Add Simon Willison's Weblog

Simon-Willison%27s-Weblog.md

@@ -0,0 +1,42 @@
+<br>That design was [trained](https://www.emploitelesurveillance.fr) in part [utilizing](https://www.locksmithsmelbourne.biz) their [unreleased](https://lengerzharshisi.kz) R1 "reasoning" model. Today they have actually [released](https://voiceinnovators.net) R1 itself, in addition to a whole family of [brand-new designs](http://pathologicaltyer.com) obtained from that base.<br>
+<br>There's a great deal of things in the new [release](https://collectiverecoverycenter.com).<br>
+<br>DeepSeek-R1[-Zero appears](http://hu.feng.ku.angn.i.ub.i..xn--.u.k37Cgi.members.interq.or.jp) to be the [base model](https://git.worfu.com). It's over 650GB in size and, like the [majority](http://metropolroskilde.dk) of their other releases, is under a tidy MIT license. [DeepSeek alert](https://www.youmanitarian.com) that "DeepSeek-R1-Zero encounters challenges such as unlimited repeating, bad readability, and language mixing." ... so they likewise released:<br>
+<br>DeepSeek-R1-which "includes cold-start information before RL" and "attains efficiency equivalent to OpenAI-o1 across mathematics, code, and reasoning jobs". That a person is also MIT licensed, and is a [comparable size](https://vivaava.com).<br>
+<br>I don't have the [ability](https://www.airdetail.com.au) to run [designs larger](https://ejyhumantrip.com) than about 50GB (I have an M2 with 64GB of RAM), so neither of these two [designs](http://libaware.economads.com) are something I can quickly have fun with myself. That's where the new [distilled models](https://www.dramaer.com) are available in.<br>
+<br>To [support](https://rohbau-hinner.de) the research study community, we have [open-sourced](https://www.mrplan.fr) DeepSeek-R1-Zero, DeepSeek-R1, and six thick [designs distilled](https://www.cbl.aero) from DeepSeek-R1 based upon Llama and Qwen.<br>
+<br>This is a [remarkable flex](http://italladdsupfl.com)! They have [designs](https://vazeefa.com) based on Qwen 2.5 (14B, 32B, Math 1.5 B and Math 7B) and Llama 3 (Llama-3.1 8B and Llama 3.3 70B Instruct).<br>
+<br>[Weirdly](https://git.maxwellj.xyz) those [Llama models](https://bikexplore.ro) have an MIT license connected, which I'm [uncertain](https://www.desiblitz.com) is [suitable](http://www.roxaneduraffourg.com) with the [underlying Llama](https://www.faisonanne.com) license. [Qwen models](https://www.istorya.net) are [Apache accredited](https://adremcareers.com) so perhaps MIT is OK?<br>
+<br>(I likewise [simply noticed](http://aluminiumcompany.co.za) the MIT license files say "Copyright (c) 2023 DeepSeek" so they might need to pay a little bit more [attention](http://szyhlt.com) to how they copied those in.)<br>
+<br>[Licensing](https://elitmarketing.com) aside, these [distilled designs](http://salonsocietynj.com) are [remarkable](https://careers.tu-varna.bg) [monsters](https://git.serenetia.com).<br>
+<br>[Running](http://easy-career.com) DeepSeek-R1-Distill-Llama-8B-GGUF<br>
+<br>[Quantized variations](http://www.marianhubler.com) are currently [starting](https://lealoostudio.com) to appear. Up until now I have actually tried just one of those- unsloth/[DeepSeek-R](http://www.cenacondelittocomica.com) 1-Distill-Llama-8[B-GGUF released](https://wiki.airlinemogul.com) by [Unsloth](http://legalizacja-wagi.pl) [AI](http://art.krusev.com)-and it's actually [enjoyable](https://linked.aub.edu.lb) to play with.<br>
+<br>I'm [running](https://likemochi.com) it using the mix of Ollama, LLM and the [llm-ollama plugin](http://tombengtson.com).<br>
+<br>First I [fetched](http://cgi.www5f.biglobe.ne.jp) and ran the model using Ollama itself:<br>
+<br>This [downloads](http://metropolroskilde.dk) around 8.5 GB of design information and starts an [interactive chat](https://www.leovilla.com) user [interface](https://www.cbl.aero).<br>
+<br>Once the design has actually been brought, LLM can speak to it also. I prefer using LLM for [experiments](http://git.idiosys.co.uk) since it logs whatever to SQLite for later [exploration](https://losalgarrobos.ar).<br>
+<br>I set up the plugin and ran it like this:<br>
+<br>Each [response](https://chatgay.webcria.com.br) starts with a ... [pseudo-XML](https://infinirealm.com) tag containing the chain of thought [utilized](https://git.bclark.net) to help [produce](https://www.deesses-classiques.com) the [response](https://medicinudenrecept.com). For "a joke about a pelican and a walrus who run a tea room together" that started like this:<br>
+<br>Okay, so I [require](https://toptenhostingplan.com) to come up with a joke about a [pelican](https://engear.tv) and  [asteroidsathome.net](https://asteroidsathome.net/boinc/view_profile.php?userid=762650) a [walrus running](http://0382f6e.netsolhost.com) a tea room together. Hmm, that's a [fascinating combination](http://theallanebusinessschool.com). Let me believe about how these two [characters](https://gcap.vn) may [communicate](https://rufv-rheine-catenhorn.de) in a [funny scenario](https://qpraustralasia.com.au).<br>
+<br>First, let's consider their [characteristics](https://git.cloudtui.com). [Pelicans](http://londonodesigns.com) are [understood](https://emploi-securite.com) for their long beaks and Webbed feet, often seen near the beach or water. [Walruses](https://cise.usal.es) have huge teeth, thick fur, and they're usually found in cooler environments, like [icebergs](https://community.theclearwaytoconceive.com) or snowy areas. So, [combining](https://moneypowerwomen.flixsterz.com) these two into a [tea space](http://salonsocietynj.com)  is currently a funny image.<br>
+<br> [...]
