aeoxxian's picture
Upload batch 1/20 (100 files)
39c4d07 verified
Raw
History Blame Contribute Delete
60.1 kB
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>GPU Power Experiment Results</title>
<style>
* { margin: 0; padding: 0; box-sizing: border-box; }
body {
font-family: 'Segoe UI', -apple-system, BlinkMacSystemFont, sans-serif;
background: #0f172a; color: #e2e8f0;
line-height: 1.6; padding: 2rem;
}
.container { max-width: 1200px; margin: 0 auto; }
h1 {
font-size: 2rem; font-weight: 700; margin-bottom: 0.5rem;
background: linear-gradient(135deg, #38bdf8, #818cf8);
-webkit-background-clip: text; -webkit-text-fill-color: transparent;
}
.subtitle { color: #94a3b8; font-size: 0.95rem; margin-bottom: 2rem; }
.meta-bar {
display: flex; gap: 2rem; background: #1e293b; border-radius: 12px;
padding: 1rem 1.5rem; margin-bottom: 2rem; flex-wrap: wrap;
}
.meta-item { font-size: 0.85rem; color: #94a3b8; }
.meta-item strong { color: #e2e8f0; }
.exp-card {
background: #1e293b; border-radius: 12px; border: 1px solid #334155;
padding: 1.5rem; margin-bottom: 1.5rem;
transition: border-color 0.2s;
}
.exp-card:hover { border-color: #38bdf8; }
.exp-header { display: flex; align-items: baseline; gap: 1rem; margin-bottom: 0.5rem; }
.exp-number {
font-size: 0.75rem; font-weight: 700; color: #38bdf8;
background: rgba(56,189,248,0.1); padding: 2px 8px; border-radius: 6px;
}
.exp-title { font-size: 1.2rem; font-weight: 600; color: #f1f5f9; }
.exp-desc { color: #94a3b8; font-size: 0.9rem; margin-bottom: 1rem; }
.exp-variable {
font-size: 0.8rem; color: #a78bfa; margin-bottom: 1rem;
display: inline-block; background: rgba(167,139,250,0.1);
padding: 2px 10px; border-radius: 6px;
}
.run-table {
width: 100%; border-collapse: collapse; font-size: 0.85rem;
}
.run-table th {
text-align: left; padding: 0.5rem 0.75rem; color: #94a3b8;
border-bottom: 1px solid #334155; font-weight: 600; font-size: 0.75rem;
text-transform: uppercase; letter-spacing: 0.5px;
}
.run-table td {
padding: 0.5rem 0.75rem; border-bottom: 1px solid #1e293b;
}
.run-table tr:hover td { background: rgba(56,189,248,0.05); }
.run-table .num { text-align: right; font-variant-numeric: tabular-nums; }
a { color: #38bdf8; text-decoration: none; }
a:hover { text-decoration: underline; }
.btn {
display: inline-block; padding: 4px 12px; border-radius: 6px;
font-size: 0.8rem; font-weight: 500; margin-right: 0.5rem;
transition: background 0.2s;
}
.btn-plot { background: rgba(56,189,248,0.15); color: #38bdf8; }
.btn-plot:hover { background: rgba(56,189,248,0.25); text-decoration: none; }
.btn-png { background: rgba(167,139,250,0.15); color: #a78bfa; }
.btn-png:hover { background: rgba(167,139,250,0.25); text-decoration: none; }
.master-section { margin-top: 3rem; }
.master-section h2 {
font-size: 1.4rem; font-weight: 600; margin-bottom: 1rem; color: #f1f5f9;
}
.master-table {
width: 100%; border-collapse: collapse; font-size: 0.8rem;
background: #1e293b; border-radius: 12px; overflow: hidden;
}
.master-table th {
padding: 0.6rem 0.75rem; text-align: left; color: #94a3b8;
border-bottom: 2px solid #334155; font-weight: 600; font-size: 0.7rem;
text-transform: uppercase; letter-spacing: 0.5px;
position: sticky; top: 0; background: #1e293b;
}
.master-table td {
padding: 0.5rem 0.75rem; border-bottom: 1px solid #0f172a;
}
.master-table tr:hover td { background: rgba(56,189,248,0.05); }
.master-table .num { text-align: right; font-variant-numeric: tabular-nums; }
.footer {
margin-top: 3rem; padding-top: 1rem; border-top: 1px solid #334155;
color: #64748b; font-size: 0.8rem; text-align: center;
}
.glossary {
background: #1e293b; border-radius: 12px; border: 1px solid #334155;
padding: 1.5rem; margin-bottom: 2rem;
}
.glossary h2 { font-size: 1.1rem; margin-bottom: 0.75rem; color: #f1f5f9; }
.glossary dl { display: grid; grid-template-columns: 180px 1fr; gap: 0.3rem 1rem; }
.glossary dt { font-weight: 600; color: #38bdf8; font-size: 0.85rem; }
.glossary dd { color: #94a3b8; font-size: 0.85rem; }
.empty-msg { color: #64748b; font-style: italic; padding: 1rem 0; }
</style>
</head>
<body>
<div class="container">
<h1>GPU Power Experiment Results</h1>
<p class="subtitle">AI Datacenter Power Profile Analysis β€” Transfer Package</p>
<div class="meta-bar">
<div class="meta-item"><strong>12</strong> Experiment Groups</div>
<div class="meta-item"><strong>31</strong> Total Runs</div>
<div class="meta-item">Hardware: <strong>NVIDIA RTX A6000 / PRO 6000</strong></div>
<div class="meta-item">Sampling: <strong>10 ms (NVML)</strong></div>
<div class="meta-item">Generated: <strong>2026-02-19 13:10</strong></div>
</div>
<div class="glossary">
<h2>Quick Reference β€” Column Definitions</h2>
<dl>
<dt>Idle Avg (W)</dt>
<dd>Average power during pre-experiment idle phase (GPU loaded but no computation).</dd>
<dt>Train Avg / Peak (W)</dt>
