Revert "Fix CUDA kernel index data type in vision/fair/pytorch3d/pytorch3d/csrc/compositing/alpha_composite.cu +10"

This reverts commit 3987612062.
Fix CUDA kernel index data type in vision/fair/pytorch3d/pytorch3d/csrc/compositing/alpha_composite.cu +10
2026-02-26 16:26:00 +08:00 · 2025-03-27 05:28:03 -07:00 · 2025-03-19 13:21:43 -07:00 · 2025-03-11 12:25:59 -07:00 · 2025-03-08 13:03:17 -08:00 · 2025-03-04 17:49:30 -08:00
209 changed files with 1401 additions and 959 deletions
--- a/.circleci/config.yml
+++ b/.circleci/config.yml
@@ -162,90 +162,6 @@ workflows:
    jobs:
      # - main:
      #     context: DOCKERHUB_TOKEN
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda113
-          context: DOCKERHUB_TOKEN
-          cu_version: cu113
-          name: linux_conda_py38_cu113_pyt1120
-          python_version: '3.8'
-          pytorch_version: 1.12.0
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda116
-          context: DOCKERHUB_TOKEN
-          cu_version: cu116
-          name: linux_conda_py38_cu116_pyt1120
-          python_version: '3.8'
-          pytorch_version: 1.12.0
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda113
-          context: DOCKERHUB_TOKEN
-          cu_version: cu113
-          name: linux_conda_py38_cu113_pyt1121
-          python_version: '3.8'
-          pytorch_version: 1.12.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda116
-          context: DOCKERHUB_TOKEN
-          cu_version: cu116
-          name: linux_conda_py38_cu116_pyt1121
-          python_version: '3.8'
-          pytorch_version: 1.12.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda116
-          context: DOCKERHUB_TOKEN
-          cu_version: cu116
-          name: linux_conda_py38_cu116_pyt1130
-          python_version: '3.8'
-          pytorch_version: 1.13.0
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda117
-          context: DOCKERHUB_TOKEN
-          cu_version: cu117
-          name: linux_conda_py38_cu117_pyt1130
-          python_version: '3.8'
-          pytorch_version: 1.13.0
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda116
-          context: DOCKERHUB_TOKEN
-          cu_version: cu116
-          name: linux_conda_py38_cu116_pyt1131
-          python_version: '3.8'
-          pytorch_version: 1.13.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda117
-          context: DOCKERHUB_TOKEN
-          cu_version: cu117
-          name: linux_conda_py38_cu117_pyt1131
-          python_version: '3.8'
-          pytorch_version: 1.13.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda117
-          context: DOCKERHUB_TOKEN
-          cu_version: cu117
-          name: linux_conda_py38_cu117_pyt200
-          python_version: '3.8'
-          pytorch_version: 2.0.0
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda118
-          context: DOCKERHUB_TOKEN
-          cu_version: cu118
-          name: linux_conda_py38_cu118_pyt200
-          python_version: '3.8'
-          pytorch_version: 2.0.0
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda117
-          context: DOCKERHUB_TOKEN
-          cu_version: cu117
-          name: linux_conda_py38_cu117_pyt201
-          python_version: '3.8'
-          pytorch_version: 2.0.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda118
-          context: DOCKERHUB_TOKEN
-          cu_version: cu118
-          name: linux_conda_py38_cu118_pyt201
-          python_version: '3.8'
-          pytorch_version: 2.0.1
      - binary_linux_conda:
          conda_docker_image: pytorch/conda-builder:cuda118
          context: DOCKERHUB_TOKEN
@@ -331,89 +247,33 @@ workflows:
          python_version: '3.8'
          pytorch_version: 2.3.1
      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda113
+          conda_docker_image: pytorch/conda-builder:cuda118
          context: DOCKERHUB_TOKEN
-          cu_version: cu113
-          name: linux_conda_py39_cu113_pyt1120
-          python_version: '3.9'
-          pytorch_version: 1.12.0
+          cu_version: cu118
+          name: linux_conda_py38_cu118_pyt240
+          python_version: '3.8'
+          pytorch_version: 2.4.0
      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda116
+          conda_docker_image: pytorch/conda-builder:cuda121
          context: DOCKERHUB_TOKEN
-          cu_version: cu116
-          name: linux_conda_py39_cu116_pyt1120
-          python_version: '3.9'
-          pytorch_version: 1.12.0
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda113
-          context: DOCKERHUB_TOKEN
-          cu_version: cu113
-          name: linux_conda_py39_cu113_pyt1121
-          python_version: '3.9'
-          pytorch_version: 1.12.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda116
-          context: DOCKERHUB_TOKEN
-          cu_version: cu116
-          name: linux_conda_py39_cu116_pyt1121
-          python_version: '3.9'
-          pytorch_version: 1.12.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda116
-          context: DOCKERHUB_TOKEN
-          cu_version: cu116
-          name: linux_conda_py39_cu116_pyt1130
-          python_version: '3.9'
-          pytorch_version: 1.13.0
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda117
-          context: DOCKERHUB_TOKEN
-          cu_version: cu117
-          name: linux_conda_py39_cu117_pyt1130
-          python_version: '3.9'
-          pytorch_version: 1.13.0
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda116
-          context: DOCKERHUB_TOKEN
-          cu_version: cu116
-          name: linux_conda_py39_cu116_pyt1131
-          python_version: '3.9'
-          pytorch_version: 1.13.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda117
-          context: DOCKERHUB_TOKEN
-          cu_version: cu117
-          name: linux_conda_py39_cu117_pyt1131
-          python_version: '3.9'
-          pytorch_version: 1.13.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda117
-          context: DOCKERHUB_TOKEN
-          cu_version: cu117
-          name: linux_conda_py39_cu117_pyt200
-          python_version: '3.9'
-          pytorch_version: 2.0.0
+          cu_version: cu121
+          name: linux_conda_py38_cu121_pyt240
+          python_version: '3.8'
+          pytorch_version: 2.4.0
      - binary_linux_conda:
          conda_docker_image: pytorch/conda-builder:cuda118
          context: DOCKERHUB_TOKEN
          cu_version: cu118
-          name: linux_conda_py39_cu118_pyt200
-          python_version: '3.9'
-          pytorch_version: 2.0.0
+          name: linux_conda_py38_cu118_pyt241
+          python_version: '3.8'
+          pytorch_version: 2.4.1
      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda117
+          conda_docker_image: pytorch/conda-builder:cuda121
          context: DOCKERHUB_TOKEN
-          cu_version: cu117
-          name: linux_conda_py39_cu117_pyt201
-          python_version: '3.9'
-          pytorch_version: 2.0.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda118
-          context: DOCKERHUB_TOKEN
-          cu_version: cu118
-          name: linux_conda_py39_cu118_pyt201
-          python_version: '3.9'
-          pytorch_version: 2.0.1
+          cu_version: cu121
+          name: linux_conda_py38_cu121_pyt241
+          python_version: '3.8'
+          pytorch_version: 2.4.1
      - binary_linux_conda:
          conda_docker_image: pytorch/conda-builder:cuda118
          context: DOCKERHUB_TOKEN
@@ -499,89 +359,33 @@ workflows:
          python_version: '3.9'
          pytorch_version: 2.3.1
      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda113
+          conda_docker_image: pytorch/conda-builder:cuda118
          context: DOCKERHUB_TOKEN
-          cu_version: cu113
-          name: linux_conda_py310_cu113_pyt1120
-          python_version: '3.10'
-          pytorch_version: 1.12.0
+          cu_version: cu118
+          name: linux_conda_py39_cu118_pyt240
+          python_version: '3.9'
+          pytorch_version: 2.4.0
      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda116
+          conda_docker_image: pytorch/conda-builder:cuda121
          context: DOCKERHUB_TOKEN
-          cu_version: cu116
-          name: linux_conda_py310_cu116_pyt1120
-          python_version: '3.10'
-          pytorch_version: 1.12.0
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda113
-          context: DOCKERHUB_TOKEN
-          cu_version: cu113
-          name: linux_conda_py310_cu113_pyt1121
-          python_version: '3.10'
-          pytorch_version: 1.12.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda116
-          context: DOCKERHUB_TOKEN
-          cu_version: cu116
-          name: linux_conda_py310_cu116_pyt1121
-          python_version: '3.10'
-          pytorch_version: 1.12.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda116
-          context: DOCKERHUB_TOKEN
-          cu_version: cu116
-          name: linux_conda_py310_cu116_pyt1130
-          python_version: '3.10'
-          pytorch_version: 1.13.0
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda117
-          context: DOCKERHUB_TOKEN
-          cu_version: cu117
-          name: linux_conda_py310_cu117_pyt1130
-          python_version: '3.10'
-          pytorch_version: 1.13.0
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda116
-          context: DOCKERHUB_TOKEN
-          cu_version: cu116
-          name: linux_conda_py310_cu116_pyt1131
-          python_version: '3.10'
-          pytorch_version: 1.13.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda117
-          context: DOCKERHUB_TOKEN
-          cu_version: cu117
-          name: linux_conda_py310_cu117_pyt1131
-          python_version: '3.10'
-          pytorch_version: 1.13.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda117
-          context: DOCKERHUB_TOKEN
-          cu_version: cu117
-          name: linux_conda_py310_cu117_pyt200
-          python_version: '3.10'
-          pytorch_version: 2.0.0
+          cu_version: cu121
+          name: linux_conda_py39_cu121_pyt240
+          python_version: '3.9'
+          pytorch_version: 2.4.0
      - binary_linux_conda:
          conda_docker_image: pytorch/conda-builder:cuda118
          context: DOCKERHUB_TOKEN
          cu_version: cu118
-          name: linux_conda_py310_cu118_pyt200
-          python_version: '3.10'
-          pytorch_version: 2.0.0
+          name: linux_conda_py39_cu118_pyt241
+          python_version: '3.9'
+          pytorch_version: 2.4.1
      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda117
+          conda_docker_image: pytorch/conda-builder:cuda121
          context: DOCKERHUB_TOKEN
-          cu_version: cu117
-          name: linux_conda_py310_cu117_pyt201
-          python_version: '3.10'
-          pytorch_version: 2.0.1
-      - binary_linux_conda:
-          conda_docker_image: pytorch/conda-builder:cuda118
-          context: DOCKERHUB_TOKEN
-          cu_version: cu118
-          name: linux_conda_py310_cu118_pyt201
-          python_version: '3.10'
-          pytorch_version: 2.0.1
+          cu_version: cu121
+          name: linux_conda_py39_cu121_pyt241
+          python_version: '3.9'
+          pytorch_version: 2.4.1
      - binary_linux_conda:
          conda_docker_image: pytorch/conda-builder:cuda118
          context: DOCKERHUB_TOKEN
@@ -666,6 +470,34 @@ workflows:
          name: linux_conda_py310_cu121_pyt231
          python_version: '3.10'
          pytorch_version: 2.3.1
+      - binary_linux_conda:
+          conda_docker_image: pytorch/conda-builder:cuda118
+          context: DOCKERHUB_TOKEN
+          cu_version: cu118
+          name: linux_conda_py310_cu118_pyt240
+          python_version: '3.10'
+          pytorch_version: 2.4.0
+      - binary_linux_conda:
+          conda_docker_image: pytorch/conda-builder:cuda121
+          context: DOCKERHUB_TOKEN
+          cu_version: cu121
+          name: linux_conda_py310_cu121_pyt240
+          python_version: '3.10'
+          pytorch_version: 2.4.0
+      - binary_linux_conda:
+          conda_docker_image: pytorch/conda-builder:cuda118
+          context: DOCKERHUB_TOKEN
+          cu_version: cu118
+          name: linux_conda_py310_cu118_pyt241
+          python_version: '3.10'
+          pytorch_version: 2.4.1
+      - binary_linux_conda:
+          conda_docker_image: pytorch/conda-builder:cuda121
+          context: DOCKERHUB_TOKEN
+          cu_version: cu121
+          name: linux_conda_py310_cu121_pyt241
+          python_version: '3.10'
+          pytorch_version: 2.4.1
      - binary_linux_conda:
          conda_docker_image: pytorch/conda-builder:cuda118
          context: DOCKERHUB_TOKEN
@@ -750,6 +582,34 @@ workflows:
          name: linux_conda_py311_cu121_pyt231
          python_version: '3.11'
          pytorch_version: 2.3.1
+      - binary_linux_conda:
+          conda_docker_image: pytorch/conda-builder:cuda118
+          context: DOCKERHUB_TOKEN
+          cu_version: cu118
+          name: linux_conda_py311_cu118_pyt240
+          python_version: '3.11'
+          pytorch_version: 2.4.0
+      - binary_linux_conda:
+          conda_docker_image: pytorch/conda-builder:cuda121
+          context: DOCKERHUB_TOKEN
+          cu_version: cu121
+          name: linux_conda_py311_cu121_pyt240
+          python_version: '3.11'
+          pytorch_version: 2.4.0
+      - binary_linux_conda:
+          conda_docker_image: pytorch/conda-builder:cuda118
+          context: DOCKERHUB_TOKEN
+          cu_version: cu118
+          name: linux_conda_py311_cu118_pyt241
+          python_version: '3.11'
+          pytorch_version: 2.4.1
+      - binary_linux_conda:
+          conda_docker_image: pytorch/conda-builder:cuda121
+          context: DOCKERHUB_TOKEN
+          cu_version: cu121
+          name: linux_conda_py311_cu121_pyt241
+          python_version: '3.11'
+          pytorch_version: 2.4.1
      - binary_linux_conda:
          conda_docker_image: pytorch/conda-builder:cuda118
          context: DOCKERHUB_TOKEN
@@ -792,6 +652,34 @@ workflows:
          name: linux_conda_py312_cu121_pyt231
          python_version: '3.12'
          pytorch_version: 2.3.1
+      - binary_linux_conda:
+          conda_docker_image: pytorch/conda-builder:cuda118
+          context: DOCKERHUB_TOKEN
+          cu_version: cu118
+          name: linux_conda_py312_cu118_pyt240
+          python_version: '3.12'
+          pytorch_version: 2.4.0
+      - binary_linux_conda:
+          conda_docker_image: pytorch/conda-builder:cuda121
+          context: DOCKERHUB_TOKEN
+          cu_version: cu121
+          name: linux_conda_py312_cu121_pyt240
+          python_version: '3.12'
+          pytorch_version: 2.4.0
+      - binary_linux_conda:
+          conda_docker_image: pytorch/conda-builder:cuda118
+          context: DOCKERHUB_TOKEN
+          cu_version: cu118
+          name: linux_conda_py312_cu118_pyt241
+          python_version: '3.12'
+          pytorch_version: 2.4.1
+      - binary_linux_conda:
+          conda_docker_image: pytorch/conda-builder:cuda121
+          context: DOCKERHUB_TOKEN
+          cu_version: cu121
+          name: linux_conda_py312_cu121_pyt241
+          python_version: '3.12'
+          pytorch_version: 2.4.1
      - binary_linux_conda_cuda:
          name: testrun_conda_cuda_py310_cu117_pyt201
          context: DOCKERHUB_TOKEN
--- a/.circleci/regenerate.py
+++ b/.circleci/regenerate.py
@@ -19,18 +19,14 @@ from packaging import version
 # The CUDA versions which have pytorch conda packages available for linux for each
 # version of pytorch.
 CONDA_CUDA_VERSIONS = {
-    "1.12.0": ["cu113", "cu116"],
-    "1.12.1": ["cu113", "cu116"],
-    "1.13.0": ["cu116", "cu117"],
-    "1.13.1": ["cu116", "cu117"],
-    "2.0.0": ["cu117", "cu118"],
-    "2.0.1": ["cu117", "cu118"],
    "2.1.0": ["cu118", "cu121"],
    "2.1.1": ["cu118", "cu121"],
    "2.1.2": ["cu118", "cu121"],
    "2.2.0": ["cu118", "cu121"],
    "2.2.2": ["cu118", "cu121"],
    "2.3.1": ["cu118", "cu121"],
+    "2.4.0": ["cu118", "cu121"],
+    "2.4.1": ["cu118", "cu121"],
 }


@@ -92,7 +88,6 @@ def workflow_pair(
    upload=False,
    filter_branch,
 ):
-
    w = []
    py = python_version.replace(".", "")
    pyt = pytorch_version.replace(".", "")
@@ -131,7 +126,6 @@ def generate_base_workflow(
    btype,
    filter_branch=None,
 ):
-
    d = {
        "name": base_workflow_name,
        "python_version": python_version,
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -0,0 +1,23 @@
+name: facebookresearch/pytorch3d/build_and_test
+on:
+  pull_request:
+    branches:
+      - main
+  push:
+    branches:
+      - main
+jobs:
+  binary_linux_conda_cuda:
+    runs-on: 4-core-ubuntu-gpu-t4
+    env:
+      PYTHON_VERSION: "3.12"
+      BUILD_VERSION: "${{ github.run_number }}"
+      PYTORCH_VERSION: "2.4.1"
+      CU_VERSION: "cu121"
+      JUST_TESTRUN: 1
+    steps:
+    - uses: actions/checkout@v4
+    - name: Build and run tests
+      run: |-
+        conda create --name env --yes --quiet conda-build
+        conda run --no-capture-output --name env python3 ./packaging/build_conda.py --use-conda-cuda
--- a/INSTALL.md
+++ b/INSTALL.md
@@ -8,11 +8,10 @@
 The core library is written in PyTorch. Several components have underlying implementation in CUDA for improved performance. A subset of these components have CPU implementations in C++/PyTorch. It is advised to use PyTorch3D with GPU support in order to use all the features.

