penkovsky_sa/llm-arch-research
1
0
Fork 0
mirror of https://github.com/pese-git/llm-arch-research.git synced 2026-01-23 21:10:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

refactor(assets): update and reorganize GPT-1 architecture diagrams

Browse Source 
- 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
This commit is contained in:
Sergey Penkovsky
2025-10-30 14:40:31 +03:00
parent 9e2796e6be
commit 21cfd79c19
13 changed files with 995 additions and 144 deletions
Download Patch File Download Diff File Expand all files Collapse all files

103
assets/drawio/gpt1.drawio → assets/drawio/gpt1-architecture.drawio
View File

@@ -1,54 +1,57 @@
<mxfile host="65bd71144e"> <mxfile host="65bd71144e">
<diagram name="GPT Architecture" id="DEYydPS-O6mnllJWumln"> <diagram name="GPT Architecture" id="DEYydPS-O6mnllJWumln">
<mxGraphModel dx="1339" dy="371" 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"> <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> <root>
<mxCell id="0"/> <mxCell id="0"/>
<mxCell id="1" parent="0"/> <mxCell id="1" parent="0"/>
<mxCell id="3" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#fff2cc;" vertex="1" parent="1"> <mxCell id="92" value="" style="group" vertex="1" connectable="0" parent="1">
<mxGeometry x="280" y="330" width="440" height="160" as="geometry"/> <mxGeometry x="40" y="320" width="1286" height="160" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="4" value="&lt;div&gt;Masked&lt;/div&gt;Multi+Head&lt;br&gt;Attention" style="rounded=0;whiteSpace=wrap;html=1;strokeColor=#6c8ebf;fillColor=#dae8fc;" vertex="1" parent="3"> <mxCell id="3" 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="92" vertex="1">
<mxGeometry x="230" width="440" height="160" as="geometry"/>
</mxCell>
<mxCell id="4" value="&lt;div&gt;Masked&lt;/div&gt;Multi+Head&lt;br&gt;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="3" vertex="1">
<mxGeometry x="51.42776556776556" y="50" width="78.97435897435898" height="60" as="geometry"/> <mxGeometry x="51.42776556776556" y="50" width="78.97435897435898" height="60" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="22" value="" style="edgeStyle=none;html=1;" edge="1" parent="3" source="5"> <mxCell id="22" value="" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" source="5" edge="1">
<mxGeometry relative="1" as="geometry"> <mxGeometry relative="1" as="geometry">
<mxPoint x="350" y="80" as="targetPoint"/> <mxPoint x="350" y="80" as="targetPoint"/>
</mxGeometry> </mxGeometry>
</mxCell> </mxCell>
<mxCell id="5" value="Feed&lt;div&gt;Forward&lt;/div&gt;&lt;div&gt;Network&lt;/div&gt;" style="rounded=0;whiteSpace=wrap;html=1;strokeColor=#9673a6;fillColor=#e1d5e7;" vertex="1" parent="3"> <mxCell id="5" value="Feed&lt;div&gt;Forward&lt;/div&gt;&lt;div&gt;Network&lt;/div&gt;" 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="3" vertex="1">
<mxGeometry x="260.9564102564102" y="50" width="71.9230769230769" height="60" as="geometry"/> <mxGeometry x="260.9564102564102" y="50" width="71.9230769230769" height="60" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="7" value="Norm" style="rounded=0;whiteSpace=wrap;html=1;" vertex="1" parent="3"> <mxCell id="7" value="Norm" style="rounded=0;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" vertex="1">
<mxGeometry x="379.997619047619" y="60" width="37.87142857142857" height="40" as="geometry"/> <mxGeometry x="379.997619047619" y="60" width="37.87142857142857" height="40" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="21" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;" edge="1" parent="3" source="12" target="5"> <mxCell id="21" 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="3" source="12" target="5" edge="1">
<mxGeometry relative="1" as="geometry"/> <mxGeometry relative="1" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="12" value="Norm" style="rounded=0;whiteSpace=wrap;html=1;" vertex="1" parent="3"> <mxCell id="12" value="Norm" style="rounded=0;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" vertex="1">
<mxGeometry x="177.14285714285714" y="60" width="41.904761904761905" height="40" as="geometry"/> <mxGeometry x="177.14285714285714" y="60" width="41.904761904761905" height="40" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="13" value="" style="endArrow=classic;html=1;exitX=0;exitY=0.5;exitDx=0;exitDy=0;edgeStyle=elbowEdgeStyle;" edge="1" parent="3" source="3" target="4"> <mxCell id="13" 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="3" source="3" target="4" edge="1">
<mxGeometry width="50" height="50" relative="1" as="geometry"> <mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="20" y="80.00000000000011" as="sourcePoint"/> <mxPoint x="20" y="80.00000000000011" as="sourcePoint"/>
<mxPoint x="229.52380952380952" y="-50" as="targetPoint"/> <mxPoint x="229.52380952380952" y="-50" as="targetPoint"/>
</mxGeometry> </mxGeometry>
</mxCell> </mxCell>
<mxCell id="14" value="" style="endArrow=classic;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;edgeStyle=orthogonalEdgeStyle;" edge="1" parent="3" source="18" target="12"> <mxCell id="14" 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="3" source="18" target="12" edge="1">
<mxGeometry width="50" height="50" relative="1" as="geometry"> <mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="155.71428571427464" y="79.99999999999989" as="sourcePoint"/> <mxPoint x="155.71428571427464" y="79.99999999999989" as="sourcePoint"/>
<mxPoint x="229.52380952380952" y="-50" as="targetPoint"/> <mxPoint x="229.52380952380952" y="-50" as="targetPoint"/>
</mxGeometry> </mxGeometry>
</mxCell> </mxCell>
<mxCell id="18" value="+" style="ellipse;whiteSpace=wrap;html=1;aspect=fixed;" vertex="1" parent="3"> <mxCell id="18" value="+" style="ellipse;whiteSpace=wrap;html=1;aspect=fixed;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" vertex="1">
<mxGeometry x="150.00428571428571" y="75" width="10" height="10" as="geometry"/> <mxGeometry x="150.00428571428571" y="75" width="10" height="10" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="19" 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;" edge="1" parent="3" source="4" target="18"> <mxCell id="19" 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="3" source="4" target="18" edge="1">
<mxGeometry width="50" height="50" relative="1" as="geometry"> <mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="213.80952380952382" y="410" as="sourcePoint"/> <mxPoint x="213.80952380952382" y="410" as="sourcePoint"/>
<mxPoint x="145.71428571428578" y="80.00000000000011" as="targetPoint"/> <mxPoint x="145.71428571428578" y="80.00000000000011" as="targetPoint"/>
</mxGeometry> </mxGeometry>
</mxCell> </mxCell>
<mxCell id="23" value="" style="endArrow=classic;html=1;entryX=0.5;entryY=1;entryDx=0;entryDy=0;" edge="1" parent="3" target="24"> <mxCell id="23" 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="3" target="24" edge="1">
<mxGeometry width="50" height="50" relative="1" as="geometry"> <mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="236.85714285714286" y="80" as="sourcePoint"/> <mxPoint x="236.85714285714286" y="80" as="sourcePoint"/>
<mxPoint x="349.7619047619048" y="85" as="targetPoint"/> <mxPoint x="349.7619047619048" y="85" as="targetPoint"/>
@@ -59,13 +62,13 @@
</Array> </Array>
</mxGeometry> </mxGeometry>
</mxCell> </mxCell>
<mxCell id="28" value="" style="edgeStyle=none;html=1;" edge="1" parent="3" source="24" target="7"> <mxCell id="28" value="" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" source="24" target="7" edge="1">
<mxGeometry relative="1" as="geometry"/> <mxGeometry relative="1" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="24" value="+" style="ellipse;whiteSpace=wrap;html=1;aspect=fixed;" vertex="1" parent="3"> <mxCell id="24" value="+" style="ellipse;whiteSpace=wrap;html=1;aspect=fixed;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" vertex="1">
