rippertnt's picture
omniparser-florence2-ov (549456030 bytes)
6de1d94 verified
Raw History Blame Contribute Delete
327 kB
<?xml version="1.0"?> <net name="Model9" version="11"> <layers> <layer id="0" name="attention_mask" type="Parameter" version="opset1"> <data shape="?,?" element_type="f32" /> <output> <port id="0" precision="FP32" names="69,attention_mask"> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="1" name="inputs_embeds" type="Parameter" version="opset1"> <data shape="?,?,?" element_type="f32" /> <output> <port id="0" precision="FP32" names="46,47,inputs_embeds"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="2" name="self.model.embed_positions.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1026, 768" offset="0" size="1575936" /> <output> <port id="0" precision="FP16" names="self.model.embed_positions.weight"> <dim>1026</dim> <dim>768</dim> </port> </output> </layer> <layer id="3" name="self.model.embed_positions.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1026</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1026</dim> <dim>768</dim> </port> </output> </layer> <layer id="4" name="31" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575936" size="8" /> <output> <port id="0" precision="I64" names="31" /> </output> </layer> <layer id="5" name="ShapeOf_154554" type="ShapeOf" version="opset3"> <data output_type="i64" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="6" name="Constant_154555" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575944" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="7" name="Constant_154556" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="8" name="Gather_154557" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I64"> <dim>3</dim> </port> <port id="1" precision="I64" /> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="I64" names="51,53" /> </output> </layer> <layer id="9" name="__module.model.embed_positions/aten::arange/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="1575952" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="10" name="__module.model.embed_positions/aten::arange/Range" type="Range" version="opset4"> <data output_type="i64" /> <input> <port id="0" precision="I64" /> <port id="1" precision="I64" /> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="I64" names="55"> <dim>-1</dim> </port> </output> </layer> <layer id="11" name="Constant_153464" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575956" size="8" /> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="12" name="Add_150171" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I64"> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="2" precision="I64"> <dim>-1</dim> </port> </output> </layer> <layer id="13" name="Constant_150173" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575944" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="14" name="Constant_154558" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575944" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="15" name="Reshape_154559" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="I64" /> <port id="1" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="2" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="16" name="Concat_150174" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="2" precision="I64"> <dim>2</dim> </port> </output> </layer> <layer id="17" name="Constant_154563" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="18" name="Constant_154564" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="19" name="Gather_154565" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I64"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="I64" names="50"> <dim>1</dim> </port> </output> </layer> <layer id="20" name="Constant_148507" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575944" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="21" name="__module.model.embed_positions/prim::ListConstruct/SequenceMark" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="2" precision="I64"> <dim>2</dim> </port> </output> </layer> <layer id="22" name="Maximum_150175" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I64"> <dim>2</dim> </port> <port id="1" precision="I64"> <dim>2</dim> </port> </input> <output> <port id="2" precision="I64"> <dim>2</dim> </port> </output> </layer> <layer id="23" name="__module.model.embed_positions/aten::add/Add_1" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="I64"> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>2</dim> </port> </input> <output> <port id="2" precision="I64" names="58,input.1"> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="24" name="__module.model.embed_positions/aten::embedding/Convert" type="Convert" version="opset1"> <data destination_type="i32" /> <input> <port id="0" precision="I64"> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="25" name="__module.model.embed_positions/aten::embedding/Constant" type="Const" version="opset1"> <data element_type="i32" shape="" offset="1575964" size="4" /> <output> <port id="0" precision="I32" /> </output> </layer> <layer id="26" name="__module.model.embed_positions/aten::embedding/Gather" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="FP32"> <dim>1026</dim> <dim>768</dim> </port> <port id="1" precision="I32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="2" precision="I32" /> </input> <output> <port id="3" precision="FP32" names="59,60,embed_pos,embed_pos.1"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="27" name="__module.model/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="61,input.3"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="28" name="__module.model.layernorm_embedding/aten::layer_norm/Multiply" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="1575968" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="29" name="__module.model.layernorm_embedding/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="30" name="Constant_153465_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="1575972" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="31" name="Constant_153465" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="32" name="__module.model.layernorm_embedding/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="33" name="Constant_153466_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="1577508" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="34" name="Constant_153466" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="35" name="__module.model.layernorm_embedding/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="65,input.5"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="36" name="self.model.layers.0.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="1579044" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.0.self_attn.q_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="37" name="self.model.layers.0.self_attn.q_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="38" name="__module.model.layers.0.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="39" name="Constant_153467_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 1" offset="2758692" size="2" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="40" name="Constant_153467" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="41" name="Multiply_153286" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="42" name="Constant_153468_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="2758694" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="43" name="Constant_153468" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="44" name="__module.model.layers.0.self_attn/aten::mul/Multiply" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="94,tensor.5"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="45" name="Constant_154744" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="46" name="__module.model.layers.0.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="114"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="47" name="__module.model.layers.0.self_attn/aten::transpose/Constant" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="48" name="__module.model.layers.0.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="115"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="49" name="ShapeOf_154574" type="ShapeOf" version="opset3"> <data output_type="i64" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="50" name="Constant_154575" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="51" name="Constant_154576" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="52" name="Gather_154577" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I64"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="I64" names="173,258,343,428,513,88"> <dim>1</dim> </port> </output> </layer> <layer id="53" name="Constant_148489" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="2760278" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="54" name="__module.model.layers.0.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="2" precision="I64" names="109,110,111,112,129,130,194,195,196,197,214,215,279,280,281,282,299,300,364,365,366,367,384,385,449,450,451,452,469,470,534,535,536,537,554,555"> <dim>1</dim> </port> </output> </layer> <layer id="55" name="Constant_131055" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="2760286" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="56" name="Constant_131056" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="2760294" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="57" name="__module.model.layers.0.self_attn/prim::ListConstruct/SequenceMark_1" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="3" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="58" name="__module.model.layers.0.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="118,query_states.1"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="59" name="self.model.layers.0.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="2760302" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.0.self_attn.k_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="60" name="self.model.layers.0.self_attn.k_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="61" name="__module.model.layers.0.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="62" name="Constant_153469_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="3939950" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="63" name="Constant_153469" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="64" name="__module.model.layers.0.self_attn.k_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="97,tensor.1"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="65" name="Constant_154745" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="66" name="__module.model.layers.0.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="99"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="67" name="__module.model.layers.0.self_attn/aten::transpose/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="68" name="__module.model.layers.0.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="100"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="69" name="__module.model.layers.0.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="120,key_states.3"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="70" name="__module.model.layers.0.self_attn/aten::bmm/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="2" precision="FP32" names="125,attn_weights.1"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="71" name="Constant_131449" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="2760278" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="72" name="Constant_154593" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575944" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="73" name="Constant_154594" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="74" name="Gather_154595" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I64"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="I64" names="175,260,345,430,515,90"> <dim>1</dim> </port> </output> </layer> <layer id="75" name="ShapeOf_154598" type="ShapeOf" version="opset3"> <data output_type="i64" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="1" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="76" name="Constant_154599" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575944" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="77" name="Constant_154600" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="78" name="Gather_154601" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I64"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="I64" names="123"> <dim>1</dim> </port> </output> </layer> <layer id="79" name="__module.model.layers.0.self_attn/prim::ListConstruct/SequenceMark_4" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="80" name="__module.model.layers.0.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="127"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="81" name="Constant_153470_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 1, 1" offset="3941486" size="2" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="82" name="Constant_153470" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="83" name="29" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575944" size="8" /> <output> <port id="0" precision="I64" names="29" /> </output> </layer> <layer id="84" name="__module.model/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I64" /> </input> <output> <port id="2" precision="FP32" names="70"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="85" name="28" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575956" size="8" /> <output> <port id="0" precision="I64" names="28" /> </output> </layer> <layer id="86" name="__module.model/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> </port> <port id="1" precision="I64" /> </input> <output> <port id="2" precision="FP32" names="71,72"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="87" name="Subtract_150245" type="Subtract" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="88" name="ShapeOf_150246" type="ShapeOf" version="opset3"> <data output_type="i64" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="89" name="Constant_154585" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="90" name="Constant_154586" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="91" name="Gather_154587" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I64"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="I64" names="67"> <dim>1</dim> </port> </output> </layer> <layer id="92" name="Constant_148513" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575944" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="93" name="__module.model/aten::size/ShapeOf" type="ShapeOf" version="opset3"> <data output_type="i64" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="I64"> <dim>2</dim> </port> </output> </layer> <layer id="94" name="Constant_148594" type="Const" version="opset1"> <data element_type="i64" shape="2" offset="3941488" size="16" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>2</dim> </port> </output> </layer> <layer id="95" name="Constant_148595" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="96" name="Gather_148596" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I64"> <dim>2</dim> </port> <port id="1" precision="I64"> <dim>2</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="I64"> <dim>2</dim> </port> </output> </layer> <layer id="97" name="__module.model/prim::ListConstruct/SequenceMark" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>2</dim> </port> </input> <output> <port id="3" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="98" name="Maximum_150247" type="Maximum" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="I64"> <dim>4</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="99" name="__module.model/aten::rsub/Subtract" type="Broadcast" version="opset3"> <data mode="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>1</dim> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="76,inverted_mask"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="100" name="__module.model/aten::to/Convert_2" type="Convert" version="opset1"> <data destination_type="boolean" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="BOOL" names="77"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="101" name="__module.model/aten::masked_fill/ConvertLike" type="Const" version="opset1"> <data element_type="f32" shape="" offset="3941504" size="4" /> <output> <port id="0" precision="FP32" /> </output> </layer> <layer id="102" name="__module.model/aten::masked_fill/Select" type="Select" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="BOOL"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32" /> <port id="2" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="3" precision="FP32" names="78,attention_mask_1"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="103" name="__module.model.layers.0.self_attn/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32" names="128,attn_weights.3"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="104" name="__module.model.layers.0.self_attn/prim::ListConstruct/SequenceMark_5" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="3" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="105" name="__module.model.layers.0.self_attn/aten::view/Reshape_4" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="132,input.7"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="106" name="__module.model.layers.0.self_attn/aten::softmax/Softmax" type="SoftMax" version="opset8"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="FP32" names="133,input.9"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="107" name="self.model.layers.0.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="3941508" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.0.self_attn.v_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="108" name="self.model.layers.0.self_attn.v_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="109" name="__module.model.layers.0.