 - Linux or macOS or Windows
- Python 3.8, 3.9 or 3.10
- PyTorch 1.12.0, 1.12.1, 1.13.0, 2.0.0, 2.0.1, 2.1.0, 2.1.1, 2.1.2, 2.2.0, 2.2.1, 2.2.2, 2.3.0 or 2.3.1.
+- Python
+- PyTorch 2.1.0, 2.1.1, 2.1.2, 2.2.0, 2.2.1, 2.2.2, 2.3.0, 2.3.1, 2.4.0 or 2.4.1.
 - torchvision that matches the PyTorch installation. You can install them together as explained at pytorch.org to make sure of this.
 - gcc & g++ ≥ 4.9
- [fvcore](https://github.com/facebookresearch/fvcore)
 - [ioPath](https://github.com/facebookresearch/iopath)
 - If CUDA is to be used, use a version which is supported by the corresponding pytorch version and at least version 9.2.
 - If CUDA older than 11.7 is to be used and you are building from source, the CUB library must be available. We recommend version 1.10.0.
@@ -22,7 +21,7 @@ The runtime dependencies can be installed by running:
 conda create -n pytorch3d python=3.9
 conda activate pytorch3d
 conda install pytorch=1.13.0 torchvision pytorch-cuda=11.6 -c pytorch -c nvidia
-conda install -c fvcore -c iopath -c conda-forge fvcore iopath
+conda install -c iopath iopath
 ```

 For the CUB build time dependency, which you only need if you have CUDA older than 11.7, if you are using conda, you can continue with
@@ -49,6 +48,7 @@ For developing on top of PyTorch3D or contributing, you will need to run the lin
 - tdqm
 - jupyter
 - imageio
+- fvcore
 - plotly
 - opencv-python

@@ -59,6 +59,7 @@ conda install jupyter
 pip install scikit-image matplotlib imageio plotly opencv-python

 # Tests/Linting
+conda install -c fvcore -c conda-forge fvcore
 pip install black usort flake8 flake8-bugbear flake8-comprehensions
 ```

@@ -97,7 +98,7 @@ version_str="".join([
    torch.version.cuda.replace(".",""),
    f"_pyt{pyt_version_str}"
 ])
-!pip install fvcore iopath
+!pip install iopath
 !pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html
 ```

--- a/dev/linter.sh
+++ b/dev/linter.sh
@@ -36,5 +36,5 @@ then

  echo "Running pyre..."
  echo "To restart/kill pyre server, run 'pyre restart' or 'pyre kill' in fbcode/"
-  ( cd ~/fbsource/fbcode; pyre -l vision/fair/pytorch3d/ )
+  ( cd ~/fbsource/fbcode; arc pyre check //vision/fair/pytorch3d/... )
 fi
--- a/dev/run_tutorials.sh
+++ b/dev/run_tutorials.sh
@@ -23,7 +23,7 @@ conda init bash
 source ~/.bashrc
 conda create -y -n myenv python=3.8 matplotlib ipython ipywidgets nbconvert
 conda activate myenv
-conda install -y -c fvcore -c iopath -c conda-forge fvcore iopath
+conda install -y -c iopath iopath
 conda install -y -c pytorch pytorch=1.6.0 cudatoolkit=10.1 torchvision
 conda install -y -c pytorch3d-nightly pytorch3d
 pip install plotly scikit-image
--- a/docs/examples/pulsar_basic.py
+++ b/docs/examples/pulsar_basic.py
@@ -10,6 +10,7 @@ This example demonstrates the most trivial, direct interface of the pulsar
 sphere renderer. It renders and saves an image with 10 random spheres.
 Output: basic.png.
 """
+
 import logging
 import math
 from os import path
--- a/docs/examples/pulsar_basic_unified.py
+++ b/docs/examples/pulsar_basic_unified.py
@@ -11,6 +11,7 @@ interface for sphere renderering. It renders and saves an image with
 10 random spheres.
 Output: basic-pt3d.png.
 """
+
 import logging
 from os import path

--- a/docs/examples/pulsar_cam.py
+++ b/docs/examples/pulsar_cam.py
@@ -14,6 +14,7 @@ distorted. Gradient-based optimization is used to converge towards the
 original camera parameters.
 Output: cam.gif.
 """
+
 import logging
 import math
 from os import path
--- a/docs/examples/pulsar_cam_unified.py
+++ b/docs/examples/pulsar_cam_unified.py
@@ -14,6 +14,7 @@ distorted. Gradient-based optimization is used to converge towards the
 original camera parameters.
 Output: cam-pt3d.gif
 """
+
 import logging
 from os import path

--- a/docs/examples/pulsar_multiview.py
+++ b/docs/examples/pulsar_multiview.py
@@ -18,6 +18,7 @@ This example is not available yet through the 'unified' interface,
 because opacity support has not landed in PyTorch3D for general data
 structures yet.
 """
+
 import logging
 import math
 from os import path
--- a/docs/examples/pulsar_optimization.py
+++ b/docs/examples/pulsar_optimization.py
@@ -13,6 +13,7 @@ The scene is initialized with random spheres. Gradient-based
 optimization is used to converge towards a faithful
 scene representation.
 """
+
 import logging
 import math

--- a/docs/examples/pulsar_optimization_unified.py
+++ b/docs/examples/pulsar_optimization_unified.py
@@ -13,6 +13,7 @@ The scene is initialized with random spheres. Gradient-based
 optimization is used to converge towards a faithful
 scene representation.
 """
+
 import logging
 import math

--- a/docs/requirements.txt
+++ b/docs/requirements.txt
@@ -5,7 +5,6 @@ sphinx_rtd_theme
 sphinx_markdown_tables
 numpy
 iopath
-fvcore
 https://download.pytorch.org/whl/cpu/torchvision-0.15.2%2Bcpu-cp311-cp311-linux_x86_64.whl
 https://download.pytorch.org/whl/cpu/torch-2.0.1%2Bcpu-cp311-cp311-linux_x86_64.whl
 omegaconf
--- a/docs/tutorials/bundle_adjustment.ipynb
+++ b/docs/tutorials/bundle_adjustment.ipynb
@@ -96,7 +96,7 @@
    "        torch.version.cuda.replace(\".\",\"\"),\n",
    "        f\"_pyt{pyt_version_str}\"\n",
    "    ])\n",
-    "    !pip install fvcore iopath\n",
+    "    !pip install iopath\n",
    "    if sys.platform.startswith(\"linux\"):\n",
    "        print(\"Trying to install wheel for PyTorch3D\")\n",
    "        !pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html\n",
--- a/docs/tutorials/camera_position_optimization_with_differentiable_rendering.ipynb
+++ b/docs/tutorials/camera_position_optimization_with_differentiable_rendering.ipynb
@@ -83,7 +83,7 @@
    "        torch.version.cuda.replace(\".\",\"\"),\n",
    "        f\"_pyt{pyt_version_str}\"\n",
    "    ])\n",
-    "    !pip install fvcore iopath\n",
+    "    !pip install iopath\n",
    "    if sys.platform.startswith(\"linux\"):\n",
    "        print(\"Trying to install wheel for PyTorch3D\")\n",
    "        !pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html\n",
--- a/docs/tutorials/dataloaders_ShapeNetCore_R2N2.ipynb
+++ b/docs/tutorials/dataloaders_ShapeNetCore_R2N2.ipynb
@@ -58,7 +58,7 @@
    "        torch.version.cuda.replace(\".\",\"\"),\n",
    "        f\"_pyt{pyt_version_str}\"\n",
    "    ])\n",
-    "    !pip install fvcore iopath\n",
+    "    !pip install iopath\n",
    "    if sys.platform.startswith(\"linux\"):\n",
    "        print(\"Trying to install wheel for PyTorch3D\")\n",
    "        !pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html\n",
--- a/docs/tutorials/deform_source_mesh_to_target_mesh.ipynb
+++ b/docs/tutorials/deform_source_mesh_to_target_mesh.ipynb
@@ -97,7 +97,7 @@
    "        torch.version.cuda.replace(\".\",\"\"),\n",
    "        f\"_pyt{pyt_version_str}\"\n",
    "    ])\n",
-    "    !pip install fvcore iopath\n",
+    "    !pip install iopath\n",
    "    if sys.platform.startswith(\"linux\"):\n",
    "        print(\"Trying to install wheel for PyTorch3D\")\n",
    "        !pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html\n",
--- a/docs/tutorials/fit_simple_neural_radiance_field.ipynb
+++ b/docs/tutorials/fit_simple_neural_radiance_field.ipynb
@@ -63,7 +63,7 @@
    "        torch.version.cuda.replace(\".\",\"\"),\n",
    "        f\"_pyt{pyt_version_str}\"\n",
    "    ])\n",
-    "    !pip install fvcore iopath\n",
+    "    !pip install iopath\n",
    "    if sys.platform.startswith(\"linux\"):\n",
    "        print(\"Trying to install wheel for PyTorch3D\")\n",
    "        !pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html\n",
--- a/docs/tutorials/fit_textured_mesh.ipynb
+++ b/docs/tutorials/fit_textured_mesh.ipynb
@@ -75,7 +75,7 @@
    "        torch.version.cuda.replace(\".\",\"\"),\n",
    "        f\"_pyt{pyt_version_str}\"\n",
    "    ])\n",
-    "    !pip install fvcore iopath\n",
+    "    !pip install iopath\n",
    "    if sys.platform.startswith(\"linux\"):\n",
    "        print(\"Trying to install wheel for PyTorch3D\")\n",
    "        !pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html\n",
--- a/docs/tutorials/fit_textured_volume.ipynb
+++ b/docs/tutorials/fit_textured_volume.ipynb
@@ -54,7 +54,7 @@
    "        torch.version.cuda.replace(\".\",\"\"),\n",
    "        f\"_pyt{pyt_version_str}\"\n",
    "    ])\n",
-    "    !pip install fvcore iopath\n",
+    "    !pip install iopath\n",
    "    if sys.platform.startswith(\"linux\"):\n",
    "        print(\"Trying to install wheel for PyTorch3D\")\n",
    "        !pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html\n",
--- a/docs/tutorials/implicitron_config_system.ipynb
+++ b/docs/tutorials/implicitron_config_system.ipynb
@@ -85,7 +85,7 @@
        "        torch.version.cuda.replace(\".\",\"\"),\n",
        "        f\"_pyt{pyt_version_str}\"\n",
        "    ])\n",
-        "    !pip install fvcore iopath\n",
+        "    !pip install iopath\n",
        "    if sys.platform.startswith(\"linux\"):\n",
        "        print(\"Trying to install wheel for PyTorch3D\")\n",
        "        !pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html\n",
--- a/docs/tutorials/implicitron_volumes.ipynb
+++ b/docs/tutorials/implicitron_volumes.ipynb
@@ -79,7 +79,7 @@
        "        torch.version.cuda.replace(\".\",\"\"),\n",
        "        f\"_pyt{pyt_version_str}\"\n",
        "    ])\n",
-        "    !pip install fvcore iopath\n",
+        "    !pip install iopath\n",
        "    if sys.platform.startswith(\"linux\"):\n",
        "        print(\"Trying to install wheel for PyTorch3D\")\n",
        "        !pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html\n",
--- a/docs/tutorials/render_colored_points.ipynb
+++ b/docs/tutorials/render_colored_points.ipynb
@@ -57,7 +57,7 @@
    "        torch.version.cuda.replace(\".\",\"\"),\n",
    "        f\"_pyt{pyt_version_str}\"\n",
    "    ])\n",
-    "    !pip install fvcore iopath\n",
+    "    !pip install iopath\n",
    "    if sys.platform.startswith(\"linux\"):\n",
    "        print(\"Trying to install wheel for PyTorch3D\")\n",
    "        !pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html\n",
--- a/docs/tutorials/render_densepose.ipynb
+++ b/docs/tutorials/render_densepose.ipynb
@@ -64,7 +64,7 @@
    "        torch.version.cuda.replace(\".\",\"\"),\n",
    "        f\"_pyt{pyt_version_str}\"\n",
    "    ])\n",
-    "    !pip install fvcore iopath\n",
+    "    !pip install iopath\n",
    "    if sys.platform.startswith(\"linux\"):\n",
    "        print(\"Trying to install wheel for PyTorch3D\")\n",
    "        !pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html\n",
--- a/docs/tutorials/render_textured_meshes.ipynb
+++ b/docs/tutorials/render_textured_meshes.ipynb
@@ -80,7 +80,7 @@
    "        torch.version.cuda.replace(\".\",\"\"),\n",
    "        f\"_pyt{pyt_version_str}\"\n",
    "    ])\n",
-    "    !pip install fvcore iopath\n",
+    "    !pip install iopath\n",
    "    if sys.platform.startswith(\"linux\"):\n",
    "        print(\"Trying to install wheel for PyTorch3D\")\n",
    "        !pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html\n",
--- a/packaging/build_conda.py
+++ b/packaging/build_conda.py
@@ -4,10 +4,11 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.

+import argparse
 import os.path
 import runpy
 import subprocess
-from typing import List
+from typing import List, Tuple

 # required env vars:
 # CU_VERSION: E.g. cu112
@@ -23,7 +24,7 @@ pytorch_major_minor = tuple(int(i) for i in PYTORCH_VERSION.split(".")[:2])
 source_root_dir = os.environ["PWD"]


-def version_constraint(version):
+def version_constraint(version) -> str:
    """
    Given version "11.3" returns " >=11.3,<11.4"
    """
@@ -32,7 +33,7 @@ def version_constraint(version):
    return f" >={version},<{upper}"


-def get_cuda_major_minor():
+def get_cuda_major_minor() -> Tuple[str, str]:
    if CU_VERSION == "cpu":
        raise ValueError("fn only for cuda builds")
    if len(CU_VERSION) != 5 or CU_VERSION[:2] != "cu":
@@ -42,11 +43,10 @@ def get_cuda_major_minor():
    return major, minor


-def setup_cuda():
+def setup_cuda(use_conda_cuda: bool) -> List[str]:
    if CU_VERSION == "cpu":
-        return
+        return []
    major, minor = get_cuda_major_minor()
-    os.environ["CUDA_HOME"] = f"/usr/local/cuda-{major}.{minor}/"
    os.environ["FORCE_CUDA"] = "1"

    basic_nvcc_flags = (
@@ -75,6 +75,15 @@ def setup_cuda():

    if os.environ.get("JUST_TESTRUN", "0") != "1":
        os.environ["NVCC_FLAGS"] = nvcc_flags
+    if use_conda_cuda:
+        os.environ["CONDA_CUDA_TOOLKIT_BUILD_CONSTRAINT1"] = "- cuda-toolkit"
+        os.environ["CONDA_CUDA_TOOLKIT_BUILD_CONSTRAINT2"] = (
+            f"- cuda-version={major}.{minor}"
+        )
+        return ["-c", f"nvidia/label/cuda-{major}.{minor}.0"]
+    else:
+        os.environ["CUDA_HOME"] = f"/usr/local/cuda-{major}.{minor}/"
+        return []


 def setup_conda_pytorch_constraint() -> List[str]:
@@ -95,7 +104,7 @@ def setup_conda_pytorch_constraint() -> List[str]:
        return ["-c", "pytorch", "-c", "nvidia"]


-def setup_conda_cudatoolkit_constraint():
+def setup_conda_cudatoolkit_constraint() -> None:
    if CU_VERSION == "cpu":
        os.environ["CONDA_CPUONLY_FEATURE"] = "- cpuonly"
        os.environ["CONDA_CUDATOOLKIT_CONSTRAINT"] = ""
@@ -116,14 +125,14 @@ def setup_conda_cudatoolkit_constraint():
    os.environ["CONDA_CUDATOOLKIT_CONSTRAINT"] = toolkit


-def do_build(start_args: List[str]):
+def do_build(start_args: List[str]) -> None:
    args = start_args.copy()

    test_flag = os.environ.get("TEST_FLAG")
    if test_flag is not None:
        args.append(test_flag)

-    args.extend(["-c", "bottler", "-c", "fvcore", "-c", "iopath", "-c", "conda-forge"])
+    args.extend(["-c", "bottler", "-c", "iopath", "-c", "conda-forge"])
    args.append("--no-anaconda-upload")
    args.extend(["--python", os.environ["PYTHON_VERSION"]])
    args.append("packaging/pytorch3d")
@@ -132,8 +141,16 @@ def do_build(start_args: List[str]):


 if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Build the conda package.")
+    parser.add_argument(
+        "--use-conda-cuda",
+        action="store_true",
+        help="get cuda from conda ignoring local cuda",
+    )
+    our_args = parser.parse_args()
+
    args = ["conda", "build"]
-    setup_cuda()
+    args += setup_cuda(use_conda_cuda=our_args.use_conda_cuda)

    init_path = source_root_dir + "/pytorch3d/__init__.py"
    build_version = runpy.run_path(init_path)["__version__"]
--- a/packaging/linux_wheels/README.md
+++ b/packaging/linux_wheels/README.md
@@ -26,6 +26,6 @@ version_str="".join([
    torch.version.cuda.replace(".",""),
    f"_pyt{pyt_version_str}"
 ])
-!pip install fvcore iopath
+!pip install iopath
 !pip install --no-index --no-cache-dir pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html
 ```
--- a/packaging/linux_wheels/inside.sh
+++ b/packaging/linux_wheels/inside.sh
@@ -144,7 +144,7 @@ do
            conda activate "$tag"
            # shellcheck disable=SC2086
            conda install -y -c pytorch $extra_channel "pytorch=$pytorch_version" "$cudatools=$CUDA_TAG"
-            pip install fvcore iopath
+            pip install iopath
            echo "python version" "$python_version" "pytorch version" "$pytorch_version" "cuda version" "$cu_version" "tag" "$tag"

            rm -rf dist
--- a/packaging/pytorch3d/meta.yaml
+++ b/packaging/pytorch3d/meta.yaml
@@ -8,10 +8,13 @@ source:
 requirements:
  build:
    - {{ compiler('c') }} # [win]
+    {{ environ.get('CONDA_CUDA_TOOLKIT_BUILD_CONSTRAINT1', '') }}
+    {{ environ.get('CONDA_CUDA_TOOLKIT_BUILD_CONSTRAINT2', '') }}
    {{ environ.get('CONDA_CUB_CONSTRAINT') }}

  host:
    - python
+    - mkl =2023  # [x86_64]
    {{ environ.get('SETUPTOOLS_CONSTRAINT') }}
    {{ environ.get('CONDA_PYTORCH_BUILD_CONSTRAINT') }}
    {{ environ.get('CONDA_PYTORCH_MKL_CONSTRAINT') }}
@@ -22,7 +25,7 @@ requirements:
    - python
    - numpy >=1.11
    - torchvision >=0.5
-    - fvcore
+    - mkl =2023  # [x86_64]
    - iopath
    {{ environ.get('CONDA_PYTORCH_CONSTRAINT') }}
    {{ environ.get('CONDA_CUDATOOLKIT_CONSTRAINT') }}
@@ -48,8 +51,11 @@ test:
    - imageio
    - hydra-core
    - accelerate
+    - matplotlib
+    - tabulate
+    - pandas
+    - sqlalchemy
  commands:
-    #pytest .
    python -m unittest discover -v -s tests -t .


--- a/projects/implicitron_trainer/experiment.py
+++ b/projects/implicitron_trainer/experiment.py
@@ -7,7 +7,7 @@

 # pyre-unsafe

-""""
+""" "
 This file is the entry point for launching experiments with Implicitron.

 Launch Training
@@ -44,6 +44,7 @@ The outputs of the experiment are saved and logged in multiple ways:
        config file.

 """
+
 import logging
 import os
 import warnings
@@ -99,7 +100,7 @@ except ModuleNotFoundError:
 no_accelerate = os.environ.get("PYTORCH3D_NO_ACCELERATE") is not None


-class Experiment(Configurable):  # pyre-ignore: 13
+class Experiment(Configurable):
    """
    This class is at the top level of Implicitron's config hierarchy. Its
    members are high-level components necessary for training an implicit rende-
@@ -120,12 +121,16 @@ class Experiment(Configurable):  # pyre-ignore: 13
            will be saved here.
    """

+    # pyre-fixme[13]: Attribute `data_source` is never initialized.
    data_source: DataSourceBase
    data_source_class_type: str = "ImplicitronDataSource"
+    # pyre-fixme[13]: Attribute `model_factory` is never initialized.
    model_factory: ModelFactoryBase
    model_factory_class_type: str = "ImplicitronModelFactory"
+    # pyre-fixme[13]: Attribute `optimizer_factory` is never initialized.
    optimizer_factory: OptimizerFactoryBase
    optimizer_factory_class_type: str = "ImplicitronOptimizerFactory"
+    # pyre-fixme[13]: Attribute `training_loop` is never initialized.
    training_loop: TrainingLoopBase
    training_loop_class_type: str = "ImplicitronTrainingLoop"

--- a/projects/implicitron_trainer/impl/model_factory.py
+++ b/projects/implicitron_trainer/impl/model_factory.py
@@ -26,7 +26,6 @@ logger = logging.getLogger(__name__)


 class ModelFactoryBase(ReplaceableBase):
-
    resume: bool = True  # resume from the last checkpoint

    def __call__(self, **kwargs) -> ImplicitronModelBase:
@@ -45,7 +44,7 @@ class ModelFactoryBase(ReplaceableBase):


@registry.register
-class ImplicitronModelFactory(ModelFactoryBase):  # pyre-ignore [13]
+class ImplicitronModelFactory(ModelFactoryBase):
    """
    A factory class that initializes an implicit rendering model.

