mirror of
https://github.com/pese-git/llm-arch-research.git
synced 2026-01-23 21:10:54 +00:00
- Renamed GPT-1 main scheme files for clarity - Added new diagram files for attention, decoder, embeddings, and forward blocks (both .drawio and .png) - Removed deprecated files (gpt11.drawio, gpt1.svg) - Updated notebooks/gpt.ipynb with relevant changes
148 lines
14 KiB
Plaintext
148 lines
14 KiB
Plaintext
<mxfile host="65bd71144e">
|
|
<diagram name="GPT Architecture" id="DEYydPS-O6mnllJWumln">
|
|
<mxGraphModel dx="1216" dy="316" grid="1" gridSize="10" guides="1" tooltips="1" connect="1" arrows="1" fold="1" page="1" pageScale="1" pageWidth="827" pageHeight="1169" math="0" shadow="0">
|
|
<root>
|
|
<mxCell id="0"/>
|
|
<mxCell id="1" parent="0"/>
|
|
<mxCell id="91" value="" style="group;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="1" vertex="1" connectable="0">
|
|
<mxGeometry x="40" y="360" width="1286" height="160" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="56" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#fff2cc;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" vertex="1">
|
|
<mxGeometry x="230" width="440" height="160" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="57" value="<div>Masked</div>Multi+Head<br>Attention" style="rounded=0;whiteSpace=wrap;html=1;strokeColor=#6c8ebf;fillColor=#dae8fc;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="56" vertex="1">
|
|
<mxGeometry x="51.42776556776556" y="50" width="78.97435897435898" height="60" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="58" value="" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="56" source="59" edge="1">
|
|
<mxGeometry relative="1" as="geometry">
|
|
<mxPoint x="350" y="80" as="targetPoint"/>
|
|
</mxGeometry>
|
|
</mxCell>
|
|
<mxCell id="59" value="Feed<div>Forward</div><div>Network</div>" style="rounded=0;whiteSpace=wrap;html=1;strokeColor=#9673a6;fillColor=#e1d5e7;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="56" vertex="1">
|
|
<mxGeometry x="260.9564102564102" y="50" width="71.9230769230769" height="60" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="60" value="Norm" style="rounded=0;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="56" vertex="1">
|
|
<mxGeometry x="379.997619047619" y="60" width="37.87142857142857" height="40" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="61" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="56" source="62" target="59" edge="1">
|
|
<mxGeometry relative="1" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="62" value="Norm" style="rounded=0;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="56" vertex="1">
|
|
<mxGeometry x="177.14285714285714" y="60" width="41.904761904761905" height="40" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="63" value="" style="endArrow=classic;html=1;exitX=0;exitY=0.5;exitDx=0;exitDy=0;edgeStyle=elbowEdgeStyle;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="56" source="56" target="57" edge="1">
|
|
<mxGeometry width="50" height="50" relative="1" as="geometry">
|
|
<mxPoint x="20" y="80.00000000000011" as="sourcePoint"/>
|
|
<mxPoint x="229.52380952380952" y="-50" as="targetPoint"/>
|
|
</mxGeometry>
|
|
</mxCell>
|
|
<mxCell id="64" value="" style="endArrow=classic;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;edgeStyle=orthogonalEdgeStyle;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="56" source="65" target="62" edge="1">
|
|
<mxGeometry width="50" height="50" relative="1" as="geometry">
|
|
<mxPoint x="155.71428571427464" y="79.99999999999989" as="sourcePoint"/>
|
|
<mxPoint x="229.52380952380952" y="-50" as="targetPoint"/>
|
|
</mxGeometry>
|
|
</mxCell>
|
|
<mxCell id="65" value="+" style="ellipse;whiteSpace=wrap;html=1;aspect=fixed;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="56" vertex="1">
|
|
<mxGeometry x="150.00428571428571" y="75" width="10" height="10" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="66" value="" style="endArrow=classic;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;edgeStyle=orthogonalEdgeStyle;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="56" source="57" target="65" edge="1">
|
|
<mxGeometry width="50" height="50" relative="1" as="geometry">
|
|
<mxPoint x="213.80952380952382" y="410" as="sourcePoint"/>
|
|
<mxPoint x="145.71428571428578" y="80.00000000000011" as="targetPoint"/>
|
|
</mxGeometry>
|
|
</mxCell>
|
|
<mxCell id="67" value="" style="endArrow=classic;html=1;entryX=0.5;entryY=1;entryDx=0;entryDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="56" target="69" edge="1">
|
|
<mxGeometry width="50" height="50" relative="1" as="geometry">
|
|
<mxPoint x="236.85714285714286" y="80" as="sourcePoint"/>
|
|
<mxPoint x="349.7619047619048" y="85" as="targetPoint"/>
|
|
<Array as="points">
|
|
<mxPoint x="236.67904761904765" y="125"/>
|
|
<mxPoint x="292.38095238095235" y="125"/>
|
|
<mxPoint x="355" y="125"/>
|
|
</Array>
|
|
</mxGeometry>
|
|
</mxCell>
|
|
<mxCell id="68" value="" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="56" source="69" target="60" edge="1">
|
|
<mxGeometry relative="1" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="69" value="+" style="ellipse;whiteSpace=wrap;html=1;aspect=fixed;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="56" vertex="1">
