o
    i                  	   @   sl   d dl mZ ddlmZ erddlmZ ddlmZmZ e	e
Zddd	d
dddddZG dd deZdS )    )TYPE_CHECKING   )HfQuantizer   )PreTrainedModel)is_quark_availableloggingzweight_quantizer.scalezbias_quantizer.scalezinput_quantizer.scalezoutput_quantizer.scalezweight_quantizer.zero_pointzbias_quantizer.zero_pointzinput_quantizer.zero_pointzoutput_quantizer.zero_point)weight_scale
bias_scaleinput_scaleoutput_scaleweight_zero_pointbias_zero_pointinput_zero_pointoutput_zero_pointc                       s~   e Zd ZdZdZdgZdZ fddZdd Zdd
dZ	dd	de
defddZdd ZdddZdddZedd Z  ZS )QuarkHfQuantizerz?
    Quark quantizer (https://quark.docs.amd.com/latest/).
    Tquarkc                    s    t  j|fi | |j| _d S N)super__init__json_export_config)selfquantization_configkwargs	__class__ d/home/ubuntu/veenaModal/venv/lib/python3.10/site-packages/transformers/quantizers/quantizer_quark.pyr   7   s   zQuarkHfQuantizer.__init__c                 O   s   t  stdd S )NzLoading a Quark quantized model requires the `quark` library but it was not found in the environment. Please refer to https://quark.docs.amd.com/latest/install.html.)r   ImportError)r   argsr   r   r   r   validate_environment<   s
   z%QuarkHfQuantizer.validate_environmentmodelr   c                 K   s,   ddl m} ||| jj| jj| jjd |S )Nr   )_map_to_quark)pack_methodcustom_mode)quark.torch.export.apir"   r   quant_configr   r#   r$   )r   r!   r   r"   r   r   r   $_process_model_before_weight_loadingB   s   z5QuarkHfQuantizer._process_model_before_weight_loading
param_namereturnc                 K      dS )NTr   )r   r!   r(   r   r   r   r   param_needs_quantizationN      z)QuarkHfQuantizer.param_needs_quantizationc                 K   sH   ddl m} |dd }|tv r||t| }||||| d S )Nr   )_load_parameter_into_model.)modeling_utilsr-   splitCHECKPOINT_KEYSreplaceto)r   r!   paramr(   param_devicer   r-   postfixr   r   r   create_quantized_paramQ   s
   z'QuarkHfQuantizer.create_quantized_paramc                 K   s   |S r   r   )r   r!   r   r   r   r   #_process_model_after_weight_loading[   r,   z4QuarkHfQuantizer._process_model_after_weight_loadingNc                 C   r*   NFr   )r   safe_serializationr   r   r   is_serializable^   r,   z QuarkHfQuantizer.is_serializablec                 C   r*   r:   r   )r   r   r   r   is_trainablea   s   zQuarkHfQuantizer.is_trainable)r!   r   r   )__name__
__module____qualname____doc__requires_calibrationrequired_packages requires_parameters_quantizationr   r    r'   strboolr+   r8   r9   r<   propertyr=   __classcell__r   r   r   r   r   *   s    



r   N)typingr   baser   r0   r   utilsr   r   
get_loggerr>   loggerr2   r   r   r   r   r   <module>   s    
