<?xml version="1.0" encoding="UTF-8"?><oembed><type>video</type><version>1.0</version><html>&lt;iframe src=&quot;https://www.loom.com/embed/79b1c49beb854bed84f8e11d0b3b0b02&quot; frameborder=&quot;0&quot; width=&quot;1600&quot; height=&quot;1200&quot; webkitallowfullscreen mozallowfullscreen allowfullscreen&gt;&lt;/iframe&gt;</html><height>1200</height><width>1600</width><provider_name>Loom</provider_name><provider_url>https://www.loom.com</provider_url><thumbnail_height>1200</thumbnail_height><thumbnail_width>1600</thumbnail_width><thumbnail_url>https://cdn.loom.com/sessions/thumbnails/79b1c49beb854bed84f8e11d0b3b0b02-95cbf7a0cb362266.gif</thumbnail_url><duration>23.516</duration><title>Token repetition issue in Mistral-7B-Instruct-v0.3 · Hugging Face</title><description>When providing a single word input &quot;vanakkam&quot; the model repeats the same word almost indefinitely in the output.
It seems to get stuck in a loop generating the same token again and again.</description></oembed>