
Janus Pro 7B is a future of multimodal AI that’s what coders believe. What’s your thought by the way?
I tried Janus pro 7b multimodal AI on my Windows system. It worked for me and was slow compared to the online version available on Hugging Face.
Read this blog to know how to access and use janus pro 7b locally, whether trying on Ubuntu or Windows system.
Also read: Top 10 AI Infrastructure Companies In The WorldTable of Contents
There are multiple ways to access Deepseek janus pro 7b for users.
Accessing online is faster and easier as you don’t need to wait for dependencies. While accessing janus pro 7b locally is a tedious job and requires coding skills.
I have explained how to install and run DeepSeek Janus Pro 7B multimodal AI model locally on Linux Ubuntu and Windows.
Developers mostly utilize Linux operating systems for testing and development. If that’s so, here’s how to run Deepseek Janus pro 7b locally.
I have used “bash” as a command line interface.
1. Check if your GPU is detected
nvidia-smi
2. If not, install CUDA & cuDNN
sudo apt update && sudo apt upgrade -y
sudo apt install -y nvidia-driver-535
3. Reboot your system
sudo reboot
1. Ensure you have Python 3.10+ installed
sudo apt install python3.10 python3.10-venv python3-pip -y
2. Create and activate a virtual environment
python3 -m venv janus_env
source janus_env/bin/activate
1. Inside your virtual environment, install the necessary Python libraries
pip install torch torchvision torchaudio transformers accelerate auto-gptq
2. Test if CUDA is working with PyTorch
python -c “import torch; print(torch.cuda.is_available())”
If it prints True, CUDA is active
1. If the model is available on Hugging Face, download it
git lfs install
huggingface-cli login # (if needed)
mkdir models && cd models
git clone https://huggingface.co/janus/janus-pro-7b
2. Once downloaded, load the model in Python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
model_path = “models/janus-pro-7b”
tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForCausalLM.from_pretrained(model_path, device_map=”auto”, torch_dtype=torch.float16)
# Run an inference test
input_text = “Explain the importance of AI in healthcare.”
input_ids = tokenizer(input_text, return_tensors=”pt”).input_ids.cuda()
output = model.generate(input_ids, max_length=150)
print(tokenizer.decode(output[0], skip_special_tokens=True))
1. If you prefer a no-code Web UI, install text-generation-webui
git clone https://github.com/oobabooga/text-generation-webui.git
cd text-generation-webui
pip install -r requirements.txt
2. Move the model files to the models directory, then start the server
python server.py –model janus-pro-7b
👉 Access the Janus pro 7b locally on the provided url.
If you are a windows user and want to try Janus Pro 7B locally utilizing hardware specifications, follow the mentioned steps.
Pre-requisites
👉 Download Anaconda from here,
👉 Email address to receive download link, and
👉 Good hardware specifications for smooth installation
Double click the installer to begin installation on your local computer. Select the path for installation and wait until its completion.
Click on the “Environment” tab and at the bottom you will find the “Create” option. Click on it and name the environment, for example; I name it “Janus” and select Python 3.10.16 and click create.
From the dropdown menu, select “not installed” and search “git” and check mark the file name with “git – Distributed version control system” and click the apply button.
Press the play icon next to Janus and tap “open terminal”. Type “cd c:\conda” to change the repository files for saving.
Go to pytorch.org and click get started. Select Pip as a Package and Compute Platform as CUDA 12.4 because it’s good with new GPUs.
Copy the command as shown and paste it back in the terminal. This will download and install PyTorch setup and help Conda to complete the installation easily.
Search for “DeepSeek AI GitHub” on the browser and open the github website. Find “Janus” which is a publicly available repository. Click to open its repositories.
Next, click on the “Code” option and copy the clone url and get back to the terminal. Type “git clone” and paste the url.
After cloning is done. Switch the folder to the downloaded repository. To do so, type “cd janus” and enter. It will change the folder to “(Janus) c:\Conda\Janus”
Head back to the DeepSeek GitHub and click on “Janus-pro” for the installation guide. Scroll down to this section “Gradio Demo” and copy the install command line code and paste it in the terminal.
Wait a few seconds to complete the installation in the terminal. Now you are almost done.
You may see “python demo/app_Januspro.py” automatically added. It is the same code available in the DeepSeek github.
Simply press enter to continue the installation.
In some cases, you may see an error saying about OSError: We couldn’t connect to ‘https://huggingface.co’ to load this file.
Tip: Try resolving this error through this link.
After the installation completes successfully, you will have the url to run it locally in a web version.
The interface and experience is similar to the original Janus pro 7b AI model. However, the response is highly dependent on the technical specifications of the hardware you are using.
In my case, it’s a bit slow. It took 4 minutes to generate an image from text.
There is a possible way to download Janus Pro 7B using Hugging Face. Create an account and search for Janus Pro 7B model with the name of Deepseek.
Now begin downloading files by tapping into the ‘files and version’ section. Next, use Git and Transformers library to clone and later install the required dependencies using Python’s transformers.
After this, follow the usage instructions on the repository page. You might require integrating the model with Python code.
Well, Janus pro 7B is an advanced multimodal AI which locally requires some technical setup. Here’s my recommendation for smooth installation and running Janus Pro 7B locally.
GPU: NVIDIA GPU with at least 12GB VRAM, however more is better.
RAM: Minimum 16GB (preferably 32GB for handling larger workloads)
Storage: At least 30GB of free space.
Though I have lower specifications, I have noticed big lags and longer waiting time for image generations. Therefore, before thinking of trying it locally, you must have recommended specifications available.
That’s all in this blog. Thanks for reading 🙂
Author’s Recommendation:
👉 Architecture Of Janus Pro 7B
👉 Janus Pro 7B vs Stable Diffusion
Yes, by setting up janus pro 7b locally you may use it actually without accessing online on hugging face or its official website.
By using github repository you can run Janus pro 7b on linux-based operating system as well as Windows based.
Yes, knowledge of Python and basic coding skills are required to install Janus pro 7b offline.
The response rate is usually slower because of the restricted hardware specification. It requires a heavy hardware system, for example 16GB of vRAM and 30+ GB of disk space and i7 or Ryzen 7 processor.
Sources:
https://github.com/deepseek-ai
https://pytorch.org/
https://huggingface.co/deepseek-ai
Disclaimer: The information written on this article is for education purposes only. We do not own them or are not partnered to these websites. For more information, read our terms and conditions.
FYI: Explore more financial tips and tricks here. For more tech tips and quick solutions, follow our Facebook page, for AI-driven insights and guides, follow our LinkedIn page.
Tuesday February 11, 2025
Wednesday January 22, 2025
Monday December 23, 2024
Friday December 20, 2024
Tuesday November 19, 2024
Tuesday November 12, 2024
Tuesday November 5, 2024
Monday October 21, 2024
Monday October 7, 2024
Friday September 20, 2024