@@ -61,6 +60,7 @@ class ImplicitronModelFactory(ModelFactoryBase):  # pyre-ignore [13]

    """

+    # pyre-fixme[13]: Attribute `model` is never initialized.
    model: ImplicitronModelBase
    model_class_type: str = "GenericModel"
    resume: bool = True
@@ -115,7 +115,9 @@ class ImplicitronModelFactory(ModelFactoryBase):  # pyre-ignore [13]
                        "cuda:%d" % 0: "cuda:%d" % accelerator.local_process_index
                    }
                model_state_dict = torch.load(
-                    model_io.get_model_path(model_path), map_location=map_location
+                    model_io.get_model_path(model_path),
+                    map_location=map_location,
+                    weights_only=True,
                )

                try:
--- a/projects/implicitron_trainer/impl/optimizer_factory.py
+++ b/projects/implicitron_trainer/impl/optimizer_factory.py
@@ -123,6 +123,7 @@ class ImplicitronOptimizerFactory(OptimizerFactoryBase):
        """
        # Get the parameters to optimize
        if hasattr(model, "_get_param_groups"):  # use the model function
+            # pyre-fixme[29]: `Union[Tensor, Module]` is not a function.
            p_groups = model._get_param_groups(self.lr, wd=self.weight_decay)
        else:
            p_groups = [
@@ -241,7 +242,7 @@ class ImplicitronOptimizerFactory(OptimizerFactoryBase):
                    map_location = {
                        "cuda:%d" % 0: "cuda:%d" % accelerator.local_process_index
                    }
-                optimizer_state = torch.load(opt_path, map_location)
+                optimizer_state = torch.load(opt_path, map_location, weights_only=True)
            else:
                raise FileNotFoundError(f"Optimizer state {opt_path} does not exist.")
        return optimizer_state
--- a/projects/implicitron_trainer/impl/training_loop.py
+++ b/projects/implicitron_trainer/impl/training_loop.py
@@ -30,13 +30,13 @@ from .utils import seed_all_random_engines
 logger = logging.getLogger(__name__)


-# pyre-fixme[13]: Attribute `evaluator` is never initialized.
 class TrainingLoopBase(ReplaceableBase):
    """
    Members:
        evaluator: An EvaluatorBase instance, used to evaluate training results.
    """

+    # pyre-fixme[13]: Attribute `evaluator` is never initialized.
    evaluator: Optional[EvaluatorBase]
    evaluator_class_type: Optional[str] = "ImplicitronEvaluator"

@@ -161,7 +161,6 @@ class ImplicitronTrainingLoop(TrainingLoopBase):
        for epoch in range(start_epoch, self.max_epochs):
            # automatic new_epoch and plotting of stats at every epoch start
            with stats:
-
                # Make sure to re-seed random generators to ensure reproducibility
                # even after restart.
                seed_all_random_engines(seed + epoch)
@@ -395,6 +394,7 @@ class ImplicitronTrainingLoop(TrainingLoopBase):
            ):
                prefix = f"e{stats.epoch}_it{stats.it[trainmode]}"
                if hasattr(model, "visualize"):
+                    # pyre-fixme[29]: `Union[Tensor, Module]` is not a function.
                    model.visualize(
                        viz,
                        visdom_env_imgs,
--- a/projects/implicitron_trainer/tests/test_experiment.py
+++ b/projects/implicitron_trainer/tests/test_experiment.py
@@ -53,12 +53,8 @@ class TestExperiment(unittest.TestCase):
        cfg.data_source_ImplicitronDataSource_args.dataset_map_provider_class_type = (
            "JsonIndexDatasetMapProvider"
        )
-        dataset_args = (
-            cfg.data_source_ImplicitronDataSource_args.dataset_map_provider_JsonIndexDatasetMapProvider_args
-        )
-        dataloader_args = (
-            cfg.data_source_ImplicitronDataSource_args.data_loader_map_provider_SequenceDataLoaderMapProvider_args
-        )
+        dataset_args = cfg.data_source_ImplicitronDataSource_args.dataset_map_provider_JsonIndexDatasetMapProvider_args
+        dataloader_args = cfg.data_source_ImplicitronDataSource_args.data_loader_map_provider_SequenceDataLoaderMapProvider_args
        dataset_args.category = "skateboard"
        dataset_args.test_restrict_sequence_id = 0
        dataset_args.dataset_root = "manifold://co3d/tree/extracted"
@@ -94,12 +90,8 @@ class TestExperiment(unittest.TestCase):
        cfg.data_source_ImplicitronDataSource_args.dataset_map_provider_class_type = (
            "JsonIndexDatasetMapProvider"
        )
-        dataset_args = (
-            cfg.data_source_ImplicitronDataSource_args.dataset_map_provider_JsonIndexDatasetMapProvider_args
-        )
-        dataloader_args = (
-            cfg.data_source_ImplicitronDataSource_args.data_loader_map_provider_SequenceDataLoaderMapProvider_args
-        )
+        dataset_args = cfg.data_source_ImplicitronDataSource_args.dataset_map_provider_JsonIndexDatasetMapProvider_args
+        dataloader_args = cfg.data_source_ImplicitronDataSource_args.data_loader_map_provider_SequenceDataLoaderMapProvider_args
        dataset_args.category = "skateboard"
        dataset_args.test_restrict_sequence_id = 0
        dataset_args.dataset_root = "manifold://co3d/tree/extracted"
@@ -111,9 +103,7 @@ class TestExperiment(unittest.TestCase):
        cfg.training_loop_ImplicitronTrainingLoop_args.max_epochs = 2
        cfg.training_loop_ImplicitronTrainingLoop_args.store_checkpoints = False
        cfg.optimizer_factory_ImplicitronOptimizerFactory_args.lr_policy = "Exponential"
-        cfg.optimizer_factory_ImplicitronOptimizerFactory_args.exponential_lr_step_size = (
-            2
-        )
+        cfg.optimizer_factory_ImplicitronOptimizerFactory_args.exponential_lr_step_size = 2

        if DEBUG:
            experiment.dump_cfg(cfg)
--- a/projects/implicitron_trainer/tests/test_optimizer_factory.py
+++ b/projects/implicitron_trainer/tests/test_optimizer_factory.py
@@ -81,8 +81,9 @@ class TestOptimizerFactory(unittest.TestCase):

    def test_param_overrides_self_param_group_assignment(self):
        pa, pb, pc = [torch.nn.Parameter(data=torch.tensor(i * 1.0)) for i in range(3)]
-        na, nb = Node(params=[pa]), Node(
-            params=[pb], param_groups={"self": "pb_self", "p1": "pb_param"}
+        na, nb = (
+            Node(params=[pa]),
+            Node(params=[pb], param_groups={"self": "pb_self", "p1": "pb_param"}),
        )
        root = Node(children=[na, nb], params=[pc], param_groups={"m1": "pb_member"})
        param_groups = self._get_param_groups(root)
--- a/projects/nerf/nerf/dataset.py
+++ b/projects/nerf/nerf/dataset.py
@@ -84,9 +84,9 @@ def get_nerf_datasets(

    if autodownload and any(not os.path.isfile(p) for p in (cameras_path, image_path)):
        # Automatically download the data files if missing.
-        download_data((dataset_name,), data_root=data_root)
+        download_data([dataset_name], data_root=data_root)

-    train_data = torch.load(cameras_path)
+    train_data = torch.load(cameras_path, weights_only=True)
    n_cameras = train_data["cameras"]["R"].shape[0]

    _image_max_image_pixels = Image.MAX_IMAGE_PIXELS
--- a/projects/nerf/nerf/stats.py
+++ b/projects/nerf/nerf/stats.py
@@ -194,7 +194,6 @@ class Stats:
        it = self.it[stat_set]

        for stat in self.log_vars:
-
            if stat not in self.stats[stat_set]:
                self.stats[stat_set][stat] = AverageMeter()

--- a/projects/nerf/test_nerf.py
+++ b/projects/nerf/test_nerf.py
@@ -24,7 +24,6 @@ CONFIG_DIR = os.path.join(os.path.dirname(os.path.realpath(__file__)), "configs"

@hydra.main(config_path=CONFIG_DIR, config_name="lego")
 def main(cfg: DictConfig):
-
    # Device on which to run.
    if torch.cuda.is_available():
        device = "cuda"
@@ -63,7 +62,7 @@ def main(cfg: DictConfig):
        raise ValueError(f"Model checkpoint {checkpoint_path} does not exist!")

    print(f"Loading checkpoint {checkpoint_path}.")
-    loaded_data = torch.load(checkpoint_path)
+    loaded_data = torch.load(checkpoint_path, weights_only=True)
    # Do not load the cached xy grid.
    # - this allows setting an arbitrary evaluation image size.
    state_dict = {
--- a/projects/nerf/tests/test_raysampler.py
+++ b/projects/nerf/tests/test_raysampler.py
@@ -42,7 +42,6 @@ class TestRaysampler(unittest.TestCase):
        cameras, rays = [], []

        for _ in range(batch_size):
-
            R = random_rotations(1)
            T = torch.randn(1, 3)
            focal_length = torch.rand(1, 2) + 0.5
--- a/projects/nerf/train_nerf.py
+++ b/projects/nerf/train_nerf.py
@@ -25,7 +25,6 @@ CONFIG_DIR = os.path.join(os.path.dirname(os.path.realpath(__file__)), "configs"

@hydra.main(config_path=CONFIG_DIR, config_name="lego")
 def main(cfg: DictConfig):
-
    # Set the relevant seeds for reproducibility.
    np.random.seed(cfg.seed)
    torch.manual_seed(cfg.seed)
@@ -77,7 +76,7 @@ def main(cfg: DictConfig):
        # Resume training if requested.
        if cfg.resume and os.path.isfile(checkpoint_path):
            print(f"Resuming from checkpoint {checkpoint_path}.")
-            loaded_data = torch.load(checkpoint_path)
+            loaded_data = torch.load(checkpoint_path, weights_only=True)
            model.load_state_dict(loaded_data["model"])
            stats = pickle.loads(loaded_data["stats"])
            print(f"   => resuming from epoch {stats.epoch}.")
@@ -219,7 +218,6 @@ def main(cfg: DictConfig):

        # Validation
        if epoch % cfg.validation_epoch_interval == 0 and epoch > 0:
-
            # Sample a validation camera/image.
            val_batch = next(val_dataloader.__iter__())
            val_image, val_camera, camera_idx = val_batch[0].values()
--- a/pytorch3d/init.py
+++ b/pytorch3d/init.py
@@ -6,4 +6,4 @@

 # pyre-unsafe

-__version__ = "0.7.6"
+__version__ = "0.7.8"
--- a/pytorch3d/common/compat.py
+++ b/pytorch3d/common/compat.py
@@ -17,7 +17,7 @@ Some functions which depend on PyTorch or Python versions.


 def meshgrid_ij(
-    *A: Union[torch.Tensor, Sequence[torch.Tensor]]
+    *A: Union[torch.Tensor, Sequence[torch.Tensor]],
 ) -> Tuple[torch.Tensor, ...]:  # pragma: no cover
    """
    Like torch.meshgrid was before PyTorch 1.10.0, i.e. with indexing set to ij
--- a/pytorch3d/csrc/ball_query/ball_query_cpu.cpp
+++ b/pytorch3d/csrc/ball_query/ball_query_cpu.cpp
@@ -7,7 +7,6 @@
 */

 #include <torch/extension.h>
-#include <queue>
 #include <tuple>

 std::tuple<at::Tensor, at::Tensor> BallQueryCpu(
--- a/pytorch3d/csrc/compositing/alpha_composite.cu
+++ b/pytorch3d/csrc/compositing/alpha_composite.cu
@@ -28,7 +28,6 @@ __global__ void alphaCompositeCudaForwardKernel(
    const at::PackedTensorAccessor64<float, 4, at::RestrictPtrTraits> alphas,
    const at::PackedTensorAccessor64<int64_t, 4, at::RestrictPtrTraits> points_idx) {
  // clang-format on
-  const int64_t batch_size = result.size(0);
  const int64_t C = features.size(0);
  const int64_t H = points_idx.size(2);
  const int64_t W = points_idx.size(3);
@@ -79,7 +78,6 @@ __global__ void alphaCompositeCudaBackwardKernel(
    const at::PackedTensorAccessor64<float, 4, at::RestrictPtrTraits> alphas,
    const at::PackedTensorAccessor64<int64_t, 4, at::RestrictPtrTraits> points_idx) {
  // clang-format on
-  const int64_t batch_size = points_idx.size(0);
  const int64_t C = features.size(0);
  const int64_t H = points_idx.size(2);
  const int64_t W = points_idx.size(3);
--- a/pytorch3d/csrc/compositing/norm_weighted_sum.cu
+++ b/pytorch3d/csrc/compositing/norm_weighted_sum.cu
@@ -28,7 +28,6 @@ __global__ void weightedSumNormCudaForwardKernel(
    const at::PackedTensorAccessor64<float, 4, at::RestrictPtrTraits> alphas,
    const at::PackedTensorAccessor64<int64_t, 4, at::RestrictPtrTraits> points_idx) {
  // clang-format on
-  const int64_t batch_size = result.size(0);
  const int64_t C = features.size(0);
  const int64_t H = points_idx.size(2);
  const int64_t W = points_idx.size(3);
@@ -92,7 +91,6 @@ __global__ void weightedSumNormCudaBackwardKernel(
    const at::PackedTensorAccessor64<float, 4, at::RestrictPtrTraits> alphas,
    const at::PackedTensorAccessor64<int64_t, 4, at::RestrictPtrTraits> points_idx) {
  // clang-format on
-  const int64_t batch_size = points_idx.size(0);
  const int64_t C = features.size(0);
  const int64_t H = points_idx.size(2);
  const int64_t W = points_idx.size(3);
--- a/pytorch3d/csrc/compositing/weighted_sum.cu
+++ b/pytorch3d/csrc/compositing/weighted_sum.cu
@@ -26,7 +26,6 @@ __global__ void weightedSumCudaForwardKernel(
    const at::PackedTensorAccessor64<float, 4, at::RestrictPtrTraits> alphas,
    const at::PackedTensorAccessor64<int64_t, 4, at::RestrictPtrTraits> points_idx) {
  // clang-format on
-  const int64_t batch_size = result.size(0);
  const int64_t C = features.size(0);
  const int64_t H = points_idx.size(2);
  const int64_t W = points_idx.size(3);
@@ -74,7 +73,6 @@ __global__ void weightedSumCudaBackwardKernel(
    const at::PackedTensorAccessor64<float, 4, at::RestrictPtrTraits> alphas,
    const at::PackedTensorAccessor64<int64_t, 4, at::RestrictPtrTraits> points_idx) {
  // clang-format on
-  const int64_t batch_size = points_idx.size(0);
  const int64_t C = features.size(0);
  const int64_t H = points_idx.size(2);
  const int64_t W = points_idx.size(3);
--- a/pytorch3d/csrc/ext.cpp
+++ b/pytorch3d/csrc/ext.cpp
@@ -99,6 +99,7 @@ PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
  m.def("marching_cubes", &MarchingCubes);

  // Pulsar.
+  // Pulsar not enabled on AMD.
 #ifdef PULSAR_LOGGING_ENABLED
  c10::ShowLogInfoToStderr();
 #endif
@@ -148,10 +149,10 @@ PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
          py::arg("gamma"),
          py::arg("max_depth"),
          py::arg("min_depth") /* = 0.f*/,
-          py::arg(
-              "bg_col") /* = at::nullopt not exposed properly in pytorch 1.1. */
+          py::arg("bg_col") /* = std::nullopt not exposed properly in
+                               pytorch 1.1. */
          ,
-          py::arg("opacity") /* = at::nullopt ... */,
+          py::arg("opacity") /* = std::nullopt ... */,
          py::arg("percent_allowed_difference") = 0.01f,
          py::arg("max_n_hits") = MAX_UINT,
          py::arg("mode") = 0)
--- a/pytorch3d/csrc/iou_box3d/iou_box3d_cpu.cpp
+++ b/pytorch3d/csrc/iou_box3d/iou_box3d_cpu.cpp
@@ -7,10 +7,7 @@
 */

 #include <torch/extension.h>
-#include <torch/torch.h>
 #include <list>
-#include <numeric>
-#include <queue>
 #include <tuple>
 #include "iou_box3d/iou_utils.h"

--- a/pytorch3d/csrc/iou_box3d/iou_utils.cuh
+++ b/pytorch3d/csrc/iou_box3d/iou_utils.cuh
@@ -461,10 +461,8 @@ __device__ inline std::tuple<float3, float3> ArgMaxVerts(
 __device__ inline bool IsCoplanarTriTri(
    const FaceVerts& tri1,
    const FaceVerts& tri2) {
-  const float3 tri1_ctr = FaceCenter({tri1.v0, tri1.v1, tri1.v2});
  const float3 tri1_n = FaceNormal({tri1.v0, tri1.v1, tri1.v2});

-  const float3 tri2_ctr = FaceCenter({tri2.v0, tri2.v1, tri2.v2});
  const float3 tri2_n = FaceNormal({tri2.v0, tri2.v1, tri2.v2});

  // Check if parallel
@@ -500,7 +498,6 @@ __device__ inline bool IsCoplanarTriPlane(
    const FaceVerts& tri,
    const FaceVerts& plane,
    const float3& normal) {
-  const float3 tri_ctr = FaceCenter({tri.v0, tri.v1, tri.v2});
  const float3 nt = FaceNormal({tri.v0, tri.v1, tri.v2});

  // check if parallel
@@ -728,7 +725,7 @@ __device__ inline int BoxIntersections(
      }
    }
    // Update the face_verts_out tris
-    num_tris = offset;
+    num_tris = min(MAX_TRIS, offset);
    for (int j = 0; j < num_tris; ++j) {
      face_verts_out[j] = tri_verts_updated[j];
    }
--- a/pytorch3d/csrc/points_to_volumes/points_to_volumes_cpu.cpp
+++ b/pytorch3d/csrc/points_to_volumes/points_to_volumes_cpu.cpp
@@ -8,9 +8,7 @@

 #include <torch/csrc/autograd/VariableTypeUtils.h>
 #include <torch/extension.h>
-#include <algorithm>
 #include <cmath>
-#include <thread>
 #include <vector>

 // In the x direction, the location {0, ..., grid_size_x - 1} correspond to
--- a/pytorch3d/csrc/pulsar/global.h
+++ b/pytorch3d/csrc/pulsar/global.h
@@ -36,11 +36,13 @@
 #pragma nv_diag_suppress 2951
 #pragma nv_diag_suppress 2967
 #else
+#if !defined(USE_ROCM)
 #pragma diag_suppress = attribute_not_allowed
 #pragma diag_suppress = 1866
 #pragma diag_suppress = 2941
 #pragma diag_suppress = 2951
 #pragma diag_suppress = 2967
+#endif //! USE_ROCM
 #endif
 #else // __CUDACC__
 #define INLINE inline
@@ -56,7 +58,9 @@
 #pragma clang diagnostic pop
 #ifdef WITH_CUDA
 #include <ATen/cuda/CUDAContext.h>
+#if !defined(USE_ROCM)
 #include <vector_functions.h>
+#endif //! USE_ROCM
 #else
 #ifndef cudaStream_t
 typedef void* cudaStream_t;
--- a/pytorch3d/csrc/pulsar/cuda/README.md
+++ b/pytorch3d/csrc/pulsar/cuda/README.md
--- a/pytorch3d/csrc/pulsar/cuda/commands.h
+++ b/pytorch3d/csrc/pulsar/cuda/commands.h
@@ -59,6 +59,11 @@ getLastCudaError(const char* errorMessage, const char* file, const int line) {
 #define SHARED __shared__
 #define ACTIVEMASK() __activemask()
 #define BALLOT(mask, val) __ballot_sync((mask), val)
+
+/* TODO (ROCM-6.2): None of the WARP_* are used anywhere and ROCM-6.2 natively
+ * supports __shfl_*. Disabling until the move to ROCM-6.2.
+ */
+#if !defined(USE_ROCM)
 /**
 * Find the cumulative sum within a warp up to the current
 * thread lane, with each mask thread contributing base.
@@ -115,6 +120,7 @@ INLINE DEVICE float3 WARP_SUM_FLOAT3(
  ret.z = WARP_SUM(group, mask, base.z);
  return ret;
 }
+#endif //! USE_ROCM

