本地部署大語言模型-DeepSeek

DeepSeek 是國內頂尖 AI 團隊「深度求索」開發的多模態大模型,具備數學推理、代碼生成等深度能力,堪稱"AI界的六邊形戰士"。

Hostease AMD 9950X/96G/3.84T NVMe/1G/5IP/RTX4090?GPU服務器提供多種計費模式。

DeepSeek-R1-32B配置

配置項

規格要求

CPU

16 核以上(如 AMD Ryzen 9950)

內存

96GB

硬盤

960GB

顯卡

24GB+ 顯存(如 RTX 4090)

# cat cpuinfo |grep name|head -1
model name?? ?: AMD Ryzen 9 9950X 16-Core Processor

# nvidia-smi?
Sun Mar ?2 14:51:30 2025 ? ? ??
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.183.01 ? ? ? ? ? ? Driver Version: 535.183.01 ? CUDA Version: 12.2 ? ? |
|-----------------------------------------+----------------------+----------------------+
| GPU ?Name ? ? ? ? ? ? ? ? Persistence-M | Bus-Id ? ? ? ?Disp.A | Volatile Uncorr. ECC |
| Fan ?Temp ? Perf ? ? ? ? ?Pwr:Usage/Cap | ? ? ? ? Memory-Usage | GPU-Util ?Compute M. |
| ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? | ? ? ? ? ? ? ? ? ? ? ?| ? ? ? ? ? ? ? MIG M. |
|=========================================+======================+======================|
| ? 0 ?NVIDIA GeForce RTX 4090 ? ? ? ?Off | 00000000:01:00.0 Off | ? ? ? ? ? ? ? ? ?Off |
| ?0% ? 27C ? ?P8 ? ? ? ? ? ? ? 5W / 450W | ? ?495MiB / 24564MiB | ? ? ?0% ? ? ?Default |
| ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? | ? ? ? ? ? ? ? ? ? ? ?| ? ? ? ? ? ? ? ? ?N/A |
+-----------------------------------------+----------------------+----------------------+
安裝步驟(略寫)

apt update
apt install nvidia-driver-535 -y
curl -fsSL https://ollama.com/install.sh | sh
ollama run deepseek-r1:32b-qwen-distill-q4_K_M --verbose

