Merge branch 'master' into ioannis@18861-CSVLogger-fails-on-remote-fs

Author: bhimrazy

.actions/assistant.py

@@ -154,8 +154,8 @@ def load_readme_description(path_dir: str, homepage: str, version: str) -> str:
 
     """
     path_readme = os.path.join(path_dir, "README.md")
-    with open(path_readme, encoding="utf-8") as fo:
-        text = fo.read()
+    with open(path_readme, encoding="utf-8") as fopen:
+        text = fopen.read()
 
     # drop images from readme
     text = text.replace(
@@ -308,17 +308,17 @@ def copy_replace_imports(
         if ext in (".pyc",):
             continue
         # Try to parse everything else
-        with open(fp, encoding="utf-8") as fo:
+        with open(fp, encoding="utf-8") as fopen:
             try:
-                lines = fo.readlines()
+                lines = fopen.readlines()
             except UnicodeDecodeError:
                 # a binary file, skip
                 print(f"Skipped replacing imports for {fp}")
                 continue
         lines = _replace_imports(lines, list(zip(source_imports, target_imports)), lightning_by=lightning_by)
         os.makedirs(os.path.dirname(fp_new), exist_ok=True)
-        with open(fp_new, "w", encoding="utf-8") as fo:
-            fo.writelines(lines)
+        with open(fp_new, "w", encoding="utf-8") as fopen:
+            fopen.writelines(lines)
 
 
 def create_mirror_package(source_dir: str, package_mapping: dict[str, str]) -> None:
@@ -341,47 +341,6 @@ def create_mirror_package(source_dir: str, package_mapping: dict[str, str]) -> N
 
 
 class AssistantCLI:
-    @staticmethod
-    def requirements_prune_pkgs(packages: Sequence[str], req_files: Sequence[str] = REQUIREMENT_FILES_ALL) -> None:
-        """Remove some packages from given requirement files."""
-        if isinstance(req_files, str):
-            req_files = [req_files]
-        for req in req_files:
-            AssistantCLI._prune_packages(req, packages)
-
-    @staticmethod
-    def _prune_packages(req_file: str, packages: Sequence[str]) -> None:
-        """Remove some packages from given requirement files."""
-        path = Path(req_file)
-        assert path.exists()
-        text = path.read_text()
-        lines = text.splitlines()
-        final = []
-        for line in lines:
-            ln_ = line.strip()
-            if not ln_ or ln_.startswith("#"):
-                final.append(line)
-                continue
-            req = list(_parse_requirements([ln_]))[0]
-            if req.name not in packages:
-                final.append(line)
-        print(final)
-        path.write_text("\n".join(final) + "\n")
-
-    @staticmethod
-    def _replace_min(fname: str) -> None:
-        with open(fname, encoding="utf-8") as fo:
-            req = fo.read().replace(">=", "==")
-        with open(fname, "w", encoding="utf-8") as fw:
-            fw.write(req)
-
-    @staticmethod
-    def replace_oldest_ver(requirement_fnames: Sequence[str] = REQUIREMENT_FILES_ALL) -> None:
-        """Replace the min package version by fixed one."""
-        for fname in requirement_fnames:
-            print(fname)
-            AssistantCLI._replace_min(fname)
-
     @staticmethod
     def copy_replace_imports(
         source_dir: str,
@@ -471,15 +430,15 @@ def convert_version2nightly(ver_file: str = "src/version.info") -> None:
         """Load the actual version and convert it to the nightly version."""
         from datetime import datetime
 
-        with open(ver_file) as fo:
-            version = fo.read().strip()
+        with open(ver_file) as fopen:
+            version = fopen.read().strip()
         # parse X.Y.Z version and prune any suffix
         vers = re.match(r"(\d+)\.(\d+)\.(\d+).*", version)
         # create timestamp  YYYYMMDD
         timestamp = datetime.now().strftime("%Y%m%d")
         version = f"{'.'.join(vers.groups())}.dev{timestamp}"
-        with open(ver_file, "w") as fo:
-            fo.write(version + os.linesep)
+        with open(ver_file, "w") as fopen:
+            fopen.write(version + os.linesep)
 
