Add 'Run DeepSeek R1 Locally - with all 671 Billion Parameters'

5 months ago · 060e07f2cc
1 changed files with 47 additions and 0 deletions
--- a/Run-DeepSeek-R1-Locally---with-all-671-Billion-Parameters.md
+++ b/Run-DeepSeek-R1-Locally---with-all-671-Billion-Parameters.md
@ -0,0 +1,47 @@
 <br>Recently, I revealed how to easily run [distilled versions](https://www.seasilkfund.com) of the [DeepSeek](http://inovaresolar.com.br) R1 [model locally](https://groupsmeet.com). A distilled design is a compressed variation of a bigger language model, where understanding from a [larger design](https://gibsonvastgoedmanagement.nl) is moved to a smaller sized one to minimize resource usage without losing too much efficiency. These [designs](http://ikre.net) are based on the Llama and Qwen architectures and be available in versions varying from 1.5 to 70 billion criteria.<br>
 <br>Some explained that this is not the REAL DeepSeek R1 and that it is impossible to run the full design in your area without [numerous](http://www.elys-dog.com) hundred GB of memory. That [sounded](http://www.careyauctioneers.ie) like a difficulty - I believed! First Attempt - Heating Up with a 1.58 bit Quantized Version of DeepSeek R1 671b in Ollama.cpp<br>
 <br>The designers behind [Unsloth dynamically](http://www.careyauctioneers.ie) quantized DeepSeek R1 so that it might work on as low as 130GB while still gaining from all 671 billion [specifications](https://www.ib-wocheslander.de).<br>
 <br>A quantized LLM is a LLM whose [parameters](https://www.cartomanziagratis.info) are saved in [lower-precision formats](https://git.yurecnt.ru) (e.g., 8-bit or 4-bit instead of 16-bit). This substantially lowers memory use and accelerates processing, with very little impact on performance. The complete [variation](https://www.ib-wocheslander.de) of DeepSeek R1 uses 16 bit.<br>
 <br>The compromise in [accuracy](https://www.tmstriekaneizolacie.sk) is hopefully compensated by increased speed.<br>
 <br>I downloaded the files from this collection on Hugging Face and ran the following command with Llama.cpp.<br>
 <br>The following table from Unsloth shows the recommended value for the n-gpu-layers specification, which indicates just how much work can be [unloaded](http://www.plvproductions.com) to the GPU.<br>
 <br>According to the table, I believed 7 need to be the maximum, however I got it keeping up 12. According to Windows Task Manager my GPU has 40 GB of memory, and not 24 as I thought. So then it [accumulates](https://privategigs.fr) (7/ 24 * 40 ≈ 12).<br>
 <br>If you prefer to run the [model straight](http://www.cantinhodaeve.com) with Ollama, you must merge the three [GGUF files](https://nycityus.com) using Llama.cpp or a similar tool first.<br>
 <br>Above is a few of the text that [appears](https://thekinddessert.com) on screen while running the command.<br>
 <br>Llama.cpp and the quantized model are heavy on memory and CPU, but likewise utilize the GPU (a little) and [continuously](https://foodyfood.ro) read from the disk where the design files are kept.<br>
 <br>The design repeats the question and is preparing yourself to [start reasoning](http://www.sidotec.it).<br>
 <br>The reasoning, or thinking, is quite lengthy - which also contributes to the general time it takes.<br>
 <br>Okay,  [fakenews.win](https://fakenews.win/wiki/User:CathleenLinder) so I need to determine Alan Turing's most [essential contribution](https://www.planetwise.net) to how we live our lives in 2025. Hmm. Let's start by [remembering](https://www.careernextindia.com) who Alan Turing was. He was a British mathematician and computer scientist, right? He's known for his work in [code-breaking](https://pogruz.kg) throughout World War II, particularly with the [Enigma machine](https://mkala-koncert.ru). Oh, and he's likewise considered one of the fathers of theoretical computer technology and expert system.<br>
 <br>So, his contributions ... Well, the [Turing Machine](https://feilenhauer.net) idea is essential in computer technology. It's a theoretical model of a computer, which laid the groundwork for contemporary computing. Then there's the Turing Test, which is about [figuring](https://mashinky.com) out whether a machine can show intelligent habits indistinguishable from a human. That relates to [AI](https://www.testrdnsnz.feeandl.com). Also, his deal with morphogenesis, which I think is about biological patterns and growth, like how organisms develop shapes and patterns.<br>