<dd>Average and peak power during model training compute (forward + backward pass).</dd>
<dt>Val Avg (W)</dt>
<dd>Average power during validation (forward pass only, no gradients).</dd>
<dt>Infer Idle (W)</dt>
<dd>Average power between inference requests ("hot idle" β€” model stays loaded).</dd>
<dt>Energy (J)</dt>
<dd>Total energy consumed across all phases of the run.</dd>
</dl>
</div>
<div class="exp-card">
<div class="exp-header">
<span class="exp-number">#00</span>
<span class="exp-title">Baseline Reference</span>
</div>
<p class="exp-desc">Standard conditions with no power throttling. Control experiment for comparison.</p>
<span class="exp-variable">Variable: None (control)</span>
<table class="run-table">
<thead><tr>
<th>Model</th><th>Batch</th><th>Pattern</th>
<th>Idle Avg (W)</th><th>Train Avg (W)</th><th>Train Peak (W)</th>
<th>Val Avg (W)</th><th>Infer Idle (W)</th><th>Energy (J)</th>
<th>Plots</th>
</tr></thead>
<tbody>
<tr>
<td>resnet50</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">52.7</td>
<td class="num">134.3</td>
<td class="num">178.0</td>
<td class="num">89.9</td>
<td class="num">78.1</td>
<td class="num">29670</td>
<td><a href="00_baseline_reference/fixed_resnet50_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="00_baseline_reference/fixed_resnet50_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
</tbody>
</table>
</div>
<div class="exp-card">
<div class="exp-header">
<span class="exp-number">#00</span>
<span class="exp-title">Smoke Test</span>
</div>
<p class="exp-desc">Quick end-to-end validation of the measurement pipeline.</p>
<span class="exp-variable">Variable: Pipeline validation</span>
<table class="run-table">
<thead><tr>
<th>Model</th><th>Batch</th><th>Pattern</th>
<th>Idle Avg (W)</th><th>Train Avg (W)</th><th>Train Peak (W)</th>
<th>Val Avg (W)</th><th>Infer Idle (W)</th><th>Energy (J)</th>
<th>Plots</th>
</tr></thead>
<tbody>
<tr>
<td>resnet18</td>
<td class="num">8</td>
<td>fixed</td>
<td class="num">52.2</td>
<td class="num">125.7</td>
<td class="num">137.2</td>
<td class="num">123.6</td>
<td class="num">23.7</td>
<td class="num">9149</td>
<td><a href="00_smoke_test/fixed_resnet18_cifar10_gpu0_bs8_ep1_req2_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="00_smoke_test/fixed_resnet18_cifar10_gpu0_bs8_ep1_req2_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
</tbody>
</table>
</div>
<div class="exp-card">
<div class="exp-header">
<span class="exp-number">#01</span>
<span class="exp-title">Batch Size Sweep</span>
</div>
<p class="exp-desc">How does batch size affect power draw? Larger batches β†’ higher GPU utilization β†’ different power profiles.</p>
<span class="exp-variable">Variable: Batch size (16 / 64 / 128)</span>
<table class="run-table">
<thead><tr>
<th>Model</th><th>Batch</th><th>Pattern</th>
<th>Idle Avg (W)</th><th>Train Avg (W)</th><th>Train Peak (W)</th>
<th>Val Avg (W)</th><th>Infer Idle (W)</th><th>Energy (J)</th>
<th>Plots</th>
</tr></thead>
<tbody>
<tr>
<td>resnet50</td>
<td class="num">128</td>
<td>fixed</td>
<td class="num">51.8</td>
<td class="num">138.1</td>
<td class="num">211.8</td>
<td class="num">98.1</td>
<td class="num">78.2</td>
<td class="num">29205</td>
<td><a href="01_batch_sweep_resnet50/fixed_resnet50_cifar10_gpu0_bs128_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="01_batch_sweep_resnet50/fixed_resnet50_cifar10_gpu0_bs128_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>resnet50</td>
<td class="num">16</td>
<td>fixed</td>
<td class="num">53.3</td>
<td class="num">139.9</td>
<td class="num">152.2</td>
<td class="num">132.5</td>
<td class="num">77.9</td>
<td class="num">32440</td>
<td><a href="01_batch_sweep_resnet50/fixed_resnet50_cifar10_gpu0_bs16_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="01_batch_sweep_resnet50/fixed_resnet50_cifar10_gpu0_bs16_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>resnet50</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">52.5</td>
<td class="num">133.4</td>
<td class="num">173.2</td>
<td class="num">83.7</td>
<td class="num">78.3</td>
<td class="num">29732</td>
<td><a href="01_batch_sweep_resnet50/fixed_resnet50_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="01_batch_sweep_resnet50/fixed_resnet50_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
</tbody>
</table>
</div>
<div class="exp-card">
<div class="exp-header">
<span class="exp-number">#02</span>
<span class="exp-title">Operator Control (Variable Inference)</span>
</div>
<p class="exp-desc">Non-uniform inference request patterns, simulating real operator traffic.</p>
<span class="exp-variable">Variable: Variable inference interval</span>
<table class="run-table">
<thead><tr>
<th>Model</th><th>Batch</th><th>Pattern</th>
<th>Idle Avg (W)</th><th>Train Avg (W)</th><th>Train Peak (W)</th>