<mxGeometry x="350.00190476190477" y="75" width="10" height="10" as="geometry"/> <mxGeometry x="350.00190476190477" y="75" width="10" height="10" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="25" value="" style="endArrow=classic;html=1;entryX=0.5;entryY=1;entryDx=0;entryDy=0;" edge="1" parent="3" target="18"> <mxCell id="25" 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="3" target="18" edge="1">
<mxGeometry width="50" height="50" relative="1" as="geometry"> <mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="34.325581395348834" y="80" as="sourcePoint"/> <mxPoint x="34.325581395348834" y="80" as="sourcePoint"/>
<mxPoint x="150.71428571428578" y="85" as="targetPoint"/> <mxPoint x="150.71428571428578" y="85" as="targetPoint"/>
@@ -76,67 +79,67 @@
</Array> </Array>
</mxGeometry> </mxGeometry>
</mxCell> </mxCell>
<mxCell id="36" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;" edge="1" parent="1" source="32" target="3"> <mxCell id="36" 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="92" source="32" target="3" edge="1">
<mxGeometry relative="1" as="geometry"/> <mxGeometry relative="1" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="32" value="+" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#f5f5f5;fontColor=#333333;strokeColor=#666666;" vertex="1" parent="1"> <mxCell id="32" value="+" style="rounded=0;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="92" vertex="1">
<mxGeometry x="140" y="330" width="110" height="160" as="geometry"/> <mxGeometry x="90" width="110" height="160" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="33" value="Token Emb" style="rounded=0;whiteSpace=wrap;html=1;strokeColor=#6c8ebf;fillColor=#dae8fc;" vertex="1" parent="1"> <mxCell id="33" 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="92" vertex="1">
<mxGeometry x="145" y="347.5" width="100" height="42.5" as="geometry"/> <mxGeometry x="95" y="17.5" width="100" height="42.5" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="34" value="Position Emb" style="rounded=0;whiteSpace=wrap;html=1;strokeColor=#9673a6;fillColor=#e1d5e7;" vertex="1" parent="1"> <mxCell id="34" 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="92" vertex="1">
<mxGeometry x="145" y="430" width="100" height="42.5" as="geometry"/> <mxGeometry x="95" y="100" width="100" height="42.5" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="46" style="edgeStyle=none;html=1;" edge="1" parent="1" source="37" target="40"> <mxCell id="46" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" source="37" target="40" edge="1">
<mxGeometry relative="1" as="geometry"/> <mxGeometry relative="1" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="37" value="Decoder" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"> <mxCell id="37" 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="92" vertex="1">
<mxGeometry x="740" y="330" width="70" height="160" as="geometry"/> <mxGeometry x="690" width="70" height="160" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="38" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;" edge="1" parent="1" source="7" target="37"> <mxCell id="38" 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="92" source="7" target="37" edge="1">
<mxGeometry relative="1" as="geometry"/> <mxGeometry relative="1" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="47" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;" edge="1" parent="1" source="40" target="44"> <mxCell id="47" 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="92" source="40" target="44" edge="1">
<mxGeometry relative="1" as="geometry"/> <mxGeometry relative="1" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="40" value="Decoder" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"> <mxCell id="40" 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="92" vertex="1">
<mxGeometry x="840" y="330" width="70" height="160" as="geometry"/> <mxGeometry x="790" width="70" height="160" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="49" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;" edge="1" parent="1" source="41" target="42"> <mxCell id="49" 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="92" source="41" target="42" edge="1">
<mxGeometry relative="1" as="geometry"/> <mxGeometry relative="1" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="41" value="Decoder" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"> <mxCell id="41" 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="92" vertex="1">
<mxGeometry x="1000" y="330" width="70" height="160" as="geometry"/> <mxGeometry x="950" width="70" height="160" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="52" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;" edge="1" parent="1" source="42" target="50"> <mxCell id="52" 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="92" source="42" target="50" edge="1">
<mxGeometry relative="1" as="geometry"/> <mxGeometry relative="1" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="42" value="Decoder" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;strokeColor=#82b366;" vertex="1" parent="1"> <mxCell id="42" 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="92" vertex="1">
<mxGeometry x="1100" y="330" width="70" height="160" as="geometry"/> <mxGeometry x="1050" width="70" height="160" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="48" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;" edge="1" parent="1" source="44" target="41"> <mxCell id="48" 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="92" source="44" target="41" edge="1">
<mxGeometry relative="1" as="geometry"/> <mxGeometry relative="1" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="44" value=".&lt;div&gt;.&lt;/div&gt;&lt;div&gt;.&lt;/div&gt;" style="rounded=1;whiteSpace=wrap;html=1;" vertex="1" parent="1"> <mxCell id="44" value=".&lt;div&gt;.&lt;/div&gt;&lt;div&gt;.&lt;/div&gt;" style="rounded=1;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" vertex="1">
<mxGeometry x="940" y="370" width="30" height="80" as="geometry"/> <mxGeometry x="890" y="40" width="30" height="80" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="53" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;" edge="1" parent="1" source="50" target="51"> <mxCell id="53" 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="92" source="50" target="51" edge="1">
<mxGeometry relative="1" as="geometry"/> <mxGeometry relative="1" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="50" value="Linear" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;fontColor=#333333;strokeColor=#666666;" vertex="1" parent="1"> <mxCell id="50" 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="92" vertex="1">
<mxGeometry x="1200" y="335" width="50" height="150" as="geometry"/> <mxGeometry x="1150" y="5" width="50" height="150" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="51" value="Softmax" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="1"> <mxCell id="51" 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="92" vertex="1">
<mxGeometry x="1286" y="335" width="50" height="150" as="geometry"/> <mxGeometry x="1236" y="5" width="50" height="150" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="54" value="Tokens" style="rounded=1;whiteSpace=wrap;html=1;" vertex="1" parent="1"> <mxCell id="54" value="Tokens" style="rounded=1;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" vertex="1">
<mxGeometry x="50" y="370" width="60" height="90" as="geometry"/> <mxGeometry y="40" width="60" height="90" as="geometry"/>
</mxCell> </mxCell>
<mxCell id="55" 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;" edge="1" parent="1" source="54"> <mxCell id="55" 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="92" source="54" edge="1">
<mxGeometry relative="1" as="geometry"> <mxGeometry relative="1" as="geometry">
<mxPoint x="92.75" y="414.6694174757282" as="sourcePoint"/> <mxPoint x="42.75" y="84.66941747572821" as="sourcePoint"/>
<mxPoint x="140" y="415.33000000000004" as="targetPoint"/> <mxPoint x="90" y="85.33000000000004" as="targetPoint"/>
</mxGeometry> </mxGeometry>
</mxCell> </mxCell>
</root> </root>