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="110" name="Constant_153471_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="5121156" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="111" name="Constant_153471" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="112" name="__module.model.layers.0.self_attn.v_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="104,tensor.3"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="113" name="Constant_154746" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="114" name="__module.model.layers.0.self_attn/aten::view/Reshape_5" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="106"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="115" name="__module.model.layers.0.self_attn/aten::transpose/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="116" name="__module.model.layers.0.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="107"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="117" name="__module.model.layers.0.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="122,value_states.3"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="118" name="__module.model.layers.0.self_attn/aten::bmm/MatMul_1" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="2" precision="FP32" names="135,attn_output.1"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="119" name="Constant_131610" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="2760278" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="120" name="Constant_131611" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="2760294" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="121" name="__module.model.layers.0.self_attn/prim::ListConstruct/SequenceMark_8" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="122" name="__module.model.layers.0.self_attn/aten::view/Reshape_6" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="137,attn_output.3"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="123" name="__module.model.layers.0.self_attn/aten::transpose/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="124" name="__module.model.layers.0.self_attn/aten::transpose/Transpose_4" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="138,attn_output.5"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="125" name="Constant_154747" type="Const" version="opset1"> <data element_type="i64" shape="3" offset="5122692" size="24" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="126" name="__module.model.layers.0.self_attn/aten::reshape/Reshape_2" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="140,input.11"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="127" name="self.model.layers.0.self_attn.out_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="5122716" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.0.self_attn.out_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="128" name="self.model.layers.0.self_attn.out_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="129" name="__module.model.layers.0.self_attn.out_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="130" name="Constant_153472_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="6302364" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="131" name="Constant_153472" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="132" name="__module.model.layers.0.self_attn.out_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="143,input.13"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="133" name="__module.model.layers.0/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="145,input.15"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="134" name="__module.model.layers.0.self_attn_layer_norm/aten::layer_norm/Multiply" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="1575968" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="135" name="__module.model.layers.0.self_attn_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="136" name="Constant_153473_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="6303900" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="137" name="Constant_153473" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="138" name="__module.model.layers.0.self_attn_layer_norm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="139" name="Constant_153474_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="6305436" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="140" name="Constant_153474" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="141" name="__module.model.layers.0.self_attn_layer_norm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="149,input.17"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="142" name="self.model.layers.0.fc1.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="3072, 768" offset="6306972" size="4718592" /> <output> <port id="0" precision="FP16" names="self.model.layers.0.fc1.weight"> <dim>3072</dim> <dim>768</dim> </port> </output> </layer> <layer id="143" name="self.model.layers.0.fc1.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>3072</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>3072</dim> <dim>768</dim> </port> </output> </layer> <layer id="144" name="__module.model.layers.0.fc1/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>3072</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="145" name="Constant_153475_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 3072" offset="11025564" size="6144" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="146" name="Constant_153475" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="147" name="__module.model.layers.0.fc1/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </input> <output> <port id="2" precision="FP32" names="152,input.19"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="148" name="__module.model.layers.0.activation_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> <data approximation_mode="ERF" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32" names="153,input.21"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="149" name="self.model.layers.0.fc2.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 3072" offset="11031708" size="4718592" /> <output> <port id="0" precision="FP16" names="self.model.layers.0.fc2.weight"> <dim>768</dim> <dim>3072</dim> </port> </output> </layer> <layer id="150" name="self.model.layers.0.fc2.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>3072</dim> </port> </output> </layer> <layer id="151" name="__module.model.layers.0.fc2/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>3072</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="152" name="Constant_153476_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="15750300" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="153" name="Constant_153476" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="154" name="__module.model.layers.0.fc2/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="157,input.25"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="155" name="__module.model.layers.0/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="159,input.27"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="156" name="__module.model.layers.0.final_layer_norm/aten::layer_norm/Multiply" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="1575968" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="157" name="__module.model.layers.0.final_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="158" name="Constant_153477_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="15751836" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="159" name="Constant_153477" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="160" name="__module.model.layers.0.final_layer_norm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="161" name="Constant_153478_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="15753372" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="162" name="Constant_153478" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="163" name="__module.model.layers.0.final_layer_norm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="163,hidden_states.7"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="164" name="self.model.layers.1.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="15754908" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.1.self_attn.q_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="165" name="self.model.layers.1.self_attn.q_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="166" name="__module.model.layers.1.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="167" name="Constant_153479_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 1" offset="2758692" size="2" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="168" name="Constant_153479" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="169" name="Multiply_153290" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="170" name="Constant_153480_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="16934556" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="171" name="Constant_153480" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="172" name="__module.model.layers.1.self_attn/aten::mul/Multiply" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="179,tensor.11"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="173" name="Constant_154748" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="174" name="__module.model.layers.1.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="199"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="175" name="__module.model.layers.1.self_attn/aten::transpose/Constant" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="176" name="__module.model.layers.1.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="200"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="177" name="__module.model.layers.1.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="203,query_states.3"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="178" name="self.model.layers.1.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="16936092" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.1.self_attn.k_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="179" name="self.model.layers.1.self_attn.k_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="180" name="__module.model.layers.1.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="181" name="Constant_153481_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="18115740" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="182" name="Constant_153481" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="183" name="__module.model.layers.1.self_attn.k_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="182,tensor.7"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="184" name="Constant_154749" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="185" name="__module.model.layers.1.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="184"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="186" name="__module.model.layers.1.self_attn/aten::transpose/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="187" name="__module.model.layers.1.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="185"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="188" name="__module.model.layers.1.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="205,key_states.7"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="189" name="__module.model.layers.1.self_attn/aten::bmm/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="2" precision="FP32" names="210,attn_weights.7"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="190" name="Constant_132360" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="2760278" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="191" name="ShapeOf_154620" type="ShapeOf" version="opset3"> <data output_type="i64" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="1" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="192" name="Constant_154621" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575944" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="193" name="Constant_154622" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="194" name="Gather_154623" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I64"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="I64" names="208"> <dim>1</dim> </port> </output> </layer> <layer id="195" name="__module.model.layers.1.self_attn/prim::ListConstruct/SequenceMark_4" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="196" name="__module.model.layers.1.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="212"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="197" name="__module.model.layers.1.self_attn/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32" names="213,attn_weights.9"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="198" name="__module.model.layers.1.self_attn/prim::ListConstruct/SequenceMark_5" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="3" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="199" name="__module.model.layers.1.self_attn/aten::view/Reshape_4" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="217,input.29"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="200" name="__module.model.layers.1.self_attn/aten::softmax/Softmax" type="SoftMax" version="opset8"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="FP32" names="218,input.31"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="201" name="self.model.layers.1.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="18117276" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.1.self_attn.v_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="202" name="self.model.layers.1.self_attn.v_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="203" name="__module.model.layers.1.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="204" name="Constant_153482_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="19296924" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="205" name="Constant_153482" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="206" name="__module.model.layers.1.self_attn.v_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="189,tensor.9"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="207" name="Constant_154750" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="208" name="__module.model.layers.1.self_attn/aten::view/Reshape_5" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="191"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="209" name="__module.model.layers.1.self_attn/aten::transpose/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="210" name="__module.model.layers.1.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="192"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="211" name="__module.model.layers.1.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="207,value_states.7"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="212" name="__module.model.layers.1.self_attn/aten::bmm/MatMul_1" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="2" precision="FP32" names="220,attn_output.7"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="213" name="__module.model.layers.1.self_attn/aten::view/Reshape_6" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="222,attn_output.9"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="214" name="__module.model.layers.1.self_attn/aten::transpose/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="215" name="__module.model.layers.1.self_attn/aten::transpose/Transpose_4" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="223,attn_output.11"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="216" name="Constant_154751" type="Const" version="opset1"> <data element_type="i64" shape="3" offset="5122692" size="24" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="217" name="__module.model.layers.1.self_attn/aten::reshape/Reshape_2" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="225,input.33"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="218" name="self.model.layers.1.self_attn.out_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="19298460" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.1.self_attn.out_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="219" name="self.model.layers.1.self_attn.out_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="220" name="__module.model.layers.1.self_attn.out_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="221" name="Constant_153483_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="20478108" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="222" name="Constant_153483" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="223" name="__module.model.layers.1.self_attn.out_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="228,input.35"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="224" name="__module.model.layers.1/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="230,input.37"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="225" name="__module.model.layers.1.self_attn_layer_norm/aten::layer_norm/Multiply" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="1575968" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="226" name="__module.model.layers.1.self_attn_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="227" name="Constant_153484_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="20479644" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="228" name="Constant_153484" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="229" name="__module.model.layers.1.self_attn_layer_norm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="230" name="Constant_153485_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="20481180" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="231" name="Constant_153485" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="232" name="__module.model.layers.1.self_attn_layer_norm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="234,input.39"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="233" name="self.model.layers.1.fc1.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="3072, 768" offset="20482716" size="4718592" /> <output> <port id="0" precision="FP16" names="self.model.layers.1.fc1.weight"> <dim>3072</dim> <dim>768</dim> </port> </output> </layer> <layer id="234" name="self.model.layers.1.fc1.