<th>Val Avg (W)</th><th>Infer Idle (W)</th><th>Energy (J)</th>
<th>Plots</th>
</tr></thead>
<tbody>
<tr>
<td>resnet18</td>
<td class="num">64</td>
<td>variable</td>
<td class="num">53.7</td>
<td class="num">128.5</td>
<td class="num">152.2</td>
<td class="num">133.8</td>
<td class="num">78.2</td>
<td class="num">34124</td>
<td><a href="02_operator_control/fixed_resnet18_cifar10_gpu0_bs64_ep5_req60_nocap_variable/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="02_operator_control/fixed_resnet18_cifar10_gpu0_bs64_ep5_req60_nocap_variable/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
</tbody>
</table>
</div>
<div class="exp-card">
<div class="exp-header">
<span class="exp-number">#03</span>
<span class="exp-title">Power Cap Sweep</span>
</div>
<p class="exp-desc">Effect of nvidia-smi power limits on performance and power. Tests whether power capping is effective.</p>
<span class="exp-variable">Variable: Power cap (345W / 460W / 575W)</span>
<table class="run-table">
<thead><tr>
<th>Model</th><th>Batch</th><th>Pattern</th>
<th>Idle Avg (W)</th><th>Train Avg (W)</th><th>Train Peak (W)</th>
<th>Val Avg (W)</th><th>Infer Idle (W)</th><th>Energy (J)</th>
<th>Plots</th>
</tr></thead>
<tbody>
<tr>
<td>resnet18</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">9.9</td>
<td class="num">111.6</td>
<td class="num">148.0</td>
<td class="num">148.6</td>
<td class="num">77.9</td>
<td class="num">27225</td>
<td><a href="03_control_cap_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_cap345W_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="03_control_cap_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_cap345W_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>resnet18</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">9.4</td>
<td class="num">118.1</td>
<td class="num">147.3</td>
<td class="num">145.8</td>
<td class="num">78.1</td>
<td class="num">27206</td>
<td><a href="03_control_cap_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_cap460W_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="03_control_cap_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_cap460W_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>resnet18</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">52.1</td>
<td class="num">102.0</td>
<td class="num">150.0</td>
<td class="num">79.7</td>
<td class="num">77.9</td>
<td class="num">28881</td>
<td><a href="03_control_cap_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_cap575W_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="03_control_cap_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_cap575W_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
</tbody>
</table>
</div>
<div class="exp-card">
<div class="exp-header">
<span class="exp-number">#04</span>
<span class="exp-title">Stepwise Power Ramp</span>
</div>
<p class="exp-desc">Gradually increasing/decreasing power limits to observe GPU adaptation behavior.</p>
<span class="exp-variable">Variable: Ramp enable</span>
<table class="run-table">
<thead><tr>
<th>Model</th><th>Batch</th><th>Pattern</th>
<th>Idle Avg (W)</th><th>Train Avg (W)</th><th>Train Peak (W)</th>
<th>Val Avg (W)</th><th>Infer Idle (W)</th><th>Energy (J)</th>
<th>Plots</th>
</tr></thead>
<tbody>
<tr>
<td>resnet18</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">54.1</td>
<td class="num">105.8</td>
<td class="num">148.3</td>
<td class="num">79.3</td>
<td class="num">77.9</td>
<td class="num">29572</td>
<td><a href="04_control_ramp/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_ramp_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="04_control_ramp/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_ramp_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
</tbody>
</table>
</div>
<div class="exp-card">
<div class="exp-header">
<span class="exp-number">#05</span>
<span class="exp-title">Inference Pattern Comparison</span>
</div>
<p class="exp-desc">Side-by-side comparison of three inference scheduling patterns and their grid impact.</p>
<span class="exp-variable">Variable: Pattern (fixed / variable / burst)</span>
<table class="run-table">
<thead><tr>
<th>Model</th><th>Batch</th><th>Pattern</th>
<th>Idle Avg (W)</th><th>Train Avg (W)</th><th>Train Peak (W)</th>
<th>Val Avg (W)</th><th>Infer Idle (W)</th><th>Energy (J)</th>
<th>Plots</th>
</tr></thead>
<tbody>
<tr>
<td>resnet18</td>
<td class="num">64</td>
<td>burst</td>
<td class="num">51.9</td>
<td class="num">102.0</td>
<td class="num">151.7</td>
<td class="num">79.5</td>
<td class="num">78.4</td>
<td class="num">28671</td>
<td><a href="05_pattern_fixed_var_burst/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_burst/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="05_pattern_fixed_var_burst/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_burst/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>resnet18</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">53.4</td>
