<?xml version="1.0"?>
<feed xmlns="http://www.w3.org/2005/Atom" xml:lang="en">
	<id>https://wiki.kunix.org/index.php?action=history&amp;feed=atom&amp;title=Boris-1.3-75M</id>
	<title>Boris-1.3-75M - Revision history</title>
	<link rel="self" type="application/atom+xml" href="https://wiki.kunix.org/index.php?action=history&amp;feed=atom&amp;title=Boris-1.3-75M"/>
	<link rel="alternate" type="text/html" href="https://wiki.kunix.org/index.php?title=Boris-1.3-75M&amp;action=history"/>
	<updated>2026-11-22T13:53:24Z</updated>
	<subtitle>Revision history for this page on the wiki</subtitle>
	<generator>MediaWiki 1.43.9</generator>
	<entry>
		<id>https://wiki.kunix.org/index.php?title=Boris-1.3-75M&amp;diff=123&amp;oldid=prev</id>
		<title>Jdj: New article with Infobox model (data from the Hugging Face model card)</title>
		<link rel="alternate" type="text/html" href="https://wiki.kunix.org/index.php?title=Boris-1.3-75M&amp;diff=123&amp;oldid=prev"/>
		<updated>2026-11-16T00:43:47Z</updated>

		<summary type="html">&lt;p&gt;New article with Infobox model (data from the Hugging Face model card)&lt;/p&gt;
&lt;p&gt;&lt;b&gt;New page&lt;/b&gt;&lt;/p&gt;&lt;div&gt;{{Infobox model&lt;br /&gt;
| developer       = [[OpenCerebral]]&lt;br /&gt;
| family          = Boris-1.3&lt;br /&gt;
| model_type      = Base (continued pretraining)&lt;br /&gt;
| base_model      = [[Boris-75M]]&lt;br /&gt;
| license         = Apache 2.0&lt;br /&gt;
| status          = Released&lt;br /&gt;
| architecture    = GPT-2 decoder (pre-LayerNorm, learned positions, tied embeddings), 9 heads, d&amp;lt;sub&amp;gt;model&amp;lt;/sub&amp;gt; 576&lt;br /&gt;
| parameters      = 77.4M&lt;br /&gt;
| layers          = 12&lt;br /&gt;
| context         = 1,024 tokens&lt;br /&gt;
| tokenizer       = GPT-NeoX-20B BPE, 50,304 (padded)&lt;br /&gt;
| training_tokens = 3.95B (1.55B base + 2.4B continued)&lt;br /&gt;
| dataset         = Base: FineWeb-Edu 60% / DCLM 40%. Continued: DCLM-Baseline 1.5B, then FineWeb-Edu 0.3B + 0.6B&lt;br /&gt;
| hardware        = 1 × RTX 3060 12GB&lt;br /&gt;
| training_time   = 14 h 49 m base + about 22 h continued&lt;br /&gt;
| compute         = ≈ 1.8 × 10&amp;lt;sup&amp;gt;18&amp;lt;/sup&amp;gt; FLOPs (6ND estimate)&lt;br /&gt;
| precision       = bf16 autocast, fp32 master weights&lt;br /&gt;
| hellaswag       = 27.57&lt;br /&gt;
| hellaswag_ref   = 27.20&lt;br /&gt;
| piqa            = 59.41&lt;br /&gt;
| piqa_ref        = 57.18&lt;br /&gt;
| winogrande      = 51.54&lt;br /&gt;
| winogrande_ref  = 49.72&lt;br /&gt;
| arc_easy        = 40.57&lt;br /&gt;
| arc_easy_ref    = 39.14&lt;br /&gt;
| arc_challenge   = 23.46&lt;br /&gt;
| arc_challenge_ref= 23.04&lt;br /&gt;
| lambada         = 18.16&lt;br /&gt;
| lambada_ref     = 15.21&lt;br /&gt;
| bench_ref_label = Boris-75M&lt;br /&gt;
| bench_note      = Zero-shot; acc_norm for HellaSwag, PIQA and ARC, acc for WinoGrande and LAMBADA&lt;br /&gt;
| bench_source    = [https://huggingface.co/opencerebral/Boris-1.3-75M model card]&lt;br /&gt;
| weights         = [https://huggingface.co/opencerebral/Boris-1.3-75M opencerebral/Boris-1.3-75M]&lt;br /&gt;
| formats         = safetensors (transformers GPT-2)&lt;br /&gt;
| website         = [https://oc.kunix.org oc.kunix.org]&lt;br /&gt;
| released        = 2026-07-26&lt;br /&gt;
| series          = Boris base models&lt;br /&gt;
| preceded_by     = [[Boris-250M]]&lt;br /&gt;
| succeeded_by    = [[Boris-1.3-125M]]&lt;br /&gt;
}}&lt;br /&gt;
&amp;#039;&amp;#039;&amp;#039;Boris-1.3-75M&amp;#039;&amp;#039;&amp;#039; is [[Boris-75M]] with 2.4B tokens of continued pretraining (one DCLM pass, then two FineWeb-Edu passes), aimed at the gaps in Boris-75M&amp;#039;s own benchmark results.&lt;br /&gt;
&lt;br /&gt;
== External links ==&lt;br /&gt;
* [https://huggingface.co/opencerebral/Boris-1.3-75M Boris-1.3-75M on Hugging Face]&lt;br /&gt;
&lt;br /&gt;
{{Stub}}&lt;/div&gt;</summary>
		<author><name>Jdj</name></author>
	</entry>
</feed>