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>3072</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>3072</dim> <dim>768</dim> </port> </output> </layer> <layer id="235" name="__module.model.layers.1.fc1/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>3072</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="236" name="Constant_153486_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 3072" offset="25201308" size="6144" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="237" name="Constant_153486" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="238" name="__module.model.layers.1.fc1/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </input> <output> <port id="2" precision="FP32" names="237,input.41"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="239" name="__module.model.layers.1.activation_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> <data approximation_mode="ERF" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32" names="238,input.43"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="240" name="self.model.layers.1.fc2.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 3072" offset="25207452" size="4718592" /> <output> <port id="0" precision="FP16" names="self.model.layers.1.fc2.weight"> <dim>768</dim> <dim>3072</dim> </port> </output> </layer> <layer id="241" name="self.model.layers.1.fc2.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>3072</dim> </port> </output> </layer> <layer id="242" name="__module.model.layers.1.fc2/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>3072</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="243" name="Constant_153487_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="29926044" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="244" name="Constant_153487" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="245" name="__module.model.layers.1.fc2/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="242,input.47"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="246" name="__module.model.layers.1/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="244,input.49"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="247" name="__module.model.layers.1.final_layer_norm/aten::layer_norm/Multiply" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="1575968" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="248" name="__module.model.layers.1.final_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="249" name="Constant_153488_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="29927580" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="250" name="Constant_153488" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="251" name="__module.model.layers.1.final_layer_norm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="252" name="Constant_153489_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="29929116" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="253" name="Constant_153489" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="254" name="__module.model.layers.1.final_layer_norm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="248,hidden_states.13"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="255" name="self.model.layers.2.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="29930652" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.2.self_attn.q_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="256" name="self.model.layers.2.self_attn.q_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="257" name="__module.model.layers.2.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="258" name="Constant_153490_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 1" offset="2758692" size="2" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="259" name="Constant_153490" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="260" name="Multiply_153294" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="261" name="Constant_153491_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="31110300" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="262" name="Constant_153491" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="263" name="__module.model.layers.2.self_attn/aten::mul/Multiply" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="264,tensor.17"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="264" name="Constant_154752" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="265" name="__module.model.layers.2.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="284"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="266" name="__module.model.layers.2.self_attn/aten::transpose/Constant" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="267" name="__module.model.layers.2.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="285"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="268" name="__module.model.layers.2.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="288,query_states.5"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="269" name="self.model.layers.2.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="31111836" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.2.self_attn.k_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="270" name="self.model.layers.2.self_attn.k_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="271" name="__module.model.layers.2.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="272" name="Constant_153492_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="32291484" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="273" name="Constant_153492" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="274" name="__module.model.layers.2.self_attn.k_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="267,tensor.13"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="275" name="Constant_154753" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="276" name="__module.model.layers.2.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="269"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="277" name="__module.model.layers.2.self_attn/aten::transpose/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="278" name="__module.model.layers.2.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="270"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="279" name="__module.model.layers.2.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="290,key_states.11"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="280" name="__module.model.layers.2.self_attn/aten::bmm/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="2" precision="FP32" names="295,attn_weights.13"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="281" name="Constant_133271" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="2760278" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="282" name="ShapeOf_154642" type="ShapeOf" version="opset3"> <data output_type="i64" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="1" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="283" name="Constant_154643" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575944" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="284" name="Constant_154644" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="285" name="Gather_154645" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I64"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="I64" names="293"> <dim>1</dim> </port> </output> </layer> <layer id="286" name="__module.model.layers.2.self_attn/prim::ListConstruct/SequenceMark_4" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="287" name="__module.model.layers.2.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="297"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="288" name="__module.model.layers.2.self_attn/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32" names="298,attn_weights.15"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="289" name="__module.model.layers.2.self_attn/prim::ListConstruct/SequenceMark_5" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="3" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="290" name="__module.model.layers.2.self_attn/aten::view/Reshape_4" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="302,input.51"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="291" name="__module.model.layers.2.self_attn/aten::softmax/Softmax" type="SoftMax" version="opset8"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="FP32" names="303,input.53"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="292" name="self.model.layers.2.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="32293020" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.2.self_attn.v_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="293" name="self.model.layers.2.self_attn.v_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="294" name="__module.model.layers.2.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="295" name="Constant_153493_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="33472668" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="296" name="Constant_153493" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="297" name="__module.model.layers.2.self_attn.v_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="274,tensor.15"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="298" name="Constant_154754" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="299" name="__module.model.layers.2.self_attn/aten::view/Reshape_5" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="276"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="300" name="__module.model.layers.2.self_attn/aten::transpose/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="301" name="__module.model.layers.2.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="277"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="302" name="__module.model.layers.2.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="292,value_states.11"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="303" name="__module.model.layers.2.self_attn/aten::bmm/MatMul_1" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="2" precision="FP32" names="305,attn_output.13"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="304" name="__module.model.layers.2.self_attn/aten::view/Reshape_6" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="307,attn_output.15"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="305" name="__module.model.layers.2.self_attn/aten::transpose/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="306" name="__module.model.layers.2.self_attn/aten::transpose/Transpose_4" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="308,attn_output.17"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="307" name="Constant_154755" type="Const" version="opset1"> <data element_type="i64" shape="3" offset="5122692" size="24" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="308" name="__module.model.layers.2.self_attn/aten::reshape/Reshape_2" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="310,input.55"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="309" name="self.model.layers.2.self_attn.out_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="33474204" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.2.self_attn.out_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="310" name="self.model.layers.2.self_attn.out_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="311" name="__module.model.layers.2.self_attn.out_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="312" name="Constant_153494_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="34653852" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="313" name="Constant_153494" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="314" name="__module.model.layers.2.self_attn.out_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="313,input.57"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="315" name="__module.model.layers.2/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="315,input.59"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="316" name="__module.model.layers.2.self_attn_layer_norm/aten::layer_norm/Multiply" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="1575968" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="317" name="__module.model.layers.2.self_attn_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="318" name="Constant_153495_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="34655388" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="319" name="Constant_153495" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="320" name="__module.model.layers.2.self_attn_layer_norm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="321" name="Constant_153496_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="34656924" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="322" name="Constant_153496" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="323" name="__module.model.layers.2.self_attn_layer_norm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="319,input.61"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="324" name="self.model.layers.2.fc1.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="3072, 768" offset="34658460" size="4718592" /> <output> <port id="0" precision="FP16" names="self.model.layers.2.fc1.weight"> <dim>3072</dim> <dim>768</dim> </port> </output> </layer> <layer id="325" name="self.model.layers.2.fc1.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>3072</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>3072</dim> <dim>768</dim> </port> </output> </layer> <layer id="326" name="__module.model.layers.2.fc1/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>3072</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="327" name="Constant_153497_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 3072" offset="39377052" size="6144" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="328" name="Constant_153497" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="329" name="__module.model.layers.2.fc1/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </input> <output> <port id="2" precision="FP32" names="322,input.63"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="330" name="__module.model.layers.2.activation_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> <data approximation_mode="ERF" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32" names="323,input.65"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="331" name="self.model.layers.2.fc2.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 3072" offset="39383196" size="4718592" /> <output> <port id="0" precision="FP16" names="self.model.layers.2.fc2.weight"> <dim>768</dim> <dim>3072</dim> </port> </output> </layer> <layer id="332" name="self.model.layers.2.fc2.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>3072</dim> </port> </output> </layer> <layer id="333" name="__module.model.layers.2.fc2/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>3072</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="334" name="Constant_153498_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="44101788" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="335" name="Constant_153498" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="336" name="__module.model.layers.2.fc2/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="327,input.69"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="337" name="__module.model.layers.2/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="329,input.71"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="338" name="__module.model.layers.2.final_layer_norm/aten::layer_norm/Multiply" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="1575968" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="339" name="__module.model.layers.2.final_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="340" name="Constant_153499_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="44103324" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="341" name="Constant_153499" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="342" name="__module.model.layers.2.final_layer_norm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="343" name="Constant_153500_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="44104860" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="344" name="Constant_153500" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="345" name="__module.model.layers.2.final_layer_norm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="333,hidden_states.19"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="346" name="self.model.layers.3.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="44106396" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.3.self_attn.q_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="347" name="self.model.layers.3.self_attn.q_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="348" name="__module.model.layers.3.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="349" name="Constant_153501_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 1" offset="2758692" size="2" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="350" name="Constant_153501" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="351" name="Multiply_153298" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="352" name="Constant_153502_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="45286044" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="353" name="Constant_153502" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="354" name="__module.model.layers.3.self_attn/aten::mul/Multiply" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="349,tensor.23"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="355" name="Constant_154756" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="356" name="__module.model.layers.3.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="369"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="357" name="__module.model.layers.3.self_attn/aten::transpose/Constant" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="358" name="__module.model.layers.3.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="370"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="359" name="__module.model.layers.3.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="373,query_states.7"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="360" name="self.model.layers.3.