     @staticmethod
     def generate_docker_tags(

.actions/requirements.txt

@@ -1,3 +1,3 @@
-jsonargparse >=4.16.0, <=4.35.0
+jsonargparse
 requests
 packaging

.azure/gpu-benchmarks.yml

@@ -46,7 +46,7 @@ jobs:
     variables:
       DEVICES: $( python -c 'print("$(Agent.Name)".split("_")[-1])' )
     container:
-      image: "pytorchlightning/pytorch_lightning:base-cuda-py3.12-torch2.5-cuda12.1.0"
+      image: "pytorchlightning/pytorch_lightning:base-cuda12.6.3-py3.12-torch2.8"
       options: "--gpus=all --shm-size=32g"
     strategy:
       matrix:
@@ -76,8 +76,13 @@ jobs:
         displayName: "Image info & NVIDIA"
 
       - bash: |
-          pip install -e .[dev] --find-links ${TORCH_URL}
-          pip install setuptools==75.6.0 jsonargparse==4.35.0
+          pip install -U -q -r .actions/requirements.txt
+          python .actions/assistant.py copy_replace_imports --source_dir="./tests" \
+            --source_import="lightning.fabric,lightning.pytorch" \
+            --target_import="lightning_fabric,pytorch_lightning"
+        displayName: "Adjust tests"
+
+      - bash: pip install -e .[dev] --find-links ${TORCH_URL}
         env:
           FREEZE_REQUIREMENTS: "1"
         displayName: "Install package"
@@ -88,17 +93,10 @@ jobs:
           python -c "import torch ; mgpu = torch.cuda.device_count() ; assert mgpu == 2, f'GPU: {mgpu}'"
         displayName: "Env details"
 
-      - bash: |
-          pip install -q -r .actions/requirements.txt
-          python .actions/assistant.py copy_replace_imports --source_dir="./tests" \
-            --source_import="lightning.fabric,lightning.pytorch" \
-            --target_import="lightning_fabric,pytorch_lightning"
-        displayName: "Adjust tests"
-
       - bash: python -m pytest parity_$(PACKAGE_NAME) -v --durations=0
         env:
           PL_RUNNING_BENCHMARKS: "1"
-          PL_RUN_CUDA_TESTS: "1"
+          RUN_ONLY_CUDA_TESTS: "1"
         workingDirectory: tests/
         displayName: "Testing: benchmarks"
 
@@ -107,7 +105,7 @@ jobs:
         # without succeeded this could run even if the job has already failed
         condition: and(succeeded(), eq(variables['PACKAGE_NAME'], 'fabric'))
         env:
-          PL_RUN_CUDA_TESTS: "1"
+          RUN_ONLY_CUDA_TESTS: "1"
           PL_RUN_STANDALONE_TESTS: "1"
         displayName: "Testing: fabric standalone tasks"
         timeoutInMinutes: "10"