<td class="num">100.1</td>
<td class="num">149.9</td>
<td class="num">80.6</td>
<td class="num">78.3</td>
<td class="num">28981</td>
<td><a href="05_pattern_fixed_var_burst/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="05_pattern_fixed_var_burst/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>resnet18</td>
<td class="num">64</td>
<td>variable</td>
<td class="num">52.3</td>
<td class="num">103.6</td>
<td class="num">150.7</td>
<td class="num">79.5</td>
<td class="num">78.0</td>
<td class="num">28852</td>
<td><a href="05_pattern_fixed_var_burst/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_variable/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="05_pattern_fixed_var_burst/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_variable/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
</tbody>
</table>
</div>
<div class="exp-card">
<div class="exp-header">
<span class="exp-number">#06</span>
<span class="exp-title">Model Architecture Scaling</span>
</div>
<p class="exp-desc">How different AI model architectures produce different power signatures. Models range from lightweight (~55W) to heavy (~310W).</p>
<span class="exp-variable">Variable: Model architecture (10 models)</span>
<table class="run-table">
<thead><tr>
<th>Model</th><th>Batch</th><th>Pattern</th>
<th>Idle Avg (W)</th><th>Train Avg (W)</th><th>Train Peak (W)</th>
<th>Val Avg (W)</th><th>Infer Idle (W)</th><th>Energy (J)</th>
<th>Plots</th>
</tr></thead>
<tbody>
<tr>
<td>convnext_tiny</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">53.6</td>
<td class="num">325.6</td>
<td class="num">514.4</td>
<td class="num">264.8</td>
<td class="num">79.1</td>
<td class="num">42655</td>
<td><a href="06_model_scaling_image/fixed_convnext_tiny_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="06_model_scaling_image/fixed_convnext_tiny_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>densenet121</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">53.7</td>
<td class="num">378.6</td>
<td class="num">436.7</td>
<td class="num">260.1</td>
<td class="num">79.2</td>
<td class="num">43719</td>
<td><a href="06_model_scaling_image/fixed_densenet121_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="06_model_scaling_image/fixed_densenet121_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>efficientnet_b0</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">53.2</td>
<td class="num">310.7</td>
<td class="num">401.6</td>
<td class="num">166.6</td>
<td class="num">78.9</td>
<td class="num">38061</td>
<td><a href="06_model_scaling_image/fixed_efficientnet_b0_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="06_model_scaling_image/fixed_efficientnet_b0_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>mobilenetv2</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">54.0</td>
<td class="num">103.0</td>
<td class="num">138.9</td>
<td class="num">81.8</td>
<td class="num">78.0</td>
<td class="num">29468</td>
<td><a href="06_model_scaling_image/fixed_mobilenetv2_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="06_model_scaling_image/fixed_mobilenetv2_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>resnet18</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">52.9</td>
<td class="num">101.9</td>
<td class="num">147.4</td>
<td class="num">80.7</td>
<td class="num">77.9</td>
<td class="num">28871</td>
<td><a href="06_model_scaling_image/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="06_model_scaling_image/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>resnet50</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">53.7</td>
<td class="num">134.7</td>
<td class="num">173.4</td>
<td class="num">92.4</td>
<td class="num">78.0</td>
<td class="num">29658</td>
<td><a href="06_model_scaling_image/fixed_resnet50_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="06_model_scaling_image/fixed_resnet50_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>resnext50_32x4d</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">53.4</td>
<td class="num">381.1</td>
<td class="num">514.3</td>
<td class="num">287.4</td>
<td class="num">79.5</td>
<td class="num">45881</td>
<td><a href="06_model_scaling_image/fixed_resnext50_32x4d_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="06_model_scaling_image/fixed_resnext50_32x4d_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>swin_t</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">51.8</td>
<td class="num">481.1</td>
<td class="num">512.4</td>
<td class="num">452.1</td>
<td class="num">80.2</td>
<td class="num">45197</td>
