v4.57.0 Branch

.circleci/create_circleci_config.py

@@ -16,10 +16,9 @@
 import argparse
 import copy
 import os
-import random
 from dataclasses import dataclass
-from typing import Any, Dict, List, Optional
-import glob
+from typing import Any, Optional
+
 import yaml
 
 
@@ -30,6 +29,7 @@
     "RUN_PIPELINE_TESTS": False,
     # will be adjust in `CircleCIJob.to_dict`.
     "RUN_FLAKY": True,
+    "DISABLE_SAFETENSORS_CONVERSION": True,
 }
 # 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, "vvv": None, "rsfE":None}
@@ -82,15 +82,15 @@ def to_dict(self):
 @dataclass
 class CircleCIJob:
     name: str
-    additional_env: Dict[str, Any] = None
-    docker_image: List[Dict[str, str]] = None
-    install_steps: List[str] = None
+    additional_env: dict[str, Any] = None
+    docker_image: list[dict[str, str]] = None
+    install_steps: list[str] = None
     marker: Optional[str] = None
     parallelism: Optional[int] = 0
     pytest_num_workers: int = 8
-    pytest_options: Dict[str, Any] = None
+    pytest_options: dict[str, Any] = None
     resource_class: Optional[str] = "xlarge"
-    tests_to_run: Optional[List[str]] = None
+    tests_to_run: Optional[list[str]] = None
     num_test_files_per_worker: Optional[int] = 10
     # This should be only used for doctest job!
     command_timeout: Optional[int] = None
@@ -130,6 +130,12 @@ def __post_init__(self):
 
     def to_dict(self):
         env = COMMON_ENV_VARIABLES.copy()
+        if self.job_name != "tests_hub":
+            # fmt: off
+            # not critical
+            env.update({"HF_TOKEN": "".join(["h", "f", "_", "H", "o", "d", "V", "u", "M", "q", "b", "R", "m", "t", "b", "z", "F", "Q", "O", "Q", "A", "J", "G", "D", "l", "V", "Q", "r", "R", "N", "w", "D", "M", "V", "C", "s", "d"])})
+            # fmt: on
+
         # Do not run tests decorated by @is_flaky on pull requests
         env['RUN_FLAKY'] = os.environ.get("CIRCLE_PULL_REQUEST", "") == ""
         env.update(self.additional_env)
@@ -149,7 +155,7 @@ 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 ""
-        junit_flags = f" -p no:warning -o junit_family=xunit1 --junitxml=test-results/junit.xml"
+        junit_flags = " -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 >> '
@@ -180,6 +186,7 @@ def to_dict(self):
             # During the CircleCI docker images build time, we might already (or not) download the data.
             # If it's done already, the files are inside the directory `/test_data/`.
             {"run": {"name": "fetch hub objects before pytest", "command": "cp -r /test_data/* . 2>/dev/null || true; python3 utils/fetch_hub_objects_for_ci.py"}},
+            {"run": {"name": "download and unzip hub cache", "command": 'curl -L -o huggingface-cache.tar.gz https://huggingface.co/datasets/hf-internal-testing/hf_hub_cache/resolve/main/huggingface-cache.tar.gz && apt-get install pigz && tar --use-compress-program="pigz -d -p 8" -xf huggingface-cache.tar.gz && mv -n hub/* /root/.cache/huggingface/hub/ && ls -la /root/.cache/huggingface/hub/'}},
             {"run": {
                 "name": "Run tests",
                 "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)"}
@@ -200,9 +207,9 @@ def to_dict(self):
                         fi"""
                 },
             },
-            {"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"}},
-            {"run": {"name": "Errors",                       "when": "always", "command": f"python3 .circleci/parse_test_outputs.py --file tests_output.txt --errors"}},
+            {"run": {"name": "Expand to show skipped tests", "when": "always", "command": "python3 .circleci/parse_test_outputs.py --file tests_output.txt --skip"}},
+            {"run": {"name": "Failed tests: show reasons",   "when": "always", "command": "python3 .circleci/parse_test_outputs.py --file tests_output.txt --fail"}},
+            {"run": {"name": "Errors",                       "when": "always", "command": "python3 .circleci/parse_test_outputs.py --file tests_output.txt --errors"}},
             {"store_test_results": {"path": "test-results"}},
             {"store_artifacts": {"path": "test-results/junit.xml"}},
             {"store_artifacts": {"path": "reports"}},