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="45287580" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.3.self_attn.k_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="361" name="self.model.layers.3.self_attn.k_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="362" name="__module.model.layers.3.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="363" name="Constant_153503_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="46467228" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="364" name="Constant_153503" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="365" name="__module.model.layers.3.self_attn.k_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="352,tensor.19"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="366" name="Constant_154757" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="367" name="__module.model.layers.3.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="354"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="368" name="__module.model.layers.3.self_attn/aten::transpose/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="369" name="__module.model.layers.3.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="355"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="370" name="__module.model.layers.3.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="375,key_states.15"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="371" name="__module.model.layers.3.self_attn/aten::bmm/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="2" precision="FP32" names="380,attn_weights.19"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="372" name="Constant_134182" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="2760278" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="373" name="ShapeOf_154664" type="ShapeOf" version="opset3"> <data output_type="i64" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="1" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="374" name="Constant_154665" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575944" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="375" name="Constant_154666" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="376" name="Gather_154667" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I64"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="I64" names="378"> <dim>1</dim> </port> </output> </layer> <layer id="377" name="__module.model.layers.3.self_attn/prim::ListConstruct/SequenceMark_4" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="378" name="__module.model.layers.3.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="382"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="379" name="__module.model.layers.3.self_attn/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32" names="383,attn_weights.21"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="380" name="__module.model.layers.3.self_attn/prim::ListConstruct/SequenceMark_5" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="3" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="381" name="__module.model.layers.3.self_attn/aten::view/Reshape_4" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="387,input.73"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="382" name="__module.model.layers.3.self_attn/aten::softmax/Softmax" type="SoftMax" version="opset8"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="FP32" names="388,input.75"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="383" name="self.model.layers.3.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="46468764" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.3.self_attn.v_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="384" name="self.model.layers.3.self_attn.v_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="385" name="__module.model.layers.3.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="386" name="Constant_153504_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="47648412" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="387" name="Constant_153504" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="388" name="__module.model.layers.3.self_attn.v_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="359,tensor.21"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="389" name="Constant_154758" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="390" name="__module.model.layers.3.self_attn/aten::view/Reshape_5" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="361"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="391" name="__module.model.layers.3.self_attn/aten::transpose/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="392" name="__module.model.layers.3.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="362"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="393" name="__module.model.layers.3.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="377,value_states.15"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="394" name="__module.model.layers.3.self_attn/aten::bmm/MatMul_1" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="2" precision="FP32" names="390,attn_output.19"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="395" name="__module.model.layers.3.self_attn/aten::view/Reshape_6" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="392,attn_output.21"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="396" name="__module.model.layers.3.self_attn/aten::transpose/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="397" name="__module.model.layers.3.self_attn/aten::transpose/Transpose_4" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="393,attn_output.23"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="398" name="Constant_154759" type="Const" version="opset1"> <data element_type="i64" shape="3" offset="5122692" size="24" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="399" name="__module.model.layers.3.self_attn/aten::reshape/Reshape_2" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="395,input.77"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="400" name="self.model.layers.3.self_attn.out_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="47649948" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.3.self_attn.out_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="401" name="self.model.layers.3.self_attn.out_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="402" name="__module.model.layers.3.self_attn.out_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="403" name="Constant_153505_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="48829596" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="404" name="Constant_153505" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="405" name="__module.model.layers.3.self_attn.out_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="398,input.79"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="406" name="__module.model.layers.3/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="400,input.81"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="407" name="__module.model.layers.3.self_attn_layer_norm/aten::layer_norm/Multiply" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="1575968" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="408" name="__module.model.layers.3.self_attn_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="409" name="Constant_153506_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="48831132" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="410" name="Constant_153506" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="411" name="__module.model.layers.3.self_attn_layer_norm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="412" name="Constant_153507_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="48832668" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="413" name="Constant_153507" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="414" name="__module.model.layers.3.self_attn_layer_norm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="404,input.83"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="415" name="self.model.layers.3.fc1.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="3072, 768" offset="48834204" size="4718592" /> <output> <port id="0" precision="FP16" names="self.model.layers.3.fc1.weight"> <dim>3072</dim> <dim>768</dim> </port> </output> </layer> <layer id="416" name="self.model.layers.3.fc1.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>3072</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>3072</dim> <dim>768</dim> </port> </output> </layer> <layer id="417" name="__module.model.layers.3.fc1/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>3072</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="418" name="Constant_153508_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 3072" offset="53552796" size="6144" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="419" name="Constant_153508" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="420" name="__module.model.layers.3.fc1/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </input> <output> <port id="2" precision="FP32" names="407,input.85"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="421" name="__module.model.layers.3.activation_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> <data approximation_mode="ERF" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32" names="408,input.87"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="422" name="self.model.layers.3.fc2.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 3072" offset="53558940" size="4718592" /> <output> <port id="0" precision="FP16" names="self.model.layers.3.fc2.weight"> <dim>768</dim> <dim>3072</dim> </port> </output> </layer> <layer id="423" name="self.model.layers.3.fc2.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>3072</dim> </port> </output> </layer> <layer id="424" name="__module.model.layers.3.fc2/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>3072</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="425" name="Constant_153509_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="58277532" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="426" name="Constant_153509" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="427" name="__module.model.layers.3.fc2/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="412,input.91"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="428" name="__module.model.layers.3/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="414,input.93"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="429" name="__module.model.layers.3.final_layer_norm/aten::layer_norm/Multiply" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="1575968" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="430" name="__module.model.layers.3.final_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="431" name="Constant_153510_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="58279068" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="432" name="Constant_153510" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="433" name="__module.model.layers.3.final_layer_norm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="434" name="Constant_153511_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="58280604" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="435" name="Constant_153511" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="436" name="__module.model.layers.3.final_layer_norm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="418,hidden_states.25"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="437" name="self.model.layers.4.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="58282140" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.4.self_attn.q_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="438" name="self.model.layers.4.self_attn.q_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="439" name="__module.model.layers.4.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="440" name="Constant_153512_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 1" offset="2758692" size="2" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="441" name="Constant_153512" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="442" name="Multiply_153302" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="443" name="Constant_153513_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="59461788" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="444" name="Constant_153513" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="445" name="__module.model.layers.4.self_attn/aten::mul/Multiply" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="434,tensor.29"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="446" name="Constant_154760" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="447" name="__module.model.layers.4.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="454"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="448" name="__module.model.layers.4.self_attn/aten::transpose/Constant" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="449" name="__module.model.layers.4.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="455"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="450" name="__module.model.layers.4.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="458,query_states.9"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="451" name="self.model.layers.4.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="59463324" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.4.self_attn.k_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="452" name="self.model.layers.4.self_attn.k_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="453" name="__module.model.layers.4.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="454" name="Constant_153514_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="60642972" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="455" name="Constant_153514" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="456" name="__module.model.layers.4.self_attn.k_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="437,tensor.25"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="457" name="Constant_154761" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="458" name="__module.model.layers.4.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="439"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="459" name="__module.model.layers.4.self_attn/aten::transpose/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="460" name="__module.model.layers.4.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="440"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="461" name="__module.model.layers.4.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="460,key_states.19"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="462" name="__module.model.layers.4.self_attn/aten::bmm/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="2" precision="FP32" names="465,attn_weights.25"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="463" name="Constant_135093" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="2760278" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="464" name="ShapeOf_154686" type="ShapeOf" version="opset3"> <data output_type="i64" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="1" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="465" name="Constant_154687" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575944" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="466" name="Constant_154688" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="467" name="Gather_154689" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I64"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="I64" names="463"> <dim>1</dim> </port> </output> </layer> <layer id="468" name="__module.model.layers.4.self_attn/prim::ListConstruct/SequenceMark_4" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="469" name="__module.model.layers.4.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="467"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="470" name="__module.model.layers.4.self_attn/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32" names="468,attn_weights.27"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="471" name="__module.model.layers.4.self_attn/prim::ListConstruct/SequenceMark_5" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="3" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="472" name="__module.model.layers.4.self_attn/aten::view/Reshape_4" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="472,input.95"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="473" name="__module.model.layers.4.self_attn/aten::softmax/Softmax" type="SoftMax" version="opset8"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="FP32" names="473,input.97"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="474" name="self.model.layers.4.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="60644508" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.4.self_attn.v_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="475" name="self.model.layers.4.self_attn.v_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="476" name="__module.model.layers.4.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="477" name="Constant_153515_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="61824156" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="478" name="Constant_153515" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="479" name="__module.model.layers.4.self_attn.v_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="444,tensor.27"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="480" name="Constant_154762" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="481" name="__module.model.layers.4.self_attn/aten::view/Reshape_5" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="446"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="482" name="__module.model.layers.4.self_attn/aten::transpose/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="483" name="__module.model.layers.4.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="447"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="484" name="__module.model.layers.4.