<td><a href="06_model_scaling_image/fixed_swin_t_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="06_model_scaling_image/fixed_swin_t_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>vgg16</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">52.1</td>
<td class="num">218.4</td>
<td class="num">287.3</td>
<td class="num">95.9</td>
<td class="num">78.0</td>
<td class="num">31082</td>
<td><a href="06_model_scaling_image/fixed_vgg16_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="06_model_scaling_image/fixed_vgg16_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>vit_b_16</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">51.7</td>
<td class="num">562.2</td>
<td class="num">585.8</td>
<td class="num">509.5</td>
<td class="num">80.2</td>
<td class="num">48888</td>
<td><a href="06_model_scaling_image/fixed_vit_b_16_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="06_model_scaling_image/fixed_vit_b_16_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
</tbody>
</table>
</div>
<div class="exp-card">
<div class="exp-header">
<span class="exp-number">#07</span>
<span class="exp-title">Dataset Comparison</span>
</div>
<p class="exp-desc">Whether the dataset (image resolution, number of classes) changes the power profile.</p>
<span class="exp-variable">Variable: Dataset (CIFAR-10 / CIFAR-100 / ImageNet)</span>
<table class="run-table">
<thead><tr>
<th>Model</th><th>Batch</th><th>Pattern</th>
<th>Idle Avg (W)</th><th>Train Avg (W)</th><th>Train Peak (W)</th>
<th>Val Avg (W)</th><th>Infer Idle (W)</th><th>Energy (J)</th>
<th>Plots</th>
</tr></thead>
<tbody>
<tr>
<td>resnet50</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">52.4</td>
<td class="num">131.6</td>
<td class="num">170.0</td>
<td class="num">85.2</td>
<td class="num">78.2</td>
<td class="num">29700</td>
<td><a href="07_dataset_cifar10_cifar100_imagenet/fixed_resnet50_cifar100_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="07_dataset_cifar10_cifar100_imagenet/fixed_resnet50_cifar100_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>resnet50</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">53.7</td>
<td class="num">135.0</td>
<td class="num">172.6</td>
<td class="num">94.8</td>
<td class="num">78.3</td>
<td class="num">29821</td>
<td><a href="07_dataset_cifar10_cifar100_imagenet/fixed_resnet50_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="07_dataset_cifar10_cifar100_imagenet/fixed_resnet50_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>resnet50</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">52.0</td>
<td class="num">479.7</td>
<td class="num">490.5</td>
<td class="num">163.7</td>
<td class="num">79.0</td>
<td class="num">338171</td>
<td><a href="07_dataset_cifar10_cifar100_imagenet/fixed_resnet50_imagenet_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="07_dataset_cifar10_cifar100_imagenet/fixed_resnet50_imagenet_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
</tbody>
</table>
</div>
<div class="exp-card">
<div class="exp-header">
<span class="exp-number">#08</span>
<span class="exp-title">Training Mode Comparison</span>
</div>
<p class="exp-desc">Fixed SGD training vs Optuna-based automated hyperparameter search.</p>
<span class="exp-variable">Variable: Training mode (fixed / automl)</span>
<table class="run-table">
<thead><tr>
<th>Model</th><th>Batch</th><th>Pattern</th>
<th>Idle Avg (W)</th><th>Train Avg (W)</th><th>Train Peak (W)</th>
<th>Val Avg (W)</th><th>Infer Idle (W)</th><th>Energy (J)</th>
<th>Plots</th>
</tr></thead>
<tbody>
<tr>
<td>resnet18</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">54.2</td>
<td class="num">102.5</td>
<td class="num">147.7</td>
<td class="num">82.2</td>
<td class="num">77.9</td>
<td class="num">28945</td>
<td><a href="08_train_modes_fixed_automl/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="08_train_modes_fixed_automl/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
</tbody>
</table>
</div>
<div class="exp-card">
<div class="exp-header">
<span class="exp-number">#09</span>
<span class="exp-title">Diffusion Models</span>
</div>
<p class="exp-desc">Stable Diffusion (SD1.5 / SDXL) power experiments.</p>
<span class="exp-variable">Variable: Diffusion model</span>
<p class="empty-msg">No completed runs in this experiment group.</p>
</div>
<div class="exp-card">
<div class="exp-header">
<span class="exp-number">#10</span>
<span class="exp-title">Checkpoint I/O Observation</span>
</div>
<p class="exp-desc">Periodic model saves create visible power spikes as GPU compute pauses for disk I/O.</p>
<span class="exp-variable">Variable: Checkpoint frequency</span>
<table class="run-table">
<thead><tr>
<th>Model</th><th>Batch</th><th>Pattern</th>
<th>Idle Avg (W)</th><th>Train Avg (W)</th><th>Train Peak (W)</th>
<th>Val Avg (W)</th><th>Infer Idle (W)</th><th>Energy (J)</th>
<th>Plots</th>
</tr></thead>
<tbody>
<tr>
<td>resnet18</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">53.5</td>
