o
    /ѹg                     @   s   d dl Z d dlZd dlZddlmZ ddlmZ ddlmZmZm	Z	m
Z
mZ ddlmZ ddlmZ dd	lmZ d
d Zdd Zdd ZG dd deZG dd deZdS )    N   )find_by_name)get_mul_node)TENSOR_NAME_QUANT_SUFFIXQuantizedValueQuantizedValueTypeattribute_to_kwarg	ms_domain   )QuantOperatorBase)	QOpMatMul)QDQOperatorBasec                 C   s0   dd | j D }t|rtj|d dkS dS )Nc                 S      g | ]	}|j d kr|qS )ZtransBname.0attr r   k/Users/admin/.pyenv/versions/3.10.0/lib/python3.10/site-packages/onnxruntime/quantization/operators/gemm.py
<listcomp>       z#is_B_transposed.<locals>.<listcomp>r   F	attributelenonnxhelperZget_attribute_value)	gemm_nodeZtransB_attributer   r   r   is_B_transposed   s   r   c                 C   s,   dd | j D }t|rtj|d S dS )Nc                 S   r   betar   r   r   r   r   r      r   zget_beta.<locals>.<listcomp>r         ?r   r   Zbeta_attributer   r   r   get_beta   s   r#   c                 C   s&   dd | j D }t|rd|d _dS )Nc                 S   r   r   r   r   r   r   r   r      r   z$set_default_beta.<locals>.<listcomp>r!   r   )r   r   fr"   r   r   r   set_default_beta   s   
r%   c                       s(   e Zd Z fddZ fddZ  ZS )QLinearGemmc                       t  || d S Nsuper__init__selfZonnx_quantizerZ	onnx_node	__class__r   r   r+   '      zQLinearGemm.__init__c              	      s  | j }|jdks
J | j|jd \}}}}}| j|jd r[| j r[| j|dg\}}}}	| j	|jd | jj
t|rBdnd}
||
d  ||
d  ||
d  n0| j|dg\}}}}	| jj|dg| jjd\}}}}|| || || |	| |r|d u rt  S d}t|jdkr| j|jd st  S | j|jd |jd |jd t| j }|jd t }|jr|jd nd}i }|jD ]}|jd	kr|t| qt|d
< g }tdD ]}||| || || g q||||g tjjd||g|fi |}|	| t|jd |||t j!|j| jj
d}|| jj"|jd < | j j#|	7  _#d S )NGemmr   r
   r   )reduce_range    Z_quantr    domainZQGemm)Z	node_typeZ
node_qtype)$nodeop_type	quantizerZ_get_quantization_paramsoutputis_input_a_initializerinputZis_per_channelZquantize_activationZquantize_weight_per_channelZweight_qTyper   appendZquantize_weightr2   extendr*   quantizer   Zquantize_bias_staticr#   r   r   r   updater   r	   ranger   r   Z	make_noder   r   ZInputZquantized_value_mapZ	new_nodes)r-   r6   Z
data_foundZoutput_scale_nameZoutput_zp_name_Zquantized_input_namesZzero_point_namesZscale_namesnodesZquant_weight_tupleZquantized_input_names_weightZzero_point_names_weightZscale_names_weightZnodes_weightZquantized_bias_nameZqgemm_outputZ
qgemm_namekwargsr   Zqgemm_inputsiZ
qgemm_nodeZq_outputr.   r   r   r>   *   s   





 


	zQLinearGemm.quantize__name__
__module____qualname__r+   r>   __classcell__r   r   r.   r   r&   &   s    r&   c                       s$   e Zd Z fddZdd Z  ZS )QDQGemmc                    r'   r(   r)   r,   r.   r   r   r+      r0   zQDQGemm.__init__c                 C   s
  | j }|jdks
J | j|jd  | js| j|jd  | jj|jd t|r,dndd\}}|r?| j	|jd | n	| j
|jd  t|jdkr| j|jd rv| j|j|jd |jd |jd t| j  t| j  d S td| j j d d S d S )	Nr1   r   r
   )Zdefault_axisr4   r   zBias of Gemm node 'zC' is not constant. Please exclude this node for better performance.)r6   r7   r8   Zquantize_activation_tensorr;   Zdisable_qdq_for_node_outputr9   Zis_tensor_per_channelr   Z"quantize_weight_tensor_per_channelZquantize_weight_tensorr   r:   Zquantize_bias_tensorr   r#   r%   loggingwarning)r-   r6   Zis_weight_per_channelZweight_axisr   r   r   r>      s*   
$zQDQGemm.quantizerE   r   r   r.   r   rJ      s    rJ   )rK   Znumpynpr   Zquant_utilsr   r   r   r   r   r   r	   Zbase_operatorr   matmulr   Zqdq_base_operatorr   r   r#   r%   r&   rJ   r   r   r   r   <module>   s    c