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="462,value_states.19"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="485" name="__module.model.layers.4.self_attn/aten::bmm/MatMul_1" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="2" precision="FP32" names="475,attn_output.25"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="486" name="__module.model.layers.4.self_attn/aten::view/Reshape_6" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="477,attn_output.27"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="487" name="__module.model.layers.4.self_attn/aten::transpose/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="488" name="__module.model.layers.4.self_attn/aten::transpose/Transpose_4" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="478,attn_output.29"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="489" name="Constant_154763" type="Const" version="opset1"> <data element_type="i64" shape="3" offset="5122692" size="24" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="490" name="__module.model.layers.4.self_attn/aten::reshape/Reshape_2" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="480,input.99"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="491" name="self.model.layers.4.self_attn.out_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="61825692" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.4.self_attn.out_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="492" name="self.model.layers.4.self_attn.out_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="493" name="__module.model.layers.4.self_attn.out_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="494" name="Constant_153516_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="63005340" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="495" name="Constant_153516" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="496" name="__module.model.layers.4.self_attn.out_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="483,input.101"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="497" name="__module.model.layers.4/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="485,input.103"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="498" name="__module.model.layers.4.self_attn_layer_norm/aten::layer_norm/Multiply" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="1575968" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="499" name="__module.model.layers.4.self_attn_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="500" name="Constant_153517_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="63006876" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="501" name="Constant_153517" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="502" name="__module.model.layers.4.self_attn_layer_norm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="503" name="Constant_153518_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="63008412" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="504" name="Constant_153518" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="505" name="__module.model.layers.4.self_attn_layer_norm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="489,input.105"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="506" name="self.model.layers.4.fc1.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="3072, 768" offset="63009948" size="4718592" /> <output> <port id="0" precision="FP16" names="self.model.layers.4.fc1.weight"> <dim>3072</dim> <dim>768</dim> </port> </output> </layer> <layer id="507" name="self.model.layers.4.fc1.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>3072</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>3072</dim> <dim>768</dim> </port> </output> </layer> <layer id="508" name="__module.model.layers.4.fc1/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>3072</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="509" name="Constant_153519_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 3072" offset="67728540" size="6144" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="510" name="Constant_153519" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="511" name="__module.model.layers.4.fc1/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </input> <output> <port id="2" precision="FP32" names="492,input.107"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="512" name="__module.model.layers.4.activation_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> <data approximation_mode="ERF" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32" names="493,input.109"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="513" name="self.model.layers.4.fc2.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 3072" offset="67734684" size="4718592" /> <output> <port id="0" precision="FP16" names="self.model.layers.4.fc2.weight"> <dim>768</dim> <dim>3072</dim> </port> </output> </layer> <layer id="514" name="self.model.layers.4.fc2.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>3072</dim> </port> </output> </layer> <layer id="515" name="__module.model.layers.4.fc2/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>3072</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="516" name="Constant_153520_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="72453276" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="517" name="Constant_153520" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="518" name="__module.model.layers.4.fc2/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="497,input.113"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="519" name="__module.model.layers.4/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="499,input.115"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="520" name="__module.model.layers.4.final_layer_norm/aten::layer_norm/Multiply" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="1575968" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="521" name="__module.model.layers.4.final_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="522" name="Constant_153521_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="72454812" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="523" name="Constant_153521" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="524" name="__module.model.layers.4.final_layer_norm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="525" name="Constant_153522_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="72456348" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="526" name="Constant_153522" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="527" name="__module.model.layers.4.final_layer_norm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="503,hidden_states.31"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="528" name="self.model.layers.5.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="72457884" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.5.self_attn.q_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="529" name="self.model.layers.5.self_attn.q_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="530" name="__module.model.layers.5.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="531" name="Constant_153523_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 1" offset="2758692" size="2" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="532" name="Constant_153523" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </output> </layer> <layer id="533" name="Multiply_153306" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="534" name="Constant_153524_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="73637532" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="535" name="Constant_153524" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="536" name="__module.model.layers.5.self_attn/aten::mul/Multiply" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="519,tensor"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="537" name="Constant_154764" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="538" name="__module.model.layers.5.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="539"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="539" name="__module.model.layers.5.self_attn/aten::transpose/Constant" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="540" name="__module.model.layers.5.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="540"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="541" name="__module.model.layers.5.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="543,query_states"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="542" name="self.model.layers.5.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="73639068" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.5.self_attn.k_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="543" name="self.model.layers.5.self_attn.k_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="544" name="__module.model.layers.5.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="545" name="Constant_153525_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="74818716" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="546" name="Constant_153525" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="547" name="__module.model.layers.5.self_attn.k_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="522,tensor.31"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="548" name="Constant_154765" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="549" name="__module.model.layers.5.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="524"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="550" name="__module.model.layers.5.self_attn/aten::transpose/Constant_1" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="551" name="__module.model.layers.5.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="525"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="552" name="__module.model.layers.5.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="545,key_states"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="553" name="__module.model.layers.5.self_attn/aten::bmm/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="2" precision="FP32" names="550,attn_weights.31"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="554" name="Constant_136004" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="2760278" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="555" name="ShapeOf_154708" type="ShapeOf" version="opset3"> <data output_type="i64" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="1" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="556" name="Constant_154709" type="Const" version="opset1"> <data element_type="i64" shape="1" offset="1575944" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>1</dim> </port> </output> </layer> <layer id="557" name="Constant_154710" type="Const" version="opset1"> <data element_type="i64" shape="" offset="1575936" size="8" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64" /> </output> </layer> <layer id="558" name="Gather_154711" type="Gather" version="opset8"> <data batch_dims="0" /> <input> <port id="0" precision="I64"> <dim>3</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64" /> </input> <output> <port id="3" precision="I64" names="548"> <dim>1</dim> </port> </output> </layer> <layer id="559" name="__module.model.layers.5.self_attn/prim::ListConstruct/SequenceMark_4" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> <port id="3" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="4" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="560" name="__module.model.layers.5.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="552"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="561" name="__module.model.layers.5.self_attn/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="2" precision="FP32" names="553,attn_weights.33"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="562" name="__module.model.layers.5.self_attn/prim::ListConstruct/SequenceMark_5" type="Concat" version="opset1"> <data axis="0" /> <input> <port id="0" precision="I64"> <dim>1</dim> </port> <port id="1" precision="I64"> <dim>1</dim> </port> <port id="2" precision="I64"> <dim>1</dim> </port> </input> <output> <port id="3" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="563" name="__module.model.layers.5.self_attn/aten::view/Reshape_4" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="557,input.117"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="564" name="__module.model.layers.5.self_attn/aten::softmax/Softmax" type="SoftMax" version="opset8"> <data axis="-1" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </input> <output> <port id="1" precision="FP32" names="558,input.119"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> </output> </layer> <layer id="565" name="self.model.layers.5.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="74820252" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.5.self_attn.v_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="566" name="self.model.layers.5.self_attn.v_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="567" name="__module.model.layers.5.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="568" name="Constant_153526_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="75999900" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="569" name="Constant_153526" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="570" name="__module.model.layers.5.self_attn.v_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="529,tensor.33"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="571" name="Constant_154766" type="Const" version="opset1"> <data element_type="i64" shape="4" offset="2760230" size="32" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>4</dim> </port> </output> </layer> <layer id="572" name="__module.model.layers.5.self_attn/aten::view/Reshape_5" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="531"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="573" name="__module.model.layers.5.self_attn/aten::transpose/Constant_3" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="574" name="__module.model.layers.5.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="532"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="575" name="__module.model.layers.5.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="547,value_states"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="576" name="__module.model.layers.5.self_attn/aten::bmm/MatMul_1" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>-1</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </input> <output> <port id="2" precision="FP32" names="560,attn_output.31"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="577" name="__module.model.layers.5.self_attn/aten::view/Reshape_6" type="Reshape" version="opset1"> <data special_zero="false" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="562,attn_output.33"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> </output> </layer> <layer id="578" name="__module.model.layers.5.self_attn/aten::transpose/Constant_4" type="Const" version="opset1"> <data element_type="i32" shape="4" offset="2760262" size="16" /> <output> <port id="0" precision="I32"> <dim>4</dim> </port> </output> </layer> <layer id="579" name="__module.model.layers.5.self_attn/aten::transpose/Transpose_4" type="Transpose" version="opset1"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>12</dim> <dim>-1</dim> <dim>64</dim> </port> <port id="1" precision="I32"> <dim>4</dim> </port> </input> <output> <port id="2" precision="FP32" names="563,attn_output"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> </output> </layer> <layer id="580" name="Constant_154767" type="Const" version="opset1"> <data element_type="i64" shape="3" offset="5122692" size="24" /> <rt_info> <attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" /> </rt_info> <output> <port id="0" precision="I64"> <dim>3</dim> </port> </output> </layer> <layer id="581" name="__module.model.layers.5.self_attn/aten::reshape/Reshape_2" type="Reshape" version="opset1"> <data special_zero="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>12</dim> <dim>64</dim> </port> <port id="1" precision="I64"> <dim>3</dim> </port> </input> <output> <port id="2" precision="FP32" names="565,input.121"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="582" name="self.model.layers.5.self_attn.out_proj.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 768" offset="76001436" size="1179648" /> <output> <port id="0" precision="FP16" names="self.model.layers.5.self_attn.out_proj.weight"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="583" name="self.model.layers.5.self_attn.out_proj.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </output> </layer> <layer id="584" name="__module.model.layers.5.self_attn.out_proj/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="585" name="Constant_153527_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="77181084" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="586" name="Constant_153527" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="587" name="__module.model.layers.5.self_attn.out_proj/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="568,input.123"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="588" name="__module.model.layers.5/aten::add/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="570,input.125"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="589" name="__module.model.layers.5.self_attn_layer_norm/aten::layer_norm/Multiply" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="1575968" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="590" name="__module.model.layers.5.self_attn_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="591" name="Constant_153528_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="77182620" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="592" name="Constant_153528" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="593" name="__module.model.layers.5.self_attn_layer_norm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="594" name="Constant_153529_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="77184156" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="595" name="Constant_153529" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="596" name="__module.model.layers.5.self_attn_layer_norm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="574,input.127"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="597" name="self.model.layers.5.fc1.