Update on "Prevent module full_backward_hook from erroring in double …

…backward"


Fixes #88312


[ghstack-poisoned]

.circleci/scripts/setup_ci_environment.sh

@@ -32,7 +32,7 @@ if ! command -v aws >/dev/null; then
 fi
 
 if [ -n "${USE_CUDA_DOCKER_RUNTIME:-}" ]; then
-  DRIVER_FN="NVIDIA-Linux-x86_64-515.57.run"
+  DRIVER_FN="NVIDIA-Linux-x86_64-515.76.run"
   wget "https://s3.amazonaws.com/ossci-linux/nvidia_driver/$DRIVER_FN"
   sudo /bin/bash "$DRIVER_FN" -s --no-drm || (sudo cat /var/log/nvidia-installer.log && false)
   nvidia-smi

.github/actions/filter-test-configs/action.yml

@@ -52,7 +52,8 @@ runs:
         .github/scripts/filter_test_configs.py \
           --test-matrix "${{ inputs.test-matrix }}" \
           --pr-number "${{ github.event.pull_request.number }}" \
-          --tag "${{ steps.parse-ref.outputs.tag }}"
+          --tag "${{ steps.parse-ref.outputs.tag }}" \
+          --event-name "${{ github.event_name }}"
 
     - name: Print the filtered test matrix
       shell: bash

.github/auto_request_review.yml

@@ -4,16 +4,18 @@ reviewers:
     symbolic-shapes:
       - ezyang
       - Chillee
-      - wconstab
       - anjali411
       - albanD
       - miladm
       - bdhirsh
+      - voznesenskym
+      - SherlockNoMad
 
   per_author:
     symbolic-shapes:
       - symbolic-shapes
       - antoniojkim
+      - wconstab
 
 files:
   # none yet, TODO: migrate CODEOWNERS here

.github/ci_commit_pins/triton.txt

@@ -1 +1 @@
-f16138d447bccc54641a9c48ffedbd449a1a40a7
+0d7e7532279e45672555e344646f5c19c3972331

.github/ci_commit_pins/vision.txt

@@ -1 +1 @@
-d95fbaf1efd5346a4afcf5b9953df75696432265
+ffd5a567eb90abf6b5555063da434d3c130d540f

.github/ci_commit_pins/xla.txt

@@ -1 +1 @@
-1e9b8bdc75114ac6c16305c970be37a1cd2fdb1c
+08121e41079319cd369f82f523f5a714a0563f9d

.github/merge_rules.yaml

@@ -241,9 +241,12 @@
   - third_party/ideep
   - caffe2/ideep/**
   - caffe2/python/ideep/**
+  - cmake/Modules/FindMKLDNN.cmake
+  - third_party/mkl-dnn.BUILD
   approved_by:
   - XiaobingSuper
-  - yanbing-j
+  - jgong5
+  - mingfeima
   mandatory_checks_name:
   - EasyCLA
   - Lint
@@ -256,6 +259,7 @@
   approved_by:
   - sanchitintel
   - chunyuan-w
+  - jgong5
   mandatory_checks_name:
   - EasyCLA
   - Lint
@@ -268,9 +272,11 @@
   - aten/src/ATen/native/quantized/cpu/**
   - aten/src/ATen/native/Convolution*.cpp
   - aten/src/ATen/native/mkldnn/**
+  - test/test_mkldnn.py
   approved_by:
   - mingfeima
   - XiaobingSuper
+  - jgong5
   mandatory_checks_name:
   - EasyCLA
   - Lint
@@ -283,7 +289,7 @@
   - test/test_mkldnn.py
   approved_by:
   - leslie-fang-intel
-  - CaoE
+  - jgong5
   mandatory_checks_name:
   - EasyCLA
   - Lint
@@ -297,7 +303,18 @@
   - test/test_autocast.py
   approved_by:
   - leslie-fang-intel
-  - CaoE
+  - jgong5
+  mandatory_checks_name:
+  - EasyCLA
+  - Lint
+  - pull
+
+- name: NNC
+  patterns:
+  - torch/csrc/jit/tensorexpr/**
+  approved_by:
+  - EikanWang
+  - jgong5
   mandatory_checks_name:
   - EasyCLA
   - Lint
@@ -308,11 +325,12 @@
   - torch/csrc/lazy/**
   - test/cpp/lazy/**
   - test/lazy/**
-  - codegen/api/lazy.py
-  - codegen/dest/lazy_ir.py
-  - codegen/dest/lazy_ts_lowering.py
-  - codegen/gen_lazy_tensor.py
+  - torchgen/api/lazy.py
+  - torchgen/dest/lazy_ir.py
+  - torchgen/dest/lazy_ts_lowering.py
+  - torchgen/gen_lazy_tensor.py
   - aten/src/ATen/native/ts_native_functions.yaml
+  - .github/ci_commit_pins/xla.txt
   approved_by:
   - alanwaketan
   - JackCaoG