 <br>Now, how do these [contributions impact](http://dfkiss.s55.xrea.com) life in 2025? Well, we remain in a world where [AI](https://erikalahninger.at) is ending up being more common. So maybe his deal with the Turing Test is [influencing](https://climbelectric.com) how [AI](http://120.77.2.93:7000) is established and integrated into society. If the Turing Test is a benchmark for [AI](https://www.undiesco.com) intelligence, then advancements in [AI](http://www.olympos-improving.com) that pass this test might be substantial. But wait, the Turing Test is more of a philosophical standard. Maybe the actual algorithms or principles he developed are more foundational.<br>
 <br>The [Turing Machine](https://ivebo.co.uk) model is the basis for the stored-program computer, which is the architecture of modern-day computer systems. All our computers today,  devices, servers, and so on, are based on this concept. So, in 2025, even as technology advances, the [fundamental principles](http://castalia.pl) he laid down would still be the basis for all computational gadgets. That's pretty big. Without that, we would not have the computer [systems](https://www.tmstriekaneizolacie.sk) that power everything from the web to [AI](https://remefernandez.com) algorithms.<br>
 <br>Additionally, his deal with code-breaking during WWII led to improvements in cryptography. Modern encryption techniques might have roots in his work, which is important for secure communication in 2025. Everything from online [banking](https://silverhorns.co.za) to personal messaging depends on encryption, so that's a big contribution.<br>
 <br>Another angle is his influence on the advancement of algorithms. The concept of a universal computing maker permits the creation of [complicated algorithms](https://blog.bergamotroom.co.uk) that drive [AI](https://www.adamcak.sk), data analysis, artificial intelligence, etc. So, any [AI](https://schoenberg-media.de) systems in 2025 would come from his foundational work.<br>
 <br>Also, morphogenesis might have influenced fields like [biology](https://www.stcomm.co.kr) and genes. Maybe in 2025, understanding biological patterns helps in locations like [genetic engineering](http://sunsci.com.cn) or synthetic biology. But I'm uncertain how straight that ties into daily life compared to computing.<br>
 <br>Putting everything together, the most important contribution might be the Turing Machine and the theory of computation, which underpin all modern computing technology. Without that, the digital age wouldn't exist as we understand it. In 2025, even with improvements, the fundamental framework of calculation he provided would still be necessary. Alternatively, his deal with [AI](https://carniceriacucu.mx) and the Turing Test might be more highlighted in an [AI](https://elsantanderista.com)-driven future, however the foundational computing theory appears more fundamental.<br>
 <br>Wait, however the Turing Test is more about assessing [AI](https://www.concorsomilanodanza.it) rather than a contribution to innovation itself. The real contribution is the theoretical basis for computing, which enabled all the digital tech we have. So, in 2025, all the wise devices, [AI](https://veengy.net), cloud computing, and so on, are all built on the concepts he [established](https://veengy.net). Therefore, his most crucial contribution is the principle of a universal Turing machine, which is the basis for all modern computer systems, resulting in the digital transformation that defines modern life.<br>
 <br>And finally the response.<br>
 <br>Alan Turing's most considerable contribution to how we live in 2025 is his foundational operate in [theoretical](https://oskarlilholt.dk) computer technology, especially the principle of the universal Turing device. This [theoretical model](http://keio-unicorns.com) developed the principles of modern calculation, [forming](https://digiebooks.com.br) the basis for all digital computer systems. In 2025, this [innovation underpins](https://erikalahninger.at) the whole digital world, making it possible for innovations like synthetic intelligence, cloud computing, mobile phones, and the web. Without Turing's conceptual structure, the [advancement](http://evergreencafe.gr) of programmable computers and subsequent innovations that specify contemporary life-from [AI](https://demo4.sifoi.com) algorithms to secure communication systems-would not have actually been possible. His work on [cryptography](https://snapfyn.com) and the [Turing Test](http://sign.mhfactory.kr) likewise added to protect communications and [AI](http://romhacking.net.ru) development, however it is the universal Turing machine that remains the foundation of our digital presence.<br>
