U T?h&@s\ddlmZddlmZmZddlmZddlmZddl m Z ee Z GdddeZ dS) ) getLogger)DictOptional)Fusion)helper) OnnxModelcsteZdZedfdd ZeedddZeeeeddd Z eeeedd d Z eeeedd d Z Z S) FusionGelu)modelcst|dddS)NGeluZErf)super__init__)selfr  __class__V/var/www/html/venv/lib/python3.8/site-packages/onnxruntime/transformers/fusion_gelu.pyr szFusionGelu.__init__)input_name_to_nodesoutput_name_to_nodecCs6||||rdS||||r$dS||||dS)N)fuse_1fuse_2fuse_3)r erf_noderrrrrfuses zFusionGelu.fuse)rrreturncCs|jd|krdS||jd}t|dks:|djdkr>dS|d}|j|dsXdS|jd|krjdS||jd}t|dks|djdkrdS|d}|j|dd|}|dkrdS|jj|ddd dkrdS|jd}|jd|jdkrdnd} ||j| krd||jd}t|dks8|djdkr Div -----> Erf --> Add --> Mul --> (B=1.4142...) (1) Pattern 2: +------------------------------------+ | | | v [root] --> Div -----> Erf --> Add --> Mul -->Mul --> (B=1.4142...) (1) (0.5) Note that constant input for Add and Mul could be first or second input: like either A=0.5 or B=0.5 is fine. rNAddMulDiv -?MbP?delta?r inputsoutputs com.microsoftT)outputlenop_typer has_constant_input match_parentfind_constant_inputinputis_safe_to_fuse_nodesnodes_to_removeextendr make_nodedomain nodes_to_addappendthis_graph_namenode_name_to_graph_namename)r rrrchildren add_after_erf mul_after_erfdivZsubgraph_inputZanothermul_halfZsubgraph_outputsubgraph_nodes fused_noderrrrsf       zFusionGelu.fuse_1c Cs(|jd|krdS||jd}t|dks:|djdkr>dS|d}|j|dsXdS|jd|krjdS||jd}t|dks|djdkrdS|d}|j|dsdS|jd|krdS||jd}t|dks|djdkrdS|d}|j|dd|}|dkrdSd} |jj|dd d dkrf|j|d d|} | dkrRdS|j| d sfdS|j|d|} | dkrdS| jd|jkrdS|||||g} | r| | |j | |jdg||sdS|j | t jd | jdg|jdgd} d| _|j | |j|j| j<dS)a& This pattern is from Keras model Fuse Gelu with Erf into one node: +------------------------------------------+ | | | v [root] --> Div -----> Erf --> Add --> Mul -->Mul (B=1.4142...) (A=1) (A=0.5) Note that constant input for Add and Mul could be first or second input: like either A=0.5 or B=0.5 is fine. rNrrrr"rrrr ZSqrtg@r r#r&T)r'r(r)r r*r+r, get_parentr-r4r.r/r0rr1r2r3r5r6r7) r rrrr8r9r:mulr;Z sqrt_node root_noder=r>rrrrksj         zFusionGelu.fuse_2c Cs|jd|krdS||jd}t|dks:|djdkr>dS|d}|j|dsXdS|jd|krjdS||jd}t|dks|djdkrdS|d}|j|dsdS|j|dd|}|dkrdS|jj|ddd }|dkrdS|j||dkrdnd|} | dkrdS|jd|kr,dS||jd}t|dksX|djdkr\dS|d} | jd| jdks| jd| jdksdS||||| g} |j | | jdg||sdS|j | t j d | jdg| jdgd } d | _|j| |j|j| j<d S)a? This pattern is from TensorFlow model Fuse Gelu with Erf into one node: +----------------------------------------------+ | | | v [root] --> Mul -----> Erf --> Add --> Mul -->Mul (A=0.7071067690849304) (B=1) (B=0.5) Note that constant input for Add and Mul could be first or second input: like either A=0.5 or B=0.5 is fine. rNrrrr"g`?rr r r#r&T)r'r(r)r r*r+r,r?r-r.r/r0rr1r2r3r4r5r6r7) r rrrr8r9r<Z first_mulirAZlast_mulr=r>rrrrs^  ,    zFusionGelu.fuse_3) __name__ __module__ __qualname__rr rrrboolrrr __classcell__rrrrrs QJrN)loggingrtypingrrZ fusion_baserZonnxrZ onnx_modelrrCloggerrrrrrs