<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[ACUITY 导出 YOLOv5nu 模型 nbg_compile error 512 — DFL 层移除后仍失败 求救！！！！]]></title><description><![CDATA[<p dir="auto">在 <strong>Orange Pi 4 Pro（全志 A733）</strong> 上使用 <strong>ACUITY 工具链（v6.30.22）</strong> 将自训练的 <strong>YOLOv5nu</strong> ONNX 模型转换为 <strong>VIPLite NPU .nb 格式</strong>时，<code>pegasus_export_ovx.sh</code> 的 <strong>nbg_compile 步骤始终报 error 512</strong>。</p>
<p dir="auto">已尝试 <strong>3 种不同的 ONNX 变体</strong>（原始含 DFL、去除 DFL 单输出、多输出分离），均无法通过编译。</p>
<h2>硬件环境</h2>
<table class="table table-bordered table-striped">
<thead>
<tr>
<th>项目</th>
<th>详情</th>
</tr>
</thead>
<tbody>
<tr>
<td>开发板</td>
<td>Orange Pi 4 Pro</td>
</tr>
<tr>
<td>SoC</td>
<td>全志 Allwinner A733</td>
</tr>
<tr>
<td>NPU</td>
<td>VIPLite NPU，3 TOPS</td>
</tr>
<tr>
<td>NPU 驱动</td>
<td>viplite v2.0.3.2</td>
</tr>
<tr>
<td>目标模型格式</td>
<td><code>.nb</code>（NBG），通过 ACUITY Docker 容器转换</td>
</tr>
<tr>
<td><a href="/assets/uploads/files/1783050677685-forum_post_nbg512.md">forum_post_nbg512.md</a></td>
<td></td>
</tr>
</tbody>
</table>
]]></description><link>https://bbs.aw-ol.com/topic/7057/acuity-导出-yolov5nu-模型-nbg_compile-error-512-dfl-层移除后仍失败-求救</link><generator>RSS for Node</generator><lastBuildDate>Wed, 29 Jul 2026 21:42:10 GMT</lastBuildDate><atom:link href="https://bbs.aw-ol.com/topic/7057.rss" rel="self" type="application/rss+xml"/><pubDate>Fri, 03 Jul 2026 03:52:02 GMT</pubDate><ttl>60</ttl></channel></rss>