413
assets/drawio/gpt1-attention.drawio Normal file
View File

@@ -0,0 +1,413 @@
<mxfile host="65bd71144e">
<diagram name="GPT Architecture" id="DEYydPS-O6mnllJWumln">
<mxGraphModel dx="2176" dy="702" 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="92" value="" style="group" parent="1" vertex="1" connectable="0">
<mxGeometry x="40" y="320" width="1286" height="160" as="geometry"/>
</mxCell>
<mxCell id="3" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#fff2cc;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;container=0;" parent="92" vertex="1">
<mxGeometry x="230" width="440" height="160" as="geometry"/>
</mxCell>
<mxCell id="36" 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="92" source="32" target="3" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="32" value="+" style="rounded=0;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="92" vertex="1">
<mxGeometry x="90" width="110" height="160" as="geometry"/>
</mxCell>
<mxCell id="33" 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="92" vertex="1">
<mxGeometry x="95" y="17.5" width="100" height="42.5" as="geometry"/>
</mxCell>
<mxCell id="34" 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="92" vertex="1">
<mxGeometry x="95" y="100" width="100" height="42.5" as="geometry"/>
</mxCell>
<mxCell id="46" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" source="37" target="40" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="37" 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="92" vertex="1">
<mxGeometry x="690" width="70" height="160" as="geometry"/>
</mxCell>
<mxCell id="38" 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="92" source="7" target="37" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="47" 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="92" source="40" target="44" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="40" 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="92" vertex="1">
<mxGeometry x="790" width="70" height="160" as="geometry"/>
</mxCell>
<mxCell id="49" 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="92" source="41" target="42" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="41" 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="92" vertex="1">
<mxGeometry x="950" width="70" height="160" as="geometry"/>
</mxCell>
<mxCell id="52" 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="92" source="42" target="50" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="42" 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="92" vertex="1">
<mxGeometry x="1050" width="70" height="160" as="geometry"/>
</mxCell>
<mxCell id="48" 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="92" source="44" target="41" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="44" value=".&lt;div&gt;.&lt;/div&gt;&lt;div&gt;.&lt;/div&gt;" style="rounded=1;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" vertex="1">
<mxGeometry x="890" y="40" width="30" height="80" as="geometry"/>
</mxCell>
<mxCell id="53" 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="92" source="50" target="51" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="50" 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="92" vertex="1">
<mxGeometry x="1150" y="5" width="50" height="150" as="geometry"/>
</mxCell>
<mxCell id="51" 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="92" vertex="1">
<mxGeometry x="1236" y="5" width="50" height="150" as="geometry"/>
</mxCell>
<mxCell id="54" value="Tokens" style="rounded=1;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" vertex="1">
<mxGeometry y="40" width="60" height="90" as="geometry"/>
</mxCell>
<mxCell id="55" 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="92" source="54" 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>
<mxCell id="4" value="&lt;div&gt;Masked&lt;/div&gt;Multi+Head&lt;br&gt;Attention" style="rounded=0;whiteSpace=wrap;html=1;strokeColor=#FF3333;fillColor=#dae8fc;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" vertex="1">
<mxGeometry x="281.42776556776556" y="50" width="78.97435897435898" height="60" as="geometry"/>
</mxCell>
<mxCell id="22" value="" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" source="5" edge="1">
<mxGeometry relative="1" as="geometry">
<mxPoint x="580" y="80" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="5" value="Feed&lt;div&gt;Forward&lt;/div&gt;&lt;div&gt;Network&lt;/div&gt;" 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="92" vertex="1">
<mxGeometry x="490.9564102564102" y="50" width="71.9230769230769" height="60" as="geometry"/>
</mxCell>
<mxCell id="7" value="Norm" style="rounded=0;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" vertex="1">
<mxGeometry x="609.9976190476191" y="60" width="37.87142857142857" height="40" as="geometry"/>
</mxCell>
<mxCell id="21" 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="92" source="12" target="5" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="12" value="Norm" style="rounded=0;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" vertex="1">
<mxGeometry x="407.1428571428571" y="60" width="41.904761904761905" height="40" as="geometry"/>
</mxCell>
<mxCell id="13" 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="92" source="3" target="4" edge="1">
<mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="250" y="80.00000000000011" as="sourcePoint"/>
<mxPoint x="459.5238095238095" y="-50" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="14" 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="92" source="18" target="12" edge="1">
<mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="385.71428571427464" y="79.99999999999989" as="sourcePoint"/>
<mxPoint x="459.5238095238095" y="-50" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="18" value="+" style="ellipse;whiteSpace=wrap;html=1;aspect=fixed;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" vertex="1">
<mxGeometry x="380.00428571428574" y="75" width="10" height="10" as="geometry"/>
</mxCell>
<mxCell id="19" 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="92" source="4" target="18" edge="1">
<mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="443.80952380952385" y="410" as="sourcePoint"/>
<mxPoint x="375.7142857142858" y="80.00000000000011" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="23" 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="92" target="24" edge="1">
<mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="466.8571428571429" y="80" as="sourcePoint"/>
<mxPoint x="579.7619047619048" y="85" as="targetPoint"/>
<Array as="points">
<mxPoint x="466.67904761904765" y="125"/>
<mxPoint x="522.3809523809523" y="125"/>
<mxPoint x="585" y="125"/>
</Array>
</mxGeometry>
</mxCell>
<mxCell id="28" value="" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" source="24" target="7" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="24" value="+" style="ellipse;whiteSpace=wrap;html=1;aspect=fixed;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" vertex="1">
<mxGeometry x="580.0019047619048" y="75" width="10" height="10" as="geometry"/>
</mxCell>
<mxCell id="25" 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="92" target="18" edge="1">
<mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="264.3255813953488" y="80" as="sourcePoint"/>
<mxPoint x="380.7142857142858" y="85" as="targetPoint"/>
<Array as="points">
<mxPoint x="264.2585825027686" y="130"/>
<mxPoint x="319.96048726467325" y="130"/>
<mxPoint x="385" y="130"/>
</Array>
</mxGeometry>
</mxCell>
<mxCell id="141" value="" style="endArrow=none;dashed=1;html=1;" edge="1" parent="92">
<mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="180" y="250" as="sourcePoint"/>
<mxPoint x="281" y="110" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="142" value="" style="endArrow=none;dashed=1;html=1;entryX=1;entryY=1;entryDx=0;entryDy=0;" edge="1" parent="1" target="4">
<mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="620" y="560" as="sourcePoint"/>
<mxPoint x="660" y="520" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="218" value="" style="group;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" connectable="0" parent="1">
<mxGeometry x="130" y="660" width="680" height="160" as="geometry"/>
</mxCell>
<mxCell id="195" style="edgeStyle=orthogonalEdgeStyle;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" edge="1" parent="218" source="143" target="147">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="196" style="edgeStyle=none;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" edge="1" parent="218" source="143" target="148">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="197" style="edgeStyle=orthogonalEdgeStyle;html=1;exitX=1;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" edge="1" parent="218" source="143" target="149">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="143" value="X" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#fff2cc;strokeColor=#d6b656;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="218">
<mxGeometry y="60" width="40" height="40" as="geometry"/>
</mxCell>
<mxCell id="147" value="W&lt;sub&gt;k&lt;/sub&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;fontColor=#333333;strokeColor=#666666;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="218">
<mxGeometry x="80" width="40" height="40" as="geometry"/>
</mxCell>
<mxCell id="199" style="edgeStyle=none;html=1;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" edge="1" parent="218" source="148" target="151">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="148" value="W&lt;sub&gt;q&lt;/sub&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;fontColor=#333333;strokeColor=#666666;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="218">
<mxGeometry x="80" y="60" width="40" height="40" as="geometry"/>
</mxCell>
<mxCell id="149" value="W&lt;sub&gt;v&lt;/sub&gt;" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f5f5f5;fontColor=#333333;strokeColor=#666666;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="218">
<mxGeometry x="80" y="120" width="40" height="40" as="geometry"/>
</mxCell>
<mxCell id="207" style="edgeStyle=orthogonalEdgeStyle;html=1;entryX=0;entryY=1;entryDx=0;entryDy=0;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" edge="1" parent="218" source="150" target="158">
<mxGeometry relative="1" as="geometry">
<Array as="points">
<mxPoint x="236" y="20"/>
<mxPoint x="236" y="50"/>
</Array>
</mxGeometry>
</mxCell>
<mxCell id="150" value="K" style="rounded=1;whiteSpace=wrap;html=1;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="218">
<mxGeometry x="158.97000000000003" width="41.03" height="40" as="geometry"/>
</mxCell>
<mxCell id="208" style="edgeStyle=orthogonalEdgeStyle;html=1;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" edge="1" parent="218" source="151" target="190">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="151" value="Q" style="rounded=1;whiteSpace=wrap;html=1;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="218">
<mxGeometry x="158.97000000000003" y="60" width="41.03" height="40" as="geometry"/>
</mxCell>
<mxCell id="214" style="edgeStyle=orthogonalEdgeStyle;html=1;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" edge="1" parent="218" source="152" target="187">
<mxGeometry relative="1" as="geometry">
<Array as="points">
<mxPoint x="600" y="140"/>
<mxPoint x="600" y="80"/>
</Array>
</mxGeometry>
</mxCell>
<mxCell id="152" value="V" style="rounded=1;whiteSpace=wrap;html=1;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="218">
<mxGeometry x="158.97000000000003" y="120" width="40" height="40" as="geometry"/>
</mxCell>
<mxCell id="215" style="edgeStyle=none;html=1;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" edge="1" parent="218" source="187">
<mxGeometry relative="1" as="geometry">
<mxPoint x="680" y="80" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="187" value="O" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#fff2cc;strokeColor=#d6b656;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="218">
<mxGeometry x="620" y="60" width="40" height="40" as="geometry"/>
</mxCell>
<mxCell id="211" style="edgeStyle=none;html=1;entryX=0;entryY=0;entryDx=0;entryDy=0;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" edge="1" parent="218" source="188" target="179">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="188" value="Scale" style="rounded=1;whiteSpace=wrap;html=1;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;direction=east;rotation=90;fillColor=#f8cecc;strokeColor=#b85450;" vertex="1" parent="218">
<mxGeometry x="370" y="37.5" width="50" height="25" as="geometry"/>
</mxCell>
<mxCell id="213" style="edgeStyle=orthogonalEdgeStyle;html=1;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" edge="1" parent="218" source="189" target="187">
<mxGeometry relative="1" as="geometry">
<Array as="points">
<mxPoint x="600" y="50"/>
<mxPoint x="600" y="80"/>
</Array>
</mxGeometry>
</mxCell>
<mxCell id="189" value="Softmax" style="rounded=1;whiteSpace=wrap;html=1;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;direction=east;rotation=90;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="218">
<mxGeometry x="530" y="37.5" width="50" height="25" as="geometry"/>
</mxCell>
<mxCell id="209" style="edgeStyle=none;html=1;entryX=0.5;entryY=1;entryDx=0;entryDy=0;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" edge="1" parent="218" source="190" target="188">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="190" value="" style="group;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" connectable="0" parent="218">
<mxGeometry x="272.5" y="10" width="80" height="80" as="geometry"/>
</mxCell>
<mxCell id="153" value="" style="whiteSpace=wrap;html=1;aspect=fixed;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry width="80" height="80" as="geometry"/>
</mxCell>
<mxCell id="154" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="155" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry x="20" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="156" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry x="40" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="157" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry x="60" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="158" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry y="20" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="159" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry x="20" y="20" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="160" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry x="40" y="20" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="161" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry x="60" y="20" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="162" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry y="40" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="163" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry x="20" y="40" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="164" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry x="40" y="40" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="165" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry x="60" y="40" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="166" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry y="60" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="167" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry x="20" y="60" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="168" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry x="40" y="60" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="169" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="190">