.circleci/parse_test_outputs.py

@@ -1,5 +1,6 @@
-import re
 import argparse
+import re
+
 
 def parse_pytest_output(file_path):
     skipped_tests = {}

.github/ISSUE_TEMPLATE/bug-report.yml

@@ -61,6 +61,7 @@ body:
           - Big Model Inference: @SunMarc
           - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber
           - kernels: @MekkCyber @drbh
+          - peft: @BenjaminBossan @githubnemo
 
         Devices/Backends:
 

.github/PULL_REQUEST_TEMPLATE.md

@@ -39,41 +39,41 @@ members/contributors who may be interested in your PR.
 
 Models:
 
-- text models: @ArthurZucker
-- vision models: @amyeroberts, @qubvel
-- speech models: @eustlb
+- text models: @ArthurZucker @Cyrilvallez
+- vision models: @yonigozlan @molbap
+- audio models: @eustlb @ebezzam @vasqu
+- multimodal models: @zucchini-nlp
 - graph models: @clefourrier
 
 Library:
 
-- flax: @gante and @Rocketknight1
 - generate: @zucchini-nlp (visual-language models) or @gante (all others)
+- continuous batching: @remi-or @ArthurZucker @McPatate
 - pipelines: @Rocketknight1
-- tensorflow: @gante and @Rocketknight1
-- tokenizers: @ArthurZucker
-- trainer: @zach-huggingface, @SunMarc and @qgallouedec
-- chat templates: @Rocketknight1
+- tokenizers: @ArthurZucker and @itazap
+- trainer: @zach-huggingface @SunMarc
+- attention: @vasqu @ArthurZucker @CyrilVallez
+- model loading (from pretrained, etc): @CyrilVallez
+- distributed: @3outeille @ArthurZucker @S1ro1
+- CIs: @ydshieh
 
 Integrations:
 
 - deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface
 - ray/raytune: @richardliaw, @amogkam
 - Big Model Inference: @SunMarc
 - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber
+- kernels: @MekkCyber @drbh
+- peft: @BenjaminBossan @githubnemo
 
-Documentation: @stevhliu
-
-HF projects:
+Devices/Backends:
 
-- accelerate: [different repo](https://github.com/huggingface/accelerate)
-- datasets: [different repo](https://github.com/huggingface/datasets)
-- diffusers: [different repo](https://github.com/huggingface/diffusers)
-- rust tokenizers: [different repo](https://github.com/huggingface/tokenizers)
+- AMD ROCm: @ivarflakstad
+- Intel XPU: @IlyasMoutawwakil
+- Ascend NPU: @ivarflakstad 
 
-Maintained examples (not research project or legacy):
+Documentation: @stevhliu
 
-- Flax: @Rocketknight1
-- PyTorch: See Models above and tag the person corresponding to the modality of the example.
-- TensorFlow: @Rocketknight1
+Research projects are not maintained and should be taken as is.
 
  -->

.github/scripts/assign_reviewers.py

@@ -13,14 +13,16 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import os
-import github
 import json
-from github import Github
+import os
 import re
 from collections import Counter
 from pathlib import Path
 
+import github
+from github import Github
+
+
 def pattern_to_regex(pattern):
     if pattern.startswith("/"):
         start_anchor = True

.github/scripts/codeowners_for_review_action

@@ -7,8 +7,8 @@ docs/ @stevhliu
 /docker/ @ydshieh @ArthurZucker
 
 # More high-level globs catch cases when specific rules later don't apply
-/src/transformers/models/*/processing* @molbap @yonigozlan @qubvel
-/src/transformers/models/*/image_processing* @qubvel
+/src/transformers/models/*/processing* @molbap @yonigozlan
+/src/transformers/models/*/image_processing* @yonigozlan
 /src/transformers/models/*/image_processing_*_fast* @yonigozlan
 
 # Owners of subsections of the library
@@ -186,65 +186,65 @@ trainer_utils.py @zach-huggingface @SunMarc
 /src/transformers/models/zamba/mod*_zamba* @ArthurZucker
 
