[deps] bump transformers to 4.49.0 (#8564)

.github/workflows/tests.yml

@@ -34,9 +34,6 @@ jobs:
         transformers:
           - null
         include:  # test backward compatibility
-          - python: "3.9"
-            os: "ubuntu-latest"
-            transformers: "4.45.0"
           - python: "3.9"
             os: "ubuntu-latest"
             transformers: "4.49.0"

README.md

@@ -449,7 +449,7 @@ huggingface-cli login
 | python       | 3.9     | 3.10      |
 | torch        | 2.0.0   | 2.6.0     |
 | torchvision  | 0.15.0  | 0.21.0    |
-| transformers | 4.45.0  | 4.50.0    |
+| transformers | 4.49.0  | 4.50.0    |
 | datasets     | 2.16.0  | 3.2.0     |
 | accelerate   | 0.34.0  | 1.2.1     |
 | peft         | 0.14.0  | 0.15.1    |
@@ -491,7 +491,7 @@ cd LLaMA-Factory
 pip install -e ".[torch,metrics]" --no-build-isolation
 ```
 
-Extra dependencies available: torch, torch-npu, metrics, deepspeed, liger-kernel, bitsandbytes, hqq, eetq, gptq, aqlm, vllm, sglang, galore, apollo, badam, adam-mini, qwen, minicpm_v, modelscope, openmind, swanlab, dev
+Extra dependencies available: torch, torch-npu, metrics, deepspeed, liger-kernel, bitsandbytes, hqq, eetq, gptq, aqlm, vllm, sglang, galore, apollo, badam, adam-mini, qwen, minicpm_v, openmind, swanlab, dev
 
 #### Install from Docker Image
 

README_zh.md

@@ -451,7 +451,7 @@ huggingface-cli login
 | python       | 3.9     | 3.10      |
 | torch        | 2.0.0   | 2.6.0     |
 | torchvision  | 0.15.0  | 0.21.0    |
-| transformers | 4.45.0  | 4.50.0    |
+| transformers | 4.49.0  | 4.50.0    |
 | datasets     | 2.16.0  | 3.2.0     |
 | accelerate   | 0.34.0  | 1.2.1     |
 | peft         | 0.14.0  | 0.15.1    |
@@ -493,7 +493,7 @@ cd LLaMA-Factory
 pip install -e ".[torch,metrics]" --no-build-isolation
 ```
 
-可选的额外依赖项：torch、torch-npu、metrics、deepspeed、liger-kernel、bitsandbytes、hqq、eetq、gptq、aqlm、vllm、sglang、galore、apollo、badam、adam-mini、qwen、minicpm_v、modelscope、openmind、swanlab、dev
+可选的额外依赖项：torch、torch-npu、metrics、deepspeed、liger-kernel、bitsandbytes、hqq、eetq、gptq、aqlm、vllm、sglang、galore、apollo、badam、adam-mini、qwen、minicpm_v、openmind、swanlab、dev
 
 #### 从镜像安装
 

docker/docker-cuda/README.md

@@ -13,7 +13,7 @@ Before running the Docker container with GPU support, you need to install the fo
    # Ubuntu/Debian
    sudo apt-get update
    sudo apt-get install docker.io
-   
+
    # Or install Docker Engine from the official repository:
    # https://docs.docker.com/engine/install/
    ```
@@ -22,7 +22,7 @@ Before running the Docker container with GPU support, you need to install the fo
    ```bash
    # Ubuntu/Debian
    sudo apt-get install docker-compose
-   
+
    # Or install the latest version:
    # https://docs.docker.com/compose/install/
    ```
@@ -33,11 +33,11 @@ Before running the Docker container with GPU support, you need to install the fo
    distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
    curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
    curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
-   
+
    # Install nvidia-container-toolkit
    sudo apt-get update
    sudo apt-get install -y nvidia-container-toolkit
-   
+
    # Restart Docker to apply changes
    sudo systemctl restart docker
    ```

requirements.txt

@@ -1,27 +1,36 @@
-transformers>=4.45.0,<=4.52.4,!=4.46.*,!=4.47.*,!=4.48.0,!=4.52.0; sys_platform != 'darwin'
+# core deps
-transformers>=4.45.0,<=4.51.3,!=4.46.*,!=4.47.*,!=4.48.0,!=4.52.0; sys_platform == 'darwin'
+transformers>=4.49.0,<=4.52.4,!=4.52.0; sys_platform != 'darwin'
+transformers>=4.49.0,<=4.51.3,!=4.52.0; sys_platform == 'darwin'
 datasets>=2.16.0,<=3.6.0
 accelerate>=1.3.0,<=1.7.0
 peft>=0.14.0,<=0.15.2
 trl>=0.8.6,<=0.9.6
 tokenizers>=0.19.0,<=0.21.1
+# gui
 gradio>=4.38.0,<=5.31.0
-scipy
+matplotlib>=3.7.0
+tyro<0.9.0
+# ops
 einops
+numpy<2.0.0
+pandas>=2.0.0
+scipy
+# model and tokenizer
 sentencepiece
 tiktoken
-protobuf
+modelscope>=1.23
-uvicorn
+hf-transfer
-fastapi
+# python
-sse-starlette
-matplotlib>=3.7.0
 fire
 omegaconf
 packaging
+protobuf
 pyyaml
-numpy<2.0.0
 pydantic<=2.10.6
-pandas>=2.0.0
+# api
+uvicorn
+fastapi
+sse-starlette
+# media
 av
 librosa
-tyro<0.9.0

setup.py

@@ -68,7 +68,6 @@ extra_require = {
         "referencing",
         "jsonschema_specifications",
     ],
-    "modelscope": ["modelscope"],
     "openmind": ["openmind"],
     "swanlab": ["swanlab"],
     "dev": ["pre-commit", "ruff", "pytest", "build"],

src/llamafactory/data/mm_plugin.py

@@ -27,6 +27,10 @@ from typing import TYPE_CHECKING, BinaryIO, Literal, Optional, TypedDict, Union
 import numpy as np
 import torch
 from transformers.image_utils import get_image_size, is_valid_image, to_numpy_array
+from transformers.models.mllama.processing_mllama import (
+    convert_sparse_cross_attention_mask_to_dense,
+    get_cross_attention_token_mask,
+)
 from typing_extensions import override
 
 from ..extras.constants import AUDIO_PLACEHOLDER, IGNORE_INDEX, IMAGE_PLACEHOLDER, VIDEO_PLACEHOLDER
@@ -51,17 +55,10 @@ if is_pyav_available():
     import av
 