<mxGeometry x="60" y="60" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="191" value="" style="group;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" connectable="0" parent="218">
<mxGeometry x="440" y="10" width="80" height="80" as="geometry"/>
</mxCell>
<mxCell id="170" value="" style="whiteSpace=wrap;html=1;aspect=fixed;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry width="80" height="80" as="geometry"/>
</mxCell>
<mxCell id="171" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="172" value="" style="rounded=0;whiteSpace=wrap;html=1;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry x="20" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="173" value="" style="rounded=0;whiteSpace=wrap;html=1;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry x="40" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="174" value="" style="rounded=0;whiteSpace=wrap;html=1;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry x="60" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="175" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry y="20" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="176" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry x="20" y="20" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="177" value="" style="rounded=0;whiteSpace=wrap;html=1;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry x="40" y="20" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="178" value="" style="rounded=0;whiteSpace=wrap;html=1;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry x="60" y="20" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="179" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry y="40" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="180" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry x="20" y="40" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="181" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry x="40" y="40" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="182" value="" style="rounded=0;whiteSpace=wrap;html=1;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry x="60" y="40" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="183" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry y="60" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="184" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry x="20" y="60" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="185" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry x="40" y="60" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="186" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" parent="191">
<mxGeometry x="60" y="60" width="20" height="20" as="geometry"/>
</mxCell>
<mxCell id="198" style="edgeStyle=none;html=1;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" edge="1" parent="218" source="147" target="150">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="200" style="edgeStyle=none;html=1;entryX=0;entryY=0.5;entryDx=0;entryDy=0;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" edge="1" parent="218" target="152">
<mxGeometry relative="1" as="geometry">
<mxPoint x="120" y="140" as="sourcePoint"/>
<mxPoint x="148.97000000000008" y="139.8599999999998" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="212" value="" style="endArrow=classic;html=1;exitX=1;exitY=0;exitDx=0;exitDy=0;entryX=0.5;entryY=1;entryDx=0;entryDy=0;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" edge="1" parent="218" source="182" target="189">
<mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="610" y="50" as="sourcePoint"/>
<mxPoint x="660" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="219" value="" style="group;movable=0;resizable=0;rotatable=0;deletable=0;editable=0;locked=1;connectable=0;" vertex="1" connectable="0" parent="1">
<mxGeometry x="289.99776556776555" y="520" width="250.00223443223445" height="90" as="geometry"/>
</mxCell>
<mxCell id="145" style="edgeStyle=none;html=1;exitX=0.5;exitY=0;exitDx=0;exitDy=0;entryX=0.5;entryY=1;entryDx=0;entryDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" edge="1" parent="219" source="133" target="144">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="133" value="Concat" style="rounded=1;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;direction=east;rotation=90;" vertex="1" parent="219">
<mxGeometry x="132.50223443223445" y="32.5" width="50" height="25" as="geometry"/>
</mxCell>
<mxCell id="136" value="" style="endArrow=classic;html=1;entryX=0.5;entryY=1;entryDx=0;entryDy=0;exitX=1;exitY=0.5;exitDx=0;exitDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" edge="1" parent="219" target="133">
<mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="108.97435897435912" y="45" as="sourcePoint"/>
<mxPoint x="250.00223443223445" y="25" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="129" value="&lt;div&gt;Masked&lt;/div&gt;Multi+Head&lt;br&gt;Attention" style="rounded=0;whiteSpace=wrap;html=1;strokeColor=#FF3333;fillColor=#dae8fc;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" vertex="1" parent="219">
<mxGeometry x="30" width="78.97435897435898" height="60" as="geometry"/>
</mxCell>
<mxCell id="130" value="&lt;div&gt;Masked&lt;/div&gt;Multi+Head&lt;br&gt;Attention" style="rounded=0;whiteSpace=wrap;html=1;strokeColor=#FF3333;fillColor=#dae8fc;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" vertex="1" parent="219">
<mxGeometry x="20" y="10" width="78.97435897435898" height="60" as="geometry"/>
</mxCell>
<mxCell id="131" value="&lt;div&gt;Masked&lt;/div&gt;Multi+Head&lt;br&gt;Attention" style="rounded=0;whiteSpace=wrap;html=1;strokeColor=#FF3333;fillColor=#dae8fc;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" vertex="1" parent="219">
<mxGeometry x="10" y="20" width="78.97435897435898" height="60" as="geometry"/>
</mxCell>
<mxCell id="132" value="&lt;div&gt;Masked&lt;/div&gt;Multi+Head&lt;br&gt;Attention" style="rounded=0;whiteSpace=wrap;html=1;strokeColor=#FF3333;fillColor=#dae8fc;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" vertex="1" parent="219">
<mxGeometry y="30" width="78.97435897435898" height="60" as="geometry"/>
</mxCell>
<mxCell id="146" style="edgeStyle=none;html=1;exitX=0.5;exitY=0;exitDx=0;exitDy=0;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" edge="1" parent="219" source="144">
<mxGeometry relative="1" as="geometry">
<mxPoint x="250.00223443223445" y="44.969696969697" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="144" value="Linear" style="rounded=1;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;direction=east;rotation=90;" vertex="1" parent="219">
<mxGeometry x="182.50223443223445" y="32.5" width="50" height="25" as="geometry"/>
</mxCell>
<mxCell id="220" value="" style="endArrow=none;dashed=1;html=1;" edge="1" parent="1">
<mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="90" y="690" as="sourcePoint"/>
<mxPoint x="290" y="610" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="221" value="" style="endArrow=none;dashed=1;html=1;" edge="1" parent="1">
<mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="370" y="610" as="sourcePoint"/>
<mxPoint x="830" y="700" as="targetPoint"/>
</mxGeometry>
</mxCell>
</root>
</mxGraphModel>
</diagram>
</mxfile>