 // Floating point.
 // #define FMUL(a, b) __fmul_rn((a), (b))
@@ -142,6 +148,7 @@ INLINE DEVICE float3 WARP_SUM_FLOAT3(
 #define FMA(x, y, z) __fmaf_rn((x), (y), (z))
 #define I2F(a) __int2float_rn(a)
 #define FRCP(x) __frcp_rn(x)
+#if !defined(USE_ROCM)
 __device__ static float atomicMax(float* address, float val) {
  int* address_as_i = (int*)address;
  int old = *address_as_i, assumed;
@@ -166,6 +173,7 @@ __device__ static float atomicMin(float* address, float val) {
  } while (assumed != old);
  return __int_as_float(old);
 }
+#endif //! USE_ROCM
 #define DMAX(a, b) FMAX(a, b)
 #define DMIN(a, b) FMIN(a, b)
 #define DSQRT(a) sqrt(a)
--- a/pytorch3d/csrc/pulsar/cuda/renderer.backward.gpu.cu
+++ b/pytorch3d/csrc/pulsar/cuda/renderer.backward.gpu.cu
--- a/pytorch3d/csrc/pulsar/cuda/renderer.backward_dbg.gpu.cu
+++ b/pytorch3d/csrc/pulsar/cuda/renderer.backward_dbg.gpu.cu
--- a/pytorch3d/csrc/pulsar/cuda/renderer.calc_gradients.gpu.cu
+++ b/pytorch3d/csrc/pulsar/cuda/renderer.calc_gradients.gpu.cu
--- a/pytorch3d/csrc/pulsar/cuda/renderer.calc_signature.gpu.cu
+++ b/pytorch3d/csrc/pulsar/cuda/renderer.calc_signature.gpu.cu
--- a/pytorch3d/csrc/pulsar/cuda/renderer.construct.gpu.cu
+++ b/pytorch3d/csrc/pulsar/cuda/renderer.construct.gpu.cu
--- a/pytorch3d/csrc/pulsar/cuda/renderer.create_selector.gpu.cu
+++ b/pytorch3d/csrc/pulsar/cuda/renderer.create_selector.gpu.cu
--- a/pytorch3d/csrc/pulsar/cuda/renderer.destruct.gpu.cu
+++ b/pytorch3d/csrc/pulsar/cuda/renderer.destruct.gpu.cu
--- a/pytorch3d/csrc/pulsar/cuda/renderer.fill_bg.gpu.cu
+++ b/pytorch3d/csrc/pulsar/cuda/renderer.fill_bg.gpu.cu
--- a/pytorch3d/csrc/pulsar/cuda/renderer.forward.gpu.cu
+++ b/pytorch3d/csrc/pulsar/cuda/renderer.forward.gpu.cu
--- a/pytorch3d/csrc/pulsar/cuda/renderer.norm_cam_gradients.gpu.cu
+++ b/pytorch3d/csrc/pulsar/cuda/renderer.norm_cam_gradients.gpu.cu
--- a/pytorch3d/csrc/pulsar/cuda/renderer.norm_sphere_gradients.gpu.cu
+++ b/pytorch3d/csrc/pulsar/cuda/renderer.norm_sphere_gradients.gpu.cu
--- a/pytorch3d/csrc/pulsar/cuda/renderer.render.gpu.cu
+++ b/pytorch3d/csrc/pulsar/cuda/renderer.render.gpu.cu
--- a/pytorch3d/csrc/pulsar/include/camera.device.h
+++ b/pytorch3d/csrc/pulsar/include/camera.device.h
@@ -14,7 +14,7 @@
 #include "./commands.h"

 namespace pulsar {
-IHD CamGradInfo::CamGradInfo() {
+IHD CamGradInfo::CamGradInfo(int x) {
  cam_pos = make_float3(0.f, 0.f, 0.f);
  pixel_0_0_center = make_float3(0.f, 0.f, 0.f);
  pixel_dir_x = make_float3(0.f, 0.f, 0.f);
--- a/pytorch3d/csrc/pulsar/include/camera.h
+++ b/pytorch3d/csrc/pulsar/include/camera.h
@@ -63,7 +63,7 @@ inline bool operator==(const CamInfo& a, const CamInfo& b) {
 };

 struct CamGradInfo {
-  HOST DEVICE CamGradInfo();
+  HOST DEVICE CamGradInfo(int = 0);
  float3 cam_pos;
  float3 pixel_0_0_center;
  float3 pixel_dir_x;
--- a/pytorch3d/csrc/pulsar/include/commands.h
+++ b/pytorch3d/csrc/pulsar/include/commands.h
@@ -24,7 +24,7 @@
 // #pragma diag_suppress = 68
 #include <ATen/cuda/CUDAContext.h>
 // #pragma pop
-#include "../cuda/commands.h"
+#include "../gpu/commands.h"
 #else
 #pragma clang diagnostic push
 #pragma clang diagnostic ignored "-Weverything"
--- a/pytorch3d/csrc/pulsar/include/math.h
+++ b/pytorch3d/csrc/pulsar/include/math.h
@@ -46,6 +46,7 @@ IHD float3 outer_product_sum(const float3& a) {
 }

 // TODO: put intrinsics here.
+#if !defined(USE_ROCM)
 IHD float3 operator+(const float3& a, const float3& b) {
  return make_float3(a.x + b.x, a.y + b.y, a.z + b.z);
 }
@@ -93,6 +94,7 @@ IHD float3 operator*(const float3& a, const float3& b) {
 IHD float3 operator*(const float& a, const float3& b) {
  return b * a;
 }
+#endif //! USE_ROCM

 INLINE DEVICE float length(const float3& v) {
  // TODO: benchmark what's faster.
--- a/pytorch3d/csrc/pulsar/include/renderer.render.device.h
+++ b/pytorch3d/csrc/pulsar/include/renderer.render.device.h
@@ -283,9 +283,15 @@ GLOBAL void render(
          (percent_allowed_difference > 0.f &&
           max_closest_possible_intersection > depth_threshold) ||
          tracker.get_n_hits() >= max_n_hits;
+#if defined(__CUDACC__) && defined(__HIP_PLATFORM_AMD__)
+      unsigned long long warp_done = __ballot(done);
+      int warp_done_bit_cnt = __popcll(warp_done);
+#else
      uint warp_done = thread_warp.ballot(done);
+      int warp_done_bit_cnt = POPC(warp_done);
+#endif //__CUDACC__ && __HIP_PLATFORM_AMD__
      if (thread_warp.thread_rank() == 0)
-        ATOMICADD_B(&n_pixels_done, POPC(warp_done));
+        ATOMICADD_B(&n_pixels_done, warp_done_bit_cnt);
      // This sync is necessary to keep n_loaded until all threads are done with
      // painting.
      thread_block.sync();
--- a/pytorch3d/csrc/pulsar/pytorch/renderer.cpp
+++ b/pytorch3d/csrc/pulsar/pytorch/renderer.cpp
@@ -213,8 +213,8 @@ std::tuple<size_t, size_t, bool, torch::Tensor> Renderer::arg_check(
    const float& gamma,
    const float& max_depth,
    float& min_depth,
-    const c10::optional<torch::Tensor>& bg_col,
-    const c10::optional<torch::Tensor>& opacity,
+    const std::optional<torch::Tensor>& bg_col,
+    const std::optional<torch::Tensor>& opacity,
    const float& percent_allowed_difference,
    const uint& max_n_hits,
    const uint& mode) {
@@ -668,8 +668,8 @@ std::tuple<torch::Tensor, torch::Tensor> Renderer::forward(
    const float& gamma,
    const float& max_depth,
    float min_depth,
-    const c10::optional<torch::Tensor>& bg_col,
-    const c10::optional<torch::Tensor>& opacity,
+    const std::optional<torch::Tensor>& bg_col,
+    const std::optional<torch::Tensor>& opacity,
    const float& percent_allowed_difference,
    const uint& max_n_hits,
    const uint& mode) {
@@ -888,14 +888,14 @@ std::tuple<torch::Tensor, torch::Tensor> Renderer::forward(
 };

 std::tuple<
-    at::optional<torch::Tensor>,
-    at::optional<torch::Tensor>,
-    at::optional<torch::Tensor>,
-    at::optional<torch::Tensor>,
-    at::optional<torch::Tensor>,
-    at::optional<torch::Tensor>,
-    at::optional<torch::Tensor>,
-    at::optional<torch::Tensor>>
+    std::optional<torch::Tensor>,
+    std::optional<torch::Tensor>,
+    std::optional<torch::Tensor>,
+    std::optional<torch::Tensor>,
+    std::optional<torch::Tensor>,
+    std::optional<torch::Tensor>,
+    std::optional<torch::Tensor>,
+    std::optional<torch::Tensor>>
 Renderer::backward(
    const torch::Tensor& grad_im,
    const torch::Tensor& image,
@@ -912,8 +912,8 @@ Renderer::backward(
    const float& gamma,
    const float& max_depth,
    float min_depth,
-    const c10::optional<torch::Tensor>& bg_col,
-    const c10::optional<torch::Tensor>& opacity,
+    const std::optional<torch::Tensor>& bg_col,
+    const std::optional<torch::Tensor>& opacity,
    const float& percent_allowed_difference,
    const uint& max_n_hits,
    const uint& mode,
@@ -922,7 +922,7 @@ Renderer::backward(
    const bool& dif_rad,
    const bool& dif_cam,
    const bool& dif_opy,
-    const at::optional<std::pair<uint, uint>>& dbg_pos) {
+    const std::optional<std::pair<uint, uint>>& dbg_pos) {
  this->ensure_on_device(this->device_tracker.device());
  size_t batch_size;
  size_t n_points;
@@ -1045,14 +1045,14 @@ Renderer::backward(
  }
  // Prepare the return value.
  std::tuple<
-      at::optional<torch::Tensor>,
-      at::optional<torch::Tensor>,
-      at::optional<torch::Tensor>,
-      at::optional<torch::Tensor>,
-      at::optional<torch::Tensor>,
-      at::optional<torch::Tensor>,
-      at::optional<torch::Tensor>,
-      at::optional<torch::Tensor>>
+      std::optional<torch::Tensor>,
+      std::optional<torch::Tensor>,
+      std::optional<torch::Tensor>,
+      std::optional<torch::Tensor>,
+      std::optional<torch::Tensor>,
+      std::optional<torch::Tensor>,
+      std::optional<torch::Tensor>,
+      std::optional<torch::Tensor>>
      ret;
  if (mode == 1 || (!dif_pos && !dif_col && !dif_rad && !dif_cam && !dif_opy)) {
    return ret;
--- a/pytorch3d/csrc/pulsar/pytorch/renderer.h
+++ b/pytorch3d/csrc/pulsar/pytorch/renderer.h
@@ -44,21 +44,21 @@ struct Renderer {
      const float& gamma,
      const float& max_depth,
      float min_depth,
-      const c10::optional<torch::Tensor>& bg_col,
-      const c10::optional<torch::Tensor>& opacity,
+      const std::optional<torch::Tensor>& bg_col,
+      const std::optional<torch::Tensor>& opacity,
      const float& percent_allowed_difference,
      const uint& max_n_hits,
      const uint& mode);

  std::tuple<
-      at::optional<torch::Tensor>,
-      at::optional<torch::Tensor>,
-      at::optional<torch::Tensor>,
-      at::optional<torch::Tensor>,
-      at::optional<torch::Tensor>,
-      at::optional<torch::Tensor>,
-      at::optional<torch::Tensor>,
-      at::optional<torch::Tensor>>
+      std::optional<torch::Tensor>,
+      std::optional<torch::Tensor>,
+      std::optional<torch::Tensor>,
+      std::optional<torch::Tensor>,
+      std::optional<torch::Tensor>,
+      std::optional<torch::Tensor>,
+      std::optional<torch::Tensor>,
+      std::optional<torch::Tensor>>
  backward(
      const torch::Tensor& grad_im,
      const torch::Tensor& image,
@@ -75,8 +75,8 @@ struct Renderer {
      const float& gamma,
      const float& max_depth,
      float min_depth,
-      const c10::optional<torch::Tensor>& bg_col,
-      const c10::optional<torch::Tensor>& opacity,
+      const std::optional<torch::Tensor>& bg_col,
+      const std::optional<torch::Tensor>& opacity,
      const float& percent_allowed_difference,
      const uint& max_n_hits,
      const uint& mode,
@@ -85,7 +85,7 @@ struct Renderer {
      const bool& dif_rad,
      const bool& dif_cam,
      const bool& dif_opy,
-      const at::optional<std::pair<uint, uint>>& dbg_pos);
+      const std::optional<std::pair<uint, uint>>& dbg_pos);

  // Infrastructure.
  /**
@@ -115,8 +115,8 @@ struct Renderer {
      const float& gamma,
      const float& max_depth,
      float& min_depth,
-      const c10::optional<torch::Tensor>& bg_col,
-      const c10::optional<torch::Tensor>& opacity,
+      const std::optional<torch::Tensor>& bg_col,
+      const std::optional<torch::Tensor>& opacity,
      const float& percent_allowed_difference,
      const uint& max_n_hits,
      const uint& mode);
--- a/pytorch3d/csrc/pulsar/pytorch/tensor_util.cpp
+++ b/pytorch3d/csrc/pulsar/pytorch/tensor_util.cpp
@@ -8,6 +8,7 @@

 #ifdef WITH_CUDA
 #include <ATen/cuda/CUDAContext.h>
+#include <c10/cuda/CUDAException.h>
 #include <cuda_runtime_api.h>
 #endif
 #include <torch/extension.h>
@@ -33,13 +34,13 @@ torch::Tensor sphere_ids_from_result_info_nograd(
          .contiguous();
  if (forw_info.device().type() == c10::DeviceType::CUDA) {
 #ifdef WITH_CUDA
-    cudaMemcpyAsync(
+    C10_CUDA_CHECK(cudaMemcpyAsync(
        result.data_ptr(),
        tmp.data_ptr(),
        sizeof(uint32_t) * tmp.size(0) * tmp.size(1) * tmp.size(2) *
            tmp.size(3),
        cudaMemcpyDeviceToDevice,
-        at::cuda::getCurrentCUDAStream());
+        at::cuda::getCurrentCUDAStream()));
 #else
    throw std::runtime_error(
        "Copy on CUDA device initiated but built "
--- a/pytorch3d/csrc/pulsar/pytorch/util.cpp
+++ b/pytorch3d/csrc/pulsar/pytorch/util.cpp
@@ -7,6 +7,7 @@
 */

 #ifdef WITH_CUDA
+#include <c10/cuda/CUDAException.h>
 #include <cuda_runtime_api.h>

 namespace pulsar {
@@ -17,7 +18,8 @@ void cudaDevToDev(
    const void* src,
    const int& size,
    const cudaStream_t& stream) {
-  cudaMemcpyAsync(trg, src, size, cudaMemcpyDeviceToDevice, stream);
+  C10_CUDA_CHECK(
+      cudaMemcpyAsync(trg, src, size, cudaMemcpyDeviceToDevice, stream));
 }

 void cudaDevToHost(
@@ -25,7 +27,8 @@ void cudaDevToHost(
    const void* src,
    const int& size,
    const cudaStream_t& stream) {
-  cudaMemcpyAsync(trg, src, size, cudaMemcpyDeviceToHost, stream);
+  C10_CUDA_CHECK(
+      cudaMemcpyAsync(trg, src, size, cudaMemcpyDeviceToHost, stream));
 }

 } // namespace pytorch
--- a/pytorch3d/csrc/rasterize_meshes/rasterize_meshes.cu
+++ b/pytorch3d/csrc/rasterize_meshes/rasterize_meshes.cu
@@ -144,7 +144,7 @@ __device__ void CheckPixelInsideFace(
  const bool zero_face_area =
      (face_area <= kEpsilon && face_area >= -1.0f * kEpsilon);

-  if (zmax < 0 || cull_backfaces && back_face || outside_bbox ||
+  if (zmax < 0 || (cull_backfaces && back_face) || outside_bbox ||
      zero_face_area) {
    return;
  }
--- a/pytorch3d/csrc/rasterize_meshes/rasterize_meshes_cpu.cpp
+++ b/pytorch3d/csrc/rasterize_meshes/rasterize_meshes_cpu.cpp
@@ -9,7 +9,6 @@
 #include <torch/extension.h>
 #include <algorithm>
 #include <list>
-#include <queue>
 #include <thread>
 #include <tuple>
 #include "ATen/core/TensorAccessor.h"
--- a/pytorch3d/csrc/sample_farthest_points/sample_farthest_points.cu
+++ b/pytorch3d/csrc/sample_farthest_points/sample_farthest_points.cu
@@ -35,8 +35,6 @@ __global__ void FarthestPointSamplingKernel(
  __shared__ int64_t selected_store;

  // Get constants
-  const int64_t N = points.size(0);
-  const int64_t P = points.size(1);
  const int64_t D = points.size(2);

  // Get batch index and thread index
--- a/pytorch3d/csrc/utils/float_math.cuh
+++ b/pytorch3d/csrc/utils/float_math.cuh
@@ -18,6 +18,8 @@ const auto vEpsilon = 1e-8;

 // Common functions and operators for float2.

+// Complex arithmetic is already defined for AMD.
+#if !defined(USE_ROCM)
 __device__ inline float2 operator-(const float2& a, const float2& b) {
  return make_float2(a.x - b.x, a.y - b.y);
 }
@@ -41,6 +43,7 @@ __device__ inline float2 operator*(const float2& a, const float2& b) {
 __device__ inline float2 operator*(const float a, const float2& b) {
  return make_float2(a * b.x, a * b.y);
 }
+#endif

 __device__ inline float FloatMin3(const float a, const float b, const float c) {
  return fminf(a, fminf(b, c));
--- a/pytorch3d/csrc/utils/geometry_utils.cuh
+++ b/pytorch3d/csrc/utils/geometry_utils.cuh
@@ -376,8 +376,6 @@ PointLineDistanceBackward(
  float tt = t_top / t_bot;
  tt = __saturatef(tt);
  const float2 p_proj = (1.0f - tt) * v0 + tt * v1;
-  const float2 d = p - p_proj;
-  const float dist = sqrt(dot(d, d));

  const float2 grad_p = -1.0f * grad_dist * 2.0f * (p_proj - p);
  const float2 grad_v0 = grad_dist * (1.0f - tt) * 2.0f * (p_proj - p);
--- a/pytorch3d/csrc/utils/warp_reduce.cuh
+++ b/pytorch3d/csrc/utils/warp_reduce.cuh
@@ -23,37 +23,51 @@ WarpReduceMin(scalar_t* min_dists, int64_t* min_idxs, const size_t tid) {
    min_idxs[tid] = min_idxs[tid + 32];
    min_dists[tid] = min_dists[tid + 32];
  }
+// AMD does not use explicit syncwarp and instead automatically inserts memory
+// fences during compilation.
+#if !defined(USE_ROCM)
  __syncwarp();
+#endif
  // s = 16
  if (min_dists[tid] > min_dists[tid + 16]) {
    min_idxs[tid] = min_idxs[tid + 16];
    min_dists[tid] = min_dists[tid + 16];
  }
+#if !defined(USE_ROCM)
  __syncwarp();
+#endif
  // s = 8
  if (min_dists[tid] > min_dists[tid + 8]) {
    min_idxs[tid] = min_idxs[tid + 8];
    min_dists[tid] = min_dists[tid + 8];
  }
+#if !defined(USE_ROCM)
  __syncwarp();
+#endif
  // s = 4
  if (min_dists[tid] > min_dists[tid + 4]) {
    min_idxs[tid] = min_idxs[tid + 4];
    min_dists[tid] = min_dists[tid + 4];
  }
+#if !defined(USE_ROCM)
  __syncwarp();
+#endif
  // s = 2
  if (min_dists[tid] > min_dists[tid + 2]) {
    min_idxs[tid] = min_idxs[tid + 2];
    min_dists[tid] = min_dists[tid + 2];
  }
+#if !defined(USE_ROCM)
  __syncwarp();
+#endif
  // s = 1
  if (min_dists[tid] > min_dists[tid + 1]) {
    min_idxs[tid] = min_idxs[tid + 1];
    min_dists[tid] = min_dists[tid + 1];
  }
+#if !defined(USE_ROCM)
  __syncwarp();
+#endif
 }

 template <typename scalar_t>
@@ -65,30 +79,42 @@ __device__ void WarpReduceMax(
    dists[tid] = dists[tid + 32];
    dists_idx[tid] = dists_idx[tid + 32];
  }
+#if !defined(USE_ROCM)
  __syncwarp();
+#endif
  if (dists[tid] < dists[tid + 16]) {
    dists[tid] = dists[tid + 16];
    dists_idx[tid] = dists_idx[tid + 16];
  }
+#if !defined(USE_ROCM)
  __syncwarp();
+#endif
  if (dists[tid] < dists[tid + 8]) {
    dists[tid] = dists[tid + 8];
    dists_idx[tid] = dists_idx[tid + 8];
  }
+#if !defined(USE_ROCM)
  __syncwarp();
+#endif
  if (dists[tid] < dists[tid + 4]) {
    dists[tid] = dists[tid + 4];
    dists_idx[tid] = dists_idx[tid + 4];
  }
+#if !defined(USE_ROCM)
  __syncwarp();
+#endif
  if (dists[tid] < dists[tid + 2]) {
    dists[tid] = dists[tid + 2];
    dists_idx[tid] = dists_idx[tid + 2];
  }
+#if !defined(USE_ROCM)
  __syncwarp();
+#endif
  if (dists[tid] < dists[tid + 1]) {
    dists[tid] = dists[tid + 1];
    dists_idx[tid] = dists_idx[tid + 1];
  }
+#if !defined(USE_ROCM)
  __syncwarp();
+#endif
 }
--- a/pytorch3d/datasets/shapenet/shapenet_core.py
+++ b/pytorch3d/datasets/shapenet/shapenet_core.py
@@ -83,7 +83,7 @@ class ShapeNetCore(ShapeNetBase):  # pragma: no cover
                ):
                    synset_set.add(synset)
                elif (synset in self.synset_inv.keys()) and (
-                    (path.isdir(path.join(data_dir, self.synset_inv[synset])))
+                    path.isdir(path.join(data_dir, self.synset_inv[synset]))
                ):
                    synset_set.add(self.synset_inv[synset])
                else:
--- a/pytorch3d/datasets/utils.py
+++ b/pytorch3d/datasets/utils.py
@@ -36,7 +36,6 @@ def collate_batched_meshes(batch: List[Dict]):  # pragma: no cover

    collated_dict["mesh"] = None
    if {"verts", "faces"}.issubset(collated_dict.keys()):
-
        textures = None
        if "textures" in collated_dict:
            textures = TexturesAtlas(atlas=collated_dict["textures"])
--- a/pytorch3d/implicitron/dataset/data_source.py
+++ b/pytorch3d/implicitron/dataset/data_source.py
@@ -41,7 +41,7 @@ class DataSourceBase(ReplaceableBase):