創建對話返回結果

>>> which is bigger,9.8 or 9.11?
<think>
Alright, so I'm trying to figure out whether 9.8 is bigger than 9.11 or not. At first glance, it seems like both numbers are pretty close to each 
other since they both start with a 9 and then have some decimal parts. But wait, maybe there's something more to it.Okay, let me break this down step by step. First, I need to understand what these decimals represent. So, 9.8 can be read as nine plus eight 
tenths, right? And 9.11 is nine plus eleven hundredths. Hmm, that makes sense. Now, the question is asking which one is bigger.Maybe it would help if I line them up vertically to compare the digits after the decimal point more easily:```9.809.11
```By adding a zero to 9.8 to make it 9.80, we can see that both numbers have two decimal places now. This should make the comparison straightforward.Now, let's look at the tenths place first:- For 9.80, the tenths digit is 8.
- For 9.11, the tenths digit is 1.Okay, so right here, it seems like 8 is greater than 1. That would mean that 9.80 is bigger because its tenths place is larger than the tenths 
place of 9.11.But wait a second—is there something else I should consider? Maybe rounding or anything like that?Well, since both numbers are already given to two decimal places and we've added an extra zero to make them consistent, I think it's safe to say 
that no further adjustment is necessary.Still, just to be thorough, let me convert these decimals into fractions to see if there's a difference in their values.For 9.80:- 0.80 as a fraction is the same as \( \frac{80}{100} \), which simplifies down to \( \frac{4}{5} \).So, 9.80 is equal to \( 9 + \frac{4}{5} = \frac{45}{5} + \frac{4}{5} = \frac{49}{5} \).For 9.11:- 0.11 as a fraction is approximately \( \frac{11}{100} \).So, 9.11 is equal to \( 9 + \frac{11}{100} = \frac{900}{100} + \frac{11}{100} = \frac{911}{100} \).Now, let's compare these two fractions:- \( \frac{49}{5} = 9.8 \)
- \( \frac{911}{100} = 9.11 \)To see which one is bigger, I can convert both to have the same denominator or find a common basis for comparison.Let's try finding equivalent fractions with the same denominator:The denominators are 5 and 100. Since 100 is a multiple of 5 (specifically, \( \frac{100}{5} = 20 \)), I'll convert both fractions to have a 
denominator of 100.For \( \frac{49}{5} \):- Multiply numerator and denominator by 20: \( \frac{49 \times 20}{5 \times 20} = \frac{980}{100} \).So, \( \frac{49}{5} = \frac{980}{100} \).Now we have:- \( \frac{980}{100} \) (which is 9.8)
- \( \frac{911}{100} \) (which is 9.11)Comparing the numerators directly, since both fractions have the same denominator:- 980 > 911This means that \( \frac{980}{100} > \frac{911}{100} \), or in other words, 9.8 is greater than 9.11.Wait a minute—this contradicts what I thought earlier where the tenths place indicated that 9.8 was bigger because of having an 8 instead of a 1 in 
the tenths place. But when converting to fractions with the same denominator, it's clear that even though both numbers have nine as their whole 
number part, the decimal parts make a difference.But according to my calculations:- 9.8 is larger than 9.11 because \( \frac{49}{5} \) (or \( \frac{980}{100} \)) is greater than \( \frac{911}{100} \).So, this further confirms that 9.8 is bigger.Is there anything else I need to consider? Maybe looking at the hundredths place?Well, since we've already converted both numbers to have the same denominator and confirmed that 9.8 has a larger numerator in the hundredths place 