148
assets/drawio/gpt1-decoder.drawio Normal file
View File

@@ -0,0 +1,148 @@
<mxfile host="65bd71144e">
<diagram name="GPT Architecture" id="DEYydPS-O6mnllJWumln">
<mxGraphModel dx="979" dy="301" 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="92" value="" style="group" parent="1" vertex="1" connectable="0">
<mxGeometry x="40" y="320" width="1286" height="160" as="geometry"/>
</mxCell>
<mxCell id="3" value="" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#fff2cc;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;strokeColor=#FF3333;" parent="92" vertex="1">
<mxGeometry x="230" width="440" height="160" as="geometry"/>
</mxCell>
<mxCell id="4" value="&lt;div&gt;Masked&lt;/div&gt;Multi+Head&lt;br&gt;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="3" vertex="1">
<mxGeometry x="51.42776556776556" y="50" width="78.97435897435898" height="60" as="geometry"/>
</mxCell>
<mxCell id="22" value="" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" source="5" edge="1">
<mxGeometry relative="1" as="geometry">
<mxPoint x="350" y="80" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="5" value="Feed&lt;div&gt;Forward&lt;/div&gt;&lt;div&gt;Network&lt;/div&gt;" 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="3" vertex="1">
<mxGeometry x="260.9564102564102" y="50" width="71.9230769230769" height="60" as="geometry"/>
</mxCell>
<mxCell id="7" value="Norm" style="rounded=0;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" vertex="1">
<mxGeometry x="379.997619047619" y="60" width="37.87142857142857" height="40" as="geometry"/>
</mxCell>
<mxCell id="21" 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="3" source="12" target="5" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="12" value="Norm" style="rounded=0;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" vertex="1">
<mxGeometry x="177.14285714285714" y="60" width="41.904761904761905" height="40" as="geometry"/>
</mxCell>
<mxCell id="13" 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="3" source="3" target="4" 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="14" 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="3" source="18" target="12" 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="18" value="+" style="ellipse;whiteSpace=wrap;html=1;aspect=fixed;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" vertex="1">
<mxGeometry x="150.00428571428571" y="75" width="10" height="10" as="geometry"/>
</mxCell>
<mxCell id="19" 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="3" source="4" target="18" 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="23" 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="3" target="24" 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="28" value="" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" source="24" target="7" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="24" value="+" style="ellipse;whiteSpace=wrap;html=1;aspect=fixed;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" vertex="1">
<mxGeometry x="350.00190476190477" y="75" width="10" height="10" as="geometry"/>
</mxCell>
<mxCell id="25" 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="3" target="18" 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="36" 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="92" source="32" target="3" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="32" value="+" style="rounded=0;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="92" vertex="1">
<mxGeometry x="90" width="110" height="160" as="geometry"/>
</mxCell>
<mxCell id="33" 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="92" vertex="1">
<mxGeometry x="95" y="17.5" width="100" height="42.5" as="geometry"/>
</mxCell>
<mxCell id="34" 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="92" vertex="1">
<mxGeometry x="95" y="100" width="100" height="42.5" as="geometry"/>
</mxCell>
<mxCell id="46" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" source="37" target="40" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="37" 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="92" vertex="1">
<mxGeometry x="690" width="70" height="160" as="geometry"/>
</mxCell>
<mxCell id="38" 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="92" source="7" target="37" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="47" 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="92" source="40" target="44" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="40" 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="92" vertex="1">
<mxGeometry x="790" width="70" height="160" as="geometry"/>
</mxCell>
<mxCell id="49" 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="92" source="41" target="42" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="41" 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="92" vertex="1">
<mxGeometry x="950" width="70" height="160" as="geometry"/>
</mxCell>
<mxCell id="52" 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="92" source="42" target="50" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="42" 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="92" vertex="1">
<mxGeometry x="1050" width="70" height="160" as="geometry"/>
</mxCell>
<mxCell id="48" 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="92" source="44" target="41" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="44" value=".&lt;div&gt;.&lt;/div&gt;&lt;div&gt;.&lt;/div&gt;" style="rounded=1;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" vertex="1">
<mxGeometry x="890" y="40" width="30" height="80" as="geometry"/>
</mxCell>
<mxCell id="53" 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="92" source="50" target="51" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="50" 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="92" vertex="1">
<mxGeometry x="1150" y="5" width="50" height="150" as="geometry"/>
</mxCell>
<mxCell id="51" 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="92" vertex="1">
<mxGeometry x="1236" y="5" width="50" height="150" as="geometry"/>
</mxCell>
<mxCell id="54" value="Tokens" style="rounded=1;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" vertex="1">
<mxGeometry y="40" width="60" height="90" as="geometry"/>
</mxCell>
<mxCell id="55" 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="92" source="54" 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>