.github/scripts/ensure_actions_will_cancel.py

@@ -42,26 +42,26 @@ def should_check(filename: Path) -> bool:
             print("ERROR: duplicate workflow name:", name, file=sys.stderr)
             errors_found = True
         names.add(name)
-
-        expected = {
-            "group": EXPECTED_GROUP,
-            "cancel-in-progress": True,
-        }
-        actual = data.get("concurrency", None)
-        if actual != expected:
+        actual = data.get("concurrency", {})
+        if not actual.get("group", "").startswith(EXPECTED_GROUP):
             print(
                 f"'concurrency' incorrect or not found in '{filename.relative_to(REPO_ROOT)}'",
                 file=sys.stderr,
             )
             print(
-                f"expected: {expected}",
+                f"concurrency group should start with {EXPECTED_GROUP} but found {actual.get('group', None)}",
                 file=sys.stderr,
             )
+            errors_found = True
+        if not actual.get("cancel-in-progress", False):
             print(
-                f"actual:   {actual}",
+                f"'concurrency' incorrect or not found in '{filename.relative_to(REPO_ROOT)}'",
+                file=sys.stderr,
+            )
+            print(
+                f"concurrency cancel-in-progress should be True but found {actual.get('cancel-in-progress', None)}",
                 file=sys.stderr,
             )
-            errors_found = True
 
     if errors_found:
         sys.exit(1)

.github/scripts/filter_test_configs.py

@@ -23,6 +23,8 @@
     "force_on_cpu",
     "functorch",
     "inductor",
+    "inductor_distributed",
+    "inductor_timm",
     "jit_legacy",
     "multigpu",
     "nogpu_AVX512",
@@ -38,6 +40,7 @@ def parse_args() -> Any:
     parser.add_argument("--test-matrix", type=str, required=True, help="the original test matrix")
     parser.add_argument("--pr-number", type=str, help="the pull request number")
     parser.add_argument("--tag", type=str, help="the associated tag if it exists")
+    parser.add_argument("--event-name", type=str, help="name of the event that triggered the job (pull, schedule, etc)")
     return parser.parse_args()
 
 
@@ -159,6 +162,10 @@ def main() -> None:
         # No PR number, no tag, we can just return the test matrix as it is
         filtered_test_matrix = test_matrix
 
+    if args.event_name == "schedule":
+        for config in filtered_test_matrix.get("include", []):
+            config["mem_leak_check"] = "mem_leak_check"
+
     # Set the filtered test matrix as the output
     set_output("test-matrix", json.dumps(filtered_test_matrix))
 

.github/scripts/generate_pytorch_version.py

@@ -23,27 +23,22 @@ def get_pytorch_root() -> Path:
 
 def get_tag() -> str:
     root = get_pytorch_root()
-    # We're on a tag
-    am_on_tag = (
-        subprocess.run(
-            ['git', 'describe', '--tags', '--exact'],
-            cwd=root,
-            stdout=subprocess.DEVNULL,
-            stderr=subprocess.DEVNULL
-        ).returncode == 0
-    )
-    tag = ""
-    if am_on_tag:
+    try:
         dirty_tag = subprocess.check_output(
-            ['git', 'describe'],
+            ['git', 'describe', '--tags', '--exact'],
             cwd=root
         ).decode('ascii').strip()
-        # Strip leading v that we typically do when we tag branches
-        # ie: v1.7.1 -> 1.7.1
-        tag = re.sub(LEADING_V_PATTERN, "", dirty_tag)
-        # Strip trailing rc pattern
-        # ie: 1.7.1-rc1 -> 1.7.1
-        tag = re.sub(TRAILING_RC_PATTERN, "", tag)
+    except subprocess.CalledProcessError:
+        return ""
+    # Strip leading v that we typically do when we tag branches
+    # ie: v1.7.1 -> 1.7.1
+    tag = re.sub(LEADING_V_PATTERN, "", dirty_tag)
+    # Strip trailing rc pattern
+    # ie: 1.7.1-rc1 -> 1.7.1
+    tag = re.sub(TRAILING_RC_PATTERN, "", tag)
+    # Ignore ciflow tags
+    if tag.startswith("ciflow/"):
+        return ""
     return tag
 
 def get_base_version() -> str: