V4.50 vs v.4.46

.circleci/config.yml

@@ -13,6 +13,7 @@ jobs:
     check_circleci_user:
         docker:
             - image: python:3.10-slim
+        resource_class: small
         parallelism: 1
         steps:
             - run: echo $CIRCLE_PROJECT_USERNAME
@@ -30,6 +31,14 @@ jobs:
         parallelism: 1
         steps:
             - checkout
+            - run: if [[ "$CIRCLE_PULL_REQUEST" == "" && "$CIRCLE_BRANCH" != "main" && "$CIRCLE_BRANCH" != *-release ]]; then echo "Not a PR, not the main branch and not a release branch, skip test!"; circleci-agent step halt; fi
+            - run: 'curl -L -H "Accept: application/vnd.github+json" -H "X-GitHub-Api-Version: 2022-11-28" https://api.github.com/repos/$CIRCLE_PROJECT_USERNAME/$CIRCLE_PROJECT_REPONAME/pulls/${CIRCLE_PULL_REQUEST##*/} >> github.txt'
+            - run: cat github.txt
+            - run: (python3 -c 'import json; from datetime import datetime; fp = open("github.txt"); data = json.load(fp); fp.close(); f = "%Y-%m-%dT%H:%M:%SZ"; created = datetime.strptime(data["created_at"], f); updated = datetime.strptime(data["updated_at"], f); s = (updated - created).total_seconds(); print(int(s))' || true) > elapsed.txt
+            - run: if [ "$(cat elapsed.txt)" == "" ]; then echo 60 > elapsed.txt; fi
+            - run: cat elapsed.txt
+            - run: if [ "$(cat elapsed.txt)" -lt "30" ]; then echo "PR is just opened, wait some actions from GitHub"; sleep 30; fi
+            - run: 'if grep -q "\"draft\": true," github.txt; then echo "draft mode, skip test!"; circleci-agent step halt; fi'
             - run: uv pip install -U -e .
             - run: echo 'export "GIT_COMMIT_MESSAGE=$(git show -s --format=%s)"' >> "$BASH_ENV" && source "$BASH_ENV"
             - run: mkdir -p test_preparation
@@ -57,15 +66,15 @@ jobs:
             - run:
                 name: "Prepare pipeline parameters"
                 command: |
-                    python utils/process_test_artifacts.py 
-            
+                    python utils/process_test_artifacts.py
+
             # To avoid too long generated_config.yaml on the continuation orb, we pass the links to the artifacts as parameters.
             # Otherwise the list of tests was just too big. Explicit is good but for that it was a limitation.
             # We used:
 
             # https://circleci.com/docs/api/v2/index.html#operation/getJobArtifacts : to get the job artifacts
             # We could not pass a nested dict, which is why we create the test_file_... parameters for every single job
-                
+
             - store_artifacts:
                 path: test_preparation/transformed_artifacts.json
             - store_artifacts:
@@ -109,7 +118,7 @@ jobs:
             - run:
                 name: "Prepare pipeline parameters"
                 command: |
-                    python utils/process_test_artifacts.py 
+                    python utils/process_test_artifacts.py
 
             # To avoid too long generated_config.yaml on the continuation orb, we pass the links to the artifacts as parameters.
             # Otherwise the list of tests was just too big. Explicit is good but for that it was a limitation.
@@ -170,7 +179,6 @@ jobs:
                   path: ~/transformers/installed.txt
             - run: python utils/check_copies.py
             - run: python utils/check_modular_conversion.py
-            - run: python utils/check_table.py
             - run: python utils/check_dummies.py
             - run: python utils/check_repo.py
             - run: python utils/check_inits.py
@@ -180,7 +188,6 @@ jobs:
             - run: make deps_table_check_updated
             - run: python utils/update_metadata.py --check-only
             - run: python utils/check_docstrings.py
-            - run: python utils/check_support_list.py
 
 workflows:
     version: 2

.circleci/create_circleci_config.py

@@ -28,21 +28,52 @@
     "TRANSFORMERS_IS_CI": True,
     "PYTEST_TIMEOUT": 120,
     "RUN_PIPELINE_TESTS": False,
-    "RUN_PT_TF_CROSS_TESTS": False,
-    "RUN_PT_FLAX_CROSS_TESTS": False,
 }
 # Disable the use of {"s": None} as the output is way too long, causing the navigation on CircleCI impractical
-COMMON_PYTEST_OPTIONS = {"max-worker-restart": 0, "dist": "loadfile", "vvv": None, "rsf":None}
+COMMON_PYTEST_OPTIONS = {"max-worker-restart": 0, "vvv": None, "rsfE":None}
 DEFAULT_DOCKER_IMAGE = [{"image": "cimg/python:3.8.12"}]
 
+# Strings that commonly appear in the output of flaky tests when they fail. These are used with `pytest-rerunfailures`
+# to rerun the tests that match these patterns.
+FLAKY_TEST_FAILURE_PATTERNS = [
+    "OSError",  # Machine/connection transient error
+    "Timeout",  # Machine/connection transient error
+    "ConnectionError",  # Connection transient error
+    "FileNotFoundError",  # Raised by `datasets` on Hub failures
+    "PIL.UnidentifiedImageError",  # Raised by `PIL.Image.open` on connection issues
+    "HTTPError",  # Also catches HfHubHTTPError
+    "AssertionError: Tensor-likes are not close!",  # `torch.testing.assert_close`, we might have unlucky random values
+    # TODO: error downloading tokenizer's `merged.txt` from hub can cause all the exceptions below. Throw and handle
+    # them under a single message.
+    "TypeError: expected str, bytes or os.PathLike object, not NoneType",
+    "TypeError: stat: path should be string, bytes, os.PathLike or integer, not NoneType",
+    "Converting from Tiktoken failed",
+    "KeyError: <class ",
+    "TypeError: not a string",
+]
+
 
 class EmptyJob:
     job_name = "empty"
 
     def to_dict(self):
+        steps = [{"run": 'ls -la'}]
+        if self.job_name == "collection_job":
+            steps.extend(
+                [
+                    "checkout",
+                    {"run": "pip install requests || true"},
+                    {"run": """while [[ $(curl --location --request GET "https://circleci.com/api/v2/workflow/$CIRCLE_WORKFLOW_ID/job" --header "Circle-Token: $CCI_TOKEN"| jq -r '.items[]|select(.name != "collection_job")|.status' | grep -c "running") -gt 0 ]]; do sleep 5; done || true"""},
+                    {"run": 'python utils/process_circleci_workflow_test_reports.py --workflow_id $CIRCLE_WORKFLOW_ID || true'},
+                    {"store_artifacts": {"path": "outputs"}},
+                    {"run": 'echo "All required jobs have now completed"'},
+                ]
+            )
+
         return {
             "docker": copy.deepcopy(DEFAULT_DOCKER_IMAGE),
-            "steps":["checkout"],
+            "resource_class": "small",
+            "steps": steps,
         }
 
 
@@ -54,9 +85,9 @@ class CircleCIJob:
     install_steps: List[str] = None
     marker: Optional[str] = None
     parallelism: Optional[int] = 0
-    pytest_num_workers: int = 12
+    pytest_num_workers: int = 8
     pytest_options: Dict[str, Any] = None
-    resource_class: Optional[str] = "2xlarge"
+    resource_class: Optional[str] = "xlarge"
     tests_to_run: Optional[List[str]] = None
     num_test_files_per_worker: Optional[int] = 10
     # This should be only used for doctest job!
@@ -112,7 +143,9 @@ def to_dict(self):
                 # Examples special case: we need to download NLTK files in advance to avoid cuncurrency issues
         timeout_cmd = f"timeout {self.command_timeout} " if self.command_timeout else ""
         marker_cmd = f"-m '{self.marker}'" if self.marker is not None else ""
-        additional_flags = f" -p no:warning -o junit_family=xunit1 --junitxml=test-results/junit.xml"
+        junit_flags = f" -p no:warning -o junit_family=xunit1 --junitxml=test-results/junit.xml"
+        joined_flaky_patterns = "|".join(FLAKY_TEST_FAILURE_PATTERNS)
+        repeat_on_failure_flags = f"--reruns 5 --reruns-delay 2 --only-rerun '({joined_flaky_patterns})'"
         parallel = f' << pipeline.parameters.{self.job_name}_parallelism >> '
         steps = [
             "checkout",
@@ -133,14 +166,14 @@ def to_dict(self):
                 "command": """dpkg-query --show --showformat='${Installed-Size}\t${Package}\n' | sort -rh | head -25 | sort -h | awk '{ package=$2; sub(".*/", "", package); printf("%.5f GB %s\n", $1/1024/1024, package)}' || true"""}
             },
             {"run": {"name": "Create `test-results` directory", "command": "mkdir test-results"}},
-            {"run": {"name": "Get files to test", "command":f'curl -L -o {self.job_name}_test_list.txt <<pipeline.parameters.{self.job_name}_test_list>>' if self.name != "pr_documentation_tests" else 'echo "Skipped"'}},
+            {"run": {"name": "Get files to test", "command":f'curl -L -o {self.job_name}_test_list.txt <<pipeline.parameters.{self.job_name}_test_list>> --header "Circle-Token: $CIRCLE_TOKEN"' if self.name != "pr_documentation_tests" else 'echo "Skipped"'}},
                         {"run": {"name": "Split tests across parallel nodes: show current parallel tests",
                     "command": f"TESTS=$(circleci tests split  --split-by=timings {self.job_name}_test_list.txt) && echo $TESTS > splitted_tests.txt && echo $TESTS | tr ' ' '\n'" if self.parallelism else f"awk '{{printf \"%s \", $0}}' {self.job_name}_test_list.txt > splitted_tests.txt"
                     }
             },
             {"run": {
                 "name": "Run tests",
-                "command": f"({timeout_cmd} python3 -m pytest {marker_cmd} -n {self.pytest_num_workers} {additional_flags} {' '.join(pytest_flags)} $(cat splitted_tests.txt) | tee tests_output.txt)"}
+                "command": f"({timeout_cmd} python3 -m pytest {marker_cmd} -n {self.pytest_num_workers} {junit_flags} {repeat_on_failure_flags} {' '.join(pytest_flags)} $(cat splitted_tests.txt) | tee tests_output.txt)"}
             },
             {"run": {"name": "Expand to show skipped tests", "when": "always", "command": f"python3 .circleci/parse_test_outputs.py --file tests_output.txt --skip"}},
             {"run": {"name": "Failed tests: show reasons",   "when": "always", "command": f"python3 .circleci/parse_test_outputs.py --file tests_output.txt --fail"}},
@@ -163,66 +196,45 @@ def job_name(self):
 
 
 # JOBS
-torch_and_tf_job = CircleCIJob(
-    "torch_and_tf",
-    docker_image=[{"image":"huggingface/transformers-torch-tf-light"}],
-    additional_env={"RUN_PT_TF_CROSS_TESTS": True},
-    marker="is_pt_tf_cross_test",
-    pytest_options={"rA": None, "durations": 0},
-)
-
-
-torch_and_flax_job = CircleCIJob(
-    "torch_and_flax",
-    additional_env={"RUN_PT_FLAX_CROSS_TESTS": True},
-    docker_image=[{"image":"huggingface/transformers-torch-jax-light"}],
-    marker="is_pt_flax_cross_test",
-    pytest_options={"rA": None, "durations": 0},
-)
-
 torch_job = CircleCIJob(
     "torch",
     docker_image=[{"image": "huggingface/transformers-torch-light"}],
     marker="not generate",
     parallelism=6,
-    pytest_num_workers=8
 )
 
 generate_job = CircleCIJob(
     "generate",
     docker_image=[{"image": "huggingface/transformers-torch-light"}],
     marker="generate",
     parallelism=6,
-    pytest_num_workers=8
 )
 
 tokenization_job = CircleCIJob(
     "tokenization",
     docker_image=[{"image": "huggingface/transformers-torch-light"}],
     parallelism=8,
-    pytest_num_workers=16
 )
 
 processor_job = CircleCIJob(
     "processors",
     docker_image=[{"image": "huggingface/transformers-torch-light"}],
     parallelism=8,
-    pytest_num_workers=6
 )
 
 tf_job = CircleCIJob(
     "tf",
     docker_image=[{"image":"huggingface/transformers-tf-light"}],
     parallelism=6,
-    pytest_num_workers=16,
 )
 
 
 flax_job = CircleCIJob(
     "flax",
     docker_image=[{"image":"huggingface/transformers-jax-light"}],
     parallelism=6,
-    pytest_num_workers=16
+    pytest_num_workers=16,
+    resource_class="2xlarge",
 )
 
 
@@ -231,7 +243,7 @@ def job_name(self):
     additional_env={"RUN_PIPELINE_TESTS": True},
     docker_image=[{"image":"huggingface/transformers-torch-light"}],
     marker="is_pipeline_test",
-    parallelism=4
+    parallelism=4,
 )
 
 
@@ -240,7 +252,7 @@ def job_name(self):
     additional_env={"RUN_PIPELINE_TESTS": True},
     docker_image=[{"image":"huggingface/transformers-tf-light"}],
     marker="is_pipeline_test",
-    parallelism=4
+    parallelism=4,
 )
 
 
@@ -257,15 +269,15 @@ def job_name(self):
     docker_image=[{"image":"huggingface/transformers-examples-torch"}],
     # TODO @ArthurZucker remove this once docker is easier to build
     install_steps=["uv venv && uv pip install . && uv pip install -r examples/pytorch/_tests_requirements.txt"],
-    pytest_num_workers=8,
+    pytest_num_workers=4,
 )
 
 
 examples_tensorflow_job = CircleCIJob(
     "examples_tensorflow",
     additional_env={"OMP_NUM_THREADS": 8},
     docker_image=[{"image":"huggingface/transformers-examples-tf"}],
-    pytest_num_workers=16,
+    pytest_num_workers=2,
 )
 
 
@@ -280,6 +292,7 @@ def job_name(self):
     ],
     marker="is_staging_test",
     pytest_num_workers=2,
+    resource_class="medium",
 )
 
 
@@ -292,13 +305,13 @@ def job_name(self):
     ],
     pytest_options={"k onnx": None},
     pytest_num_workers=1,
+    resource_class="small",
 )
 
 
 exotic_models_job = CircleCIJob(
     "exotic_models",
     docker_image=[{"image":"huggingface/transformers-exotic-models"}],
-    pytest_num_workers=12,
     parallelism=4,
     pytest_options={"durations": 100},
 )
@@ -317,7 +330,6 @@ def job_name(self):
     docker_image=[{"image": "huggingface/transformers-torch-light"}],
     marker="not generate",
     parallelism=6,
-    pytest_num_workers=8,
 )
 
 
@@ -345,13 +357,14 @@ def job_name(self):
     pytest_num_workers=1,
 )
 
-REGULAR_TESTS = [torch_and_tf_job, torch_and_flax_job, torch_job, tf_job, flax_job, hub_job, onnx_job, tokenization_job, processor_job, generate_job, non_model_job] # fmt: skip
+REGULAR_TESTS = [torch_job, tf_job, flax_job, hub_job, onnx_job, tokenization_job, processor_job, generate_job, non_model_job] # fmt: skip
 EXAMPLES_TESTS = [examples_torch_job, examples_tensorflow_job]
 PIPELINE_TESTS = [pipelines_torch_job, pipelines_tf_job]
 REPO_UTIL_TESTS = [repo_utils_job]
 DOC_TESTS = [doc_test_job]
 ALL_TESTS = REGULAR_TESTS + EXAMPLES_TESTS + PIPELINE_TESTS + REPO_UTIL_TESTS + DOC_TESTS + [custom_tokenizers_job] + [exotic_models_job]  # fmt: skip
 
+
 def create_circleci_config(folder=None):
     if folder is None:
         folder = os.getcwd()
@@ -361,7 +374,13 @@ def create_circleci_config(folder=None):
 
     if len(jobs) == 0:
         jobs = [EmptyJob()]
-    print("Full list of job name inputs", {j.job_name + "_test_list":{"type":"string", "default":''} for j in jobs})
+    else:
+        print("Full list of job name inputs", {j.job_name + "_test_list":{"type":"string", "default":''} for j in jobs})
+        # Add a job waiting all the test jobs and aggregate their test summary files at the end
+        collection_job = EmptyJob()
+        collection_job.job_name = "collection_job"
+        jobs = [collection_job] + jobs
+
     config = {
         "version": "2.1",
         "parameters": {
@@ -371,9 +390,14 @@ def create_circleci_config(folder=None):
             **{j.job_name + "_test_list":{"type":"string", "default":''} for j in jobs},
             **{j.job_name + "_parallelism":{"type":"integer", "default":1} for j in jobs},
         },
-        "jobs" : {j.job_name: j.to_dict() for j in jobs},
-        "workflows": {"version": 2, "run_tests": {"jobs": [j.job_name for j in jobs]}}
+        "jobs": {j.job_name: j.to_dict() for j in jobs}
     }
+    if "CIRCLE_TOKEN" in os.environ:
+        # For private forked repo. (e.g. new model addition)
+        config["workflows"] = {"version": 2, "run_tests": {"jobs": [{j.job_name: {"context": ["TRANSFORMERS_CONTEXT"]}} for j in jobs]}}
+    else:
+        # For public repo. (e.g. `transformers`)
+        config["workflows"] = {"version": 2, "run_tests": {"jobs": [j.job_name for j in jobs]}}
     with open(os.path.join(folder, "generated_config.yml"), "w") as f:
         f.write(yaml.dump(config, sort_keys=False, default_flow_style=False).replace("' << pipeline", " << pipeline").replace(">> '", " >>"))
 

.github/ISSUE_TEMPLATE/bug-report.yml

@@ -38,12 +38,12 @@ body:
 
           - text models: @ArthurZucker
           - vision models: @amyeroberts, @qubvel
-          - speech models: @ylacombe, @eustlb
+          - speech models: @eustlb
           - graph models: @clefourrier
 
         Library:
 
-          - flax: @sanchit-gandhi
+          - flax: @gante and @Rocketknight1
           - generate: @zucchini-nlp (visual-language models) or @gante (all others)
           - pipelines: @Rocketknight1
           - tensorflow: @gante and @Rocketknight1
@@ -72,7 +72,7 @@ body:
 
         Maintained examples (not research project or legacy):
 
-          - Flax: @sanchit-gandhi
+          - Flax: @Rocketknight1
           - PyTorch: See Models above and tag the person corresponding to the modality of the example.
           - TensorFlow: @Rocketknight1
 
@@ -106,6 +106,7 @@ body:
       label: Reproduction
       description: |
         Please provide a code sample that reproduces the problem you ran into. It can be a Colab link or just a code snippet.
+        Please include relevant config information with your code, for example your Trainers, TRL, Peft, and DeepSpeed configs.
         If you have code snippets, error messages, stack traces please provide them here as well.
         Important! Use code tags to correctly format your code. See https://help.github.com/en/github/writing-on-github/creating-and-highlighting-code-blocks#syntax-highlighting
         Do not use screenshots, as they are hard to read and (more importantly) don't allow others to copy-and-paste your code.