as well (\( \frac{980}{100} \) vs. \( \frac{911}{100} \)), I think we can confidently say that 9.8 is indeed bigger than 9.11.Additionally, thinking about it on a number line might help visualize this:- If you imagine the numbers from 9 to 10:- 9.8 would be closer to 10.- 9.11 would be much closer to 9.Therefore, 9.8 is further along towards 10 and thus larger than 9.11.But just to ensure there's no mistake in the fraction conversion, let me double-check:- \( \frac{49}{5} = 9.8 \) since \( 49 ÷ 5 = 9.8 \).That seems correct because 5 times 9 is 45, and 49 minus 45 leaves a remainder of 4, so \( \frac{4}{5} = 0.8 \), making it 9.8.And for \( \frac{911}{100} \):- \( 911 ÷ 100 = 9.11 \).Yes, that's accurate because dividing by 100 moves the decimal point two places to the left.So everything checks out mathematically.Is there any alternative way to think about this?Perhaps considering money as an analogy:- If you have $9.80 and $9.11:- You'd obviously have more with $9.80 than with $9.11.This real-world scenario aligns with the mathematical conclusion that 9.8 is bigger.Alternatively, think of it in terms of measurements:- If something measures 9.8 centimeters and another measures 9.11 centimeters:- The first one would be longer than the second one.All these analogies reinforce the idea that 9.8 is indeed larger than 9.11.So, after carefully analyzing through different methods—decimal comparison, fraction conversion, number line visualization, and real-life 
analogies—it becomes clear that 9.8 is bigger than 9.11.
</think>To determine which number is larger between \( 9.8 \) and \( 9.11 \), we can follow these steps:1. **Understand the Numbers**:- \( 9.8 \) is equivalent to \( 9.80 \).- Both numbers have a whole number part of 9.2. **Compare the Decimal Parts**:- The decimal part of \( 9.8 \) (or \( 9.80 \)) is 0.80.- The decimal part of \( 9.11 \) is 0.11.- Since \( 0.80 > 0.11 \), the decimal part of \( 9.8 \) is larger.3. **Conclusion**:- Therefore, \( 9.8 \) is bigger than \( 9.11 \).Final Answer:  
\[
\boxed{9.8}
\]eval count:           501 token(s)
eval duration:        12.478s
eval rate:            40.15 tokens/s>>> 

Web頁面返回對話,2個數字比大小結果

本文來自互聯網用戶投稿,該文觀點僅代表作者本人,不代表本站立場。本站僅提供信息存儲空間服務,不擁有所有權,不承擔相關法律責任。
如若轉載,請注明出處:http://www.pswp.cn/news/896755.shtml
繁體地址,請注明出處:http://hk.pswp.cn/news/896755.shtml
英文地址,請注明出處:http://en.pswp.cn/news/896755.shtml

如若內容造成侵權/違法違規/事實不符,請聯系多彩編程網進行投訴反饋email:809451989@qq.com,一經查實,立即刪除!

相關文章

[AI機器人] Web-AI-Robot機器人前瞻版--比奇堡海之霸凱倫

文章目錄 簡述開源Web-AI-Robot 項目-比奇堡-海之霸-凱倫 技術架構效果預覽 簡述 本項目配合前端項目bikini_bottom_karen_ui運行&#xff0c;來源于柒杉工作室&#xff08;截止2025.2&#xff0c;目前我自己&#xff09;。 打造一個只需要在瀏覽器上運行的AI智能機器人&#…

250302-綠聯NAS通過Docker配置SearXNG及適配Open-WebUI的yaml配置

A. 配置Docker中的代理 綠聯NAS簡單解決docker無法獲取鏡像-不用軟路由 - 嗶哩嗶哩 B. 下載官網對應的鏡像 群暉NAS用docker搭建SearXNG元搜索引擎_嗶哩嗶哩_bilibili C. 修改默認省略的參數&#xff0c;只配置Base_URL&#xff0c;刪除其它默認的空缺項 searxng-docker/REA…

java容器 LIst、set、Map

Java容器中的List、Set、Map是核心數據結構&#xff0c;各自適用于不同的場景 一、List&#xff08;有序、可重復&#xff09; List接口代表有序集合&#xff0c;允許元素重復和通過索引訪問&#xff0c;主要實現類包括&#xff1a; ArrayList 底層結構&#xff1a;動態數組…

3471. 找出最大的幾近缺失整數