@registry.register
-class ImplicitronDataSource(DataSourceBase):  # pyre-ignore[13]
+class ImplicitronDataSource(DataSourceBase):
    """
    Represents the data used in Implicitron. This is the only implementation
    of DataSourceBase provided.
@@ -52,8 +52,11 @@ class ImplicitronDataSource(DataSourceBase):  # pyre-ignore[13]
        data_loader_map_provider_class_type: identifies type for data_loader_map_provider.
    """

+    # pyre-fixme[13]: Attribute `dataset_map_provider` is never initialized.
    dataset_map_provider: DatasetMapProviderBase
+    # pyre-fixme[13]: Attribute `dataset_map_provider_class_type` is never initialized.
    dataset_map_provider_class_type: str
+    # pyre-fixme[13]: Attribute `data_loader_map_provider` is never initialized.
    data_loader_map_provider: DataLoaderMapProviderBase
    data_loader_map_provider_class_type: str = "SequenceDataLoaderMapProvider"

--- a/pytorch3d/implicitron/dataset/frame_data.py
+++ b/pytorch3d/implicitron/dataset/frame_data.py
@@ -26,7 +26,7 @@ from typing import (
 import numpy as np
 import torch

-from pytorch3d.implicitron.dataset import types
+from pytorch3d.implicitron.dataset import orm_types, types
 from pytorch3d.implicitron.dataset.utils import (
    adjust_camera_to_bbox_crop_,
    adjust_camera_to_image_scale_,
@@ -48,8 +48,12 @@ from pytorch3d.implicitron.dataset.utils import (
 from pytorch3d.implicitron.tools.config import registry, ReplaceableBase
 from pytorch3d.renderer.camera_utils import join_cameras_as_batch
 from pytorch3d.renderer.cameras import CamerasBase, PerspectiveCameras
+from pytorch3d.structures.meshes import join_meshes_as_batch, Meshes
 from pytorch3d.structures.pointclouds import join_pointclouds_as_batch, Pointclouds

+FrameAnnotationT = types.FrameAnnotation | orm_types.SqlFrameAnnotation
+SequenceAnnotationT = types.SequenceAnnotation | orm_types.SqlSequenceAnnotation
+

@dataclass
 class FrameData(Mapping[str, Any]):
@@ -122,9 +126,9 @@ class FrameData(Mapping[str, Any]):
        meta: A dict for storing additional frame information.
    """

-    frame_number: Optional[torch.LongTensor]
-    sequence_name: Union[str, List[str]]
-    sequence_category: Union[str, List[str]]
+    frame_number: Optional[torch.LongTensor] = None
+    sequence_name: Union[str, List[str]] = ""
+    sequence_category: Union[str, List[str]] = ""
    frame_timestamp: Optional[torch.Tensor] = None
    image_size_hw: Optional[torch.LongTensor] = None
    effective_image_size_hw: Optional[torch.LongTensor] = None
@@ -155,7 +159,7 @@ class FrameData(Mapping[str, Any]):
        new_params = {}
        for field_name in iter(self):
            value = getattr(self, field_name)
-            if isinstance(value, (torch.Tensor, Pointclouds, CamerasBase)):
+            if isinstance(value, (torch.Tensor, Pointclouds, CamerasBase, Meshes)):
                new_params[field_name] = value.to(*args, **kwargs)
            else:
                new_params[field_name] = value
@@ -276,6 +280,7 @@ class FrameData(Mapping[str, Any]):
            image_size_hw=tuple(self.effective_image_size_hw),  # pyre-ignore
        )
        crop_bbox_xywh = bbox_xyxy_to_xywh(clamp_bbox_xyxy)
+        self.crop_bbox_xywh = crop_bbox_xywh

        if self.fg_probability is not None:
            self.fg_probability = crop_around_box(
@@ -416,7 +421,6 @@ class FrameData(Mapping[str, Any]):
            for f in fields(elem):
                if not f.init:
                    continue
-
                list_values = override_fields.get(
                    f.name, [getattr(d, f.name) for d in batch]
                )
@@ -425,7 +429,7 @@ class FrameData(Mapping[str, Any]):
                    if all(list_value is not None for list_value in list_values)
                    else None
                )
-            return cls(**collated)
+            return type(elem)(**collated)

        elif isinstance(elem, Pointclouds):
            return join_pointclouds_as_batch(batch)
@@ -433,8 +437,10 @@ class FrameData(Mapping[str, Any]):
        elif isinstance(elem, CamerasBase):
            # TODO: don't store K; enforce working in NDC space
            return join_cameras_as_batch(batch)
+        elif isinstance(elem, Meshes):
+            return join_meshes_as_batch(batch)
        else:
-            return torch.utils.data._utils.collate.default_collate(batch)
+            return torch.utils.data.dataloader.default_collate(batch)


 FrameDataSubtype = TypeVar("FrameDataSubtype", bound=FrameData)
@@ -453,8 +459,8 @@ class FrameDataBuilderBase(ReplaceableBase, Generic[FrameDataSubtype], ABC):
    @abstractmethod
    def build(
        self,
-        frame_annotation: types.FrameAnnotation,
-        sequence_annotation: types.SequenceAnnotation,
+        frame_annotation: FrameAnnotationT,
+        sequence_annotation: SequenceAnnotationT,
        *,
        load_blobs: bool = True,
        **kwargs,
@@ -540,8 +546,8 @@ class GenericFrameDataBuilder(FrameDataBuilderBase[FrameDataSubtype], ABC):

    def build(
        self,
-        frame_annotation: types.FrameAnnotation,
-        sequence_annotation: types.SequenceAnnotation,
+        frame_annotation: FrameAnnotationT,
+        sequence_annotation: SequenceAnnotationT,
        *,
        load_blobs: bool = True,
        **kwargs,
@@ -585,58 +591,81 @@ class GenericFrameDataBuilder(FrameDataBuilderBase[FrameDataSubtype], ABC):
            ),
        )

-        fg_mask_np: Optional[np.ndarray] = None
+        dataset_root = self.dataset_root
        mask_annotation = frame_annotation.mask
-        if mask_annotation is not None:
-            if load_blobs and self.load_masks:
-                fg_mask_np, mask_path = self._load_fg_probability(frame_annotation)
+        depth_annotation = frame_annotation.depth
+        image_path: str | None = None
+        mask_path: str | None = None
+        depth_path: str | None = None
+        pcl_path: str | None = None
+        if dataset_root is not None:  # set all paths even if we won’t load blobs
+            if frame_annotation.image.path is not None:
+                image_path = os.path.join(dataset_root, frame_annotation.image.path)
+                frame_data.image_path = image_path
+
+            if mask_annotation is not None and mask_annotation.path:
+                mask_path = os.path.join(dataset_root, mask_annotation.path)
                frame_data.mask_path = mask_path
+
+            if depth_annotation is not None and depth_annotation.path is not None:
+                depth_path = os.path.join(dataset_root, depth_annotation.path)
+                frame_data.depth_path = depth_path
+
+            if point_cloud is not None:
+                pcl_path = os.path.join(dataset_root, point_cloud.path)
+                frame_data.sequence_point_cloud_path = pcl_path
+
+        fg_mask_np: np.ndarray | None = None
+        bbox_xywh: tuple[float, float, float, float] | None = None
+
+        if mask_annotation is not None:
+            if load_blobs and self.load_masks and mask_path:
+                fg_mask_np = self._load_fg_probability(frame_annotation, mask_path)
                frame_data.fg_probability = safe_as_tensor(fg_mask_np, torch.float)

            bbox_xywh = mask_annotation.bounding_box_xywh
-            if bbox_xywh is None and fg_mask_np is not None:
-                bbox_xywh = get_bbox_from_mask(fg_mask_np, self.box_crop_mask_thr)
-
-            frame_data.bbox_xywh = safe_as_tensor(bbox_xywh, torch.float)

        if frame_annotation.image is not None:
            image_size_hw = safe_as_tensor(frame_annotation.image.size, torch.long)
            frame_data.image_size_hw = image_size_hw  # original image size
            # image size after crop/resize
            frame_data.effective_image_size_hw = image_size_hw
-            image_path = None
-            dataset_root = self.dataset_root
-            if frame_annotation.image.path is not None and dataset_root is not None:
-                image_path = os.path.join(dataset_root, frame_annotation.image.path)
-                frame_data.image_path = image_path

            if load_blobs and self.load_images:
                if image_path is None:
                    raise ValueError("Image path is required to load images.")

-                image_np = load_image(self._local_path(image_path))
+                no_mask = fg_mask_np is None  # didn’t read the mask file
+                image_np = load_image(
+                    self._local_path(image_path), try_read_alpha=no_mask
+                )
+                if image_np.shape[0] == 4:  # RGBA image
+                    if no_mask:
+                        fg_mask_np = image_np[3:]
+                        frame_data.fg_probability = safe_as_tensor(
+                            fg_mask_np, torch.float
+                        )
+
+                    image_np = image_np[:3]
+
                frame_data.image_rgb = self._postprocess_image(
                    image_np, frame_annotation.image.size, frame_data.fg_probability
                )

-        if (
-            load_blobs
-            and self.load_depths
-            and frame_annotation.depth is not None
-            and frame_annotation.depth.path is not None
-        ):
-            (
-                frame_data.depth_map,
-                frame_data.depth_path,
-                frame_data.depth_mask,
-            ) = self._load_mask_depth(frame_annotation, fg_mask_np)
+        if bbox_xywh is None and fg_mask_np is not None:
+            bbox_xywh = get_bbox_from_mask(fg_mask_np, self.box_crop_mask_thr)
+        frame_data.bbox_xywh = safe_as_tensor(bbox_xywh, torch.float)
+
+        if load_blobs and self.load_depths and depth_path is not None:
+            frame_data.depth_map, frame_data.depth_mask = self._load_mask_depth(
+                frame_annotation, depth_path, fg_mask_np
+            )

        if load_blobs and self.load_point_clouds and point_cloud is not None:
-            pcl_path = self._fix_point_cloud_path(point_cloud.path)
+            assert pcl_path is not None
            frame_data.sequence_point_cloud = load_pointcloud(
                self._local_path(pcl_path), max_points=self.max_points
            )
-            frame_data.sequence_point_cloud_path = pcl_path

        if frame_annotation.viewpoint is not None:
            frame_data.camera = self._get_pytorch3d_camera(frame_annotation)
@@ -652,18 +681,14 @@ class GenericFrameDataBuilder(FrameDataBuilderBase[FrameDataSubtype], ABC):

        return frame_data

-    def _load_fg_probability(
-        self, entry: types.FrameAnnotation
-    ) -> Tuple[np.ndarray, str]:
-        assert self.dataset_root is not None and entry.mask is not None
-        full_path = os.path.join(self.dataset_root, entry.mask.path)
-        fg_probability = load_mask(self._local_path(full_path))
+    def _load_fg_probability(self, entry: FrameAnnotationT, path: str) -> np.ndarray:
+        fg_probability = load_mask(self._local_path(path))
        if fg_probability.shape[-2:] != entry.image.size:
            raise ValueError(
                f"bad mask size: {fg_probability.shape[-2:]} vs {entry.image.size}!"
            )

-        return fg_probability, full_path
+        return fg_probability

    def _postprocess_image(
        self,
@@ -684,14 +709,14 @@ class GenericFrameDataBuilder(FrameDataBuilderBase[FrameDataSubtype], ABC):

    def _load_mask_depth(
        self,
-        entry: types.FrameAnnotation,
+        entry: FrameAnnotationT,
+        path: str,
        fg_mask: Optional[np.ndarray],
-    ) -> Tuple[torch.Tensor, str, torch.Tensor]:
+    ) -> tuple[torch.Tensor, torch.Tensor]:
        entry_depth = entry.depth
        dataset_root = self.dataset_root
        assert dataset_root is not None
-        assert entry_depth is not None and entry_depth.path is not None
-        path = os.path.join(dataset_root, entry_depth.path)
+        assert entry_depth is not None
        depth_map = load_depth(self._local_path(path), entry_depth.scale_adjustment)

        if self.mask_depths:
@@ -705,11 +730,11 @@ class GenericFrameDataBuilder(FrameDataBuilderBase[FrameDataSubtype], ABC):
        else:
            depth_mask = (depth_map > 0.0).astype(np.float32)

-        return torch.tensor(depth_map), path, torch.tensor(depth_mask)
+        return torch.tensor(depth_map), torch.tensor(depth_mask)

    def _get_pytorch3d_camera(
        self,
-        entry: types.FrameAnnotation,
+        entry: FrameAnnotationT,
    ) -> PerspectiveCameras:
        entry_viewpoint = entry.viewpoint
        assert entry_viewpoint is not None
@@ -738,19 +763,6 @@ class GenericFrameDataBuilder(FrameDataBuilderBase[FrameDataSubtype], ABC):
            T=torch.tensor(entry_viewpoint.T, dtype=torch.float)[None],
        )

-    def _fix_point_cloud_path(self, path: str) -> str:
-        """
-        Fix up a point cloud path from the dataset.
-        Some files in Co3Dv2 have an accidental absolute path stored.
-        """
-        unwanted_prefix = (
-            "/large_experiments/p3/replay/datasets/co3d/co3d45k_220512/export_v23/"
-        )
-        if path.startswith(unwanted_prefix):
-            path = path[len(unwanted_prefix) :]
-        assert self.dataset_root is not None
-        return os.path.join(self.dataset_root, path)
-
    def _local_path(self, path: str) -> str:
        if self.path_manager is None:
            return path
--- a/pytorch3d/implicitron/dataset/json_index_dataset_map_provider.py
+++ b/pytorch3d/implicitron/dataset/json_index_dataset_map_provider.py
@@ -66,7 +66,7 @@ _NEED_CONTROL: Tuple[str, ...] = (


@registry.register
-class JsonIndexDatasetMapProvider(DatasetMapProviderBase):  # pyre-ignore [13]
+class JsonIndexDatasetMapProvider(DatasetMapProviderBase):
    """
    Generates the training / validation and testing dataset objects for
    a dataset laid out on disk like Co3D, with annotations in json files.
@@ -95,6 +95,7 @@ class JsonIndexDatasetMapProvider(DatasetMapProviderBase):  # pyre-ignore [13]
        path_manager_factory_class_type: The class type of `path_manager_factory`.
    """

+    # pyre-fixme[13]: Attribute `category` is never initialized.
    category: str
    task_str: str = "singlesequence"
    dataset_root: str = _CO3D_DATASET_ROOT
@@ -104,8 +105,10 @@ class JsonIndexDatasetMapProvider(DatasetMapProviderBase):  # pyre-ignore [13]
    test_restrict_sequence_id: int = -1
    assert_single_seq: bool = False
    only_test_set: bool = False
+    # pyre-fixme[13]: Attribute `dataset` is never initialized.
    dataset: JsonIndexDataset
    dataset_class_type: str = "JsonIndexDataset"
+    # pyre-fixme[13]: Attribute `path_manager_factory` is never initialized.
    path_manager_factory: PathManagerFactory
    path_manager_factory_class_type: str = "PathManagerFactory"

--- a/pytorch3d/implicitron/dataset/json_index_dataset_map_provider_v2.py
+++ b/pytorch3d/implicitron/dataset/json_index_dataset_map_provider_v2.py
@@ -56,7 +56,7 @@ logger = logging.getLogger(__name__)


@registry.register
-class JsonIndexDatasetMapProviderV2(DatasetMapProviderBase):  # pyre-ignore [13]
+class JsonIndexDatasetMapProviderV2(DatasetMapProviderBase):
    """
    Generates the training, validation, and testing dataset objects for
    a dataset laid out on disk like CO3Dv2, with annotations in gzipped json files.
@@ -171,7 +171,9 @@ class JsonIndexDatasetMapProviderV2(DatasetMapProviderBase):  # pyre-ignore [13]
        path_manager_factory_class_type: The class type of `path_manager_factory`.
    """

+    # pyre-fixme[13]: Attribute `category` is never initialized.
    category: str
+    # pyre-fixme[13]: Attribute `subset_name` is never initialized.
    subset_name: str
    dataset_root: str = _CO3DV2_DATASET_ROOT

@@ -183,8 +185,10 @@ class JsonIndexDatasetMapProviderV2(DatasetMapProviderBase):  # pyre-ignore [13]
    n_known_frames_for_test: int = 0

    dataset_class_type: str = "JsonIndexDataset"
+    # pyre-fixme[13]: Attribute `dataset` is never initialized.
    dataset: JsonIndexDataset

+    # pyre-fixme[13]: Attribute `path_manager_factory` is never initialized.
    path_manager_factory: PathManagerFactory
    path_manager_factory_class_type: str = "PathManagerFactory"

@@ -218,7 +222,6 @@ class JsonIndexDatasetMapProviderV2(DatasetMapProviderBase):  # pyre-ignore [13]
        self.dataset_map = dataset_map

    def _load_category(self, category: str) -> DatasetMap:
-
        frame_file = os.path.join(self.dataset_root, category, "frame_annotations.jgz")
        sequence_file = os.path.join(
            self.dataset_root, category, "sequence_annotations.jgz"
--- a/pytorch3d/implicitron/dataset/load_llff.py
+++ b/pytorch3d/implicitron/dataset/load_llff.py
@@ -75,7 +75,6 @@ def _minify(basedir, path_manager, factors=(), resolutions=()):
 def _load_data(
    basedir, factor=None, width=None, height=None, load_imgs=True, path_manager=None
 ):
-
    poses_arr = np.load(
        _local_path(path_manager, os.path.join(basedir, "poses_bounds.npy"))
    )
@@ -164,7 +163,6 @@ def ptstocam(pts, c2w):


 def poses_avg(poses):
-
    hwf = poses[0, :3, -1:]

    center = poses[:, :3, 3].mean(0)
@@ -192,7 +190,6 @@ def render_path_spiral(c2w, up, rads, focal, zdelta, zrate, rots, N):


 def recenter_poses(poses):
-
    poses_ = poses + 0
    bottom = np.reshape([0, 0, 0, 1.0], [1, 4])
    c2w = poses_avg(poses)
@@ -256,7 +253,6 @@ def spherify_poses(poses, bds):
    new_poses = []

    for th in np.linspace(0.0, 2.0 * np.pi, 120):
-
        camorigin = np.array([radcircle * np.cos(th), radcircle * np.sin(th), zh])
        up = np.array([0, 0, -1.0])

@@ -311,7 +307,6 @@ def load_llff_data(
    path_zflat=False,
    path_manager=None,
 ):
-
    poses, bds, imgs = _load_data(
        basedir, factor=factor, path_manager=path_manager
    )  # factor=8 downsamples original imgs by 8x
--- a/pytorch3d/implicitron/dataset/orm_types.py
+++ b/pytorch3d/implicitron/dataset/orm_types.py
@@ -4,6 +4,8 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.

