modeling_perception_lm_int4.py

from transformers.models.perception_lm.modeling_perception_lm import PerceptionLMForConditionalGeneration
from .int4_quant import quantize_model_to_int4
class PerceptionLMForConditionalGenerationInt4(PerceptionLMForConditionalGeneration):
    def __init__(self, config):
        super().__init__(config)
        exclude = getattr(config, "int4_exclude_patterns", [])
        quantize_model_to_int4(self, name_exclude_patterns=tuple(exclude))