|
|
<mxGeometry x="350.00190476190477" y="75" width="10" height="10" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="70" value="" style="endArrow=classic;html=1;entryX=0.5;entryY=1;entryDx=0;entryDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="56" target="65" edge="1">
|
|
<mxGeometry width="50" height="50" relative="1" as="geometry">
|
|
<mxPoint x="34.325581395348834" y="80" as="sourcePoint"/>
|
|
<mxPoint x="150.71428571428578" y="85" as="targetPoint"/>
|
|
<Array as="points">
|
|
<mxPoint x="34.25858250276859" y="130"/>
|
|
<mxPoint x="89.96048726467328" y="130"/>
|
|
<mxPoint x="155" y="130"/>
|
|
</Array>
|
|
</mxGeometry>
|
|
</mxCell>
|
|
<mxCell id="71" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" source="72" target="56" edge="1">
|
|
<mxGeometry relative="1" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="72" value="+" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#f5f5f5;fontColor=#333333;strokeColor=#FF3333;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" vertex="1">
|
|
<mxGeometry x="90" width="110" height="160" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="73" value="Token Emb" style="rounded=0;whiteSpace=wrap;html=1;strokeColor=#6c8ebf;fillColor=#dae8fc;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" vertex="1">
|
|
<mxGeometry x="95" y="17.5" width="100" height="42.5" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="74" value="Position Emb" style="rounded=0;whiteSpace=wrap;html=1;strokeColor=#9673a6;fillColor=#e1d5e7;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" vertex="1">
|
|
<mxGeometry x="95" y="100" width="100" height="42.5" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="75" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" source="76" target="79" edge="1">
|
|
<mxGeometry relative="1" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="76" value="Decoder" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" vertex="1">
|
|
<mxGeometry x="690" width="70" height="160" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="77" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" source="60" target="76" edge="1">
|
|
<mxGeometry relative="1" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="78" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" source="79" target="85" edge="1">
|
|
<mxGeometry relative="1" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="79" value="Decoder" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" vertex="1">
|
|
<mxGeometry x="790" width="70" height="160" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="80" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" source="81" target="83" edge="1">
|
|
<mxGeometry relative="1" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="81" value="Decoder" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" vertex="1">
|
|
<mxGeometry x="950" width="70" height="160" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="82" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" source="83" target="87" edge="1">
|
|
<mxGeometry relative="1" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="83" value="Decoder" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" vertex="1">
|
|
<mxGeometry x="1050" width="70" height="160" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="84" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" source="85" target="81" edge="1">
|
|
<mxGeometry relative="1" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="85" value=".<div>.</div><div>.</div>" style="rounded=1;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" vertex="1">
|
|
<mxGeometry x="890" y="40" width="30" height="80" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="86" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" source="87" target="88" edge="1">
|
|
<mxGeometry relative="1" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="87" value="Linear" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;fontColor=#333333;strokeColor=#666666;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" vertex="1">
|
|
<mxGeometry x="1150" y="5" width="50" height="150" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="88" value="Softmax" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#e1d5e7;strokeColor=#9673a6;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" vertex="1">
|
|
<mxGeometry x="1236" y="5" width="50" height="150" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="89" value="Tokens" style="rounded=1;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" vertex="1">
|
|
<mxGeometry y="40" width="60" height="90" as="geometry"/>
|
|
</mxCell>
|
|
<mxCell id="90" style="edgeStyle=none;html=1;entryX=-0.025;entryY=0.538;entryDx=0;entryDy=0;entryPerimeter=0;exitX=1;exitY=0.5;exitDx=0;exitDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="91" source="89" edge="1">
|
|
<mxGeometry relative="1" as="geometry">
|
|
<mxPoint x="42.75" y="84.66941747572821" as="sourcePoint"/>
|
|
<mxPoint x="90" y="85.33000000000004" as="targetPoint"/>
|
|
</mxGeometry>
|
|
</mxCell>
|
|
</root>
|
|
</mxGraphModel>
|
|
</diagram>
|
|
</mxfile> |