<td class="num">β€”</td>
<td class="num">β€”</td>
<td class="num">79.4</td>
<td class="num">77.9</td>
<td class="num">29034</td>
<td><a href="10_checkpoint_iteration_observe/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="10_checkpoint_iteration_observe/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
</tbody>
</table>
</div>
<div class="exp-card">
<div class="exp-header">
<span class="exp-number">#11</span>
<span class="exp-title">Clock Frequency Sweep</span>
</div>
<p class="exp-desc">Effect of locking GPU SM clocks at different frequencies. Lower clocks β†’ less power, slower execution.</p>
<span class="exp-variable">Variable: Clock lock (1005 / 1500 / 2100 MHz)</span>
<table class="run-table">
<thead><tr>
<th>Model</th><th>Batch</th><th>Pattern</th>
<th>Idle Avg (W)</th><th>Train Avg (W)</th><th>Train Peak (W)</th>
<th>Val Avg (W)</th><th>Infer Idle (W)</th><th>Energy (J)</th>
<th>Plots</th>
</tr></thead>
<tbody>
<tr>
<td>resnet18</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">52.9</td>
<td class="num">64.4</td>
<td class="num">79.1</td>
<td class="num">54.1</td>
<td class="num">53.5</td>
<td class="num">21503</td>
<td><a href="11_control_clock_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_clk1005MHz_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="11_control_clock_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_clk1005MHz_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>resnet18</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">57.9</td>
<td class="num">70.9</td>
<td class="num">88.1</td>
<td class="num">59.3</td>
<td class="num">57.9</td>
<td class="num">23260</td>
<td><a href="11_control_clock_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_clk1500MHz_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="11_control_clock_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_clk1500MHz_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
<tr>
<td>resnet18</td>
<td class="num">64</td>
<td>fixed</td>
<td class="num">61.2</td>
<td class="num">76.3</td>
<td class="num">98.1</td>
<td class="num">63.1</td>
<td class="num">61.8</td>
<td class="num">24797</td>
<td><a href="11_control_clock_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_clk2100MHz_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š Interactive</a> <a href="11_control_clock_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_clk2100MHz_fixed/plots/gpu_metrics.png" class="btn btn-png">πŸ–ΌοΈ Static</a></td>
</tr>
</tbody>
</table>
</div>
<div class="master-section">
<h2>Master Summary β€” All Runs</h2>
<div style="overflow-x: auto;">
<table class="master-table">
<thead><tr>
<th>Group</th><th>Model</th><th>Batch</th><th>Pattern</th>
<th>Idle (W)</th><th>Train Avg (W)</th><th>Train Peak (W)</th>
<th>Val (W)</th><th>Infer Idle (W)</th><th>Energy (J)</th><th>Plot</th>
</tr></thead>
<tbody>
<tr>
<td>Baseline Reference</td><td>resnet50</td><td class="num">64</td>
<td>fixed</td>
<td class="num">52.7</td><td class="num">134.3</td>
<td class="num">178.0</td><td class="num">89.9</td>
<td class="num">78.1</td>
<td class="num">29670</td>
<td><a href="00_baseline_reference/fixed_resnet50_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Smoke Test</td><td>resnet18</td><td class="num">8</td>
<td>fixed</td>
<td class="num">52.2</td><td class="num">125.7</td>
<td class="num">137.2</td><td class="num">123.6</td>
<td class="num">23.7</td>
<td class="num">9149</td>
<td><a href="00_smoke_test/fixed_resnet18_cifar10_gpu0_bs8_ep1_req2_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Batch Size Sweep</td><td>resnet50</td><td class="num">128</td>
<td>fixed</td>
<td class="num">51.8</td><td class="num">138.1</td>
<td class="num">211.8</td><td class="num">98.1</td>
<td class="num">78.2</td>
<td class="num">29205</td>
<td><a href="01_batch_sweep_resnet50/fixed_resnet50_cifar10_gpu0_bs128_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Batch Size Sweep</td><td>resnet50</td><td class="num">16</td>
<td>fixed</td>
<td class="num">53.3</td><td class="num">139.9</td>
<td class="num">152.2</td><td class="num">132.5</td>
<td class="num">77.9</td>
<td class="num">32440</td>
<td><a href="01_batch_sweep_resnet50/fixed_resnet50_cifar10_gpu0_bs16_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Batch Size Sweep</td><td>resnet50</td><td class="num">64</td>
<td>fixed</td>
<td class="num">52.5</td><td class="num">133.4</td>
<td class="num">173.2</td><td class="num">83.7</td>
<td class="num">78.3</td>
<td class="num">29732</td>
<td><a href="01_batch_sweep_resnet50/fixed_resnet50_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Operator Control (Variable Inference)</td><td>resnet18</td><td class="num">64</td>
<td>variable</td>
<td class="num">53.7</td><td class="num">128.5</td>
<td class="num">152.2</td><td class="num">133.8</td>