.azure/gpu-tests-fabric.yml

@@ -48,7 +48,7 @@ jobs:
       DEVICES: $( python -c 'print("$(Agent.Name)".split("_")[-1])' )
       FREEZE_REQUIREMENTS: "1"
       PIP_CACHE_DIR: "/var/tmp/pip"
-      PL_RUN_CUDA_TESTS: "1"
+      RUN_ONLY_CUDA_TESTS: "1"
     container:
       image: $(image)
       # default shm size is 64m. Increase it to avoid:
@@ -57,16 +57,16 @@ jobs:
     strategy:
       matrix:
         "Fabric | oldest":
-          image: "pytorchlightning/pytorch_lightning:base-cuda-py3.10-torch2.1-cuda12.1.1"
+          image: "pytorchlightning/pytorch_lightning:base-cuda12.1.1-py3.10-torch2.1"
           PACKAGE_NAME: "fabric"
         "Fabric | latest":
-          image: "pytorchlightning/pytorch_lightning:base-cuda-py3.12-torch2.6-cuda12.4.1"
-          PACKAGE_NAME: "fabric"
-        "Fabric | future":
-          image: "pytorchlightning/pytorch_lightning:base-cuda-py3.12-torch2.7-cuda12.6.3"
+          image: "pytorchlightning/pytorch_lightning:base-cuda12.6.3-py3.12-torch2.8"
           PACKAGE_NAME: "fabric"
+        #"Fabric | future":
+        #  image: "pytorchlightning/pytorch_lightning:base-cuda12.6.3-py3.12-torch2.7"
+        #  PACKAGE_NAME: "fabric"
         "Lightning | latest":
-          image: "pytorchlightning/pytorch_lightning:base-cuda-py3.12-torch2.6-cuda12.4.1"
+          image: "pytorchlightning/pytorch_lightning:base-cuda12.6.3-py3.12-torch2.8"
           PACKAGE_NAME: "lightning"
     workspace:
       clean: all
@@ -78,8 +78,6 @@ jobs:
           echo "##vso[task.setvariable variable=TORCH_URL]https://download.pytorch.org/whl/cu${cuda_ver}/torch_stable.html"
           scope=$(python -c 'n = "$(PACKAGE_NAME)" ; print(dict(fabric="lightning_fabric").get(n, n))')
           echo "##vso[task.setvariable variable=COVERAGE_SOURCE]$scope"
-          python_ver=$(python -c "import sys; print(f'{sys.version_info.major}{sys.version_info.minor}')")
-          echo "##vso[task.setvariable variable=PYTHON_VERSION_MM]$python_ver"
         displayName: "set env. vars"
       - bash: |
           echo "##vso[task.setvariable variable=TORCH_URL]https://download.pytorch.org/whl/test/cu${CUDA_VERSION_MM}"
@@ -100,6 +98,20 @@ jobs:
           pip list
         displayName: "Image info & NVIDIA"
 
+      - bash: |
+          set -ex
+          pip install "cython<3.0" wheel  # for compatibility
+          pip install -U "lightning-utilities[cli]"
+          cd requirements/fabric
+          # replace range by pin minimal requirements
+          python -m lightning_utilities.cli requirements set-oldest --req_files "['base.txt', 'strategies.txt']"
+          # drop deepspeed since it is not supported by our minimal Torch requirements
+          python -m lightning_utilities.cli requirements prune-pkgs --packages deepspeed --req_files strategies.txt
+          # uninstall deepspeed since some older docker images have it pre-installed
+          pip uninstall -y deepspeed
+        condition: contains(variables['Agent.JobName'], 'oldest')
+        displayName: "setting oldest dependencies"
+
       - bash: |
           PYTORCH_VERSION=$(python -c "import torch; print(torch.__version__.split('+')[0])")
           pip install -q wget packaging
@@ -109,11 +121,22 @@ jobs:
           done
         displayName: "Adjust dependencies"
 
+      - bash: |
+          pip install -U -q -r .actions/requirements.txt
+          python .actions/assistant.py copy_replace_imports --source_dir="./tests/tests_fabric" \
+            --source_import="lightning.fabric" \
+            --target_import="lightning_fabric"
+          python .actions/assistant.py copy_replace_imports --source_dir="./examples/fabric" \
+            --source_import="lightning.fabric" \
+            --target_import="lightning_fabric"
+        # without succeeded this could run even if the job has already failed
+        condition: and(succeeded(), eq(variables['PACKAGE_NAME'], 'fabric'))
+        displayName: "Adjust tests & examples"
+
       - bash: |
           set -e
           extra=$(python -c "print({'lightning': 'fabric-'}.get('$(PACKAGE_NAME)', ''))")
-          pip install -e ".[${extra}dev]" pytest-timeout -U --extra-index-url="${TORCH_URL}"
-          pip install setuptools==75.6.0 jsonargparse==4.35.0
+          pip install -e ".[${extra}dev]" -U --upgrade-strategy=eager --extra-index-url="${TORCH_URL}"
         displayName: "Install package & dependencies"
 