 # Vision models
-/src/transformers/models/beit/mod*_beit* @amyeroberts @qubvel
-/src/transformers/models/bit/mod*_bit* @amyeroberts @qubvel
-/src/transformers/models/conditional_detr/mod*_conditional_detr* @amyeroberts @qubvel
-/src/transformers/models/convnext/mod*_convnext* @amyeroberts @qubvel
-/src/transformers/models/convnextv2/mod*_convnextv2* @amyeroberts @qubvel
-/src/transformers/models/cvt/mod*_cvt* @amyeroberts @qubvel
-/src/transformers/models/deformable_detr/mod*_deformable_detr* @amyeroberts @qubvel
-/src/transformers/models/deit/mod*_deit* @amyeroberts @qubvel
-/src/transformers/models/depth_anything/mod*_depth_anything* @amyeroberts @qubvel
-/src/transformers/models/depth_anything_v2/mod*_depth_anything_v2* @amyeroberts @qubvel
-/src/transformers/models/deta/mod*_deta* @amyeroberts @qubvel
-/src/transformers/models/detr/mod*_detr* @amyeroberts @qubvel
-/src/transformers/models/dinat/mod*_dinat* @amyeroberts @qubvel
-/src/transformers/models/dinov2/mod*_dinov2* @amyeroberts @qubvel
-/src/transformers/models/dinov2_with_registers/mod*_dinov2_with_registers* @amyeroberts @qubvel
-/src/transformers/models/dit/mod*_dit* @amyeroberts @qubvel
-/src/transformers/models/dpt/mod*_dpt* @amyeroberts @qubvel
-/src/transformers/models/efficientformer/mod*_efficientformer* @amyeroberts @qubvel
-/src/transformers/models/efficientnet/mod*_efficientnet* @amyeroberts @qubvel
-/src/transformers/models/focalnet/mod*_focalnet* @amyeroberts @qubvel
-/src/transformers/models/glpn/mod*_glpn* @amyeroberts @qubvel
-/src/transformers/models/hiera/mod*_hiera* @amyeroberts @qubvel
-/src/transformers/models/ijepa/mod*_ijepa* @amyeroberts @qubvel
-/src/transformers/models/imagegpt/mod*_imagegpt* @amyeroberts @qubvel
-/src/transformers/models/levit/mod*_levit* @amyeroberts @qubvel
-/src/transformers/models/mask2former/mod*_mask2former* @amyeroberts @qubvel
-/src/transformers/models/maskformer/mod*_maskformer* @amyeroberts @qubvel
-/src/transformers/models/mobilenet_v1/mod*_mobilenet_v1* @amyeroberts @qubvel
-/src/transformers/models/mobilenet_v2/mod*_mobilenet_v2* @amyeroberts @qubvel
-/src/transformers/models/mobilevit/mod*_mobilevit* @amyeroberts @qubvel
-/src/transformers/models/mobilevitv2/mod*_mobilevitv2* @amyeroberts @qubvel
-/src/transformers/models/nat/mod*_nat* @amyeroberts @qubvel
-/src/transformers/models/poolformer/mod*_poolformer* @amyeroberts @qubvel
-/src/transformers/models/pvt/mod*_pvt* @amyeroberts @qubvel
-/src/transformers/models/pvt_v2/mod*_pvt_v2* @amyeroberts @qubvel
-/src/transformers/models/regnet/mod*_regnet* @amyeroberts @qubvel
-/src/transformers/models/resnet/mod*_resnet* @amyeroberts @qubvel
-/src/transformers/models/rt_detr/mod*_rt_detr* @amyeroberts @qubvel
-/src/transformers/models/segformer/mod*_segformer* @amyeroberts @qubvel
-/src/transformers/models/seggpt/mod*_seggpt* @amyeroberts @qubvel
-/src/transformers/models/superpoint/mod*_superpoint* @amyeroberts @qubvel
-/src/transformers/models/swiftformer/mod*_swiftformer* @amyeroberts @qubvel
-/src/transformers/models/swin/mod*_swin* @amyeroberts @qubvel
-/src/transformers/models/swinv2/mod*_swinv2* @amyeroberts @qubvel
-/src/transformers/models/swin2sr/mod*_swin2sr* @amyeroberts @qubvel
-/src/transformers/models/table_transformer/mod*_table_transformer* @amyeroberts @qubvel
-/src/transformers/models/textnet/mod*_textnet* @amyeroberts @qubvel
-/src/transformers/models/timm_wrapper/mod*_timm_wrapper* @amyeroberts @qubvel
-/src/transformers/models/upernet/mod*_upernet* @amyeroberts @qubvel
-/src/transformers/models/van/mod*_van* @amyeroberts @qubvel
-/src/transformers/models/vit/mod*_vit* @amyeroberts @qubvel
-/src/transformers/models/vit_hybrid/mod*_vit_hybrid* @amyeroberts @qubvel
-/src/transformers/models/vitdet/mod*_vitdet* @amyeroberts @qubvel
-/src/transformers/models/vit_mae/mod*_vit_mae* @amyeroberts @qubvel
-/src/transformers/models/vitmatte/mod*_vitmatte* @amyeroberts @qubvel
-/src/transformers/models/vit_msn/mod*_vit_msn* @amyeroberts @qubvel
-/src/transformers/models/vitpose/mod*_vitpose* @amyeroberts @qubvel
-/src/transformers/models/yolos/mod*_yolos* @amyeroberts @qubvel