148
assets/drawio/gpt1-embeddings.drawio Normal file
View File

@@ -0,0 +1,148 @@
<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="&lt;div&gt;Masked&lt;/div&gt;Multi+Head&lt;br&gt;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&lt;div&gt;Forward&lt;/div&gt;&lt;div&gt;Network&lt;/div&gt;" 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=".&lt;div&gt;.&lt;/div&gt;&lt;div&gt;.&lt;/div&gt;" 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>

192
assets/drawio/gpt1-forward.drawio Normal file
View File

@@ -0,0 +1,192 @@
<mxfile host="65bd71144e">
<diagram name="GPT Architecture" id="DEYydPS-O6mnllJWumln">
<mxGraphModel dx="2176" dy="1029" 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="107" value="" style="group" vertex="1" connectable="0" parent="1">
<mxGeometry x="120" y="170" width="1286" height="265" as="geometry"/>
</mxCell>
<mxCell id="92" value="" style="group" parent="107" vertex="1" connectable="0">
<mxGeometry width="1286" height="160" as="geometry"/>
</mxCell>
<mxCell id="3" 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="92" vertex="1">
<mxGeometry x="230" width="440" height="160" as="geometry"/>
</mxCell>
<mxCell id="4" value="&lt;div&gt;Masked&lt;/div&gt;Multi+Head&lt;br&gt;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="3" vertex="1">
<mxGeometry x="51.42776556776556" y="50" width="78.97435897435898" height="60" as="geometry"/>
</mxCell>
<mxCell id="22" value="" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" source="5" edge="1">
<mxGeometry relative="1" as="geometry">
<mxPoint x="350" y="80" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="5" value="Feed&lt;div&gt;Forward&lt;/div&gt;&lt;div&gt;Network&lt;/div&gt;" style="rounded=0;whiteSpace=wrap;html=1;strokeColor=#FF3333;fillColor=#e1d5e7;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" vertex="1">
<mxGeometry x="260.9564102564102" y="50" width="71.9230769230769" height="60" as="geometry"/>
</mxCell>
<mxCell id="7" value="Norm" style="rounded=0;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" vertex="1">
<mxGeometry x="379.997619047619" y="60" width="37.87142857142857" height="40" as="geometry"/>
</mxCell>
<mxCell id="21" 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="3" source="12" target="5" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="12" value="Norm" style="rounded=0;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" vertex="1">
<mxGeometry x="177.14285714285714" y="60" width="41.904761904761905" height="40" as="geometry"/>
</mxCell>
<mxCell id="13" 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="3" source="3" target="4" 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="14" 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="3" source="18" target="12" 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="18" value="+" style="ellipse;whiteSpace=wrap;html=1;aspect=fixed;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" vertex="1">
<mxGeometry x="150.00428571428571" y="75" width="10" height="10" as="geometry"/>
</mxCell>
<mxCell id="19" 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="3" source="4" target="18" 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="23" 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="3" target="24" 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="28" value="" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" source="24" target="7" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="24" value="+" style="ellipse;whiteSpace=wrap;html=1;aspect=fixed;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="3" vertex="1">
<mxGeometry x="350.00190476190477" y="75" width="10" height="10" as="geometry"/>
</mxCell>
<mxCell id="25" 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="3" target="18" 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="104" value="" style="endArrow=none;dashed=1;html=1;" edge="1" parent="3">
<mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="200" y="200" as="sourcePoint"/>
<mxPoint x="260.96000000000004" y="110" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="36" 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="92" source="32" target="3" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="32" value="+" style="rounded=0;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="92" vertex="1">
<mxGeometry x="90" width="110" height="160" as="geometry"/>
</mxCell>
<mxCell id="33" 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="92" vertex="1">
<mxGeometry x="95" y="17.5" width="100" height="42.5" as="geometry"/>
</mxCell>
<mxCell id="34" 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="92" vertex="1">
<mxGeometry x="95" y="100" width="100" height="42.5" as="geometry"/>
</mxCell>
<mxCell id="46" style="edgeStyle=none;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" source="37" target="40" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="37" 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="92" vertex="1">
<mxGeometry x="690" width="70" height="160" as="geometry"/>
</mxCell>
<mxCell id="38" 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="92" source="7" target="37" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="47" 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="92" source="40" target="44" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="40" 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="92" vertex="1">
<mxGeometry x="790" width="70" height="160" as="geometry"/>
</mxCell>
<mxCell id="49" 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="92" source="41" target="42" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="41" 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="92" vertex="1">
<mxGeometry x="950" width="70" height="160" as="geometry"/>
</mxCell>
<mxCell id="52" 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="92" source="42" target="50" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="42" 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="92" vertex="1">
<mxGeometry x="1050" width="70" height="160" as="geometry"/>
</mxCell>
<mxCell id="48" 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="92" source="44" target="41" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="44" value=".&lt;div&gt;.&lt;/div&gt;&lt;div&gt;.&lt;/div&gt;" style="rounded=1;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" vertex="1">
<mxGeometry x="890" y="40" width="30" height="80" as="geometry"/>
</mxCell>
<mxCell id="53" 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="92" source="50" target="51" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="50" 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="92" vertex="1">
<mxGeometry x="1150" y="5" width="50" height="150" as="geometry"/>
</mxCell>
<mxCell id="51" 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="92" vertex="1">
<mxGeometry x="1236" y="5" width="50" height="150" as="geometry"/>
</mxCell>
<mxCell id="54" value="Tokens" style="rounded=1;whiteSpace=wrap;html=1;movable=1;resizable=1;rotatable=1;deletable=1;editable=1;locked=0;connectable=1;" parent="92" vertex="1">
<mxGeometry y="40" width="60" height="90" as="geometry"/>
</mxCell>
<mxCell id="55" 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="92" source="54" 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>
<mxCell id="105" value="" style="endArrow=none;dashed=1;html=1;exitX=1;exitY=1;exitDx=0;exitDy=0;" edge="1" parent="107" source="5">
<mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint x="660" y="140" as="sourcePoint"/>
<mxPoint x="620" y="190" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="106" value="" style="group" vertex="1" connectable="0" parent="107">
<mxGeometry x="450" y="195" width="170" height="70" as="geometry"/>
</mxCell>
<mxCell id="100" value="" style="edgeStyle=none;html=1;" edge="1" parent="106" source="93" target="99">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="93" value="Linear" style="rounded=1;whiteSpace=wrap;html=1;rotation=90;container=0;fillColor=#f5f5f5;fontColor=#333333;strokeColor=#666666;" vertex="1" parent="106">
<mxGeometry x="-5" y="20" width="70" height="30" as="geometry"/>
</mxCell>
<mxCell id="96" value="" style="endArrow=classic;html=1;entryX=0.5;entryY=1;entryDx=0;entryDy=0;" edge="1" parent="106" target="93">
<mxGeometry width="50" height="50" relative="1" as="geometry">
<mxPoint y="35" as="sourcePoint"/>
<mxPoint y="35.00999999999999" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="102" style="edgeStyle=none;html=1;exitX=0.5;exitY=0;exitDx=0;exitDy=0;" edge="1" parent="106" source="98">
<mxGeometry relative="1" as="geometry">
<mxPoint x="170" y="35.09433962264154" as="targetPoint"/>
</mxGeometry>
</mxCell>
<mxCell id="98" value="Linear" style="rounded=1;whiteSpace=wrap;html=1;rotation=90;container=0;fillColor=#f5f5f5;fontColor=#333333;strokeColor=#666666;" vertex="1" parent="106">
<mxGeometry x="100" y="20" width="70" height="30" as="geometry"/>
</mxCell>
<mxCell id="101" value="" style="edgeStyle=none;html=1;" edge="1" parent="106" source="99" target="98">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="99" value="ReLU" style="rounded=1;whiteSpace=wrap;html=1;rotation=90;container=0;fillColor=#e1d5e7;strokeColor=#9673a6;" vertex="1" parent="106">
<mxGeometry x="50" y="20" width="70" height="30" as="geometry"/>
</mxCell>
</root>
</mxGraphModel>
</diagram>
</mxfile>