3471. 找出最大的幾近缺失整數 class Solution:# 輔助方法&#xff0c;判斷第三種情況&#xff0c;只有首位兩個元素有可能為最大幾近缺失數def f(self,nums,x):return -1 if x in nums else xdef largestInteger(self, nums: List[int], k: int) -> int:n len(nums)if k …

【異常錯誤】No module named ‘taming.modules.vqvae‘

錯誤&#xff1a; File "/mnt/d/Pycharm_workspace/text2image/OmniGen-version/OmniGen/latentDiffusion/ldm/models/autoencoder.py", line 6, in <module> from taming.modules.vqvae.quantize import VectorQuantizer2 as VectorQuantizer ModuleNotF…

快檢查達夢庫怎么了

扁鵲的弟弟來了 要求5分鐘定位達夢數據庫問題 #!/bin/bash## content 實例個數 告警日志 實例狀態 用戶連接 活動會話 鎖 集群狀態 服務狀態 磁盤空間 cpu mem 偵聽及日志 ## scope 單機、DW、DSC Linux 多實例 ## example 將腳本保存為d.sh&#xff0c;用root用執行&#…

C++20中`constexpr`的顯著增強

文章目錄 1. **更多標準庫函數支持constexpr**2. **支持動態內存分配**3. **支持虛函數和多態**4. **支持try-catch異常處理**5. **更靈活的控制流**6. **支持std::initializer_list**7. **支持修改union活躍成員**8. **允許更多類型的非類型模板參數**總結 C20對 constexpr進…

Tomcat 亂碼問題徹底解決

1. 終端亂碼問題 找到 tomcat 安裝目錄下的 conf —> logging.properties .修改ConsoleHandler.endcoding GBK &#xff08;如果在idea中設置了UTF-8字符集&#xff0c;這里就不需要修改&#xff09; 2. CMD命令窗口設置編碼 參考&#xff1a;WIN10的cmd查看編碼方式&…

以太坊測試網

文章目錄 什么是測試網如何使用測試網獲取測試以太幣 什么是測試網 測試網&#xff08;Testnet&#xff09;是一個模擬以太坊主網&#xff08;Mainnet&#xff09;行為的區塊鏈網絡。它允許開發人員和用戶在不使用真實資金的情況下測試智能合約和應用程序。雖然測試網上的代幣…

算法隨筆_62: 買賣股票的最佳時機

上一篇:算法隨筆_61:二進制求和-CSDN博客 題目描述如下: 給定一個數組 prices &#xff0c;它的第 i 個元素 prices[i] 表示一支給定股票第 i 天的價格。 你只能選擇 某一天 買入這只股票&#xff0c;并選擇在 未來的某一個不同的日子 賣出該股票。設計一個算法來計算你所能獲…

騰訊混元文生圖大模型(Hunyuan-DiT)與Stable Diffusion(SD)對比分析

騰訊混元文生圖大模型&#xff08;Hunyuan-DiT&#xff09;與Stable Diffusion&#xff08;SD&#xff09;對比分析 騰訊混元文生圖大模型&#xff08;Hunyuan-DiT&#xff09;與Stable Diffusion&#xff08;SD&#xff09;作為當前文生圖領域的兩大代表模型&#xff0c;各自…

【HTML—前端快速入門】HTML 基礎

準備工作 vscode下載 百度網盤 Subline Text 下載 Sublime Text下載 百度網盤 vscode 下載 Sublime Text 是一款輕量好用的文本編輯器&#xff0c;我們在寫前端代碼時&#xff0c;使用 Sublime Text 打開比使用記事本打開&#xff0c;得到的代碼體驗更好&#xff0c;比 vscode…

基于單片機的GPS定位系統設計

1 系統硬件 1.1單片機模塊 單片機的種類和型號可以說是有成百上千種&#xff0c;很多大的公司和企業都生產開發自己的單片機芯片&#xff0c;并且廣泛應用于各種產品。Intel、 philips、 摩托羅拉、凌陽、宏晶等等種類繁多。大體上可以分為51系列單片機和非51系列單片機。 其…

對大模型輸出的 logits 進行處理,從而控制文本的生成

對大模型輸出的 logits 進行處理&#xff0c;從而控制文本的生成 flyfish 在文本生成任務中&#xff0c;模型輸出的 logits 代表了每個詞被選為下一個生成詞的未歸一化概率得分。通過對 logits 進行處理&#xff0c;可以精確地控制文本的生成 基本原理 在每一步生成過程中&…

Reids緩存穿透、緩存雪崩和緩存擊穿

Redis緩存中常見的三個問題&#xff1a;緩存穿透、緩存雪崩和緩存擊穿。這些問題在使用Redis作為緩存時經常遇到&#xff0c;但通過合理的策略可以有效解決。我會用簡單易懂的方式來講解&#xff0c;幫助你理解這些問題的原理和解決方案。 1. 緩存穿透 1.1 什么是緩存穿透&…

附錄-Python — 包下載緩慢,配置下載鏡像

1??命令行配置 pip config set global.index-url http://mirrors.aliyun.com/pypi/simple/ pip config set install.trusted-host mirrors.aliyun.com 2??配置文件配置 1、打開文件夾&#xff0c;輸入 %APPDATA% 回車 2、打開 %APPDATA% 路徑&#xff0c;并在此路徑下新建…

VS 2019 免費版 下載與安裝 教程說明

推薦大家直接轉到第13步&#xff0c;點擊鏈接即可下載VS2019版本 1.VS官網 2.登錄賬號 3.在搜索欄輸入“2019” 4.點擊2019這個標題 5.點擊“下載” 6.選擇合適的版本下載 7.打開下載文件&#xff08;若下載過程總是轉圈圈&#xff0c;則換個網絡下載即可&#xff09; 8.安…

介紹 torch-mlir 從 pytorch 生態到 mlir 生態

一、引言 The Torch-MLIR project provides core infrastructure for bridging the PyTorch ecosystem and the MLIR ecosystem. For example, Torch-MLIR enables PyTorch models to be lowered to a few different MLIR dialects. Torch-MLIR does not attempt to provide a…

Java并發編程之ConcurrentHashMap的原理和使用

ConcurrentHashMap(CHM)是Java為解決高并發場景下哈希表性能瓶頸而設計的線程安全容器,其核心目標在于: 線程安全?:避免多線程操作導致的數據不一致問題?;高吞吐量?:通過細粒度鎖和無鎖化設計降低線程競爭?;動態擴展?:支持自動擴容與數據結構優化(如鏈表轉紅黑樹…

AbMole揭秘傷口愈合:IGF-1-SP1-CD248信號通路的新發現

科學家們揭示了一條新的信號通路——IGF-1-SP1-CD248&#xff0c;這一發現為理解傷口愈合障礙提供了新的視角&#xff0c;并為未來的研究開辟了新方向。 研究背景 糖尿病患者的傷口愈合是一個長期存在的挑戰。據統計&#xff0c;約15%的糖尿病患者會遭受慢性傷口的困擾&#…