+# pyre-unsafe
+
 # This functionality requires SQLAlchemy 2.0 or later.

 import math
--- a/pytorch3d/implicitron/dataset/rendered_mesh_dataset_map_provider.py
+++ b/pytorch3d/implicitron/dataset/rendered_mesh_dataset_map_provider.py
@@ -32,7 +32,7 @@ from .utils import DATASET_TYPE_KNOWN


@registry.register
-class RenderedMeshDatasetMapProvider(DatasetMapProviderBase):  # pyre-ignore [13]
+class RenderedMeshDatasetMapProvider(DatasetMapProviderBase):
    """
    A simple single-scene dataset based on PyTorch3D renders of a mesh.
    Provides `num_views` renders of the mesh as train, with no val
@@ -76,6 +76,7 @@ class RenderedMeshDatasetMapProvider(DatasetMapProviderBase):  # pyre-ignore [13
    resolution: int = 128
    use_point_light: bool = True
    gpu_idx: Optional[int] = 0
+    # pyre-fixme[13]: Attribute `path_manager_factory` is never initialized.
    path_manager_factory: PathManagerFactory
    path_manager_factory_class_type: str = "PathManagerFactory"

--- a/pytorch3d/implicitron/dataset/single_sequence_dataset.py
+++ b/pytorch3d/implicitron/dataset/single_sequence_dataset.py
@@ -83,7 +83,6 @@ class SingleSceneDataset(DatasetBase, Configurable):
        return self.eval_batches


-# pyre-fixme[13]: Uninitialized attribute
 class SingleSceneDatasetMapProviderBase(DatasetMapProviderBase):
    """
    Base for provider of data for one scene from LLFF or blender datasets.
@@ -100,8 +99,11 @@ class SingleSceneDatasetMapProviderBase(DatasetMapProviderBase):
            testing frame.
    """

+    # pyre-fixme[13]: Attribute `base_dir` is never initialized.
    base_dir: str
+    # pyre-fixme[13]: Attribute `object_name` is never initialized.
    object_name: str
+    # pyre-fixme[13]: Attribute `path_manager_factory` is never initialized.
    path_manager_factory: PathManagerFactory
    path_manager_factory_class_type: str = "PathManagerFactory"
    n_known_frames_for_test: Optional[int] = None
--- a/pytorch3d/implicitron/dataset/sql_dataset.py
+++ b/pytorch3d/implicitron/dataset/sql_dataset.py
@@ -4,11 +4,15 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.

+# pyre-unsafe
+
 import hashlib
 import json
 import logging
 import os
-from dataclasses import dataclass
+
+import urllib
+from dataclasses import dataclass, Field, field
 from typing import (
    Any,
    ClassVar,
@@ -29,17 +33,18 @@ import sqlalchemy as sa
 import torch
 from pytorch3d.implicitron.dataset.dataset_base import DatasetBase

-from pytorch3d.implicitron.dataset.frame_data import (  # noqa
+from pytorch3d.implicitron.dataset.frame_data import (
    FrameData,
-    FrameDataBuilder,
+    FrameDataBuilder,  # noqa
    FrameDataBuilderBase,
 )
+
 from pytorch3d.implicitron.tools.config import (
    registry,
    ReplaceableBase,
    run_auto_creation,
 )
-from sqlalchemy.orm import Session
+from sqlalchemy.orm import scoped_session, Session, sessionmaker

 from .orm_types import SqlFrameAnnotation, SqlSequenceAnnotation

@@ -51,7 +56,7 @@ _SET_LISTS_TABLE: str = "set_lists"


@registry.register
-class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
+class SqlIndexDataset(DatasetBase, ReplaceableBase):
    """
    A dataset with annotations stored as SQLite tables. This is an index-based dataset.
    The length is returned after all sequence and frame filters are applied (see param
@@ -88,6 +93,7 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
            engine verbatim. Don’t expose it to end users of your application!
        pick_categories: Restrict the dataset to the given list of categories.
        pick_sequences: A Sequence of sequence names to restrict the dataset to.
+        pick_sequences_sql_clause: Custom SQL WHERE clause to constrain sequence annotations.
        exclude_sequences: A Sequence of the names of the sequences to exclude.
        limit_sequences_per_category_to: Limit the dataset to the first up to N
            sequences within each category (applies after all other sequence filters
@@ -102,9 +108,16 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
            more frames than that; applied after other frame-level filters.
        seed: The seed of the random generator sampling `n_frames_per_sequence`
            random frames per sequence.
+        preload_metadata: If True, the metadata is preloaded into memory.
+        precompute_seq_to_idx: If True, precomputes the mapping from sequence name to indices.
+        scoped_session: If True, allows different parts of the code to share
+            a global session to access the database.
    """

    frame_annotations_type: ClassVar[Type[SqlFrameAnnotation]] = SqlFrameAnnotation
+    sequence_annotations_type: ClassVar[Type[SqlSequenceAnnotation]] = (
+        SqlSequenceAnnotation
+    )

    sqlite_metadata_file: str = ""
    dataset_root: Optional[str] = None
@@ -117,6 +130,7 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
    pick_categories: Tuple[str, ...] = ()

    pick_sequences: Tuple[str, ...] = ()
+    pick_sequences_sql_clause: Optional[str] = None
    exclude_sequences: Tuple[str, ...] = ()
    limit_sequences_per_category_to: int = 0
    limit_sequences_to: int = 0
@@ -124,12 +138,22 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
    n_frames_per_sequence: int = -1
    seed: int = 0
    remove_empty_masks_poll_whole_table_threshold: int = 300_000
+    preload_metadata: bool = False
+    precompute_seq_to_idx: bool = False
    # we set it manually in the constructor
-    # _index: pd.DataFrame = field(init=False)
+    _index: pd.DataFrame = field(init=False, metadata={"omegaconf_ignore": True})
+    _sql_engine: sa.engine.Engine = field(
+        init=False, metadata={"omegaconf_ignore": True}
+    )
+    eval_batches: Optional[List[Any]] = field(
+        init=False, metadata={"omegaconf_ignore": True}
+    )

-    frame_data_builder: FrameDataBuilderBase
+    frame_data_builder: FrameDataBuilderBase  # pyre-ignore[13]
    frame_data_builder_class_type: str = "FrameDataBuilder"

+    scoped_session: bool = False
+
    def __post_init__(self) -> None:
        if sa.__version__ < "2.0":
            raise ImportError("This class requires SQL Alchemy 2.0 or later")
@@ -138,19 +162,28 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
            raise ValueError("sqlite_metadata_file must be set")

        if self.dataset_root:
-            frame_builder_type = self.frame_data_builder_class_type
-            getattr(self, f"frame_data_builder_{frame_builder_type}_args")[
-                "dataset_root"
-            ] = self.dataset_root
+            frame_args = f"frame_data_builder_{self.frame_data_builder_class_type}_args"
+            getattr(self, frame_args)["dataset_root"] = self.dataset_root
+            getattr(self, frame_args)["path_manager"] = self.path_manager

        run_auto_creation(self)
-        self.frame_data_builder.path_manager = self.path_manager

-        # pyre-ignore  # NOTE: sqlite-specific args (read-only mode).
+        if self.path_manager is not None:
+            self.sqlite_metadata_file = self.path_manager.get_local_path(
+                self.sqlite_metadata_file
+            )
+            self.subset_lists_file = self.path_manager.get_local_path(
+                self.subset_lists_file
+            )
+
+        # NOTE: sqlite-specific args (read-only mode).
        self._sql_engine = sa.create_engine(
-            f"sqlite:///file:{self.sqlite_metadata_file}?mode=ro&uri=true"
+            f"sqlite:///file:{urllib.parse.quote(self.sqlite_metadata_file)}?mode=ro&uri=true"
        )

+        if self.preload_metadata:
+            self._sql_engine = self._preload_database(self._sql_engine)
+
        sequences = self._get_filtered_sequences_if_any()

        if self.subsets:
@@ -166,16 +199,29 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
        if len(index) == 0:
            raise ValueError(f"There are no frames in the subsets: {self.subsets}!")

-        self._index = index.set_index(["sequence_name", "frame_number"])  # pyre-ignore
+        self._index = index.set_index(["sequence_name", "frame_number"])

-        self.eval_batches = None  # pyre-ignore
+        self.eval_batches = None
        if self.eval_batches_file:
            self.eval_batches = self._load_filter_eval_batches()

        logger.info(str(self))

+        if self.scoped_session:
+            self._session_factory = sessionmaker(bind=self._sql_engine)  # pyre-ignore
+
+        if self.precompute_seq_to_idx:
+            # This is deprecated and will be removed in the future.
+            # After we backport https://github.com/facebookresearch/uco3d/pull/3
+            logger.warning(
+                "Using precompute_seq_to_idx is deprecated and will be removed in the future."
+            )
+            self._index["rowid"] = np.arange(len(self._index))
+            groupby = self._index.groupby("sequence_name", sort=False)["rowid"]
+            self._seq_to_indices = dict(groupby.apply(list))  # pyre-ignore
+            del self._index["rowid"]
+
    def __len__(self) -> int:
-        # pyre-ignore[16]
        return len(self._index)

    def __getitem__(self, frame_idx: Union[int, Tuple[str, int]]) -> FrameData:
@@ -232,12 +278,18 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
            self.frame_annotations_type.frame_number
            == int(frame),  # cast from np.int64
        )
-        seq_stmt = sa.select(SqlSequenceAnnotation).where(
-            SqlSequenceAnnotation.sequence_name == seq
+        seq_stmt = sa.select(self.sequence_annotations_type).where(
+            self.sequence_annotations_type.sequence_name == seq
        )
-        with Session(self._sql_engine) as session:
-            entry = session.scalars(stmt).one()
-            seq_metadata = session.scalars(seq_stmt).one()
+        if self.scoped_session:
+            # pyre-ignore
+            with scoped_session(self._session_factory)() as session:
+                entry = session.scalars(stmt).one()
+                seq_metadata = session.scalars(seq_stmt).one()
+        else:
+            with Session(self._sql_engine) as session:
+                entry = session.scalars(stmt).one()
+                seq_metadata = session.scalars(seq_stmt).one()

        assert entry.image.path == self._index.loc[(seq, frame), "_image_path"]

@@ -250,7 +302,6 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
        return frame_data

    def __str__(self) -> str:
-        # pyre-ignore[16]
        return f"SqlIndexDataset #frames={len(self._index)}"

    def sequence_names(self) -> Iterable[str]:
@@ -260,9 +311,10 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
    # override
    def category_to_sequence_names(self) -> Dict[str, List[str]]:
        stmt = sa.select(
-            SqlSequenceAnnotation.category, SqlSequenceAnnotation.sequence_name
+            self.sequence_annotations_type.category,
+            self.sequence_annotations_type.sequence_name,
        ).where(  # we limit results to sequences that have frames after all filters
-            SqlSequenceAnnotation.sequence_name.in_(self.sequence_names())
+            self.sequence_annotations_type.sequence_name.in_(self.sequence_names())
        )
        with self._sql_engine.connect() as connection:
            cat_to_seqs = pd.read_sql(stmt, connection)
@@ -335,17 +387,31 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
        rows = self._index.index.get_loc(seq_name)
        if isinstance(rows, slice):
            assert rows.stop is not None, "Unexpected result from pandas"
-            rows = range(rows.start or 0, rows.stop, rows.step or 1)
+            rows_seq = range(rows.start or 0, rows.stop, rows.step or 1)
        else:
-            rows = np.where(rows)[0]
+            rows_seq = list(np.where(rows)[0])

        index_slice, idx = self._get_frame_no_coalesced_ts_by_row_indices(
-            rows, seq_name, subset_filter
+            rows_seq, seq_name, subset_filter
        )
        index_slice["idx"] = idx

        yield from index_slice.itertuples(index=False)

+    # override
+    def sequence_indices_in_order(
+        self, seq_name: str, subset_filter: Optional[Sequence[str]] = None
+    ) -> Iterator[int]:
+        """Same as `sequence_frames_in_order` but returns the iterator over
+        only dataset indices.
+        """
+        if self.precompute_seq_to_idx and subset_filter is None:
+            # pyre-ignore
+            yield from self._seq_to_indices[seq_name]
+        else:
+            for _, _, idx in self.sequence_frames_in_order(seq_name, subset_filter):
+                yield idx
+
    # override
    def get_eval_batches(self) -> Optional[List[Any]]:
        """
@@ -379,11 +445,35 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
            or self.limit_sequences_to > 0
            or self.limit_sequences_per_category_to > 0
            or len(self.pick_sequences) > 0
+            or self.pick_sequences_sql_clause is not None
            or len(self.exclude_sequences) > 0
            or len(self.pick_categories) > 0
            or self.n_frames_per_sequence > 0
        )

+    def _preload_database(
+        self, source_engine: sa.engine.base.Engine
+    ) -> sa.engine.base.Engine:
+        destination_engine = sa.create_engine("sqlite:///:memory:")
+        metadata = sa.MetaData()
+        metadata.reflect(bind=source_engine)
+        metadata.create_all(bind=destination_engine)
+
+        with source_engine.connect() as source_conn:
+            with destination_engine.connect() as destination_conn:
+                for table_obj in metadata.tables.values():
+                    # Select all rows from the source table
+                    source_rows = source_conn.execute(table_obj.select())
+
+                    # Insert rows into the destination table
+                    for row in source_rows:
+                        destination_conn.execute(table_obj.insert().values(row))
+
+                    # Commit the changes for each table
+                    destination_conn.commit()
+
+        return destination_engine
+
    def _get_filtered_sequences_if_any(self) -> Optional[pd.Series]:
        # maximum possible filter (if limit_sequences_per_category_to == 0):
        # WHERE category IN 'self.pick_categories'
@@ -396,19 +486,22 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
            *self._get_pick_filters(),
            *self._get_exclude_filters(),
        ]
+        if self.pick_sequences_sql_clause:
+            print("Applying the custom SQL clause.")
+            where_conditions.append(sa.text(self.pick_sequences_sql_clause))

        def add_where(stmt):
            return stmt.where(*where_conditions) if where_conditions else stmt

        if self.limit_sequences_per_category_to <= 0:
-            stmt = add_where(sa.select(SqlSequenceAnnotation.sequence_name))
+            stmt = add_where(sa.select(self.sequence_annotations_type.sequence_name))
        else:
            subquery = sa.select(
-                SqlSequenceAnnotation.sequence_name,
+                self.sequence_annotations_type.sequence_name,
                sa.func.row_number()
                .over(
                    order_by=sa.text("ROWID"),  # NOTE: ROWID is SQLite-specific
-                    partition_by=SqlSequenceAnnotation.category,
+                    partition_by=self.sequence_annotations_type.category,
                )
                .label("row_number"),
            )
@@ -444,31 +537,34 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
            return []

        logger.info(f"Limiting dataset to categories: {self.pick_categories}")
-        return [SqlSequenceAnnotation.category.in_(self.pick_categories)]
+        return [self.sequence_annotations_type.category.in_(self.pick_categories)]

    def _get_pick_filters(self) -> List[sa.ColumnElement]:
        if not self.pick_sequences:
            return []

        logger.info(f"Limiting dataset to sequences: {self.pick_sequences}")
-        return [SqlSequenceAnnotation.sequence_name.in_(self.pick_sequences)]
+        return [self.sequence_annotations_type.sequence_name.in_(self.pick_sequences)]

    def _get_exclude_filters(self) -> List[sa.ColumnOperators]:
        if not self.exclude_sequences:
            return []

        logger.info(f"Removing sequences from the dataset: {self.exclude_sequences}")
-        return [SqlSequenceAnnotation.sequence_name.notin_(self.exclude_sequences)]
+        return [
+            self.sequence_annotations_type.sequence_name.notin_(self.exclude_sequences)
+        ]

    def _load_subsets_from_json(self, subset_lists_path: str) -> pd.DataFrame:
-        assert self.subsets is not None
+        subsets = self.subsets
+        assert subsets is not None
        with open(subset_lists_path, "r") as f:
            subset_to_seq_frame = json.load(f)

        seq_frame_list = sum(
            (
                [(*row, subset) for row in subset_to_seq_frame[subset]]
-                for subset in self.subsets
+                for subset in subsets
            ),
            [],
        )
@@ -522,7 +618,7 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
                stmt = sa.select(
                    self.frame_annotations_type.sequence_name,
                    self.frame_annotations_type.frame_number,
-                ).where(self.frame_annotations_type._mask_mass == 0)
+                ).where(self.frame_annotations_type._mask_mass == 0)  # pyre-ignore[16]
                with Session(self._sql_engine) as session:
                    to_remove = session.execute(stmt).all()

@@ -586,7 +682,7 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
        stmt = sa.select(
            self.frame_annotations_type.sequence_name,
            self.frame_annotations_type.frame_number,
-            self.frame_annotations_type._image_path,
+            self.frame_annotations_type._image_path,  # pyre-ignore[16]
            sa.null().label("subset"),
        )
        where_conditions = []
@@ -600,7 +696,7 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
            logger.info("  excluding samples with empty masks")
            where_conditions.append(
                sa.or_(
-                    self.frame_annotations_type._mask_mass.is_(None),
+                    self.frame_annotations_type._mask_mass.is_(None),  # pyre-ignore[16]
                    self.frame_annotations_type._mask_mass != 0,
                )
            )
@@ -634,7 +730,9 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
        assert self.eval_batches_file
        logger.info(f"Loading eval batches from {self.eval_batches_file}")

-        if not os.path.isfile(self.eval_batches_file):
+        if (
+            self.path_manager and not self.path_manager.isfile(self.eval_batches_file)
+        ) or (not self.path_manager and not os.path.isfile(self.eval_batches_file)):
            # The batch indices file does not exist.
            # Most probably the user has not specified the root folder.
            raise ValueError(
@@ -642,7 +740,8 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
                + "Please specify a correct dataset_root folder."
            )

-        with open(self.eval_batches_file, "r") as f:
+        eval_batches_file = self._local_path(self.eval_batches_file)
+        with open(eval_batches_file, "r") as f:
            eval_batches = json.load(f)

        # limit the dataset to sequences to allow multiple evaluations in one file
@@ -726,9 +825,15 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
            self.frame_annotations_type.sequence_name == seq_name,
            self.frame_annotations_type.frame_number.in_(frames),
        )
+        frame_no_ts = None

-        with self._sql_engine.connect() as connection:
-            frame_no_ts = pd.read_sql_query(stmt, connection)
+        if self.scoped_session:
+            stmt_text = str(stmt.compile(compile_kwargs={"literal_binds": True}))
+            with scoped_session(self._session_factory)() as session:  # pyre-ignore
+                frame_no_ts = pd.read_sql_query(stmt_text, session.connection())
+        else:
+            with self._sql_engine.connect() as connection:
+                frame_no_ts = pd.read_sql_query(stmt, connection)

        if len(frame_no_ts) != len(index_slice):
            raise ValueError(
@@ -758,11 +863,18 @@ class SqlIndexDataset(DatasetBase, ReplaceableBase):  # pyre-ignore
            prefixes=["TEMP"],  # NOTE SQLite specific!
        )

+    @classmethod
+    def pre_expand(cls) -> None:
+        # remove dataclass annotations that are not meant to be init params
+        # because they cause troubles for OmegaConf
+        for attr, attr_value in list(cls.__dict__.items()):  # need to copy as we mutate
+            if isinstance(attr_value, Field) and attr_value.metadata.get(
+                "omegaconf_ignore", False
+            ):
+                delattr(cls, attr)
+                del cls.__annotations__[attr]
+

 def _seq_name_to_seed(seq_name) -> int:
    """Generates numbers in [0, 2 ** 28)"""
    return int(hashlib.sha1(seq_name.encode("utf-8")).hexdigest()[:7], 16)
-
-
-def _safe_as_tensor(data, dtype):
-    return torch.tensor(data, dtype=dtype) if data is not None else None
--- a/pytorch3d/implicitron/dataset/sql_dataset_provider.py
+++ b/pytorch3d/implicitron/dataset/sql_dataset_provider.py
@@ -4,6 +4,8 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.

