))^2\n",
+    "        # loss_b = mean (D_b(y) \u2212 1)^2 + mean D_b(G(x))^2\n",
     "        loss_b = compute_loss_discriminator(decision_real_b, decision_fake_b)\n",
     "      \n",
     "        # total discriminators loss:\n",

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/examples/references/classification/imagenet/main.py

@@ -6,8 +6,7 @@ import fire
 import torch
 
 try:
-    from torch.amp import autocast
-    from torch.cuda.amp import GradScaler
+    from torch.amp import autocast, GradScaler
 except ImportError:
     raise RuntimeError("Please, use recent PyTorch version, e.g. >=1.12.0")
 

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/examples/references/segmentation/pascal_voc2012/main.py

@@ -6,8 +6,7 @@ import fire
 import torch
 
 try:
-    from torch.amp import autocast
-    from torch.cuda.amp import GradScaler
+    from torch.amp import autocast, GradScaler
 except ImportError:
     raise RuntimeError("Please, use recent PyTorch version, e.g. >=1.12.0")
 

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/examples/transformers/main.py

@@ -7,8 +7,7 @@ import torch
 import torch.nn as nn
 import torch.optim as optim
 import utils
-from torch.amp import autocast
-from torch.cuda.amp import GradScaler
+from torch.amp import autocast, GradScaler
 
 import ignite
 import ignite.distributed as idist

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/ignite/__init__.py

@@ -6,4 +6,4 @@ import ignite.handlers
 import ignite.metrics
 import ignite.utils
 
-__version__ = "0.6.0.dev20250924"
+__version__ = "0.6.0.dev20251106"

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/ignite/distributed/comp_models/native.py

@@ -178,7 +178,7 @@ if has_native_dist_support:
             c: Counter = Counter(hostnames)
             sizes = torch.tensor([0] + list(c.values()))
             cumsum_sizes = torch.cumsum(sizes, dim=0)
-            node_rank = (rank // cumsum_sizes[1:]).clamp(0, 1).sum().item()
+            node_rank = cast(int, (rank // cumsum_sizes[1:]).clamp(0, 1).sum().item())
             local_rank = rank - cumsum_sizes[node_rank].item()
             return int(local_rank), node_rank
 

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/ignite/engine/__init__.py

@@ -133,11 +133,11 @@ def supervised_training_step_amp(
     prepare_batch: Callable = _prepare_batch,
     model_transform: Callable[[Any], Any] = lambda output: output,
     output_transform: Callable[[Any, Any, Any, torch.Tensor], Any] = lambda x, y, y_pred, loss: loss.item(),
-    scaler: Optional["torch.cuda.amp.GradScaler"] = None,
+    scaler: Optional["torch.amp.GradScaler"] = None,
     gradient_accumulation_steps: int = 1,
     model_fn: Callable[[torch.nn.Module, Any], Any] = lambda model, x: model(x),
 ) -> Callable:
-    """Factory function for supervised training using ``torch.cuda.amp``.
+    """Factory function for supervised training using ``torch.amp``.
 
     Args:
         model: the model to train.
@@ -170,7 +170,7 @@ def supervised_training_step_amp(
             model = ...
             optimizer = ...
             loss_fn = ...
-            scaler = torch.cuda.amp.GradScaler(2**10)
+            scaler = torch.amp.GradScaler('cuda', 2**10)
 
             update_fn = supervised_training_step_amp(model, optimizer, loss_fn, 'cuda', scaler=scaler)
             trainer = Engine(update_fn)
@@ -393,8 +393,8 @@ def supervised_training_step_tpu(
 