<td class="num">78.2</td>
<td class="num">34124</td>
<td><a href="02_operator_control/fixed_resnet18_cifar10_gpu0_bs64_ep5_req60_nocap_variable/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Power Cap Sweep</td><td>resnet18</td><td class="num">64</td>
<td>fixed</td>
<td class="num">9.9</td><td class="num">111.6</td>
<td class="num">148.0</td><td class="num">148.6</td>
<td class="num">77.9</td>
<td class="num">27225</td>
<td><a href="03_control_cap_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_cap345W_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Power Cap Sweep</td><td>resnet18</td><td class="num">64</td>
<td>fixed</td>
<td class="num">9.4</td><td class="num">118.1</td>
<td class="num">147.3</td><td class="num">145.8</td>
<td class="num">78.1</td>
<td class="num">27206</td>
<td><a href="03_control_cap_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_cap460W_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Power Cap Sweep</td><td>resnet18</td><td class="num">64</td>
<td>fixed</td>
<td class="num">52.1</td><td class="num">102.0</td>
<td class="num">150.0</td><td class="num">79.7</td>
<td class="num">77.9</td>
<td class="num">28881</td>
<td><a href="03_control_cap_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_cap575W_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Stepwise Power Ramp</td><td>resnet18</td><td class="num">64</td>
<td>fixed</td>
<td class="num">54.1</td><td class="num">105.8</td>
<td class="num">148.3</td><td class="num">79.3</td>
<td class="num">77.9</td>
<td class="num">29572</td>
<td><a href="04_control_ramp/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_ramp_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Inference Pattern Comparison</td><td>resnet18</td><td class="num">64</td>
<td>burst</td>
<td class="num">51.9</td><td class="num">102.0</td>
<td class="num">151.7</td><td class="num">79.5</td>
<td class="num">78.4</td>
<td class="num">28671</td>
<td><a href="05_pattern_fixed_var_burst/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_burst/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Inference Pattern Comparison</td><td>resnet18</td><td class="num">64</td>
<td>fixed</td>
<td class="num">53.4</td><td class="num">100.1</td>
<td class="num">149.9</td><td class="num">80.6</td>
<td class="num">78.3</td>
<td class="num">28981</td>
<td><a href="05_pattern_fixed_var_burst/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Inference Pattern Comparison</td><td>resnet18</td><td class="num">64</td>
<td>variable</td>
<td class="num">52.3</td><td class="num">103.6</td>
<td class="num">150.7</td><td class="num">79.5</td>
<td class="num">78.0</td>
<td class="num">28852</td>
<td><a href="05_pattern_fixed_var_burst/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_variable/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Model Architecture Scaling</td><td>convnext_tiny</td><td class="num">64</td>
<td>fixed</td>
<td class="num">53.6</td><td class="num">325.6</td>
<td class="num">514.4</td><td class="num">264.8</td>
<td class="num">79.1</td>
<td class="num">42655</td>
<td><a href="06_model_scaling_image/fixed_convnext_tiny_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Model Architecture Scaling</td><td>densenet121</td><td class="num">64</td>
<td>fixed</td>
<td class="num">53.7</td><td class="num">378.6</td>
<td class="num">436.7</td><td class="num">260.1</td>
<td class="num">79.2</td>
<td class="num">43719</td>
<td><a href="06_model_scaling_image/fixed_densenet121_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Model Architecture Scaling</td><td>efficientnet_b0</td><td class="num">64</td>
<td>fixed</td>
<td class="num">53.2</td><td class="num">310.7</td>
<td class="num">401.6</td><td class="num">166.6</td>
<td class="num">78.9</td>
<td class="num">38061</td>
<td><a href="06_model_scaling_image/fixed_efficientnet_b0_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Model Architecture Scaling</td><td>mobilenetv2</td><td class="num">64</td>
<td>fixed</td>
<td class="num">54.0</td><td class="num">103.0</td>
<td class="num">138.9</td><td class="num">81.8</td>
<td class="num">78.0</td>
<td class="num">29468</td>
<td><a href="06_model_scaling_image/fixed_mobilenetv2_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Model Architecture Scaling</td><td>resnet18</td><td class="num">64</td>
<td>fixed</td>
<td class="num">52.9</td><td class="num">101.9</td>
<td class="num">147.4</td><td class="num">80.7</td>
<td class="num">77.9</td>
<td class="num">28871</td>
<td><a href="06_model_scaling_image/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Model Architecture Scaling</td><td>resnet50</td><td class="num">64</td>
<td>fixed</td>
<td class="num">53.7</td><td class="num">134.7</td>
<td class="num">173.4</td><td class="num">92.4</td>