+# pyre-unsafe
+

 import logging
 import os
@@ -43,7 +45,7 @@ logger = logging.getLogger(__name__)


@registry.register
-class SqlIndexDatasetMapProvider(DatasetMapProviderBase):  # pyre-ignore [13]
+class SqlIndexDatasetMapProvider(DatasetMapProviderBase):
    """
    Generates the training, validation, and testing dataset objects for
    a dataset laid out on disk like SQL-CO3D, with annotations in an SQLite data base.
@@ -193,9 +195,9 @@ class SqlIndexDatasetMapProvider(DatasetMapProviderBase):  # pyre-ignore [13]

    # this is a mould that is never constructed, used to build self._dataset_map values
    dataset_class_type: str = "SqlIndexDataset"
-    dataset: SqlIndexDataset
+    dataset: SqlIndexDataset  # pyre-ignore [13]

-    path_manager_factory: PathManagerFactory
+    path_manager_factory: PathManagerFactory  # pyre-ignore [13]
    path_manager_factory_class_type: str = "PathManagerFactory"

    def __post_init__(self):
@@ -282,8 +284,14 @@ class SqlIndexDatasetMapProvider(DatasetMapProviderBase):  # pyre-ignore [13]
                logger.info(f"Val dataset: {str(val_dataset)}")

            logger.debug("Extracting test dataset.")
-            eval_batches_file = self._get_lists_file("eval_batches")
-            del common_dataset_kwargs["eval_batches_file"]
+            if self.eval_batches_path is None:
+                eval_batches_file = None
+            else:
+                eval_batches_file = self._get_lists_file("eval_batches")
+
+            if "eval_batches_file" in common_dataset_kwargs:
+                common_dataset_kwargs.pop("eval_batches_file", None)
+
            test_dataset = dataset_type(
                **common_dataset_kwargs,
                subsets=self._get_subsets(self.test_subsets, True),
--- a/pytorch3d/implicitron/dataset/utils.py
+++ b/pytorch3d/implicitron/dataset/utils.py
@@ -87,6 +87,15 @@ def is_train_frame(
 def get_bbox_from_mask(
    mask: np.ndarray, thr: float, decrease_quant: float = 0.05
 ) -> Tuple[int, int, int, int]:
+    # these corner cases need to be handled in order to avoid an infinite loop
+    if mask.size == 0:
+        warnings.warn("Empty mask is provided for bbox extraction.", stacklevel=1)
+        return 0, 0, 1, 1
+
+    if not mask.min() >= 0.0:
+        warnings.warn("Negative values in the mask for bbox extraction.", stacklevel=1)
+        mask = mask.clip(min=0.0)
+
    # bbox in xywh
    masks_for_box = np.zeros_like(mask)
    while masks_for_box.sum() <= 1.0:
@@ -134,7 +143,15 @@ T = TypeVar("T", bound=torch.Tensor)
 def bbox_xyxy_to_xywh(xyxy: T) -> T:
    wh = xyxy[2:] - xyxy[:2]
    xywh = torch.cat([xyxy[:2], wh])
-    return xywh  # pyre-ignore
+    return xywh  # pyre-ignore[7]
+
+
+def bbox_xywh_to_xyxy(xywh: T, clamp_size: float | int | None = None) -> T:
+    wh = xywh[2:]
+    if clamp_size is not None:
+        wh = wh.clamp(min=clamp_size)
+    xyxy = torch.cat([xywh[:2], xywh[:2] + wh])
+    return xyxy  # pyre-ignore[7]


 def get_clamp_bbox(
@@ -180,16 +197,6 @@ def rescale_bbox(
    return bbox * rel_size


-def bbox_xywh_to_xyxy(
-    xywh: torch.Tensor, clamp_size: Optional[int] = None
-) -> torch.Tensor:
-    xyxy = xywh.clone()
-    if clamp_size is not None:
-        xyxy[2:] = torch.clamp(xyxy[2:], clamp_size)
-    xyxy[2:] += xyxy[:2]
-    return xyxy
-
-
 def get_1d_bounds(arr: np.ndarray) -> Tuple[int, int]:
    nz = np.flatnonzero(arr)
    return nz[0], nz[-1] + 1
@@ -201,18 +208,24 @@ def resize_image(
    image_width: Optional[int],
    mode: str = "bilinear",
 ) -> Tuple[torch.Tensor, float, torch.Tensor]:
-
    if isinstance(image, np.ndarray):
        image = torch.from_numpy(image)

-    if image_height is None or image_width is None:
+    if (
+        image_height is None
+        or image_width is None
+        or image.shape[-2] == 0
+        or image.shape[-1] == 0
+    ):
        # skip the resizing
        return image, 1.0, torch.ones_like(image[:1])
+
    # takes numpy array or tensor, returns pytorch tensor
    minscale = min(
        image_height / image.shape[-2],
        image_width / image.shape[-1],
    )
+
    imre = torch.nn.functional.interpolate(
        image[None],
        scale_factor=minscale,
@@ -220,6 +233,7 @@ def resize_image(
        align_corners=False if mode == "bilinear" else None,
        recompute_scale_factor=True,
    )[0]
+
    imre_ = torch.zeros(image.shape[0], image_height, image_width)
    imre_[:, 0 : imre.shape[1], 0 : imre.shape[2]] = imre
    mask = torch.zeros(1, image_height, image_width)
@@ -232,9 +246,21 @@ def transpose_normalize_image(image: np.ndarray) -> np.ndarray:
    return im.astype(np.float32) / 255.0


-def load_image(path: str) -> np.ndarray:
+def load_image(
+    path: str, try_read_alpha: bool = False, pil_format: str = "RGB"
+) -> np.ndarray:
+    """
+    Load an image from a path and return it as a numpy array.
+    If try_read_alpha is True, the image is read as RGBA and the alpha channel is
+    returned as the fourth channel.
+    Otherwise, the image is read as RGB and a three-channel image is returned.
+    """
    with Image.open(path) as pil_im:
-        im = np.array(pil_im.convert("RGB"))
+        # Check if the image has an alpha channel
+        if try_read_alpha and pil_im.mode == "RGBA":
+            im = np.array(pil_im)
+        else:
+            im = np.array(pil_im.convert(pil_format))

    return transpose_normalize_image(im)

@@ -329,6 +355,7 @@ def adjust_camera_to_bbox_crop_(

    focal_length_px, principal_point_px = _convert_ndc_to_pixels(
        camera.focal_length[0],
+        # pyre-fixme[29]: `Union[(self: TensorBase, indices: Union[None, slice[Any, A...
        camera.principal_point[0],
        image_size_wh,
    )
@@ -341,6 +368,7 @@ def adjust_camera_to_bbox_crop_(
    )

    camera.focal_length = focal_length[None]
+    # pyre-fixme[16]: `PerspectiveCameras` has no attribute `principal_point`.
    camera.principal_point = principal_point_cropped[None]


@@ -348,9 +376,11 @@ def adjust_camera_to_image_scale_(
    camera: PerspectiveCameras,
    original_size_wh: torch.Tensor,
    new_size_wh: torch.LongTensor,
+    # pyre-fixme[7]: Expected `PerspectiveCameras` but got implicit return value of `None`.
 ) -> PerspectiveCameras:
    focal_length_px, principal_point_px = _convert_ndc_to_pixels(
        camera.focal_length[0],
+        # pyre-fixme[29]: `Union[(self: TensorBase, indices: Union[None, slice[Any, A...
        camera.principal_point[0],
        original_size_wh,
    )
@@ -367,7 +397,8 @@ def adjust_camera_to_image_scale_(
        image_size_wh_output,
    )
    camera.focal_length = focal_length_scaled[None]
-    camera.principal_point = principal_point_scaled[None]  # pyre-ignore
+    # pyre-fixme[16]: `PerspectiveCameras` has no attribute `principal_point`.
+    camera.principal_point = principal_point_scaled[None]  # pyre-ignore[16]


 # NOTE this cache is per-worker; they are implemented as processes.
--- a/pytorch3d/implicitron/evaluation/evaluate_new_view_synthesis.py
+++ b/pytorch3d/implicitron/evaluation/evaluate_new_view_synthesis.py
@@ -299,7 +299,6 @@ def eval_batch(
    )

    for loss_fg_mask, name_postfix in zip((mask_crop, mask_fg), ("_masked", "_fg")):
-
        loss_mask_now = mask_crop * loss_fg_mask

        for rgb_metric_name, rgb_metric_fun in zip(
--- a/pytorch3d/implicitron/models/feature_extractor/resnet_feature_extractor.py
+++ b/pytorch3d/implicitron/models/feature_extractor/resnet_feature_extractor.py
@@ -106,7 +106,7 @@ class ResNetFeatureExtractor(FeatureExtractorBase):
            self.layers = torch.nn.ModuleList()
            self.proj_layers = torch.nn.ModuleList()
            for stage in range(self.max_stage):
-                stage_name = f"layer{stage+1}"
+                stage_name = f"layer{stage + 1}"
                feature_name = self._get_resnet_stage_feature_name(stage)
                if (stage + 1) in self.stages:
                    if (
@@ -139,12 +139,18 @@ class ResNetFeatureExtractor(FeatureExtractorBase):
        self.stages = set(self.stages)  # convert to set for faster "in"

    def _get_resnet_stage_feature_name(self, stage) -> str:
-        return f"res_layer_{stage+1}"
+        return f"res_layer_{stage + 1}"

    def _resnet_normalize_image(self, img: torch.Tensor) -> torch.Tensor:
+        # pyre-fixme[58]: `-` is not supported for operand types `Tensor` and
+        #  `Union[Tensor, Module]`.
+        # pyre-fixme[58]: `/` is not supported for operand types `Tensor` and
+        #  `Union[Tensor, Module]`.
        return (img - self._resnet_mean) / self._resnet_std

    def get_feat_dims(self) -> int:
+        # pyre-fixme[29]: `Union[(self: TensorBase) -> Tensor, Tensor, Module]` is
+        #  not a function.
        return sum(self._feat_dim.values())

    def forward(
@@ -183,7 +189,12 @@ class ResNetFeatureExtractor(FeatureExtractorBase):
            else:
                imgs_normed = imgs_resized
            #  is not a function.
+            # pyre-fixme[29]: `Union[Tensor, Module]` is not a function.
            feats = self.stem(imgs_normed)
+            # pyre-fixme[6]: For 1st argument expected `Iterable[_T1]` but got
+            #  `Union[Tensor, Module]`.
+            # pyre-fixme[6]: For 2nd argument expected `Iterable[_T2]` but got
+            #  `Union[Tensor, Module]`.
            for stage, (layer, proj) in enumerate(zip(self.layers, self.proj_layers)):
                feats = layer(feats)
                # just a sanity check below
--- a/pytorch3d/implicitron/models/generic_model.py
+++ b/pytorch3d/implicitron/models/generic_model.py
@@ -65,7 +65,7 @@ logger = logging.getLogger(__name__)


@registry.register
-class GenericModel(ImplicitronModelBase):  # pyre-ignore: 13
+class GenericModel(ImplicitronModelBase):
    """
    GenericModel is a wrapper for the neural implicit
    rendering and reconstruction pipeline which consists
@@ -226,34 +226,42 @@ class GenericModel(ImplicitronModelBase):  # pyre-ignore: 13

    # ---- global encoder settings
    global_encoder_class_type: Optional[str] = None
+    # pyre-fixme[13]: Attribute `global_encoder` is never initialized.
    global_encoder: Optional[GlobalEncoderBase]

    # ---- raysampler
    raysampler_class_type: str = "AdaptiveRaySampler"
+    # pyre-fixme[13]: Attribute `raysampler` is never initialized.
    raysampler: RaySamplerBase

    # ---- renderer configs
    renderer_class_type: str = "MultiPassEmissionAbsorptionRenderer"
+    # pyre-fixme[13]: Attribute `renderer` is never initialized.
    renderer: BaseRenderer

    # ---- image feature extractor settings
    # (This is only created if view_pooler is enabled)
+    # pyre-fixme[13]: Attribute `image_feature_extractor` is never initialized.
    image_feature_extractor: Optional[FeatureExtractorBase]
    image_feature_extractor_class_type: Optional[str] = None
    # ---- view pooler settings
    view_pooler_enabled: bool = False
+    # pyre-fixme[13]: Attribute `view_pooler` is never initialized.
    view_pooler: Optional[ViewPooler]

    # ---- implicit function settings
    implicit_function_class_type: str = "NeuralRadianceFieldImplicitFunction"
    # This is just a model, never constructed.
    # The actual implicit functions live in self._implicit_functions
+    # pyre-fixme[13]: Attribute `implicit_function` is never initialized.
    implicit_function: ImplicitFunctionBase

    # ----- metrics
+    # pyre-fixme[13]: Attribute `view_metrics` is never initialized.
    view_metrics: ViewMetricsBase
    view_metrics_class_type: str = "ViewMetrics"

+    # pyre-fixme[13]: Attribute `regularization_metrics` is never initialized.
    regularization_metrics: RegularizationMetricsBase
    regularization_metrics_class_type: str = "RegularizationMetrics"

@@ -470,6 +478,8 @@ class GenericModel(ImplicitronModelBase):  # pyre-ignore: 13
            )
        custom_args["global_code"] = global_code

+        # pyre-fixme[29]: `Union[(self: Tensor) -> Any, Tensor, Module]` is not a
+        #  function.
        for func in self._implicit_functions:
            func.bind_args(**custom_args)

@@ -492,6 +502,8 @@ class GenericModel(ImplicitronModelBase):  # pyre-ignore: 13
        # Unbind the custom arguments to prevent pytorch from storing
        # large buffers of intermediate results due to points in the
        # bound arguments.
+        # pyre-fixme[29]: `Union[(self: Tensor) -> Any, Tensor, Module]` is not a
+        #  function.
        for func in self._implicit_functions:
            func.unbind_args()

--- a/pytorch3d/implicitron/models/global_encoder/autodecoder.py
+++ b/pytorch3d/implicitron/models/global_encoder/autodecoder.py
@@ -71,6 +71,7 @@ class Autodecoder(Configurable, torch.nn.Module):
        return key_map

    def calculate_squared_encoding_norm(self) -> Optional[torch.Tensor]:
+        # pyre-fixme[16]: Item `Tensor` of `Tensor | Module` has no attribute `weight`.
        return (self._autodecoder_codes.weight**2).mean()

    def get_encoding_dim(self) -> int:
@@ -95,6 +96,7 @@ class Autodecoder(Configurable, torch.nn.Module):
                # pyre-fixme[9]: x has type `Union[List[str], LongTensor]`; used as
                #  `Tensor`.
                x = torch.tensor(
+                    # pyre-fixme[29]: `Union[(self: TensorBase, indices: Union[None, ...
                    [self._key_map[elem] for elem in x],
                    dtype=torch.long,
                    device=next(self.parameters()).device,
@@ -102,6 +104,7 @@ class Autodecoder(Configurable, torch.nn.Module):
            except StopIteration:
                raise ValueError("Not enough n_instances in the autodecoder") from None

+        # pyre-fixme[29]: `Union[Tensor, Module]` is not a function.
        return self._autodecoder_codes(x)

    def _load_key_map_hook(
--- a/pytorch3d/implicitron/models/global_encoder/global_encoder.py
+++ b/pytorch3d/implicitron/models/global_encoder/global_encoder.py
@@ -59,12 +59,13 @@ class GlobalEncoderBase(ReplaceableBase):

 # TODO: probabilistic embeddings?
@registry.register
-class SequenceAutodecoder(GlobalEncoderBase, torch.nn.Module):  # pyre-ignore: 13
+class SequenceAutodecoder(GlobalEncoderBase, torch.nn.Module):
    """
    A global encoder implementation which provides an autodecoder encoding
    of the frame's sequence identifier.
    """

+    # pyre-fixme[13]: Attribute `autodecoder` is never initialized.
    autodecoder: Autodecoder

    def __post_init__(self):
@@ -121,6 +122,7 @@ class HarmonicTimeEncoder(GlobalEncoderBase, torch.nn.Module):
        if frame_timestamp.shape[-1] != 1:
            raise ValueError("Frame timestamp's last dimensions should be one.")
        time = frame_timestamp / self.time_divisor
+        # pyre-fixme[29]: `Union[Tensor, Module]` is not a function.
        return self._harmonic_embedding(time)

    def calculate_squared_encoding_norm(self) -> Optional[torch.Tensor]:
--- a/pytorch3d/implicitron/models/implicit_function/decoding_functions.py
+++ b/pytorch3d/implicitron/models/implicit_function/decoding_functions.py
@@ -232,9 +232,14 @@ class MLPWithInputSkips(Configurable, torch.nn.Module):
            # if the skip tensor is None, we use `x` instead.
            z = x
        skipi = 0
+        # pyre-fixme[6]: For 1st argument expected `Iterable[_T]` but got
+        #  `Union[Tensor, Module]`.
        for li, layer in enumerate(self.mlp):
+            # pyre-fixme[58]: `in` is not supported for right operand type
+            #  `Union[Tensor, Module]`.
            if li in self._input_skips:
                if self._skip_affine_trans:
+                    # pyre-fixme[29]: `Union[(self: TensorBase, indices: Union[None, ...
                    y = self._apply_affine_layer(self.skip_affines[skipi], y, z)
                else:
                    y = torch.cat((y, z), dim=-1)
@@ -244,7 +249,6 @@ class MLPWithInputSkips(Configurable, torch.nn.Module):