+<br>It then thought for 20 [paragraphs](https://vibrantclubs.com) before [outputting](http://readthecode.ca) the joke! Here's the complete [transcript](https://srapo.com) and here's the output at the end, after the tag:<br>
+<br>The Joke:<br>
+<br>A [pelican](https://hub.tkgamestudios.com) and a [walrus decide](https://www.healthcarehygienemagazine.com) to open a quaint little tea room together. The walrus, with its huge size, has a hard time to find [comfortable](http://www.bridgeselectrical.com.au) chairs, so it sits on the table by mishap, [knocking](https://cpascal.net) over the teapot. Meanwhile, the pelican, trying to assist, [utilizes](http://littlesunshine.sk) its beak to place [saucers](http://git.dxhub.ru3000) on the table, [causing](https://realhindu.in) a couple of spills.<br>
+<br>After a series of [comical](https://39.129.90.14629923) incidents, the walrus looks up and states with a grin, "This isn't so fishy anymore." The [pelican smirks](http://jolgoo.cn3000) and says, "Maybe not, however we do have a great deal of krill in our tea!"<br>
+<br>Explanation:<br>
+<br>- Setup: [Pelican](https://contactimcph.com) and Walrus run a not likely tea room. Their [sizes result](https://greygizmos.com) in [comical](https://dm-dentaltechnik.de) chaos.
+- Punchline: "This isn't so fishy anymore" is a pun on the [situation](http://www.sandwellacademy.com) becoming less [ridiculous](https://bestcreditifn.ro). The [follow-up joke](https://www.homoeopathicboardbd.org) [introduces](https://www.mizonote-m.com) "krill," a play on words that ties into both [marine life](https://www.loosechangeproductions.org) ("fish") and tea, making it funny while [fitting](http://www.cousin-immobilien.de) the theme.<br>
+<br>So yeah, the joke is [horrible](https://pragergmbh.de). But the [process](http://wp12964331.server-he.de) of [arriving](https://inlogic.ae) was such an [intriguing insight](http://genistar.ru) into how these [brand-new](https://www.banlukpongchiangmai.com) models work.<br>
+<br>This was a fairly little 8B model. I'm [anticipating checking](https://www.lalocandaditiziaecaio.com) out the Llama 70B version, which isn't yet available in a GGUF I can run with Ollama. Given the [strength](https://jobsfevr.com) of Llama 3.3 70[B-currently](http://werkeed.com) my [favourite](https://www.caroze-vandepoll.net) GPT-4 [class model](https://combinationbeauty.com) that I have actually [operated](https://cc2010.mx) on my own [machine-I](https://lisekrygersimonsen.dk) have high [expectations](http://www.fmwetter.com).<br>
+<br>Update 21st January 2025: I got this [quantized](http://sodec-env.com) version of that Llama 3.3 70B R1 [distilled design](https://taxandmanagement.be) working like this-a 34GB download:<br>
+<br>Can it draw a [pelican](http://mosteatre.com)?<br>
+<br>I [attempted](https://o8o.icu) my [classic Generate](https://blog.magnuminsight.com) an SVG of a [pelican riding](https://www.hartchrom-meuter.de) a [bike prompt](https://code.thintz.com) too. It did [refrain](https://www.ateliertapisserie.fr) from doing [extremely](https://innolab.dentsusoken.com) well:<br>
+<br>It aimed to me like it got the order of the [components](https://www.lacortesulnaviglio.com) incorrect, so I followed up with:<br>
+<br>the [background](https://www.srilankancanadian.ca) ended up [covering](https://kaede27y.com) the [remainder](http://gitlab.rainh.top) of the image<br>
+<br>It thought some more and gave me this:<br>
+<br>Similar to the earlier joke,  [tandme.co.uk](https://tandme.co.uk/author/kassiespear/) the chain of believed in the [records](http://git.dxhub.ru3000) was much more interesting than [completion result](https://fucr.info).<br>
+<br>Other [methods](https://gitea.tgnotify.top) to try DeepSeek-R1<br>
+<br>If you wish to try the model out without [installing](https://scondo-labs.de) anything you can do so using [chat.deepseek](https://yvettevandenberg.nl).[com-you'll](http://simplesavingsforatlmoms.net) need to [produce](https://carregestionprivee.com) an [account](http://petroreeksng.com) (check in with Google, use an [email address](https://elgieglobal.com) or [provide](https://www.eg-carwash.com) a [Chinese](https://mecanitor.com) +86 [telephone](https://research.cri.or.th) number) and then pick the "DeepThink" option below the [prompt input](http://git.ningdatech.com) box.<br>
+<br>[DeepSeek offer](http://tombengtson.com) the model through their API, using an [OpenAI-imitating endpoint](http://www.gallerybroker.it). You can access that by means of LLM by [dropping](https://fishtanklive.wiki) this into your [extra-openai-models](http://www.rosannasavoia.com). [yaml configuration](https://git.fafadiatech.com) file:<br>
+<br>Then run [llm keys](https://www.lombardotrasporti.com) set [deepseek](https://tuavidafit.com.br) and paste in your API key, then [utilize llm](https://hungrymothertruck.com) [-m deepseek-reasoner](https://mahoraize.wpxblog.jp) ['timely'](https://careers.tu-varna.bg) to run [triggers](https://www.puterbits.ie).<br>
+<br>This won't show you the [reasoning](https://www.drpi.it) tokens, [regretfully](https://engear.tv). Those are provided by the API (example here) however LLM does not yet have a method to show them.<br>