<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Yestino - The Signal · Parametrize FP4</title><link>https://yestino.com/entities/parametrize-fp4-3e04fa</link><description>Every event involving Parametrize FP4</description><language>en</language><atom:link href="https://yestino.com/entities/parametrize-fp4-3e04fa/feed.xml" rel="self" type="application/rss+xml"/><item><title>trunk/d2e122ddd079e2386d88ca97681fd1d5a9c28060: Parametrize FP4 scaled_mm tests by recipe instead of platform. (#197740)</title><link>https://yestino.com/events/trunk-d2e122ddd079e2386d88ca97681fd1d5a9c28060-parametrize-f-d6007f</link><guid isPermaLink="true">https://yestino.com/events/trunk-d2e122ddd079e2386d88ca97681fd1d5a9c28060-parametrize-f-d6007f</guid><pubDate>Fri, 25 Sep 2026 04:14:42 GMT</pubDate><description>test_blockwise_nvfp4_compile picked its recipe off torch.version.hip, so on ROCm it ran MXFP4 under an NVFP4 name; ScaledBlas.cpp rejects NVFP4 on ROCm, so the…
Sources: PyTorch Releases</description></item></channel></rss>