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="3072, 768" offset="77185692" size="4718592" /> <output> <port id="0" precision="FP16" names="self.model.layers.5.fc1.weight"> <dim>3072</dim> <dim>768</dim> </port> </output> </layer> <layer id="598" name="self.model.layers.5.fc1.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>3072</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>3072</dim> <dim>768</dim> </port> </output> </layer> <layer id="599" name="__module.model.layers.5.fc1/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>3072</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="600" name="Constant_153530_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 3072" offset="81904284" size="6144" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="601" name="Constant_153530" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="602" name="__module.model.layers.5.fc1/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>3072</dim> </port> </input> <output> <port id="2" precision="FP32" names="577,input.129"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="603" name="__module.model.layers.5.activation_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> <data approximation_mode="ERF" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32" names="578,input.131"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> </output> </layer> <layer id="604" name="self.model.layers.5.fc2.weight_compressed" type="Const" version="opset1"> <data element_type="f16" shape="768, 3072" offset="81910428" size="4718592" /> <output> <port id="0" precision="FP16" names="self.model.layers.5.fc2.weight"> <dim>768</dim> <dim>3072</dim> </port> </output> </layer> <layer id="605" name="self.model.layers.5.fc2.weight" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>768</dim> <dim>3072</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>768</dim> <dim>3072</dim> </port> </output> </layer> <layer id="606" name="__module.model.layers.5.fc2/aten::linear/MatMul" type="MatMul" version="opset1"> <data transpose_a="false" transpose_b="true" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>3072</dim> </port> <port id="1" precision="FP32"> <dim>768</dim> <dim>3072</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="607" name="Constant_153531_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="86629020" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="608" name="Constant_153531" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="609" name="__module.model.layers.5.fc2/aten::linear/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="582,input.135"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="610" name="__module.model.layers.5/aten::add/Add_1" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="584,input"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="611" name="__module.model.layers.5.final_layer_norm/aten::layer_norm/Multiply" type="Const" version="opset1"> <data element_type="i32" shape="1" offset="1575968" size="4" /> <output> <port id="0" precision="I32"> <dim>1</dim> </port> </output> </layer> <layer id="612" name="__module.model.layers.5.final_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6"> <data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="I32"> <dim>1</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="613" name="Constant_153532_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="86630556" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="614" name="Constant_153532" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="615" name="__module.model.layers.5.final_layer_norm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="616" name="Constant_153533_compressed" type="Const" version="opset1"> <data element_type="f16" shape="1, 1, 768" offset="86632092" size="1536" /> <output> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="617" name="Constant_153533" type="Convert" version="opset1"> <data destination_type="f32" /> <rt_info> <attribute name="decompression" version="0" /> </rt_info> <input> <port id="0" precision="FP16"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </output> </layer> <layer id="618" name="__module.model.layers.5.final_layer_norm/aten::layer_norm/Add" type="Add" version="opset1"> <data auto_broadcast="numpy" /> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> <port id="1" precision="FP32"> <dim>1</dim> <dim>1</dim> <dim>768</dim> </port> </input> <output> <port id="2" precision="FP32" names="last_hidden_state"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </output> </layer> <layer id="619" name="Result_136360" type="Result" version="opset1" output_names="last_hidden_state"> <input> <port id="0" precision="FP32"> <dim>-1</dim> <dim>-1</dim> <dim>768</dim> </port> </input> </layer> </layers> <edges> <edge from-layer="0" from-port="0" to-layer="84" to-port="0" /> <edge from-layer="0" from-port="0" to-layer="93" to-port="0" /> <edge from-layer="1" from-port="0" to-layer="5" to-port="0" /> <edge from-layer="1" from-port="0" to-layer="27" to-port="0" /> <edge from-layer="2" from-port="0" to-layer="3" to-port="0" /> <edge from-layer="3" from-port="1" to-layer="26" to-port="0" /> <edge from-layer="4" from-port="0" to-layer="10" to-port="0" /> <edge from-layer="5" from-port="1" to-layer="8" to-port="0" /> <edge from-layer="5" from-port="1" to-layer="19" to-port="0" /> <edge from-layer="6" from-port="0" to-layer="8" to-port="1" /> <edge from-layer="7" from-port="0" to-layer="8" to-port="2" /> <edge from-layer="8" from-port="3" to-layer="10" to-port="1" /> <edge from-layer="8" from-port="3" to-layer="15" to-port="0" /> <edge from-layer="9" from-port="0" to-layer="10" to-port="2" /> <edge from-layer="10" from-port="3" to-layer="12" to-port="0" /> <edge from-layer="11" from-port="0" to-layer="12" to-port="1" /> <edge from-layer="12" from-port="2" to-layer="23" to-port="0" /> <edge from-layer="13" from-port="0" to-layer="16" to-port="0" /> <edge from-layer="14" from-port="0" to-layer="15" to-port="1" /> <edge from-layer="15" from-port="2" to-layer="16" to-port="1" /> <edge from-layer="16" from-port="2" to-layer="22" to-port="0" /> <edge from-layer="17" from-port="0" to-layer="19" to-port="1" /> <edge from-layer="18" from-port="0" to-layer="19" to-port="2" /> <edge from-layer="19" from-port="3" to-layer="21" to-port="0" /> <edge from-layer="20" from-port="0" to-layer="21" to-port="1" /> <edge from-layer="21" from-port="2" to-layer="22" to-port="1" /> <edge from-layer="22" from-port="2" to-layer="23" to-port="1" /> <edge from-layer="23" from-port="2" to-layer="24" to-port="0" /> <edge from-layer="24" from-port="1" to-layer="26" to-port="1" /> <edge from-layer="25" from-port="0" to-layer="26" to-port="2" /> <edge from-layer="26" from-port="3" to-layer="27" to-port="1" /> <edge from-layer="27" from-port="2" to-layer="29" to-port="0" /> <edge from-layer="28" from-port="0" to-layer="29" to-port="1" /> <edge from-layer="29" from-port="2" to-layer="32" to-port="0" /> <edge from-layer="30" from-port="0" to-layer="31" to-port="0" /> <edge from-layer="31" from-port="1" to-layer="32" to-port="1" /> <edge from-layer="32" from-port="2" to-layer="35" to-port="0" /> <edge from-layer="33" from-port="0" to-layer="34" to-port="0" /> <edge from-layer="34" from-port="1" to-layer="35" to-port="1" /> <edge from-layer="35" from-port="2" to-layer="38" to-port="0" /> <edge from-layer="35" from-port="2" to-layer="61" to-port="0" /> <edge from-layer="35" from-port="2" to-layer="133" to-port="0" /> <edge from-layer="35" from-port="2" to-layer="109" to-port="0" /> <edge from-layer="35" from-port="2" to-layer="49" to-port="0" /> <edge from-layer="36" from-port="0" to-layer="37" to-port="0" /> <edge from-layer="37" from-port="1" to-layer="38" to-port="1" /> <edge from-layer="38" from-port="2" to-layer="41" to-port="0" /> <edge from-layer="39" from-port="0" to-layer="40" to-port="0" /> <edge from-layer="40" from-port="1" to-layer="41" to-port="1" /> <edge from-layer="41" from-port="2" to-layer="44" to-port="0" /> <edge from-layer="42" from-port="0" to-layer="43" to-port="0" /> <edge from-layer="43" from-port="1" to-layer="44" to-port="1" /> <edge from-layer="44" from-port="2" to-layer="46" to-port="0" /> <edge from-layer="45" from-port="0" to-layer="46" to-port="1" /> <edge from-layer="46" from-port="2" to-layer="48" to-port="0" /> <edge from-layer="47" from-port="0" to-layer="48" to-port="1" /> <edge from-layer="48" from-port="2" to-layer="58" to-port="0" /> <edge from-layer="49" from-port="1" to-layer="52" to-port="0" /> <edge from-layer="49" from-port="1" to-layer="74" to-port="0" /> <edge from-layer="50" from-port="0" to-layer="52" to-port="1" /> <edge from-layer="51" from-port="0" to-layer="52" to-port="2" /> <edge from-layer="52" from-port="3" to-layer="54" to-port="0" /> <edge from-layer="52" from-port="3" to-layer="468" to-port="0" /> <edge from-layer="52" from-port="3" to-layer="377" to-port="0" /> <edge from-layer="52" from-port="3" to-layer="286" to-port="0" /> <edge from-layer="52" from-port="3" to-layer="195" to-port="0" /> <edge from-layer="52" from-port="3" to-layer="121" to-port="0" /> <edge from-layer="52" from-port="3" to-layer="79" to-port="0" /> <edge from-layer="52" from-port="3" to-layer="559" to-port="0" /> <edge from-layer="53" from-port="0" to-layer="54" to-port="1" /> <edge from-layer="54" from-port="2" to-layer="57" to-port="0" /> <edge from-layer="54" from-port="2" to-layer="380" to-port="0" /> <edge from-layer="54" from-port="2" to-layer="289" to-port="0" /> <edge from-layer="54" from-port="2" to-layer="198" to-port="0" /> <edge from-layer="54" from-port="2" to-layer="471" to-port="0" /> <edge from-layer="54" from-port="2" to-layer="104" to-port="0" /> <edge from-layer="54" from-port="2" to-layer="562" to-port="0" /> <edge from-layer="55" from-port="0" to-layer="57" to-port="1" /> <edge from-layer="56" from-port="0" to-layer="57" to-port="2" /> <edge from-layer="57" from-port="3" to-layer="359" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="117" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="575" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="177" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="188" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="211" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="268" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="279" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="302" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="370" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="393" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="450" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="461" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="484" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="541" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="552" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="58" to-port="1" /> <edge from-layer="57" from-port="3" to-layer="69" to-port="1" /> <edge from-layer="58" from-port="2" to-layer="70" to-port="0" /> <edge from-layer="59" from-port="0" to-layer="60" to-port="0" /> <edge from-layer="60" from-port="1" to-layer="61" to-port="1" /> <edge from-layer="61" from-port="2" to-layer="64" to-port="0" /> <edge from-layer="62" from-port="0" to-layer="63" to-port="0" /> <edge from-layer="63" from-port="1" to-layer="64" to-port="1" /> <edge from-layer="64" from-port="2" to-layer="66" to-port="0" /> <edge from-layer="65" from-port="0" to-layer="66" to-port="1" /> <edge from-layer="66" from-port="2" to-layer="68" to-port="0" /> <edge from-layer="67" from-port="0" to-layer="68" to-port="1" /> <edge from-layer="68" from-port="2" to-layer="69" to-port="0" /> <edge from-layer="69" from-port="2" to-layer="70" to-port="1" /> <edge from-layer="69" from-port="2" to-layer="75" to-port="0" /> <edge from-layer="70" from-port="2" to-layer="80" to-port="0" /> <edge from-layer="71" from-port="0" to-layer="79" to-port="1" /> <edge from-layer="72" from-port="0" to-layer="74" to-port="1" /> <edge from-layer="73" from-port="0" to-layer="74" to-port="2" /> <edge from-layer="74" from-port="3" to-layer="471" to-port="1" /> <edge from-layer="74" from-port="3" to-layer="198" to-port="1" /> <edge from-layer="74" from-port="3" to-layer="79" to-port="2" /> <edge from-layer="74" from-port="3" to-layer="286" to-port="2" /> <edge from-layer="74" from-port="3" to-layer="289" to-port="1" /> <edge from-layer="74" from-port="3" to-layer="377" to-port="2" /> <edge from-layer="74" from-port="3" to-layer="195" to-port="2" /> <edge from-layer="74" from-port="3" to-layer="380" to-port="1" /> <edge from-layer="74" from-port="3" to-layer="468" to-port="2" /> <edge from-layer="74" from-port="3" to-layer="559" to-port="2" /> <edge from-layer="74" from-port="3" to-layer="562" to-port="1" /> <edge from-layer="74" from-port="3" to-layer="104" to-port="1" /> <edge from-layer="74" from-port="3" to-layer="121" to-port="2" /> <edge from-layer="75" from-port="1" to-layer="78" to-port="0" /> <edge from-layer="76" from-port="0" to-layer="78" to-port="1" /> <edge from-layer="77" from-port="0" to-layer="78" to-port="2" /> <edge from-layer="78" from-port="3" to-layer="104" to-port="2" /> <edge from-layer="78" from-port="3" to-layer="79" to-port="3" /> <edge from-layer="79" from-port="4" to-layer="80" to-port="1" /> <edge from-layer="80" from-port="2" to-layer="103" to-port="0" /> <edge from-layer="81" from-port="0" to-layer="82" to-port="0" /> <edge from-layer="82" from-port="1" to-layer="87" to-port="0" /> <edge from-layer="83" from-port="0" to-layer="84" to-port="1" /> <edge from-layer="84" from-port="2" to-layer="86" to-port="0" /> <edge from-layer="85" from-port="0" to-layer="86" to-port="1" /> <edge from-layer="86" from-port="2" to-layer="87" to-port="1" /> <edge from-layer="87" from-port="2" to-layer="99" to-port="0" /> <edge from-layer="87" from-port="2" to-layer="88" to-port="0" /> <edge from-layer="88" from-port="1" to-layer="98" to-port="0" /> <edge from-layer="88" from-port="1" to-layer="91" to-port="0" /> <edge from-layer="89" from-port="0" to-layer="91" to-port="1" /> <edge from-layer="90" from-port="0" to-layer="91" to-port="2" /> <edge from-layer="91" from-port="3" to-layer="97" to-port="0" /> <edge from-layer="92" from-port="0" to-layer="97" to-port="1" /> <edge from-layer="93" from-port="1" to-layer="96" to-port="0" /> <edge from-layer="94" from-port="0" to-layer="96" to-port="1" /> <edge from-layer="95" from-port="0" to-layer="96" to-port="2" /> <edge from-layer="96" from-port="3" to-layer="97" to-port="2" /> <edge from-layer="97" from-port="3" to-layer="98" to-port="1" /> <edge from-layer="98" from-port="2" to-layer="99" to-port="1" /> <edge from-layer="99" from-port="2" to-layer="102" to-port="2" /> <edge from-layer="99" from-port="2" to-layer="100" to-port="0" /> <edge from-layer="100" from-port="1" to-layer="102" to-port="0" /> <edge from-layer="101" from-port="0" to-layer="102" to-port="1" /> <edge from-layer="102" from-port="3" to-layer="470" to-port="1" /> <edge from-layer="102" from-port="3" to-layer="379" to-port="1" /> <edge from-layer="102" from-port="3" to-layer="197" to-port="1" /> <edge from-layer="102" from-port="3" to-layer="561" to-port="1" /> <edge from-layer="102" from-port="3" to-layer="288" to-port="1" /> <edge from-layer="102" from-port="3" to-layer="103" to-port="1" /> <edge from-layer="103" from-port="2" to-layer="105" to-port="0" /> <edge from-layer="104" from-port="3" to-layer="105" to-port="1" /> <edge from-layer="105" from-port="2" to-layer="106" to-port="0" /> <edge from-layer="106" from-port="1" to-layer="118" to-port="0" /> <edge from-layer="107" from-port="0" to-layer="108" to-port="0" /> <edge from-layer="108" from-port="1" to-layer="109" to-port="1" /> <edge from-layer="109" from-port="2" to-layer="112" to-port="0" /> <edge from-layer="110" from-port="0" to-layer="111" to-port="0" /> <edge from-layer="111" from-port="1" to-layer="112" to-port="1" /> <edge from-layer="112" from-port="2" to-layer="114" to-port="0" /> <edge from-layer="113" from-port="0" to-layer="114" to-port="1" /> <edge from-layer="114" from-port="2" to-layer="116" to-port="0" /> <edge from-layer="115" from-port="0" to-layer="116" to-port="1" /> <edge from-layer="116" from-port="2" to-layer="117" to-port="0" /> <edge from-layer="117" from-port="2" to-layer="118" to-port="1" /> <edge from-layer="118" from-port="2" to-layer="122" to-port="0" /> <edge from-layer="119" from-port="0" to-layer="121" to-port="1" /> <edge from-layer="120" from-port="0" to-layer="121" to-port="3" /> <edge from-layer="121" from-port="4" to-layer="395" to-port="1" /> <edge from-layer="121" from-port="4" to-layer="213" to-port="1" /> <edge from-layer="121" from-port="4" to-layer="304" to-port="1" /> <edge from-layer="121" from-port="4" to-layer="577" to-port="1" /> <edge from-layer="121" from-port="4" to-layer="122" to-port="1" /> <edge from-layer="121" from-port="4" to-layer="486" to-port="1" /> <edge from-layer="122" from-port="2" to-layer="124" to-port="0" /> <edge from-layer="123" from-port="0" to-layer="124" to-port="1" /> <edge from-layer="124" from-port="2" to-layer="126" to-port="0" /> <edge from-layer="125" from-port="0" to-layer="126" to-port="1" /> <edge from-layer="126" from-port="2" to-layer="129" to-port="0" /> <edge from-layer="127" from-port="0" to-layer="128" to-port="0" /> <edge from-layer="128" from-port="1" to-layer="129" to-port="1" /> <edge from-layer="129" from-port="2" to-layer="132" to-port="0" /> <edge from-layer="130" from-port="0" to-layer="131" to-port="0" /> <edge from-layer="131" from-port="1" to-layer="132" to-port="1" /> <edge from-layer="132" from-port="2" to-layer="133" to-port="1" /> <edge from-layer="133" from-port="2" to-layer="135" to-port="0" /> <edge from-layer="134" from-port="0" to-layer="135" to-port="1" /> <edge from-layer="135" from-port="2" to-layer="138" to-port="0" /> <edge from-layer="136" from-port="0" to-layer="137" to-port="0" /> <edge from-layer="137" from-port="1" to-layer="138" to-port="1" /> <edge from-layer="138" from-port="2" to-layer="141" to-port="0" /> <edge from-layer="139" from-port="0" to-layer="140" to-port="0" /> <edge from-layer="140" from-port="1" to-layer="141" to-port="1" /> <edge from-layer="141" from-port="2" to-layer="155" to-port="0" /> <edge