-/src/transformers/models/zoedepth/mod*_zoedepth* @amyeroberts @qubvel
+/src/transformers/models/beit/mod*_beit* @yonigozlan @molbap
+/src/transformers/models/bit/mod*_bit* @yonigozlan @molbap
+/src/transformers/models/conditional_detr/mod*_conditional_detr* @yonigozlan @molbap
+/src/transformers/models/convnext/mod*_convnext* @yonigozlan @molbap
+/src/transformers/models/convnextv2/mod*_convnextv2* @yonigozlan @molbap
+/src/transformers/models/cvt/mod*_cvt* @yonigozlan @molbap
+/src/transformers/models/deformable_detr/mod*_deformable_detr* @yonigozlan @molbap
+/src/transformers/models/deit/mod*_deit* @yonigozlan @molbap
+/src/transformers/models/depth_anything/mod*_depth_anything* @yonigozlan @molbap
+/src/transformers/models/depth_anything_v2/mod*_depth_anything_v2* @yonigozlan @molbap
+/src/transformers/models/deta/mod*_deta* @yonigozlan @molbap
+/src/transformers/models/detr/mod*_detr* @yonigozlan @molbap
+/src/transformers/models/dinat/mod*_dinat* @yonigozlan @molbap
+/src/transformers/models/dinov2/mod*_dinov2* @yonigozlan @molbap
+/src/transformers/models/dinov2_with_registers/mod*_dinov2_with_registers* @yonigozlan @molbap
+/src/transformers/models/dit/mod*_dit* @yonigozlan @molbap
+/src/transformers/models/dpt/mod*_dpt* @yonigozlan @molbap
+/src/transformers/models/efficientformer/mod*_efficientformer* @yonigozlan @molbap
+/src/transformers/models/efficientnet/mod*_efficientnet* @yonigozlan @molbap
+/src/transformers/models/focalnet/mod*_focalnet* @yonigozlan @molbap
+/src/transformers/models/glpn/mod*_glpn* @yonigozlan @molbap
+/src/transformers/models/hiera/mod*_hiera* @yonigozlan @molbap
+/src/transformers/models/ijepa/mod*_ijepa* @yonigozlan @molbap
+/src/transformers/models/imagegpt/mod*_imagegpt* @yonigozlan @molbap
+/src/transformers/models/levit/mod*_levit* @yonigozlan @molbap
+/src/transformers/models/mask2former/mod*_mask2former* @yonigozlan @molbap
+/src/transformers/models/maskformer/mod*_maskformer* @yonigozlan @molbap
+/src/transformers/models/mobilenet_v1/mod*_mobilenet_v1* @yonigozlan @molbap
+/src/transformers/models/mobilenet_v2/mod*_mobilenet_v2* @yonigozlan @molbap
+/src/transformers/models/mobilevit/mod*_mobilevit* @yonigozlan @molbap
+/src/transformers/models/mobilevitv2/mod*_mobilevitv2* @yonigozlan @molbap
+/src/transformers/models/nat/mod*_nat* @yonigozlan @molbap
+/src/transformers/models/poolformer/mod*_poolformer* @yonigozlan @molbap
+/src/transformers/models/pvt/mod*_pvt* @yonigozlan @molbap
+/src/transformers/models/pvt_v2/mod*_pvt_v2* @yonigozlan @molbap
+/src/transformers/models/regnet/mod*_regnet* @yonigozlan @molbap
+/src/transformers/models/resnet/mod*_resnet* @yonigozlan @molbap
+/src/transformers/models/rt_detr/mod*_rt_detr* @yonigozlan @molbap
+/src/transformers/models/segformer/mod*_segformer* @yonigozlan @molbap
+/src/transformers/models/seggpt/mod*_seggpt* @yonigozlan @molbap
+/src/transformers/models/superpoint/mod*_superpoint* @yonigozlan @molbap
+/src/transformers/models/swiftformer/mod*_swiftformer* @yonigozlan @molbap
+/src/transformers/models/swin/mod*_swin* @yonigozlan @molbap
+/src/transformers/models/swinv2/mod*_swinv2* @yonigozlan @molbap
+/src/transformers/models/swin2sr/mod*_swin2sr* @yonigozlan @molbap
+/src/transformers/models/table_transformer/mod*_table_transformer* @yonigozlan @molbap
+/src/transformers/models/textnet/mod*_textnet* @yonigozlan @molbap
+/src/transformers/models/timm_wrapper/mod*_timm_wrapper* @yonigozlan @molbap
+/src/transformers/models/upernet/mod*_upernet* @yonigozlan @molbap
+/src/transformers/models/van/mod*_van* @yonigozlan @molbap
+/src/transformers/models/vit/mod*_vit* @yonigozlan @molbap
+/src/transformers/models/vit_hybrid/mod*_vit_hybrid* @yonigozlan @molbap
+/src/transformers/models/vitdet/mod*_vitdet* @yonigozlan @molbap
+/src/transformers/models/vit_mae/mod*_vit_mae* @yonigozlan @molbap
+/src/transformers/models/vitmatte/mod*_vitmatte* @yonigozlan @molbap
+/src/transformers/models/vit_msn/mod*_vit_msn* @yonigozlan @molbap
+/src/transformers/models/vitpose/mod*_vitpose* @yonigozlan @molbap
+/src/transformers/models/yolos/mod*_yolos* @yonigozlan @molbap
+/src/transformers/models/zoedepth/mod*_zoedepth* @yonigozlan @molbap
 