 
 def _check_arg(
-    on_tpu: bool, on_mps: bool, amp_mode: Optional[str], scaler: Optional[Union[bool, "torch.cuda.amp.GradScaler"]]
-) -> Tuple[Optional[str], Optional["torch.cuda.amp.GradScaler"]]:
+    on_tpu: bool, on_mps: bool, amp_mode: Optional[str], scaler: Optional[Union[bool, "torch.amp.GradScaler"]]
+) -> Tuple[Optional[str], Optional["torch.amp.GradScaler"]]:
     """Checking tpu, mps, amp and GradScaler instance combinations."""
     if on_mps and amp_mode:
         raise ValueError("amp_mode cannot be used with mps device. Consider using amp_mode=None or device='cuda'.")
@@ -410,9 +410,9 @@ def _check_arg(
             raise ValueError(f"scaler argument is {scaler}, but amp_mode is {amp_mode}. Consider using amp_mode='amp'.")
         elif amp_mode == "amp" and isinstance(scaler, bool):
             try:
-                from torch.cuda.amp import GradScaler
+                from torch.amp import GradScaler
             except ImportError:
-                raise ImportError("Please install torch>=1.6.0 to use scaler argument.")
+                raise ImportError("Please install torch>=2.3.1 to use scaler argument.")
             scaler = GradScaler(enabled=True)
 
     if on_tpu:
@@ -434,7 +434,7 @@ def create_supervised_trainer(
     output_transform: Callable[[Any, Any, Any, torch.Tensor], Any] = lambda x, y, y_pred, loss: loss.item(),
     deterministic: bool = False,
     amp_mode: Optional[str] = None,
-    scaler: Union[bool, "torch.cuda.amp.GradScaler"] = False,
+    scaler: Union[bool, "torch.amp.GradScaler"] = False,
     gradient_accumulation_steps: int = 1,
     model_fn: Callable[[torch.nn.Module, Any], Any] = lambda model, x: model(x),
 ) -> Engine:
@@ -459,7 +459,7 @@ def create_supervised_trainer(
             :class:`~ignite.engine.deterministic.DeterministicEngine`, otherwise :class:`~ignite.engine.engine.Engine`
             (default: False).
         amp_mode: can be ``amp`` or ``apex``, model and optimizer will be casted to float16 using
-            `torch.cuda.amp <https://pytorch.org/docs/stable/amp.html>`_ for ``amp`` and
+            `torch.amp <https://pytorch.org/docs/stable/amp.html>`_ for ``amp`` and
             using `apex <https://nvidia.github.io/apex>`_ for ``apex``. (default: None)
         scaler: GradScaler instance for gradient scaling if `torch>=1.6.0`
             and ``amp_mode`` is ``amp``. If ``amp_mode`` is ``apex``, this argument will be ignored.

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/ignite/handlers/__init__.py

@@ -6,6 +6,7 @@ from ignite.handlers.checkpoint import Checkpoint, DiskSaver, ModelCheckpoint
 from ignite.handlers.clearml_logger import ClearMLLogger
 from ignite.handlers.early_stopping import EarlyStopping
 from ignite.handlers.ema_handler import EMAHandler
+from ignite.handlers.fbresearch_logger import FBResearchLogger
 from ignite.handlers.lr_finder import FastaiLRFinder
 from ignite.handlers.mlflow_logger import MLflowLogger
 from ignite.handlers.neptune_logger import NeptuneLogger
@@ -64,6 +65,7 @@ __all__ = [
     "CyclicalScheduler",
     "create_lr_scheduler_with_warmup",
     "FastaiLRFinder",
+    "FBResearchLogger",
     "EMAHandler",
     "BasicTimeProfiler",
     "HandlersTimeProfiler",

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/ignite/handlers/checkpoint.py

@@ -21,10 +21,21 @@ else:
 
 import ignite.distributed as idist
 from ignite.base import Serializable
-from ignite.engine import Engine, Events
+from ignite.engine import Engine, Events, EventEnum
 from ignite.utils import _tree_apply2, _tree_map
 
-__all__ = ["Checkpoint", "DiskSaver", "ModelCheckpoint", "BaseSaveHandler"]
+__all__ = ["Checkpoint", "DiskSaver", "ModelCheckpoint", "BaseSaveHandler", "CheckpointEvents"]
+
+
+class CheckpointEvents(EventEnum):
+    """Events fired by :class:`~ignite.handlers.checkpoint.Checkpoint`
+
+    - SAVED_CHECKPOINT : triggered when checkpoint handler has saved objects
+
+    .. versionadded:: 0.5.3
+    """
+
+    SAVED_CHECKPOINT = "saved_checkpoint"
 
 
 class BaseSaveHandler(metaclass=ABCMeta):
@@ -264,6 +275,29 @@ class Checkpoint(Serializable):
                 to_save, save_handler=DiskSaver('/tmp/models', create_dir=True, **kwargs), n_saved=2
             )
 