 
-if is_transformers_version_greater_than("4.45.0"):
-    from transformers.models.mllama.processing_mllama import (
-        convert_sparse_cross_attention_mask_to_dense,
-        get_cross_attention_token_mask,
-    )
-
-
 if is_transformers_version_greater_than("4.52.0"):
     from transformers.image_utils import make_flat_list_of_images
     from transformers.video_utils import make_batched_videos
-elif is_transformers_version_greater_than("4.49.0"):
+else:
     from transformers.image_utils import make_batched_videos, make_flat_list_of_images
 
 

src/llamafactory/extras/misc.py

@@ -35,7 +35,6 @@ from transformers.utils import (
 from transformers.utils.versions import require_version
 
 from . import logging
-from .packages import is_transformers_version_greater_than
 
 
 _is_fp16_available = is_torch_npu_available() or is_torch_cuda_available()
@@ -94,15 +93,11 @@ def check_version(requirement: str, mandatory: bool = False) -> None:
 
 def check_dependencies() -> None:
     r"""Check the version of the required packages."""
-    check_version(
+    check_version("transformers>=4.49.0,<=4.52.4,!=4.52.0")
-        "transformers>=4.45.0,<=4.52.4,!=4.46.0,!=4.46.1,!=4.46.2,!=4.46.3,!=4.47.0,!=4.47.1,!=4.48.0,!=4.52.0"
-    )
     check_version("datasets>=2.16.0,<=3.6.0")
     check_version("accelerate>=1.3.0,<=1.7.0")
     check_version("peft>=0.14.0,<=0.15.2")
     check_version("trl>=0.8.6,<=0.9.6")
-    if is_transformers_version_greater_than("4.46.0") and not is_transformers_version_greater_than("4.48.1"):
-        logger.warning_rank0_once("There are known bugs in transformers v4.46.0-v4.48.0, please use other versions.")
 
 
 def calculate_tps(dataset: list[dict[str, Any]], metrics: dict[str, float], stage: Literal["sft", "rm"]) -> float:

src/llamafactory/hparams/parser.py

@@ -348,7 +348,7 @@ def get_train_args(args: Optional[Union[dict[str, Any], list[str]]] = None) -> _
     if finetuning_args.finetuning_type == "lora":
         # https://github.com/huggingface/transformers/blob/v4.50.0/src/transformers/trainer.py#L782
         training_args.label_names = training_args.label_names or ["labels"]
-    
+
     if "swanlab" in training_args.report_to and finetuning_args.use_swanlab:
         training_args.report_to.remove("swanlab")
 

src/llamafactory/model/loader.py

@@ -19,6 +19,7 @@ import torch
 from transformers import (
     AutoConfig,
     AutoModelForCausalLM,
+    AutoModelForImageTextToText,
     AutoModelForSeq2SeqLM,
     AutoModelForTextToWaveform,
     AutoModelForVision2Seq,
@@ -29,7 +30,6 @@ from trl import AutoModelForCausalLMWithValueHead
 
 from ..extras import logging
 from ..extras.misc import count_parameters, skip_check_imports, try_download_model_from_other_hub
-from ..extras.packages import is_transformers_version_greater_than
 from .adapter import init_adapter
 from .model_utils.liger_kernel import apply_liger_kernel
 from .model_utils.misc import register_autoclass
@@ -39,10 +39,6 @@ from .model_utils.valuehead import load_valuehead_params
 from .patcher import patch_config, patch_model, patch_processor, patch_tokenizer, patch_valuehead_model
 
 
-if is_transformers_version_greater_than("4.46.0"):
-    from transformers import AutoModelForImageTextToText
-
-
 if TYPE_CHECKING:
     from transformers import PretrainedConfig, PreTrainedModel, PreTrainedTokenizer, ProcessorMixin
 
@@ -162,10 +158,7 @@ def load_model(
         else:
             if type(config) in AutoModelForVision2Seq._model_mapping.keys():  # image-text
                 load_class = AutoModelForVision2Seq
-            elif (
+            elif type(config) in AutoModelForImageTextToText._model_mapping.keys():  # image-text
-                is_transformers_version_greater_than("4.46.0")
-                and type(config) in AutoModelForImageTextToText._model_mapping.keys()
-            ):  # image-text
                 load_class = AutoModelForImageTextToText
             elif type(config) in AutoModelForSeq2SeqLM._model_mapping.keys():  # audio-text
                 load_class = AutoModelForSeq2SeqLM

tests/data/test_mm_plugin.py

@@ -238,7 +238,6 @@ def test_llama4_plugin():
     _check_plugin(**check_inputs)
 
 
-@pytest.mark.skipif(not is_transformers_version_greater_than("4.47.0"), reason="Requires transformers>=4.47.0")
 def test_llava_plugin():
     image_seqlen = 576
     tokenizer_module = _load_tokenizer_module(model_name_or_path="llava-hf/llava-1.5-7b-hf")