73
assets/drawio/gpt11.drawio
View File

@@ -1,73 +0,0 @@
<mxfile host="65bd71144e">
<diagram name="GPT-1 Architecture" id="0">
<mxGraphModel dx="569" dy="414" 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="2" value="Токены (индексы)" style="rounded=0;whiteSpace=wrap;html=1;" parent="1" vertex="1">
<mxGeometry x="40" y="140" width="110" height="40" as="geometry"/>
</mxCell>
<mxCell id="3" value="Token Embeddings" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;" parent="1" vertex="1">
<mxGeometry x="170" y="100" width="110" height="40" as="geometry"/>
</mxCell>
<mxCell id="4" value="Position Embeddings" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#dae8fc;" parent="1" vertex="1">
<mxGeometry x="170" y="180" width="110" height="40" as="geometry"/>
</mxCell>
<mxCell id="5" value="Сумма + Dropout" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#b9e97a;" parent="1" vertex="1">
<mxGeometry x="310" y="140" width="120" height="40" as="geometry"/>
</mxCell>
<mxCell id="6" style="edgeStyle=orthogonalEdgeStyle;endArrow=block;html=1;" parent="1" source="2" target="3" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="7" style="edgeStyle=orthogonalEdgeStyle;endArrow=none;html=1;dashed=1;" parent="1" source="2" target="4" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="8" style="endArrow=block;html=1;" parent="1" source="3" target="5" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="9" style="endArrow=block;html=1;" parent="1" source="4" target="5" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="10" value="N× Decoder блок" style="rounded=1;whiteSpace=wrap;html=1;strokeColor=#cfbc96;fillColor=#fbe7b0;" parent="1" vertex="1">
<mxGeometry x="460" y="80" width="280" height="160" as="geometry"/>
</mxCell>
<mxCell id="16" value="Multi-Head Attention (h × HeadAttention, Concat, Linear, Dropout)" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#f8cecc;" parent="10" vertex="1">
<mxGeometry x="10" y="10" width="120" height="45" as="geometry"/>
</mxCell>
<mxCell id="17" value="Residual + LayerNorm 1" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;" parent="10" vertex="1">
<mxGeometry x="150" y="20" width="110" height="25" as="geometry"/>
</mxCell>
<mxCell id="18" value="FeedForward (Linear-ReLU-Linear-Dropout)" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#dae8fc;" parent="10" vertex="1">
<mxGeometry x="10" y="95" width="120" height="45" as="geometry"/>
</mxCell>
<mxCell id="19" value="Residual + LayerNorm 2" style="rounded=1;whiteSpace=wrap;html=1;fillColor=#d5e8d4;" parent="10" vertex="1">
<mxGeometry x="150" y="105" width="110" height="25" as="geometry"/>
</mxCell>
<mxCell id="20" style="endArrow=block;html=1;" parent="10" source="16" target="17" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="21" style="endArrow=block;html=1;" parent="10" source="17" target="18" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="22" style="endArrow=block;html=1;" parent="10" source="18" target="19" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="11" style="endArrow=block;html=1;" parent="1" source="5" target="10" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="12" value="Linear (emb→vocab)" style="rounded=0;whiteSpace=wrap;html=1;fillColor=#e1d5e7;" parent="1" vertex="1">
<mxGeometry x="760" y="140" width="120" height="40" as="geometry"/>
</mxCell>
<mxCell id="13" style="endArrow=block;html=1;" parent="1" source="10" target="12" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<mxCell id="14" value="Логиты [batch, seq, vocab]" style="rounded=0;whiteSpace=wrap;html=1;" parent="1" vertex="1">
<mxGeometry x="900" y="140" width="120" height="40" as="geometry"/>
</mxCell>
<mxCell id="15" style="endArrow=block;html=1;" parent="1" source="12" target="14" edge="1">
<mxGeometry relative="1" as="geometry"/>
</mxCell>
</root>
</mxGraphModel>
</diagram>
</mxfile>

0
assets/models/gpt1.png → assets/models/gpt1-architecture.png
View File

Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 46 KiB

After

Width:  |  Height:  |  Size: 46 KiB

BIN
assets/models/gpt1-attention.png Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 124 KiB

BIN
assets/models/gpt1-decoder.png Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 50 KiB

BIN
assets/models/gpt1-embeddings.png Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 50 KiB

BIN
assets/models/gpt1-forward.png Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 62 KiB

1
assets/models/gpt1.svg
View File

File diff suppressed because one or more lines are too long
Side by Side

Before

Width:  |  Height:  |  Size: 30 KiB

61
notebooks/gpt.ipynb
View File

File diff suppressed because one or more lines are too long