Dria Node Update || DeepSeek & OpenAI Model Setup

5 min readFeb 2, 2025

Overview: Dria is the only synthetic data infrastructure that you can balance data quality, diversity, and complexity all together in a single interface.

Note: This guide explains how to update the Dria node and choose the best model for frequent pings/steps/requests. Check out my previous article below.)

Dria node Tutorial

Dria is the only synthetic data infrastructure that you can balance data quality, diversity, and complexity all…

cryptonodehindi.medium.com

Requirements:

Hardware:

To learn about hardware specifications such as required CPU and RAM, please refer to node specifications.
As per the hardware requirements, you need an optimized VPS with a good CPU and sufficient cores. Since you are running the Ollama and Deepseek models, TP should be more than 25.
You can buy VPS4 and VDS from Contabo. Check out my affiliate link below — it will redirect you to the Contabo page. No setup is required.
Affiliate link: https://linktr.ee/rajeshCNH

VPS with the Best Price-to-Performance Ratio in the US | Contabo

VPS with a lot of computing power for less. All instances have truckloads of RAM (starts from 6 GB), disk space, and 32…

www.dpbolvw.net

Software:

Depending on your chosen AI models, you may need to install some software. However, some models don’t require installation — just an API key.

Note: Kindly use the paid versions of OpenAI and OpenRouter. Just top up around $5.

Gemini(FREE): Powered by Google’s API. Get your Google API here.
OpenRouter(Paid):Get the API here.
OpenAI(Paid): Get your API here.

How to choose the Best model?

Just visit the DRIA Synthetic Data dashboard and check which model has completed the highest number of tasks, as shown in the image below

As per the current status on the above dashboard, these models are at the top of the list, completing the highest number of tasks.
llama3.1:8b-instruct-fp16 — Required GPU
qwen2.5:7b-instruct-fp16 — Required High-Performance CPU
llama3.1:70b-instruct-q4_0 — Required High-Performance CPU
gemini-1.5-flash — Google Free.
deepseek-r1:1.5b — Required above 20 TPS- try it in VPS3 or VPS4
llama3.2:1b — Required High-Performance CPU
gpt-4o — OpenAI (Paid)
o1-mini — Open AI (Paid)
qwen/qwq-32b-preview — OpenRouter(PAID)

Kindly choose the model according to the latest task completion data.

Let’s start the setup (New-Setup):

Note: This guide is for new users only. If you are already running, check the updated guide below for existing users.

Step 1️⃣: Execute the below-automated script which has created by me:

wget -q https://raw.githubusercontent.com/CryptonodesHindi/Automated_script/refs/heads/main/driacnh.sh&& chmod +x driacnh.sh&&./driacnh.sh

Once the script is completed. Then follow the next step.

Step 2️⃣: Create a dedicated screen.

screen -S dria

Step 3️⃣: Navigate to the directory:

cd dkn-compute-node

Run Dria Node:

Option 1: Go with the default and choose the model manually.

./dkn-compute-launcher

Option 2: Go with the selected model I prefer based on the highest number of task completions.

./dkn-compute-launcher -m=gpt-4o -m=deepseek-r1:1.5b -m=gemini-1.5-flash -m=o1-mini -m=qwen/qwq-32b-preview -m=qwen/qwen-2.5-72b-instruct

Once executed, it will prompt you to enter your DKN wallet Secret Key (your MetaMask private key without the 0x).

After that, it will prompt you to enter your OpenAI or OpenRouter API key. Just provide it and proceed.
If it asks for Jina or other API keys, those are optional — just press Enter to skip them.
Now your node will start Downloading Model files and Testing them each model must pass its test and it only depends on your system specification

Now, check the logs to verify if it is receiving pings or not.
Finally, safely exit the screen by pressing Ctrl + A + D 🔴

Already running? Just update. New users ignore:

Step 1️⃣:
Go to the screen and stop the script by pressing Ctrl + C 🔴

screen -r dria

Step 2️⃣:
Remove the directory and delete all data 🗑. Run the below command:

cd ..&& rm -rf dkn-compute-no*

Note: You can reboot your VPS to avoid any errors during setup. However, this will impact your running node and Docker containers.

Step 3️⃣: Execute the script:

wget -q https://raw.githubusercontent.com/CryptonodesHindi/Automated_script/refs/heads/main/driacnh.sh&& chmod +x driacnh.sh&&./driacnh.sh

Step 4️⃣: Navigate to the directory:

cd dkn-compute-node

Step 5️⃣: Run/start the Dria node:

./dkn-compute-launcher -m=gpt-4o -m=deepseek-r1:1.5b -m=gemini-1.5-flash -m=o1-mini -m=qwen/qwq-32b-preview -m=qwen/qwen-2.5-72b-instruct

Now your node will start Downloading Model files and Testing them each model must pass its test and it only depends on your system specification

Now, check the logs to verify if it is receiving pings or not.
Finally, safely exit the screen by pressing Ctrl + A + D 🔴

Ollama status:

You can check the status of your Ollama by using the command below.
Make sure it is running. If it shows ‘exited,’ it means your Ollama is not running.