@registry.register
-# pyre-fixme[13]: Attribute `network` is never initialized.
 class MLPDecoder(DecoderFunctionBase):
    """
    Decoding function which uses `MLPWithIputSkips` to convert the embedding to output.
@@ -272,6 +276,7 @@ class MLPDecoder(DecoderFunctionBase):

    input_dim: int = 3
    param_groups: Dict[str, str] = field(default_factory=lambda: {})
+    # pyre-fixme[13]: Attribute `network` is never initialized.
    network: MLPWithInputSkips

    def __post_init__(self):
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
bottler	62a2031dd4	Revert "Fix CUDA kernel index data type in vision/fair/pytorch3d/pytorch3d/csrc/compositing/alpha_composite.cu +10" This reverts commit `3987612062`.	2025-03-27 05:28:03 -07:00
Richard Barnes	3987612062	Fix CUDA kernel index data type in vision/fair/pytorch3d/pytorch3d/csrc/compositing/alpha_composite.cu +10 Summary: CUDA kernel variables matching the type `(thread\|block\|grid).(Idx\|Dim).(x\|y\|z)` [have the data type `uint`](https://docs.nvidia.com/cuda/cuda-c-programming-guide/#built-in-variables). Many programmers mistakenly use implicit casts to turn these data types into `int`. In fact, the [CUDA Programming Guide](https://docs.nvidia.com/cuda/cuda-c-programming-guide/) it self is inconsistent and incorrect in its use of data types in programming examples. The result of these implicit casts is that our kernels may give unexpected results when exposed to large datasets, i.e., those exceeding >~2B items. While we now have linters in place to prevent simple mistakes (D71236150), our codebase has many problematic instances. This diff fixes some of them. Reviewed By: dtolnay Differential Revision: D71355356 fbshipit-source-id: cea44891416d9efd2f466d6c45df4e36008fa036	2025-03-19 13:21:43 -07:00
Alexandros Benetatos	06a76ef8dd	Correct "fast" `matrix_to_axis_angle` near pi (#1953 ) Summary: A continuation of https://github.com/facebookresearch/pytorch3d/issues/1948 -- this commit fixes a small numerical issue with `matrix_to_axis_angle(..., fast=True)` near `pi`. bottler feel free to check this out, it's a single-line change. Pull Request resolved: https://github.com/facebookresearch/pytorch3d/pull/1953 Reviewed By: MichaelRamamonjisoa Differential Revision: D70088251 Pulled By: bottler fbshipit-source-id: 54cc7f946283db700cec2cd5575cf918456b7f32	2025-03-11 12:25:59 -07:00
Richard Barnes	21205730d9	Fix unused-variable issues, mostly relating to AMD/HIP Reviewed By: meyering Differential Revision: D70845538 fbshipit-source-id: 8e52b5e1f1d96b86404fc3b8cbc6fb952e2cb1a6	2025-03-08 13:03:17 -08:00
Richard Barnes	7e09505538	Enable `-Wunused-value` in vision/PACKAGE +1 Summary: This diff enables compilation warning flags for the directory in question. Further details are in [this workplace post](https://fb.workplace.com/permalink.php?story_fbid=pfbid02XaWNiCVk69r1ghfvDVpujB8Hr9Y61uDvNakxiZFa2jwiPHscVdEQwCBHrmWZSyMRl&id=100051201402394). This is a low-risk diff. There are no run-time effects and the diff has already been observed to compile locally. If the code compiles, it work; test errors are spurious. Differential Revision: D70282347 fbshipit-source-id: e2fa55c002d7124b13450c812165d244b8a53f4e	2025-03-04 17:49:30 -08:00
Nicholas Ormrod	20bd8b33f6	facebook-unused-include-check in fbcode/vision Summary: Remove headers flagged by facebook-unused-include-check over fbcode.vision. + format and autodeps This is a codemod. It was automatically generated and will be landed once it is approved and tests are passing in sandcastle. You have been added as a reviewer by Sentinel or Butterfly. Autodiff project: uiv Autodiff partition: fbcode.vision Autodiff bookmark: ad.uiv.fbcode.vision Reviewed By: dtolnay Differential Revision: D70403619 fbshipit-source-id: d109c15774eeb3d809875f75fa2a26ed20d7f9a6	2025-02-28 18:08:12 -08:00
alex-bene	7a3c0cbc9d	Increase performance for conversions including axis angles (#1948 ) Summary: This is an extension of https://github.com/facebookresearch/pytorch3d/issues/1544 with various speed, stability, and readability improvements. (I could not find a way to make a commit to the existing PR). This PR is still based on the [Rodrigues' rotation formula](https://en.wikipedia.org/wiki/Rotation_formalisms_in_three_dimensions#Rotation_matrix_%E2%86%94_Euler_axis/angle). The motivation is the same; this change speeds up the conversions up to 10x, depending on the device, batch size, etc. ### Notes - As the angles get very close to `π`, the existing implementation and the proposed one start to differ. However, (my understanding is that) this is not a problem as the axis can not be stably inferred from the rotation matrix in this case in general. - bottler , I tried to follow similar conventions as existing functions to deal with weird angles, let me know if something needs to be changed to merge this. Pull Request resolved: https://github.com/facebookresearch/pytorch3d/pull/1948 Reviewed By: MichaelRamamonjisoa Differential Revision: D69193009 Pulled By: bottler fbshipit-source-id: e5ed34b45b625114ec4419bb89e22a6aefad4eeb	2025-02-07 07:37:42 -08:00
Roman Shapovalov	215590b497	In FrameDataBuilder, set all path even if we don’t load blobs Summary: This is a somewhat not BC change: some None paths will be replaced by metadata paths, even when they were not used for data loading. Moreover, removing the legacy fix to the paths in the old CO3D release. Reviewed By: bottler Differential Revision: D69048238 fbshipit-source-id: 2a8b26d7b9f5e2adf39c65888b5863a5a9de1996	2025-02-06 09:41:44 -08:00
Antoine Toisoul	43cd681d4f	Updates to Implicitron dataset, metrics and tools Summary: Update Pytorch3D to be able to run assetgen (see later diffs in the stack) Reviewed By: shapovalov Differential Revision: D65942513 fbshipit-source-id: 1d01141c9f7e106608fa591be6e0d3262cb5944f	2025-01-27 09:43:42 -08:00
Roman Shapovalov	42a4a7d432	Generalising SqlIndexDataset to support subtypes of SqlSequenceAnnotation Summary: We did not often extend sequence-level metadata but now for applications like text-to-3D/video, we need to store captions and similar. Reviewed By: bottler Differential Revision: D68269926 fbshipit-source-id: f8af308adce51863d719a335d85cd2558943bd4c	2025-01-20 03:39:06 -08:00
generatedunixname89002005307016	699bc671ca	Add missing Pyre mode headers] [batch:3/1531] [shard:41/N] Differential Revision: D68316763 fbshipit-source-id: fb3e1e1a17786f6f681f1b11b48b4efd7a8ac311	2025-01-17 12:41:56 -08:00
Roman Shapovalov	49cf5a0f37	Loading fg probability from the alpha channel of image_rgb Summary: It is often easier to store the mask together with RGB, especially for renders. The logic in this diff: * if load_mask and mask_path provided, take the mask from mask_path, * otherwise, check if the image has the alpha channel and take it as a mask. Reviewed By: antoinetlc Differential Revision: D68160212 fbshipit-source-id: d9b6779f90027a4987ba96800983f441edff9c74	2025-01-15 11:53:30 -08:00
Roman Shapovalov	89b851e64c	Refactor a utility function for bbox conversion Summary: This function makes it easier to extend FrameData class with new channels; brushing it up a bit. Reviewed By: bottler Differential Revision: D67816470 fbshipit-source-id: 6575415c864d0f539e283889760cd2331bf226a7	2025-01-06 04:17:57 -08:00
Roman Shapovalov	5247f6ad74	Fixing type hints in FrameData Summary: As subj Reviewed By: bottler Differential Revision: D67791200 fbshipit-source-id: c2db01c94718102618f4c8bc5c5130c65ee1d81f	2025-01-06 04:17:57 -08:00
Roman Shapovalov	e41aff47db	Adding default values to FrameData for internal usage Summary: Ensuring all fields in FrameData have defaults. Reviewed By: bottler Differential Revision: D67762780 fbshipit-source-id: b680d29a1a11689850905978df544cdb4eb7ddcd	2025-01-06 04:17:57 -08:00
Roman Shapovalov	64a5bfadc8	Adding SQL Dataset related files to the build script Summary: Now that we have SQLAlchemy 2.0, we can fully use them. Reviewed By: bottler Differential Revision: D66920096 fbshipit-source-id: 25c0ea1c4f7361e66348035519627dc961b9e6e6	2024-12-23 16:05:26 -08:00
Thomas Polasek	055ab3a2e3	Convert directory fbcode/vision to use the Ruff Formatter Summary: Converts the directory specified to use the Ruff formatter in pyfmt ruff_dog If this diff causes merge conflicts when rebasing, please run `hg status -n -0 --change . -I '*/.{py,pyi}' \| xargs -0 arc pyfmt` on your diff, and amend any changes before rebasing onto latest. That should help reduce or eliminate any merge conflicts. allow-large-files Reviewed By: bottler Differential Revision: D66472063 fbshipit-source-id: 35841cb397e4f8e066e2159550d2f56b403b1bef	2024-11-26 02:38:20 -08:00
Edward Yang	f6c2ca6bfc	Prepare for "Fix type-safety of `torch.nn.Module` instances": wave 2 Summary: See D52890934 Reviewed By: malfet, r-barnes Differential Revision: D66245100 fbshipit-source-id: 019058106ac7eaacf29c1c55912922ea55894d23	2024-11-21 11:08:51 -08:00
Jeremy Reizenstein	e20cbe9b0e	test fixes and lints Summary: - followup recent pyre change D63415925 - make tests remove temporary files - weights_only=True in torch.load - lint fixes 3 test fixes from VRehnberg in https://github.com/facebookresearch/pytorch3d/issues/1914 - imageio channels fix - frozen decorator in test_config - load_blobs positional Reviewed By: MichaelRamamonjisoa Differential Revision: D66162167 fbshipit-source-id: 7737e174691b62f1708443a4fae07343cec5bfeb	2024-11-20 09:15:51 -08:00
Jeremy Reizenstein	c17e6f947a	run CI tests on main Reviewed By: MichaelRamamonjisoa Differential Revision: D66162168 fbshipit-source-id: 90268c1925fa9439b876df143035c9d3c3a74632	2024-11-20 05:06:52 -08:00
Yann Noutary	91c9f34137	Add safeguard in case num_tris diverges Summary: This PR fixes adds a safeguard preventing num_tris to overflow in `MAX_TRIS`-length arrays. The update rule of `num_tris` is bounded : - max(num_tris(t)) = 2num_tris(t-1) - num_tris(0) = 12 - t <= 6 So : - max(num_tris) = 2^612 - max(num_tris) = 768 Reviewed By: bottler Differential Revision: D66162573 fbshipit-source-id: e269a79c75c6cc33306986b1f1256cffbe96c730	2024-11-20 01:24:28 -08:00
Jeremy Reizenstein	81d82980bc	Fix ogl test hang Summary: See https://github.com/facebookresearch/pytorch3d/issues/1908 Reviewed By: MichaelRamamonjisoa Differential Revision: D65280253 fbshipit-source-id: ec05902c5f2f7eb9ddd92bda0045cc3564b8c091	2024-11-06 11:40:42 -08:00
Jeremy Reizenstein	8fe6934885	fix subdivide_meshes with empty mesh #1788 Summary: Simplify code fixes https://github.com/facebookresearch/pytorch3d/issues/1788 Reviewed By: MichaelRamamonjisoa Differential Revision: D61847675 fbshipit-source-id: 48400875d1d885bb3615bc9f4b3c7c3d822b67e7	2024-11-06 11:40:26 -08:00
bottler	c434957b2a	Run tests in github action (#1896 ) Summary: Pull Request resolved: https://github.com/facebookresearch/pytorch3d/pull/1896 Reviewed By: MichaelRamamonjisoa Differential Revision: D65272512 Pulled By: bottler fbshipit-source-id: 3bcfab43acd2d6be5444ff25178381510ddac015	2024-11-06 11:15:34 -08:00
Jeremy Reizenstein	dd2a11b5fc	Fix OFF for new numpy errors Summary: Error messages have changed around numpy version 2, making existing code fail. Reviewed By: MichaelRamamonjisoa Differential Revision: D65280674 fbshipit-source-id: b3ae613ea8f0f4ae20fb6e5e816314b8c10e6c65	2024-11-06 11:13:59 -08:00
Richard Barnes	9563ef79ca	c10::optional -> std::optional in some files Reviewed By: jermenkoo Differential Revision: D65425234 fbshipit-source-id: 1e7707d6b6aab640cc1fdd3bd71a3b50f77a0909	2024-11-04 12:03:51 -08:00
generatedunixname89002005287564	008c7ab58c	Pre-silence Pyre Errors for upcoming upgrade] [batch:67/603] [shard:3/N] Reviewed By: MaggieMoss Differential Revision: D65290095 fbshipit-source-id: ced87d096aa8939700de5599ce6984cd7ae93912	2024-10-31 16:26:25 -07:00
Jeremy Reizenstein	9eaed4c495	Fix K>1 in multimap UV sampling Summary: Fixes https://github.com/facebookresearch/pytorch3d/issues/1897 "Wrong dimension on gather". Reviewed By: cijose Differential Revision: D65280675 fbshipit-source-id: 1d587036887972bb2a2ea56d40df19cbf1aeb6cc	2024-10-31 16:05:10 -07:00
Richard Barnes	e13848265d	`at::optional` -> `std::optional` (#1170 ) Summary: Pull Request resolved: https://github.com/pytorch/ao/pull/1170 Reviewed By: gineshidalgo99 Differential Revision: D64938040 fbshipit-source-id: 57f98b90676ad0164a6975ea50e4414fd85ae6c4	2024-10-25 06:37:57 -07:00
generatedunixname89002005307016	58566963d6	Add type error suppressions for upcoming upgrade Reviewed By: MaggieMoss Differential Revision: D64502797 fbshipit-source-id: cee9a54dfa8a005d5912b895d0bd094f352c5c6f	2024-10-16 19:22:01 -07:00
Suresh Babu Kolla	e17ed5cd50	Hipify Pulsar for PyTorch3D Summary: - Hipified Pytorch Pulsar - Created separate target for Pulsar tests and enabled RE testing - Pytorch3D full test suite requires additional work like fixing EGL dependencies on AMD Reviewed By: danzimm Differential Revision: D61339912 fbshipit-source-id: 0d10bc966e4de4a959f3834a386bad24e449dc1f	2024-10-09 14:38:42 -07:00
Richard Barnes	8ed0c7a002	c10::optional -> std::optional Summary: `c10::optional` is an alias for `std::optional`. Let's remove the alias and use the real thing. Reviewed By: meyering Differential Revision: D63402341 fbshipit-source-id: 241383e7ca4b2f3f1f9cac3af083056123dfd02b	2024-10-03 14:38:37 -07:00
Richard Barnes	2da913c7e6	c10::optional -> std::optional Summary: `c10::optional` is an alias for `std::optional`. Let's remove the alias and use the real thing. Reviewed By: palmje Differential Revision: D63409387 fbshipit-source-id: fb6db59a14db9e897e2e6b6ad378f33bf2af86e8	2024-10-02 11:09:29 -07:00
generatedunixname89002005307016	fca83e6369	Convert .pyre_configuration.local to fast by default architecture] [batch:23/263] [shard:3/N] [A] Reviewed By: connernilsen Differential Revision: D63415925 fbshipit-source-id: c3e28405c70f9edcf8c21457ac4faf7315b07322	2024-09-25 17:34:03 -07:00
Jeremy Reizenstein	75ebeeaea0	update version to 0.7.8 Summary: as title Reviewed By: das-intensity Differential Revision: D62588556 fbshipit-source-id: 55bae19dd1df796e83179cd29d805fcd871b6d23	2024-09-13 02:31:49 -07:00
Jeremy Reizenstein	ab793177c6	remove pytorch2.0 builds Summary: these are failing in ci Reviewed By: das-intensity Differential Revision: D62594666 fbshipit-source-id: 5e3a7441be2978803dc2d3e361365e0fffa7ad3b	2024-09-13 02:07:25 -07:00
Jeremy Reizenstein	9acdd67b83	fix obj material indexing bug #1368 Summary: Make the negative index actually not an error fixes https://github.com/facebookresearch/pytorch3d/issues/1368 Reviewed By: das-intensity Differential Revision: D62177991 fbshipit-source-id: e5ed433bde1f54251c4d4b6db073c029cbe87343	2024-09-13 02:00:49 -07:00
Nicholas Dahm	3f428d9981	pytorch 2.4.0 + 2.4.1 Summary: Apparently pytorch 2.4 is now supported as per [this closed issue](https://github.com/facebookresearch/pytorch3d/issues/1863). Added the `2.4.0` & `2.4.1` versions to `regenerate.py` then ran that as per the `README_fb.md` which generated `config.yml` changes. Reviewed By: bottler Differential Revision: D62517831 fbshipit-source-id: 002e276dfe2fa078136ff2f6c747d937abbadd1a	2024-09-11 15:09:43 -07:00
Josh Fromm	05cbea115a	Hipify Pytorch3D (#1851 ) Summary: X-link: https://github.com/pytorch/pytorch/pull/133343 X-link: https://github.com/fairinternal/pytorch3d/pull/45 Pull Request resolved: https://github.com/facebookresearch/pytorch3d/pull/1851 Enables pytorch3d to build on AMD. An important part of enabling this was not compiling the Pulsar backend when the target is AMD. There are simply too many kernel incompatibilites to make it work (I tried haha). Fortunately, it doesnt seem like most modern applications of pytorch3d rely on Pulsar. We should be able to unlock most of pytorch3d's goodness on AMD without it. Reviewed By: bottler, houseroad Differential Revision: D61171993 fbshipit-source-id: fd4aee378a3568b22676c5bf2b727c135ff710af	2024-08-15 16:18:22 -07:00
generatedunixname89002005307016	38afdcfc68	upgrade pyre version in `fbcode/vision` - batch 2 Reviewed By: bottler Differential Revision: D60992234 fbshipit-source-id: 899db6ed590ef966ff651c11027819e59b8401a3	2024-08-09 02:07:45 -07:00
Christine Sun	1e0b1d9c72	Remove Python versions from Install.md Summary: To avoid the installation instructions for PyTorch3D becoming out-of-date, instead of specifying certain Python versions, update to just `Python`. Reader will understand it has to be a Python version compatible with GitHub. Reviewed By: bottler Differential Revision: D60919848 fbshipit-source-id: 5e974970a0db3d3d32fae44e5dd30cbc1ce237a9	2024-08-07 13:46:31 -07:00
Rebecca Chen (Python)	44702fdb4b	Add "max" point reduction for chamfer distance Summary: * Adds a "max" option for the point_reduction input to the chamfer_distance function. * When combining the x and y directions, maxes the losses instead of summing them when point_reduction="max". * Moves batch reduction to happen after the directions are combined. * Adds test_chamfer_point_reduction_max and test_single_directional_chamfer_point_reduction_max tests. Fixes https://github.com/facebookresearch/pytorch3d/issues/1838 Reviewed By: bottler Differential Revision: D60614661 fbshipit-source-id: 7879816acfda03e945bada951b931d2c522756eb	2024-08-02 10:46:07 -07:00
Jeremy Reizenstein	7edaee71a9	allow matrix_to_quaternion onnx export Summary: Attempt to allow torch.onnx.dynamo_export(matrix_to_quaternion) to work. Differential Revision: D59812279 fbshipit-source-id: 4497e5b543bec9d5c2bdccfb779d154750a075ad	2024-07-16 11:30:20 -07:00
Roman Shapovalov	d0d0e02007	Fix: setting FrameData.crop_bbox_xywh for backwards compatibility Summary: This diff is fixing a backwards compatibility issue in PyTorch3D's dataset API. The code ensures that the `crop_bbox_xywh` attribute is set when box_crop flag is on. This is an implementation detail that people should not really use, however some people depend on this behaviour. Reviewed By: bottler Differential Revision: D59777449 fbshipit-source-id: b875e9eb909038b8629ccdade87661bb2c39d529	2024-07-16 02:21:13 -07:00
Jeremy Reizenstein	4df110b0a9	remove fvcore dependency Summary: This is not actually needed and is causing a conda-forge confusion to do with python_abi - which needs users to have `-c conda-forge` when they install pytorch3d. Reviewed By: patricklabatut Differential Revision: D59587930 fbshipit-source-id: 961ae13a62e1b2b2ce6d8781db38bd97eca69e65	2024-07-11 04:35:38 -07:00
Huy Do	51fd114d8b	Forward fix internal pyre failure from D58983461 Summary: X-link: https://github.com/pytorch/pytorch/pull/129525 Somehow, using underscore alias of some builtin types breaks pyre Reviewed By: malfet, clee2000 Differential Revision: D59029768 fbshipit-source-id: cfa2171b66475727b9545355e57a8297c1dc0bc6	2024-06-27 07:35:18 -07:00
Jeremy Reizenstein	89653419d0	version 0.7.7 Summary: New version Reviewed By: MichaelRamamonjisoa Differential Revision: D58668979 fbshipit-source-id: 195eaf83e4da51a106ef72e38dbb98c51c51724c	2024-06-25 06:59:24 -07:00
Jeremy Reizenstein	7980854d44	require pytorch 2.0+ Summary: Problems with timeouts on old builds. Reviewed By: MichaelRamamonjisoa Differential Revision: D58819435 fbshipit-source-id: e1976534a102ad3841f3b297c772e916aeea12cb	2024-06-21 08:15:17 -07:00
Jeremy Reizenstein	51d7c06ddd	MKL version fix in CI (#1820 ) Summary: Fix for "undefined symbol: iJIT_NotifyEvent" build issue, Pull Request resolved: https://github.com/facebookresearch/pytorch3d/pull/1820 Reviewed By: MichaelRamamonjisoa Differential Revision: D58685326 fbshipit-source-id: 48b54367c00851cc6fbb111ca98d69a2ace8361b	2024-06-21 08:15:17 -07:00
Sergii Dymchenko	00c36ec01c	Update deprecated PyTorch functions in fbcode/vision Reviewed By: bottler Differential Revision: D58762015 fbshipit-source-id: a0d05fe63a88d33e3f7783b5a7b2a476dd3a7449	2024-06-20 14:06:28 -07:00