from-layer="141" from-port="2" to-layer="144" to-port="0" /> <edge from-layer="142" from-port="0" to-layer="143" to-port="0" /> <edge from-layer="143" from-port="1" to-layer="144" to-port="1" /> <edge from-layer="144" from-port="2" to-layer="147" to-port="0" /> <edge from-layer="145" from-port="0" to-layer="146" to-port="0" /> <edge from-layer="146" from-port="1" to-layer="147" to-port="1" /> <edge from-layer="147" from-port="2" to-layer="148" to-port="0" /> <edge from-layer="148" from-port="1" to-layer="151" to-port="0" /> <edge from-layer="149" from-port="0" to-layer="150" to-port="0" /> <edge from-layer="150" from-port="1" to-layer="151" to-port="1" /> <edge from-layer="151" from-port="2" to-layer="154" to-port="0" /> <edge from-layer="152" from-port="0" to-layer="153" to-port="0" /> <edge from-layer="153" from-port="1" to-layer="154" to-port="1" /> <edge from-layer="154" from-port="2" to-layer="155" to-port="1" /> <edge from-layer="155" from-port="2" to-layer="157" to-port="0" /> <edge from-layer="156" from-port="0" to-layer="157" to-port="1" /> <edge from-layer="157" from-port="2" to-layer="160" to-port="0" /> <edge from-layer="158" from-port="0" to-layer="159" to-port="0" /> <edge from-layer="159" from-port="1" to-layer="160" to-port="1" /> <edge from-layer="160" from-port="2" to-layer="163" to-port="0" /> <edge from-layer="161" from-port="0" to-layer="162" to-port="0" /> <edge from-layer="162" from-port="1" to-layer="163" to-port="1" /> <edge from-layer="163" from-port="2" to-layer="166" to-port="0" /> <edge from-layer="163" from-port="2" to-layer="180" to-port="0" /> <edge from-layer="163" from-port="2" to-layer="224" to-port="0" /> <edge from-layer="163" from-port="2" to-layer="203" to-port="0" /> <edge from-layer="164" from-port="0" to-layer="165" to-port="0" /> <edge from-layer="165" from-port="1" to-layer="166" to-port="1" /> <edge from-layer="166" from-port="2" to-layer="169" to-port="0" /> <edge from-layer="167" from-port="0" to-layer="168" to-port="0" /> <edge from-layer="168" from-port="1" to-layer="169" to-port="1" /> <edge from-layer="169" from-port="2" to-layer="172" to-port="0" /> <edge from-layer="170" from-port="0" to-layer="171" to-port="0" /> <edge from-layer="171" from-port="1" to-layer="172" to-port="1" /> <edge from-layer="172" from-port="2" to-layer="174" to-port="0" /> <edge from-layer="173" from-port="0" to-layer="174" to-port="1" /> <edge from-layer="174" from-port="2" to-layer="176" to-port="0" /> <edge from-layer="175" from-port="0" to-layer="176" to-port="1" /> <edge from-layer="176" from-port="2" to-layer="177" to-port="0" /> <edge from-layer="177" from-port="2" to-layer="189" to-port="0" /> <edge from-layer="178" from-port="0" to-layer="179" to-port="0" /> <edge from-layer="179" from-port="1" to-layer="180" to-port="1" /> <edge from-layer="180" from-port="2" to-layer="183" to-port="0" /> <edge from-layer="181" from-port="0" to-layer="182" to-port="0" /> <edge from-layer="182" from-port="1" to-layer="183" to-port="1" /> <edge from-layer="183" from-port="2" to-layer="185" to-port="0" /> <edge from-layer="184" from-port="0" to-layer="185" to-port="1" /> <edge from-layer="185" from-port="2" to-layer="187" to-port="0" /> <edge from-layer="186" from-port="0" to-layer="187" to-port="1" /> <edge from-layer="187" from-port="2" to-layer="188" to-port="0" /> <edge from-layer="188" from-port="2" to-layer="191" to-port="0" /> <edge from-layer="188" from-port="2" to-layer="189" to-port="1" /> <edge from-layer="189" from-port="2" to-layer="196" to-port="0" /> <edge from-layer="190" from-port="0" to-layer="195" to-port="1" /> <edge from-layer="191" from-port="1" to-layer="194" to-port="0" /> <edge from-layer="192" from-port="0" to-layer="194" to-port="1" /> <edge from-layer="193" from-port="0" to-layer="194" to-port="2" /> <edge from-layer="194" from-port="3" to-layer="198" to-port="2" /> <edge from-layer="194" from-port="3" to-layer="195" to-port="3" /> <edge from-layer="195" from-port="4" to-layer="196" to-port="1" /> <edge from-layer="196" from-port="2" to-layer="197" to-port="0" /> <edge from-layer="197" from-port="2" to-layer="199" to-port="0" /> <edge from-layer="198" from-port="3" to-layer="199" to-port="1" /> <edge from-layer="199" from-port="2" to-layer="200" to-port="0" /> <edge from-layer="200" from-port="1" to-layer="212" to-port="0" /> <edge from-layer="201" from-port="0" to-layer="202" to-port="0" /> <edge from-layer="202" from-port="1" to-layer="203" to-port="1" /> <edge from-layer="203" from-port="2" to-layer="206" to-port="0" /> <edge from-layer="204" from-port="0" to-layer="205" to-port="0" /> <edge from-layer="205" from-port="1" to-layer="206" to-port="1" /> <edge from-layer="206" from-port="2" to-layer="208" to-port="0" /> <edge from-layer="207" from-port="0" to-layer="208" to-port="1" /> <edge from-layer="208" from-port="2" to-layer="210" to-port="0" /> <edge from-layer="209" from-port="0" to-layer="210" to-port="1" /> <edge from-layer="210" from-port="2" to-layer="211" to-port="0" /> <edge from-layer="211" from-port="2" to-layer="212" to-port="1" /> <edge from-layer="212" from-port="2" to-layer="213" to-port="0" /> <edge from-layer="213" from-port="2" to-layer="215" to-port="0" /> <edge from-layer="214" from-port="0" to-layer="215" to-port="1" /> <edge from-layer="215" from-port="2" to-layer="217" to-port="0" /> <edge from-layer="216" from-port="0" to-layer="217" to-port="1" /> <edge from-layer="217" from-port="2" to-layer="220" to-port="0" /> <edge from-layer="218" from-port="0" to-layer="219" to-port="0" /> <edge from-layer="219" from-port="1" to-layer="220" to-port="1" /> <edge from-layer="220" from-port="2" to-layer="223" to-port="0" /> <edge from-layer="221" from-port="0" to-layer="222" to-port="0" /> <edge from-layer="222" from-port="1" to-layer="223" to-port="1" /> <edge from-layer="223" from-port="2" to-layer="224" to-port="1" /> <edge from-layer="224" from-port="2" to-layer="226" to-port="0" /> <edge from-layer="225" from-port="0" to-layer="226" to-port="1" /> <edge from-layer="226" from-port="2" to-layer="229" to-port="0" /> <edge from-layer="227" from-port="0" to-layer="228" to-port="0" /> <edge from-layer="228" from-port="1" to-layer="229" to-port="1" /> <edge from-layer="229" from-port="2" to-layer="232" to-port="0" /> <edge from-layer="230" from-port="0" to-layer="231" to-port="0" /> <edge from-layer="231" from-port="1" to-layer="232" to-port="1" /> <edge from-layer="232" from-port="2" to-layer="246" to-port="0" /> <edge from-layer="232" from-port="2" to-layer="235" to-port="0" /> <edge from-layer="233" from-port="0" to-layer="234" to-port="0" /> <edge from-layer="234" from-port="1" to-layer="235" to-port="1" /> <edge from-layer="235" from-port="2" to-layer="238" to-port="0" /> <edge from-layer="236" from-port="0" to-layer="237" to-port="0" /> <edge from-layer="237" from-port="1" to-layer="238" to-port="1" /> <edge from-layer="238" from-port="2" to-layer="239" to-port="0" /> <edge from-layer="239" from-port="1" to-layer="242" to-port="0" /> <edge from-layer="240" from-port="0" to-layer="241" to-port="0" /> <edge from-layer="241" from-port="1" to-layer="242" to-port="1" /> <edge from-layer="242" from-port="2" to-layer="245" to-port="0" /> <edge from-layer="243" from-port="0" to-layer="244" to-port="0" /> <edge from-layer="244" from-port="1" to-layer="245" to-port="1" /> <edge from-layer="245" from-port="2" to-layer="246" to-port="1" /> <edge from-layer="246" from-port="2" to-layer="248" to-port="0" /> <edge from-layer="247" from-port="0" to-layer="248" to-port="1" /> <edge from-layer="248" from-port="2" to-layer="251" to-port="0" /> <edge from-layer="249" from-port="0" to-layer="250" to-port="0" /> <edge from-layer="250" from-port="1" to-layer="251" to-port="1" /> <edge from-layer="251" from-port="2" to-layer="254" to-port="0" /> <edge from-layer="252" from-port="0" to-layer="253" to-port="0" /> <edge from-layer="253" from-port="1" to-layer="254" to-port="1" /> <edge from-layer="254" from-port="2" to-layer="271" to-port="0" /> <edge from-layer="254" from-port="2" to-layer="294" to-port="0" /> <edge from-layer="254" from-port="2" to-layer="315" to-port="0" /> <edge from-layer="254" from-port="2" to-layer="257" to-port="0" /> <edge from-layer="255" from-port="0" to-layer="256" to-port="0" /> <edge from-layer="256" from-port="1" to-layer="257" to-port="1" /> <edge from-layer="257" from-port="2" to-layer="260" to-port="0" /> <edge from-layer="258" from-port="0" to-layer="259" to-port="0" /> <edge from-layer="259" from-port="1" to-layer="260" to-port="1" /> <edge from-layer="260" from-port="2" to-layer="263" to-port="0" /> <edge from-layer="261" from-port="0" to-layer="262" to-port="0" /> <edge from-layer="262" from-port="1" to-layer="263" to-port="1" /> <edge from-layer="263" from-port="2" to-layer="265" to-port="0" /> <edge from-layer="264" from-port="0" to-layer="265" to-port="1" /> <edge from-layer="265" from-port="2" to-layer="267" to-port="0" /> <edge from-layer="266" from-port="0" to-layer="267" to-port="1" /> <edge from-layer="267" from-port="2" to-layer="268" to-port="0" /> <edge from-layer="268" from-port="2" to-layer="280" to-port="0" /> <edge from-layer="269" from-port="0" to-layer="270" to-port="0" /> <edge from-layer="270" from-port="1" to-layer="271" to-port="1" /> <edge from-layer="271" from-port="2" to-layer="274" to-port="0" /> <edge from-layer="272" from-port="0" to-layer="273" to-port="0" /> <edge from-layer="273" from-port="1" to-layer="274" to-port="1" /> <edge from-layer="274" from-port="2" to-layer="276" to-port="0" /> <edge from-layer="275" from-port="0" to-layer="276" to-port="1" /> <edge from-layer="276" from-port="2" to-layer="278" to-port="0" /> <edge from-layer="277" from-port="0" to-layer="278" to-port="1" /> <edge from-layer="278" from-port="2" to-layer="279" to-port="0" /> <edge from-layer="279" from-port="2" to-layer="282" to-port="0" /> <edge from-layer="279" from-port="2" to-layer="280" to-port="1" /> <edge from-layer="280" from-port="2" to-layer="287" to-port="0" /> <edge from-layer="281" from-port="0" to-layer="286" to-port="1" /> <edge from-layer="282" from-port="1" to-layer="285" to-port="0" /> <edge from-layer="283" from-port="0" to-layer="285" to-port="1" /> <edge from-layer="284" from-port="0" to-layer="285" to-port="2" /> <edge from-layer="285" from-port="3" to-layer="286" to-port="3" /> <edge from-layer="285" from-port="3" to-layer="289" to-port="2" /> <edge from-layer="286" from-port="4" to-layer="287" to-port="1" /> <edge from-layer="287" from-port="2" to-layer="288" to-port="0" /> <edge from-layer="288" from-port="2" to-layer="290" to-port="0" /> <edge from-layer="289" from-port="3" to-layer="290" to-port="1" /> <edge from-layer="290" from-port="2" to-layer="291" to-port="0" /> <edge from-layer="291" from-port="1" to-layer="303" to-port="0" /> <edge from-layer="292" from-port="0" to-layer="293" to-port="0" /> <edge from-layer="293" from-port="1" to-layer="294" to-port="1" /> <edge from-layer="294" from-port="2" to-layer="297" to-port="0" /> <edge from-layer="295" from-port="0" to-layer="296" to-port="0" /> <edge from-layer="296" from-port="1" to-layer="297" to-port="1" /> <edge from-layer="297" from-port="2" to-layer="299" to-port="0" /> <edge from-layer="298" from-port="0" to-layer="299" to-port="1" /> <edge from-layer="299" from-port="2" to-layer="301" to-port="0" /> <edge from-layer="300" from-port="0" to-layer="301" to-port="1" /> <edge from-layer="301" from-port="2" to-layer="302" to-port="0" /> <edge from-layer="302" from-port="2" to-layer="303" to-port="1" /> <edge from-layer="303" from-port="2" to-layer="304" to-port="0" /> <edge from-layer="304" from-port="2" to-layer="306" to-port="0" /> <edge from-layer="305" from-port="0" to-layer="306" to-port="1" /> <edge from-layer="306" from-port="2" to-layer="308" to-port="0" /> <edge from-layer="307" from-port="0" to-layer="308" to-port="1" /> <edge from-layer="308" from-port="2" to-layer="311" to-port="0" /> <edge from-layer="309" from-port="0" to-layer="310" to-port="0" /> <edge from-layer="310" from-port="1" to-layer="311" to-port="1" /> <edge from-layer="311" from-port="2" to-layer="314" to-port="0" /> <edge from-layer="312" from-port="0" to-layer="313" to-port="0" /> <edge from-layer="313" from-port="1" to-layer="314" to-port="1" /> <edge from-layer="314" from-port="2" to-layer="315" to-port="1" /> <edge from-layer="315" from-port="2" to-layer="317" to-port="0" /> <edge from-layer="316" from-port="0" to-layer="317" to-port="1" /> <edge from-layer="317" from-port="2" to-layer="320" to-port="0" /> <edge from-layer="318" from-port="0" to-layer="319" to-port="0" /> <edge from-layer="319" from-port="1" to-layer="320" to-port="1" /> <edge from-layer="320" from-port="2" to-layer="323" to-port="0" /> <edge from-layer="321" from-port="0" to-layer="322" to-port="0" /> <edge from-layer="322" from-port="1" to-layer="323" to-port="1" /> <edge from-layer="323" from-port="2" to-layer="326" to-port="0" /> <edge from-layer="323" from-port="2" to-layer="337" to-port="0" /> <edge from-layer="324" from-port="0" to-layer="325" to-port="0" /> <edge from-layer="325" from-port="1" to-layer="326" to-port="1" /> <edge from-layer="326" from-port="2" to-layer="329" to-port="0" /> <edge from-layer="327" from-port="0" to-layer="328" to-port="0" /> <edge from-layer="328" from-port="1" to-layer="329" to-port="1" /> <edge from-layer="329" from-port="2" to-layer="330" to-port="0" /> <edge from-layer="330" from-port="1" to-layer="333" to-port="0" /> <edge from-layer="331" from-port="0" to-layer="332" to-port="0" /> <edge from-layer="332" from-port="1" to-layer="333" to-port="1" /> <edge from-layer="333" from-port="2" to-layer="336" to-port="0" /> <edge from-layer="334" from-port="0" to-layer="335" to-port="0" /> <edge from-layer="335" from-port="1" to-layer="336" to-port="1" /> <edge from-layer="336" from-port="2" to-layer="337" to-port="1" /> <edge from-layer="337" from-port="2" to-layer="339" to-port="0" /> <edge from-layer="338" from-port="0" to-layer="339" to-port="1" /> <edge from-layer="339" from-port="2" to-layer="342" to-port="0" /> <edge from-layer="340" from-port="0" to-layer="341" to-port="0" /> <edge from-layer="341" from-port="1" to-layer="342" to-port="1" /> <edge from-layer="342" from-port="2" to-layer="345" to-port="0" /> <edge from-layer="343" from-port="0" to-layer="344" to-port="0" /> <edge from-layer="344" from-port="1" to-layer="345" to-port="1" /> <edge from-layer="345" from-port="2" to-layer="362" to-port="0" /> <edge from-layer="345" from-port="2" to-layer="385" to-port="0" /> <edge from-layer="345" from-port="2" to-layer="348" to-port="0" /> <edge from-layer="345" from-port="2" to-layer="406" to-port="0" /> <edge from-layer="346" from-port="0" to-layer="347" to-port="0" /> <edge from-layer="347" from-port="1" to-layer="348" to-port="1" /> <edge from-layer="348" from-port="2" to-layer="351" to-port="0" /> <edge from-layer="349" from-port="0" to-layer="350" to-port="0" /> <edge from-layer="350" from-port="1" to-layer="351" to-port="1" /> <edge from-layer="351" from-port="2" to-layer="354" to-port="0" /> <edge from-layer="352" from-port="0" to-layer="353" to-port="0" /> <edge from-layer="353" from-port="1" to-layer="354" to-port="1" /> <edge from-layer="354" from-port="2" to-layer="356" to-port="0" /> <edge from-layer="355" from-port="0" to-layer="356" to-port="1" /> <edge from-layer="356" from-port="2" to-layer="358" to-port="0" /> <edge from-layer="357" from-port="0" to-layer="358" to-port="1" /> <edge from-layer="358" from-port="2" to-layer="359" to-port="0" /> <edge from-layer="359" from-port="2" to-layer="371" to-port="0" /> <edge from-layer="360" from-port="0" to-layer="361" to-port="0" /> <edge from-layer="361" from-port="1" to-layer="362" to-port="1" /> <edge from-layer="362" from-port="2" to-layer="365" to-port="0" /> <edge from-layer="363" from-port="0" to-layer="364" to-port="0" /> <edge from-layer="364" from-port="1" to-layer="365" to-port="1" /> <edge from-layer="365" from-port="2" to-layer="367" to-port="0" /> <edge from-layer="366" from-port="0" to-layer="367" to-port="1" /> <edge from-layer="367" from-port="2" to-layer="369" to-port="0" /> <edge from-layer="368" from-port="0" to-layer="369" to-port="1" /> <edge from-layer="369" from-port="2" to-layer="370" to-port="0" /> <edge from-layer="370" from-port="2" to-layer="371" to-port="1" /> <edge from-layer="370" from-port="2" to-layer="373" to-port="0" /> <edge from-layer="371" from-port="2" to-layer="378" to-port="0" /> <edge from-layer="372" from-port="0" to-layer="377" to-port="1" /> <edge from-layer="373" from-port="1" to-layer="376" to-port="0" /> <edge from-layer="374" from-port="0" to-layer="376" to-port="1" /> <edge from-layer="375" from-port="0" to-layer="376" to-port="2" /> <edge from-layer="376" from-port="3" to-layer="377" to-port="3" /> <edge from-layer="376" from-port="3" to-layer="380" to-port="2" /> <edge from-layer="377" from-port="4" to-layer="378" to-port="1" /> <edge from-layer="378" from-port="2" to-layer="379" to-port="0" /> <edge from-layer="379" from-port="2" to-layer="381" to-port="0" /> <edge from-layer="380" from-port="3" to-layer="381" to-port="1" /> <edge from-layer="381" from-port="2" to-layer="382" to-port="0" /> <edge from-layer="382" from-port="1" to-layer="394" to-port="0" /> <edge from-layer="383" from-port="0" to-layer="384" to-port="0" /> <edge from-layer="384" from-port="1" to-layer="385" to-port="1" /> <edge from-layer="385" from-port="2" to-layer="388" to-port="0" /> <edge from-layer="386" from-port="0" to-layer="387" to-port="0" /> <edge from-layer="387" from-port="1" to-layer="388" to-port="1" /> <edge from-layer="388" from-port="2" to-layer="390" to-port="0" /> <edge from-layer="389" from-port="0" to-layer="390" to-port="1" /> <edge from-layer="390" from-port="2" to-layer="392" to-port="0" /> <edge from-layer="391" from-port="0" to-layer="392" to-port="1" /> <edge from-layer="392" from-port="2" to-layer="393" to-port="0" /> <edge from-layer="393" from-port="2" to-layer="394" to-port="1" /> <edge from-layer="394" from-port="2" to-layer="395" to-port="0" /> <edge from-layer="395" from-port="2" to-layer="397" to-port="0" /> <edge from-layer="396" from-port="0" to-layer="397" to-port="1" /> <edge from-layer="397" from-port="2" to-layer="399" to-port="0" /> <edge from-layer="398" from-port="0" to-layer="399" to-port="1" /> <edge