+        Respond to checkpoint events:
+
+        .. code-block:: python
+
+            from ignite.handlers import Checkpoint
+            from ignite.engine import Engine, Events
+
+            checkpoint_handler = Checkpoint(
+                {'model': model, 'optimizer': optimizer},
+                save_dir,
+                n_saved=2
+            )
+
+            @trainer.on(Checkpoint.SAVED_CHECKPOINT)
+            def on_checkpoint_saved(engine):
+                print(f"Checkpoint saved at epoch {engine.state.epoch}")
+
+            trainer.add_event_handler(Events.EPOCH_COMPLETED, checkpoint_handler)
+
+    Attributes:
+        SAVED_CHECKPOINT: Alias of ``SAVED_CHECKPOINT`` from
+            :class:`~ignite.handlers.checkpoint.CheckpointEvents`.
+
     .. versionchanged:: 0.4.3
 
         - Checkpoint can save model with same filename.
@@ -274,8 +308,13 @@ class Checkpoint(Serializable):
         - `score_name` can be used to define `score_function` automatically without providing `score_function`.
         - `save_handler` automatically saves to disk if path to directory is provided.
         - `save_on_rank` saves objects on this rank in a distributed configuration.
+
+    .. versionchanged:: 0.5.3
+
+        - Added ``SAVED_CHECKPOINT`` class attribute.
     """
 
+    SAVED_CHECKPOINT = CheckpointEvents.SAVED_CHECKPOINT
     Item = NamedTuple("Item", [("priority", int), ("filename", str)])
     _state_dict_all_req_keys = ("_saved",)
 
@@ -400,6 +439,8 @@ class Checkpoint(Serializable):
             return new > self._saved[0].priority
 
     def __call__(self, engine: Engine) -> None:
+        if not engine.has_registered_events(CheckpointEvents.SAVED_CHECKPOINT):
+            engine.register_events(*CheckpointEvents)
         global_step = None
         if self.global_step_transform is not None:
             global_step = self.global_step_transform(engine, engine.last_event_name)
@@ -460,11 +501,11 @@ class Checkpoint(Serializable):
             if self.include_self:
                 # Now that we've updated _saved, we can add our own state_dict.
                 checkpoint["checkpointer"] = self.state_dict()
-
             try:
                 self.save_handler(checkpoint, filename, metadata)
             except TypeError:
                 self.save_handler(checkpoint, filename)
+            engine.fire_event(CheckpointEvents.SAVED_CHECKPOINT)
 
     def _setup_checkpoint(self) -> Dict[str, Any]:
         if self.to_save is not None:

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/ignite/handlers/fbresearch_logger.py

@@ -7,7 +7,7 @@ import torch
 
 from ignite import utils
 from ignite.engine import Engine, Events
-from ignite.handlers import Timer
+from ignite.handlers.timing import Timer
 
 MB = 1024.0 * 1024.0
 
@@ -154,7 +154,7 @@ class FBResearchLogger:
         if torch.cuda.is_available():
             cuda_max_mem = f"GPU Max Mem: {torch.cuda.max_memory_allocated() / MB:.0f} MB"
 
-        current_iter = engine.state.iteration % (engine.state.epoch_length + 1)
+        current_iter = ((engine.state.iteration - 1) % engine.state.epoch_length) + 1
         iter_avg_time = self.iter_timer.value()
 
         eta_seconds = iter_avg_time * (engine.state.epoch_length - current_iter)

{pytorch_ignite-0.6.0.dev20250924 → pytorch_ignite-0.6.0.dev20251106}/tests/ignite/distributed/test_auto.py

@@ -311,7 +311,7 @@ def test_dist_proxy_sampler():
         DistributedProxySampler(None)
 
     with pytest.raises(TypeError, match=r"Argument sampler should have length"):
-        DistributedProxySampler(Sampler([1]))
+        DistributedProxySampler(Sampler())
 
     with pytest.raises(TypeError, match=r"Argument sampler must not be a distributed sampler already"):
         DistributedProxySampler(DistributedSampler(sampler, num_replicas=num_replicas, rank=0))