<td class="num">78.0</td>
<td class="num">29658</td>
<td><a href="06_model_scaling_image/fixed_resnet50_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Model Architecture Scaling</td><td>resnext50_32x4d</td><td class="num">64</td>
<td>fixed</td>
<td class="num">53.4</td><td class="num">381.1</td>
<td class="num">514.3</td><td class="num">287.4</td>
<td class="num">79.5</td>
<td class="num">45881</td>
<td><a href="06_model_scaling_image/fixed_resnext50_32x4d_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Model Architecture Scaling</td><td>swin_t</td><td class="num">64</td>
<td>fixed</td>
<td class="num">51.8</td><td class="num">481.1</td>
<td class="num">512.4</td><td class="num">452.1</td>
<td class="num">80.2</td>
<td class="num">45197</td>
<td><a href="06_model_scaling_image/fixed_swin_t_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Model Architecture Scaling</td><td>vgg16</td><td class="num">64</td>
<td>fixed</td>
<td class="num">52.1</td><td class="num">218.4</td>
<td class="num">287.3</td><td class="num">95.9</td>
<td class="num">78.0</td>
<td class="num">31082</td>
<td><a href="06_model_scaling_image/fixed_vgg16_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Model Architecture Scaling</td><td>vit_b_16</td><td class="num">64</td>
<td>fixed</td>
<td class="num">51.7</td><td class="num">562.2</td>
<td class="num">585.8</td><td class="num">509.5</td>
<td class="num">80.2</td>
<td class="num">48888</td>
<td><a href="06_model_scaling_image/fixed_vit_b_16_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Dataset Comparison</td><td>resnet50</td><td class="num">64</td>
<td>fixed</td>
<td class="num">52.4</td><td class="num">131.6</td>
<td class="num">170.0</td><td class="num">85.2</td>
<td class="num">78.2</td>
<td class="num">29700</td>
<td><a href="07_dataset_cifar10_cifar100_imagenet/fixed_resnet50_cifar100_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Dataset Comparison</td><td>resnet50</td><td class="num">64</td>
<td>fixed</td>
<td class="num">53.7</td><td class="num">135.0</td>
<td class="num">172.6</td><td class="num">94.8</td>
<td class="num">78.3</td>
<td class="num">29821</td>
<td><a href="07_dataset_cifar10_cifar100_imagenet/fixed_resnet50_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Dataset Comparison</td><td>resnet50</td><td class="num">64</td>
<td>fixed</td>
<td class="num">52.0</td><td class="num">479.7</td>
<td class="num">490.5</td><td class="num">163.7</td>
<td class="num">79.0</td>
<td class="num">338171</td>
<td><a href="07_dataset_cifar10_cifar100_imagenet/fixed_resnet50_imagenet_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Training Mode Comparison</td><td>resnet18</td><td class="num">64</td>
<td>fixed</td>
<td class="num">54.2</td><td class="num">102.5</td>
<td class="num">147.7</td><td class="num">82.2</td>
<td class="num">77.9</td>
<td class="num">28945</td>
<td><a href="08_train_modes_fixed_automl/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Checkpoint I/O Observation</td><td>resnet18</td><td class="num">64</td>
<td>fixed</td>
<td class="num">53.5</td><td class="num">β€”</td>
<td class="num">β€”</td><td class="num">79.4</td>
<td class="num">77.9</td>
<td class="num">29034</td>
<td><a href="10_checkpoint_iteration_observe/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Clock Frequency Sweep</td><td>resnet18</td><td class="num">64</td>
<td>fixed</td>
<td class="num">52.9</td><td class="num">64.4</td>
<td class="num">79.1</td><td class="num">54.1</td>
<td class="num">53.5</td>
<td class="num">21503</td>
<td><a href="11_control_clock_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_clk1005MHz_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Clock Frequency Sweep</td><td>resnet18</td><td class="num">64</td>
<td>fixed</td>
<td class="num">57.9</td><td class="num">70.9</td>
<td class="num">88.1</td><td class="num">59.3</td>
<td class="num">57.9</td>
<td class="num">23260</td>
<td><a href="11_control_clock_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_clk1500MHz_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
<tr>
<td>Clock Frequency Sweep</td><td>resnet18</td><td class="num">64</td>
<td>fixed</td>
<td class="num">61.2</td><td class="num">76.3</td>
<td class="num">98.1</td><td class="num">63.1</td>
<td class="num">61.8</td>
<td class="num">24797</td>
<td><a href="11_control_clock_sweep/fixed_resnet18_cifar10_gpu0_bs64_ep1_req10000_nocap_clk2100MHz_fixed/plots/gpu_power_interactive.html" class="btn btn-plot">πŸ“Š</a></td>
</tr>
</tbody>
</table>
</div>
</div>
<div class="footer">
<p>GPU Power Experiment Transfer Package Β· Generated 2026-02-19 13:10</p>
<p>Each run folder contains: config.json Β· phase_power_summary.csv Β· gpu_samples.csv Β· transition_points.csv Β· gpu_power_interactive.html Β· gpu_metrics.png</p>
</div>
</div>
</body>
</html>