from-layer="399" from-port="2" to-layer="402" to-port="0" /> <edge from-layer="400" from-port="0" to-layer="401" to-port="0" /> <edge from-layer="401" from-port="1" to-layer="402" to-port="1" /> <edge from-layer="402" from-port="2" to-layer="405" to-port="0" /> <edge from-layer="403" from-port="0" to-layer="404" to-port="0" /> <edge from-layer="404" from-port="1" to-layer="405" to-port="1" /> <edge from-layer="405" from-port="2" to-layer="406" to-port="1" /> <edge from-layer="406" from-port="2" to-layer="408" to-port="0" /> <edge from-layer="407" from-port="0" to-layer="408" to-port="1" /> <edge from-layer="408" from-port="2" to-layer="411" to-port="0" /> <edge from-layer="409" from-port="0" to-layer="410" to-port="0" /> <edge from-layer="410" from-port="1" to-layer="411" to-port="1" /> <edge from-layer="411" from-port="2" to-layer="414" to-port="0" /> <edge from-layer="412" from-port="0" to-layer="413" to-port="0" /> <edge from-layer="413" from-port="1" to-layer="414" to-port="1" /> <edge from-layer="414" from-port="2" to-layer="417" to-port="0" /> <edge from-layer="414" from-port="2" to-layer="428" to-port="0" /> <edge from-layer="415" from-port="0" to-layer="416" to-port="0" /> <edge from-layer="416" from-port="1" to-layer="417" to-port="1" /> <edge from-layer="417" from-port="2" to-layer="420" to-port="0" /> <edge from-layer="418" from-port="0" to-layer="419" to-port="0" /> <edge from-layer="419" from-port="1" to-layer="420" to-port="1" /> <edge from-layer="420" from-port="2" to-layer="421" to-port="0" /> <edge from-layer="421" from-port="1" to-layer="424" to-port="0" /> <edge from-layer="422" from-port="0" to-layer="423" to-port="0" /> <edge from-layer="423" from-port="1" to-layer="424" to-port="1" /> <edge from-layer="424" from-port="2" to-layer="427" to-port="0" /> <edge from-layer="425" from-port="0" to-layer="426" to-port="0" /> <edge from-layer="426" from-port="1" to-layer="427" to-port="1" /> <edge from-layer="427" from-port="2" to-layer="428" to-port="1" /> <edge from-layer="428" from-port="2" to-layer="430" to-port="0" /> <edge from-layer="429" from-port="0" to-layer="430" to-port="1" /> <edge from-layer="430" from-port="2" to-layer="433" to-port="0" /> <edge from-layer="431" from-port="0" to-layer="432" to-port="0" /> <edge from-layer="432" from-port="1" to-layer="433" to-port="1" /> <edge from-layer="433" from-port="2" to-layer="436" to-port="0" /> <edge from-layer="434" from-port="0" to-layer="435" to-port="0" /> <edge from-layer="435" from-port="1" to-layer="436" to-port="1" /> <edge from-layer="436" from-port="2" to-layer="476" to-port="0" /> <edge from-layer="436" from-port="2" to-layer="497" to-port="0" /> <edge from-layer="436" from-port="2" to-layer="439" to-port="0" /> <edge from-layer="436" from-port="2" to-layer="453" to-port="0" /> <edge from-layer="437" from-port="0" to-layer="438" to-port="0" /> <edge from-layer="438" from-port="1" to-layer="439" to-port="1" /> <edge from-layer="439" from-port="2" to-layer="442" to-port="0" /> <edge from-layer="440" from-port="0" to-layer="441" to-port="0" /> <edge from-layer="441" from-port="1" to-layer="442" to-port="1" /> <edge from-layer="442" from-port="2" to-layer="445" to-port="0" /> <edge from-layer="443" from-port="0" to-layer="444" to-port="0" /> <edge from-layer="444" from-port="1" to-layer="445" to-port="1" /> <edge from-layer="445" from-port="2" to-layer="447" to-port="0" /> <edge from-layer="446" from-port="0" to-layer="447" to-port="1" /> <edge from-layer="447" from-port="2" to-layer="449" to-port="0" /> <edge from-layer="448" from-port="0" to-layer="449" to-port="1" /> <edge from-layer="449" from-port="2" to-layer="450" to-port="0" /> <edge from-layer="450" from-port="2" to-layer="462" to-port="0" /> <edge from-layer="451" from-port="0" to-layer="452" to-port="0" /> <edge from-layer="452" from-port="1" to-layer="453" to-port="1" /> <edge from-layer="453" from-port="2" to-layer="456" to-port="0" /> <edge from-layer="454" from-port="0" to-layer="455" to-port="0" /> <edge from-layer="455" from-port="1" to-layer="456" to-port="1" /> <edge from-layer="456" from-port="2" to-layer="458" to-port="0" /> <edge from-layer="457" from-port="0" to-layer="458" to-port="1" /> <edge from-layer="458" from-port="2" to-layer="460" to-port="0" /> <edge from-layer="459" from-port="0" to-layer="460" to-port="1" /> <edge from-layer="460" from-port="2" to-layer="461" to-port="0" /> <edge from-layer="461" from-port="2" to-layer="462" to-port="1" /> <edge from-layer="461" from-port="2" to-layer="464" to-port="0" /> <edge from-layer="462" from-port="2" to-layer="469" to-port="0" /> <edge from-layer="463" from-port="0" to-layer="468" to-port="1" /> <edge from-layer="464" from-port="1" to-layer="467" to-port="0" /> <edge from-layer="465" from-port="0" to-layer="467" to-port="1" /> <edge from-layer="466" from-port="0" to-layer="467" to-port="2" /> <edge from-layer="467" from-port="3" to-layer="468" to-port="3" /> <edge from-layer="467" from-port="3" to-layer="471" to-port="2" /> <edge from-layer="468" from-port="4" to-layer="469" to-port="1" /> <edge from-layer="469" from-port="2" to-layer="470" to-port="0" /> <edge from-layer="470" from-port="2" to-layer="472" to-port="0" /> <edge from-layer="471" from-port="3" to-layer="472" to-port="1" /> <edge from-layer="472" from-port="2" to-layer="473" to-port="0" /> <edge from-layer="473" from-port="1" to-layer="485" to-port="0" /> <edge from-layer="474" from-port="0" to-layer="475" to-port="0" /> <edge from-layer="475" from-port="1" to-layer="476" to-port="1" /> <edge from-layer="476" from-port="2" to-layer="479" to-port="0" /> <edge from-layer="477" from-port="0" to-layer="478" to-port="0" /> <edge from-layer="478" from-port="1" to-layer="479" to-port="1" /> <edge from-layer="479" from-port="2" to-layer="481" to-port="0" /> <edge from-layer="480" from-port="0" to-layer="481" to-port="1" /> <edge from-layer="481" from-port="2" to-layer="483" to-port="0" /> <edge from-layer="482" from-port="0" to-layer="483" to-port="1" /> <edge from-layer="483" from-port="2" to-layer="484" to-port="0" /> <edge from-layer="484" from-port="2" to-layer="485" to-port="1" /> <edge from-layer="485" from-port="2" to-layer="486" to-port="0" /> <edge from-layer="486" from-port="2" to-layer="488" to-port="0" /> <edge from-layer="487" from-port="0" to-layer="488" to-port="1" /> <edge from-layer="488" from-port="2" to-layer="490" to-port="0" /> <edge from-layer="489" from-port="0" to-layer="490" to-port="1" /> <edge from-layer="490" from-port="2" to-layer="493" to-port="0" /> <edge from-layer="491" from-port="0" to-layer="492" to-port="0" /> <edge from-layer="492" from-port="1" to-layer="493" to-port="1" /> <edge from-layer="493" from-port="2" to-layer="496" to-port="0" /> <edge from-layer="494" from-port="0" to-layer="495" to-port="0" /> <edge from-layer="495" from-port="1" to-layer="496" to-port="1" /> <edge from-layer="496" from-port="2" to-layer="497" to-port="1" /> <edge from-layer="497" from-port="2" to-layer="499" to-port="0" /> <edge from-layer="498" from-port="0" to-layer="499" to-port="1" /> <edge from-layer="499" from-port="2" to-layer="502" to-port="0" /> <edge from-layer="500" from-port="0" to-layer="501" to-port="0" /> <edge from-layer="501" from-port="1" to-layer="502" to-port="1" /> <edge from-layer="502" from-port="2" to-layer="505" to-port="0" /> <edge from-layer="503" from-port="0" to-layer="504" to-port="0" /> <edge from-layer="504" from-port="1" to-layer="505" to-port="1" /> <edge from-layer="505" from-port="2" to-layer="519" to-port="0" /> <edge from-layer="505" from-port="2" to-layer="508" to-port="0" /> <edge from-layer="506" from-port="0" to-layer="507" to-port="0" /> <edge from-layer="507" from-port="1" to-layer="508" to-port="1" /> <edge from-layer="508" from-port="2" to-layer="511" to-port="0" /> <edge from-layer="509" from-port="0" to-layer="510" to-port="0" /> <edge from-layer="510" from-port="1" to-layer="511" to-port="1" /> <edge from-layer="511" from-port="2" to-layer="512" to-port="0" /> <edge from-layer="512" from-port="1" to-layer="515" to-port="0" /> <edge from-layer="513" from-port="0" to-layer="514" to-port="0" /> <edge from-layer="514" from-port="1" to-layer="515" to-port="1" /> <edge from-layer="515" from-port="2" to-layer="518" to-port="0" /> <edge from-layer="516" from-port="0" to-layer="517" to-port="0" /> <edge from-layer="517" from-port="1" to-layer="518" to-port="1" /> <edge from-layer="518" from-port="2" to-layer="519" to-port="1" /> <edge from-layer="519" from-port="2" to-layer="521" to-port="0" /> <edge from-layer="520" from-port="0" to-layer="521" to-port="1" /> <edge from-layer="521" from-port="2" to-layer="524" to-port="0" /> <edge from-layer="522" from-port="0" to-layer="523" to-port="0" /> <edge from-layer="523" from-port="1" to-layer="524" to-port="1" /> <edge from-layer="524" from-port="2" to-layer="527" to-port="0" /> <edge from-layer="525" from-port="0" to-layer="526" to-port="0" /> <edge from-layer="526" from-port="1" to-layer="527" to-port="1" /> <edge from-layer="527" from-port="2" to-layer="544" to-port="0" /> <edge from-layer="527" from-port="2" to-layer="567" to-port="0" /> <edge from-layer="527" from-port="2" to-layer="530" to-port="0" /> <edge from-layer="527" from-port="2" to-layer="588" to-port="0" /> <edge from-layer="528" from-port="0" to-layer="529" to-port="0" /> <edge from-layer="529" from-port="1" to-layer="530" to-port="1" /> <edge from-layer="530" from-port="2" to-layer="533" to-port="0" /> <edge from-layer="531" from-port="0" to-layer="532" to-port="0" /> <edge from-layer="532" from-port="1" to-layer="533" to-port="1" /> <edge from-layer="533" from-port="2" to-layer="536" to-port="0" /> <edge from-layer="534" from-port="0" to-layer="535" to-port="0" /> <edge from-layer="535" from-port="1" to-layer="536" to-port="1" /> <edge from-layer="536" from-port="2" to-layer="538" to-port="0" /> <edge from-layer="537" from-port="0" to-layer="538" to-port="1" /> <edge from-layer="538" from-port="2" to-layer="540" to-port="0" /> <edge from-layer="539" from-port="0" to-layer="540" to-port="1" /> <edge from-layer="540" from-port="2" to-layer="541" to-port="0" /> <edge from-layer="541" from-port="2" to-layer="553" to-port="0" /> <edge from-layer="542" from-port="0" to-layer="543" to-port="0" /> <edge from-layer="543" from-port="1" to-layer="544" to-port="1" /> <edge from-layer="544" from-port="2" to-layer="547" to-port="0" /> <edge from-layer="545" from-port="0" to-layer="546" to-port="0" /> <edge from-layer="546" from-port="1" to-layer="547" to-port="1" /> <edge from-layer="547" from-port="2" to-layer="549" to-port="0" /> <edge from-layer="548" from-port="0" to-layer="549" to-port="1" /> <edge from-layer="549" from-port="2" to-layer="551" to-port="0" /> <edge from-layer="550" from-port="0" to-layer="551" to-port="1" /> <edge from-layer="551" from-port="2" to-layer="552" to-port="0" /> <edge from-layer="552" from-port="2" to-layer="553" to-port="1" /> <edge from-layer="552" from-port="2" to-layer="555" to-port="0" /> <edge from-layer="553" from-port="2" to-layer="560" to-port="0" /> <edge from-layer="554" from-port="0" to-layer="559" to-port="1" /> <edge from-layer="555" from-port="1" to-layer="558" to-port="0" /> <edge from-layer="556" from-port="0" to-layer="558" to-port="1" /> <edge from-layer="557" from-port="0" to-layer="558" to-port="2" /> <edge from-layer="558" from-port="3" to-layer="559" to-port="3" /> <edge from-layer="558" from-port="3" to-layer="562" to-port="2" /> <edge from-layer="559" from-port="4" to-layer="560" to-port="1" /> <edge from-layer="560" from-port="2" to-layer="561" to-port="0" /> <edge from-layer="561" from-port="2" to-layer="563" to-port="0" /> <edge from-layer="562" from-port="3" to-layer="563" to-port="1" /> <edge from-layer="563" from-port="2" to-layer="564" to-port="0" /> <edge from-layer="564" from-port="1" to-layer="576" to-port="0" /> <edge from-layer="565" from-port="0" to-layer="566" to-port="0" /> <edge from-layer="566" from-port="1" to-layer="567" to-port="1" /> <edge from-layer="567" from-port="2" to-layer="570" to-port="0" /> <edge from-layer="568" from-port="0" to-layer="569" to-port="0" /> <edge from-layer="569" from-port="1" to-layer="570" to-port="1" /> <edge from-layer="570" from-port="2" to-layer="572" to-port="0" /> <edge from-layer="571" from-port="0" to-layer="572" to-port="1" /> <edge from-layer="572" from-port="2" to-layer="574" to-port="0" /> <edge from-layer="573" from-port="0" to-layer="574" to-port="1" /> <edge from-layer="574" from-port="2" to-layer="575" to-port="0" /> <edge from-layer="575" from-port="2" to-layer="576" to-port="1" /> <edge from-layer="576" from-port="2" to-layer="577" to-port="0" /> <edge from-layer="577" from-port="2" to-layer="579" to-port="0" /> <edge from-layer="578" from-port="0" to-layer="579" to-port="1" /> <edge from-layer="579" from-port="2" to-layer="581" to-port="0" /> <edge from-layer="580" from-port="0" to-layer="581" to-port="1" /> <edge from-layer="581" from-port="2" to-layer="584" to-port="0" /> <edge from-layer="582" from-port="0" to-layer="583" to-port="0" /> <edge from-layer="583" from-port="1" to-layer="584" to-port="1" /> <edge from-layer="584" from-port="2" to-layer="587" to-port="0" /> <edge from-layer="585" from-port="0" to-layer="586" to-port="0" /> <edge from-layer="586" from-port="1" to-layer="587" to-port="1" /> <edge from-layer="587" from-port="2" to-layer="588" to-port="1" /> <edge from-layer="588" from-port="2" to-layer="590" to-port="0" /> <edge from-layer="589" from-port="0" to-layer="590" to-port="1" /> <edge from-layer="590" from-port="2" to-layer="593" to-port="0" /> <edge from-layer="591" from-port="0" to-layer="592" to-port="0" /> <edge from-layer="592" from-port="1" to-layer="593" to-port="1" /> <edge from-layer="593" from-port="2" to-layer="596" to-port="0" /> <edge from-layer="594" from-port="0" to-layer="595" to-port="0" /> <edge from-layer="595" from-port="1" to-layer="596" to-port="1" /> <edge from-layer="596" from-port="2" to-layer="599" to-port="0" /> <edge from-layer="596" from-port="2" to-layer="610" to-port="0" /> <edge from-layer="597" from-port="0" to-layer="598" to-port="0" /> <edge from-layer="598" from-port="1" to-layer="599" to-port="1" /> <edge from-layer="599" from-port="2" to-layer="602" to-port="0" /> <edge from-layer="600" from-port="0" to-layer="601" to-port="0" /> <edge from-layer="601" from-port="1" to-layer="602" to-port="1" /> <edge from-layer="602" from-port="2" to-layer="603" to-port="0" /> <edge from-layer="603" from-port="1" to-layer="606" to-port="0" /> <edge from-layer="604" from-port="0" to-layer="605" to-port="0" /> <edge from-layer="605" from-port="1" to-layer="606" to-port="1" /> <edge from-layer="606" from-port="2" to-layer="609" to-port="0" /> <edge from-layer="607" from-port="0" to-layer="608" to-port="0" /> <edge from-layer="608" from-port="1" to-layer="609" to-port="1" /> <edge from-layer="609" from-port="2" to-layer="610" to-port="1" /> <edge from-layer="610" from-port="2" to-layer="612" to-port="0" /> <edge from-layer="611" from-port="0" to-layer="612" to-port="1" /> <edge from-layer="612" from-port="2" to-layer="615" to-port="0" /> <edge from-layer="613" from-port="0" to-layer="614" to-port="0" /> <edge from-layer="614" from-port="1" to-layer="615" to-port="1" /> <edge from-layer="615" from-port="2" to-layer="618" to-port="0" /> <edge from-layer="616" from-port="0" to-layer="617" to-port="0" /> <edge from-layer="617" from-port="1" to-layer="618" to-port="1" /> <edge from-layer="618" from-port="2" to-layer="619" to-port="0" /> </edges> <rt_info> <info name="OpenVINO Runtime" value="2026.3.1-22476-759c5a6ab8c-releases/2026/3" /> <Runtime_version value="2026.3.1-22476-759c5a6ab8c-releases/2026/3" /> <conversion_parameters> <framework value="pytorch" /> <is_python_object value="True" /> </conversion_parameters> </rt_info> </net>
<?xml version="1.0"?>
<net name="Model9" version="11">
<layers>
<layer id="0" name="attention_mask" type="Parameter" version="opset1">
<data shape="?,?" element_type="f32" />
<output>
<port id="0" precision="FP32" names="69,attention_mask">
<dim>-1</dim>
<dim>-1</dim>
</port>
</output>
</layer>
<layer id="1" name="inputs_embeds" type="Parameter" version="opset1">
<data shape="?,?,?" element_type="f32" />
<output>
<port id="0" precision="FP32" names="46,47,inputs_embeds">
<dim>-1</dim>
<dim>-1</dim>
<dim>-1</dim>
</port>
</output>
</layer>
<layer id="2" name="self.model.embed_positions.weight_compressed" type="Const" version="opset1">
<data element_type="f16" shape="1026, 768" offset="0" size="1575936" />
<output>
<port id="0" precision="FP16" names="self.model.embed_positions.weight">
<dim>1026</dim>
<dim>768</dim>
</port>
</output>
</layer>
<layer id="3" name="self.model.embed_positions.weight" type="Convert" version="opset1">
<data destination_type="f32" />
<rt_info>
<attribute name="decompression" version="0" />
</rt_info>
<input>
<port id="0" precision="FP16">
<dim>1026</dim>
<dim>768</dim>
</port>
</input>
<output>
<port id="1" precision="FP32">
<dim>1026</dim>
<dim>768</dim>
</port>
</output>
</layer>
<layer id="4" name="31" type="Const" version="opset1">
<data element_type="i64" shape="" offset="1575936" size="8" />
<output>
<port id="0" precision="I64" names="31" />
</output>
</layer>
<layer id="5" name="ShapeOf_154554" type="ShapeOf" version="opset3">
<data output_type="i64" />
<input>
<port id="0" precision="FP32">
<dim>-1</dim>
<dim>-1</dim>
<dim>-1</dim>
</port>
</input>
<output>
<port id="1" precision="I64">
<dim>3</dim>
</port>
</output>
</layer>
<layer id="6" name="Constant_154555" type="Const" version="opset1">
<data element_type="i64" shape="" offset="1575944" size="8" />
<rt_info>
<attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" />
</rt_info>
<output>
<port id="0" precision="I64" />
</output>
</layer>
<layer id="7" name="Constant_154556" type="Const" version="opset1">
<data element_type="i64" shape="" offset="1575936" size="8" />
<rt_info>
<attribute name="DisablePrecisionConversion" version="0" value="dynamic:f16" />
</rt_info>
<output>
<port id="0" precision="I64" />
</output>
</layer>
<layer id="8" name="Gather_154557" type="Gather" version="opset8">
<data batch_dims="0" />
<input>
<port id="0" precision="I64">
<dim>3</dim>
</port>
<port id="1" precision="I64" />
<port id="2" precision="I64" />
</input>
<output>
<port id="3" precision="I64" names="51,53" />
</output>