       - bash: |
@@ -130,18 +153,6 @@ jobs:
         condition: and(succeeded(), eq(variables['PACKAGE_NAME'], 'fabric'))
         displayName: "Testing: Fabric doctests"
 
-      - bash: |
-          pip install -q -r .actions/requirements.txt
-          python .actions/assistant.py copy_replace_imports --source_dir="./tests/tests_fabric" \
-            --source_import="lightning.fabric" \
-            --target_import="lightning_fabric"
-          python .actions/assistant.py copy_replace_imports --source_dir="./examples/fabric" \
-            --source_import="lightning.fabric" \
-            --target_import="lightning_fabric"
-        # without succeeded this could run even if the job has already failed
-        condition: and(succeeded(), eq(variables['PACKAGE_NAME'], 'fabric'))
-        displayName: "Adjust tests & examples"
-
       - bash: python -m coverage run --source ${COVERAGE_SOURCE} -m pytest tests_fabric/ -v --durations=50
         workingDirectory: tests/
         displayName: "Testing: fabric standard"

.azure/gpu-tests-pytorch.yml

@@ -50,23 +50,23 @@ jobs:
     strategy:
       matrix:
         "PyTorch | oldest":
-          image: "pytorchlightning/pytorch_lightning:base-cuda-py3.10-torch2.1-cuda12.1.1"
+          image: "pytorchlightning/pytorch_lightning:base-cuda12.1.1-py3.10-torch2.1"
           PACKAGE_NAME: "pytorch"
         "PyTorch | latest":
-          image: "pytorchlightning/pytorch_lightning:base-cuda-py3.12-torch2.6-cuda12.4.1"
-          PACKAGE_NAME: "pytorch"
-        "PyTorch | future":
-          image: "pytorchlightning/pytorch_lightning:base-cuda-py3.12-torch2.7-cuda12.6.3"
+          image: "pytorchlightning/pytorch_lightning:base-cuda12.6.3-py3.12-torch2.8"
           PACKAGE_NAME: "pytorch"
+        #"PyTorch | future":
+        #  image: "pytorchlightning/pytorch_lightning:base-cuda12.6.3-py3.12-torch2.7"
+        #  PACKAGE_NAME: "pytorch"
         "Lightning | latest":
-          image: "pytorchlightning/pytorch_lightning:base-cuda-py3.12-torch2.6-cuda12.4.1"
+          image: "pytorchlightning/pytorch_lightning:base-cuda12.6.3-py3.12-torch2.8"
           PACKAGE_NAME: "lightning"
     pool: lit-rtx-3090
     variables:
       DEVICES: $( python -c 'print("$(Agent.Name)".split("_")[-1])' )
       FREEZE_REQUIREMENTS: "1"
       PIP_CACHE_DIR: "/var/tmp/pip"
-      PL_RUN_CUDA_TESTS: "1"
+      RUN_ONLY_CUDA_TESTS: "1"
     container:
       image: $(image)
       # default shm size is 64m. Increase it to avoid:
@@ -82,8 +82,6 @@ jobs:
           echo "##vso[task.setvariable variable=TORCH_URL]https://download.pytorch.org/whl/cu${cuda_ver}/torch_stable.html"
           scope=$(python -c 'n = "$(PACKAGE_NAME)" ; print(dict(pytorch="pytorch_lightning").get(n, n))')
           echo "##vso[task.setvariable variable=COVERAGE_SOURCE]$scope"
-          python_ver=$(python -c "import sys; print(f'{sys.version_info.major}{sys.version_info.minor}')")
-          echo "##vso[task.setvariable variable=PYTHON_VERSION_MM]$python_ver"
         displayName: "set env. vars"
       - bash: |
           echo "##vso[task.setvariable variable=TORCH_URL]https://download.pytorch.org/whl/test/cu${CUDA_VERSION_MM}"
@@ -104,6 +102,20 @@ jobs:
           pip list
         displayName: "Image info & NVIDIA"
 