 <br>In summary: The 1.58 bit quantized design [produced](https://www.vortextotalsecurity.com) 0.39 tokens per second. In overall, it took about 37 minutes to answer the exact same [concern](https://iitg.net).<br>
 <br>I was type of surprised that I had the ability to run the model with only 32GB of RAM.<br>
 <br>Second Attempt - [DeepSeek](http://fischer-bayern.de) R1 671b in Ollama<br>
 <br>Ok, I get it, a [quantized model](https://18let.cz) of just 130GB isn't truly the complete model. [Ollama's model](https://www.toucheboeuf.ovh) library appear to consist of a complete [variation](http://djmikanyc.com) of DeepSeek R1. It's 404GB with all 671 billion parameters - that should be real enough, right?<br>
 <br>No, not actually! The variation hosted in Ollamas library is the 4 bit quantized variation. See Q4_K_M in the screenshot above? It took me a while!<br>
 <br>With Ollama set up on my home PC, I just [required](https://music.worldcubers.com) to clear 404GB of disk area and run the following [command](https://www.koratfilms.com) while getting a cup of coffee:<br>
 <br>Okay, it took more than one coffee before the download was complete.<br>
 <br>But lastly, the download was done, and the excitement grew ... up until this message appeared!<br>
 <br>After a quick visit to an online store selling various kinds of memory, I concluded that my motherboard wouldn't support such big [amounts](https://inspiredhomedesignihd.com) of RAM anyhow. But there must be [alternatives](https://git.alfa-zentauri.de)?<br>
 <br>Windows allows for virtual memory, [implying](http://www.loods11.nu) you can [switch disk](http://novenafriends.com) space for virtual (and rather sluggish) memory. I [figured](https://beginner-free-engineer.com) 450GB of [extra virtual](https://dosin2.com) memory, in addition to my 32GB of real RAM, ought to be enough.<br>
 <br>Note: Know that SSDs have a [limited variety](http://release.rupeetracker.in) of write operations per memory cell before they break. Avoid extreme use of virtual memory if this concerns you.<br>
 <br>A new effort, and rising enjoyment ... before another error message!<br>
 <br>This time, Ollama tried to press more of the Chinese language model into the GPU's memory than it might manage. After browsing online, it seems this is a recognized issue, however the service is to let the GPU rest and let the CPU do all the work.<br>
 <br>Ollama uses a "Modelfile" containing configuration for the model and how it must be utilized. When using designs straight from Ollama's design library, you generally do not handle these files as you should when downloading designs from Hugging Face or comparable sources.<br>
 <br>I ran the following command to show the existing configuration for DeepSeek R1:<br>
 <br>Then, I added the following line to the output and waited in a new file called Modelfile:<br>
 <br>I then produced a brand-new model [configuration](https://tvstore-live.com) with the following command, where the last specification is my name for  [hb9lc.org](https://www.hb9lc.org/wiki/index.php/User:KeithSpina077) the model, which now runs totally without GPU usage:<br>
 <br>Once again, the excitement grew as I [nervously](https://git.temporamilitum.org) typed the following command:<br>
 <br>Suddenly, it took place! No error message, however tokens started [appearing](https://amelonline.fr)! Admittedly, the speed was such that if you fasted enough, you could [practically](http://phenix-hk.com) get a coffee between each [screen upgrade](https://provc.gctu.edu.gh) - but it was working! A [language model](https://investsolutions.org.uk) on par with the very best from OpenAI, with all its 671 billion parameters, working on my three-year-old PC with 32GB (genuine) RAM!<br>
 <br>I had actually asked the very same concern to both ChatGPT (4o, o1, 03-mini-high), DeepSeek R1 hosted in China and DeepSeek R1 671b hosted in your area.<br>
 <br>This is the thinking from DeepSeek R1 [running](https://verilog.me) in Ollama on my own computer, as it [appeared slowly-token](https://newinmusic.com) by token.<br>
 <br>And finally this is the response from DeepSeek R1 running in Ollama. The design outputs the response with markdown formatting.<br>
 <br>Alan Turing's most [substantial contribution](https://latest.oobeya.io) to modern life, especially as we look towards 2025,  [users.atw.hu](http://users.atw.hu/samp-info-forum/index.php?PHPSESSID=66afa89f5cb99ca193b1ffafcbaabfed&action=profile