 # Audio models
 /src/transformers/models/audio_spectrogram_transformer/mod*_audio_spectrogram_transformer* @eustlb
@@ -304,7 +304,7 @@ trainer_utils.py @zach-huggingface @SunMarc
 /src/transformers/models/donut/mod*_donut* @zucchini-nlp
 /src/transformers/models/flava/mod*_flava* @zucchini-nlp
 /src/transformers/models/git/mod*_git* @zucchini-nlp
-/src/transformers/models/grounding_dino/mod*_grounding_dino* @qubvel
+/src/transformers/models/grounding_dino/mod*_grounding_dino* @yonigozlan
 /src/transformers/models/groupvit/mod*_groupvit* @zucchini-nlp
 /src/transformers/models/idefics/mod*_idefics* @zucchini-nlp
 /src/transformers/models/idefics2/mod*_idefics2* @zucchini-nlp
@@ -326,10 +326,10 @@ trainer_utils.py @zach-huggingface @SunMarc
 /src/transformers/models/mgp_str/mod*_mgp_str* @zucchini-nlp
 /src/transformers/models/mllama/mod*_mllama* @zucchini-nlp
 /src/transformers/models/nougat/mod*_nougat* @NielsRogge
-/src/transformers/models/omdet_turbo/mod*_omdet_turbo* @qubvel @yonigozlan
+/src/transformers/models/omdet_turbo/mod*_omdet_turbo* @yonigozlan
 /src/transformers/models/oneformer/mod*_oneformer* @zucchini-nlp
-/src/transformers/models/owlvit/mod*_owlvit* @qubvel
-/src/transformers/models/owlv2/mod*_owlv2* @qubvel
+/src/transformers/models/owlvit/mod*_owlvit* @yonigozlan
+/src/transformers/models/owlv2/mod*_owlv2* @yonigozlan
 /src/transformers/models/paligemma/mod*_paligemma* @zucchini-nlp @molbap
 /src/transformers/models/perceiver/mod*_perceiver* @zucchini-nlp
 /src/transformers/models/pix2struct/mod*_pix2struct* @zucchini-nlp