image-match/python/py/Lib/site-packages/transformers/quantizers/quantizer_vptq.py

# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import TYPE_CHECKING

from .base import HfQuantizer


if TYPE_CHECKING:
    from ..modeling_utils import PreTrainedModel
    from ..utils.quantization_config import VptqConfig

from ..utils import is_accelerate_available, is_torch_available, is_vptq_available, logging
from ..utils.quantization_config import QuantizationConfigMixin


if is_torch_available():
    import torch

logger = logging.get_logger(__name__)


class VptqHfQuantizer(HfQuantizer):
    """
    Quantizer of the VPTQ method. Enables the loading of prequantized models.
    """

    requires_calibration = True
    quantization_config: "VptqConfig"

    def __init__(self, quantization_config: QuantizationConfigMixin, **kwargs):
        super().__init__(quantization_config, **kwargs)

    def validate_environment(self, *args, **kwargs):
        if not is_accelerate_available():
            raise ImportError("Using `vptq` quantization requires Accelerate: `pip install accelerate`")

        if not is_vptq_available():
            raise ImportError("Using `vptq` quantization requires VPTQ>=0.0.4: `pip install -U vptq`")

        if not torch.cuda.is_available():
            raise RuntimeError("GPU is required to run VTPQ quantized model.")

    def _process_model_before_weight_loading(
        self,
        model: "PreTrainedModel",
        **kwargs,
    ):
        from ..integrations import replace_with_vptq_linear

        self.modules_to_not_convert = self.get_modules_to_not_convert(
            model, self.quantization_config.modules_to_not_convert, model._keep_in_fp32_modules
        )
        replace_with_vptq_linear(
            model,
            quantization_config=self.quantization_config,
            modules_to_not_convert=self.modules_to_not_convert,
        )

    @property
    def is_trainable(self) -> bool:
        return False

    def is_serializable(self):
        return True