+      - bash: |
+          set -ex
+          pip install "cython<3.0" wheel  # for compatibility
+          pip install -U "lightning-utilities[cli]"
+          cd requirements/pytorch
+          # replace range by pin minimal requirements
+          python -m lightning_utilities.cli requirements set-oldest --req_files "['base.txt', 'extra.txt', 'strategies.txt', 'examples.txt']"
+          # drop deepspeed since it is not supported by our minimal Torch requirements
+          python -m lightning_utilities.cli requirements prune-pkgs --packages deepspeed --req_files strategies.txt
+          # uninstall deepspeed since some older docker images have it pre-installed
+          pip uninstall -y deepspeed
+        condition: contains(variables['Agent.JobName'], 'oldest')
+        displayName: "setting oldest dependencies"
+
       - bash: |
           PYTORCH_VERSION=$(python -c "import torch; print(torch.__version__.split('+')[0])")
           pip install -q wget packaging
@@ -113,11 +125,22 @@ jobs:
           done
         displayName: "Adjust dependencies"
 
+      - bash: |
+          pip install -U -q -r .actions/requirements.txt
+          python .actions/assistant.py copy_replace_imports --source_dir="./tests/tests_pytorch" \
+            --source_import="lightning.fabric,lightning.pytorch" \
+            --target_import="lightning_fabric,pytorch_lightning"
+          python .actions/assistant.py copy_replace_imports --source_dir="./examples/pytorch/basics" \
+            --source_import="lightning.fabric,lightning.pytorch" \
+            --target_import="lightning_fabric,pytorch_lightning"
+        # without succeeded this could run even if the job has already failed
+        condition: and(succeeded(), eq(variables['PACKAGE_NAME'], 'pytorch'))
+        displayName: "Adjust tests & examples"
+
       - bash: |
           set -e
           extra=$(python -c "print({'lightning': 'pytorch-'}.get('$(PACKAGE_NAME)', ''))")
-          pip install -e ".[${extra}dev]" pytest-timeout -U --extra-index-url="${TORCH_URL}"
-          pip install setuptools==75.6.0 jsonargparse==4.35.0
+          pip install -e ".[${extra}dev]" -U --upgrade-strategy=eager --extra-index-url="${TORCH_URL}"
         displayName: "Install package & dependencies"
 
       - bash: pip uninstall -y lightning
@@ -144,17 +167,6 @@ jobs:
         condition: and(succeeded(), eq(variables['PACKAGE_NAME'], 'pytorch'))
         displayName: "Testing: PyTorch doctests"
 
-      - bash: |
-          python .actions/assistant.py copy_replace_imports --source_dir="./tests/tests_pytorch" \
-            --source_import="lightning.fabric,lightning.pytorch" \
-            --target_import="lightning_fabric,pytorch_lightning"
-          python .actions/assistant.py copy_replace_imports --source_dir="./examples/pytorch/basics" \
-            --source_import="lightning.fabric,lightning.pytorch" \
-            --target_import="lightning_fabric,pytorch_lightning"
-        # without succeeded this could run even if the job has already failed
-        condition: and(succeeded(), eq(variables['PACKAGE_NAME'], 'pytorch'))
-        displayName: "Adjust tests & examples"
-
       - bash: |
           bash .actions/pull_legacy_checkpoints.sh
           cd tests/legacy

.github/CODEOWNERS

@@ -18,3 +18,6 @@
 /README.md                           @williamfalcon @lantiga
 /src/pytorch_lightning/__about__.py  @williamfalcon @lantiga @borda
 /src/lightning_fabric/__about__.py   @williamfalcon @lantiga @borda
+
+/src/lightning/fabric/loggers        @williamfalcon
+/src/lightning/pytorch/loggers       @williamfalcon