https://github.com/pelayo-felgueroso/tensorflow-gpu-setup
Step-by-step guide to installing TensorFlow with GPU support on Conda.
https://github.com/pelayo-felgueroso/tensorflow-gpu-setup
artificial-intelligence cuda deep-learning gpu machine-learning nvidia nvidia-gpu setup-guide tensorflow
Last synced: about 2 months ago
JSON representation
Step-by-step guide to installing TensorFlow with GPU support on Conda.
- Host: GitHub
- URL: https://github.com/pelayo-felgueroso/tensorflow-gpu-setup
- Owner: pelayo-felgueroso
- Created: 2025-03-16T09:36:09.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2025-03-16T09:42:21.000Z (3 months ago)
- Last Synced: 2025-04-06T19:39:31.930Z (about 2 months ago)
- Topics: artificial-intelligence, cuda, deep-learning, gpu, machine-learning, nvidia, nvidia-gpu, setup-guide, tensorflow
- Homepage:
- Size: 5.86 KB
- Stars: 2
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# tensorflow-gpu-setup
Step-by-step guide to installing TensorFlow with GPU support on Conda.
**After days of struggling, I finally found a solution that works.**
I've seen countless Reddit and YouTube posts from people saying that **TensorFlow won’t run on their GPU**, and that tutorials don’t work due to **version conflicts**. Many guides are outdated or miss crucial details, leading to frustration.
After experiencing the same issues, I found a solution using **Python virtual environments**. This ensures TensorFlow runs in an **isolated** setup, fully compatible with **CUDA and cuDNN**, while preventing conflicts with other projects.
**My specs:**
* **OS:** Windows 11
* **CPU:** Intel Core i7-11800H
* **GPU:** Nvidia GeForce RTX 3060 Laptop GPU
* **Driver Version:** 572.16
* **RAM:** 16GB
* **Python Version:** 3.12.6 (global) but using **Python 3.10** in Conda
* **CUDA Version:** 12.3 (global) but using **CUDA 11.2** in Conda
* **cuDNN Version:** 8.1
# Step-by-Step Installation:
# 1. Install Miniconda (if you don’t have it)
Download `.exe` file:
[Miniconda3 Windows 64-bit](https://repo.anaconda.com/miniconda/Miniconda3-latest-Windows-x86_64.exe)
Or Download the Miniconda installer by yourself here:
[Miniconda installer link](https://docs.anaconda.com/miniconda/)
During installation, DO NOT check "Add Miniconda to PATH" to avoid conflicts with other Python versions.
Complete the installation and restart your computer.
After installing Miniconda, open **CMD** or **PowerShell** and run:
conda --version
If you see something like:
conda 25.1.1
Miniconda is installed correctly.
# 2. Create a Virtual Environment with Python 3.10
Open **Anaconda Prompt** or **PowerShell** and run:
conda create --name tf-2.10 python=3.10
Once created, initiate it:
conda init
Once initiated, close and reopen **PowerShell,** then activate it:
conda activate tf-2.10
# 3. Fix NumPy Version to Avoid Import Errors
TensorFlow 2.10 does **not** support NumPy 2.x. If you installed it already, downgrade it:
pip install numpy==1.23.5
# 4. Install TensorFlow 2.10 (Compatible with GPU)
pip install tensorflow==2.10
**Note:** Newer TensorFlow versions **(2.11+) dropped support for CUDA 11**, so **2.10 is the last version that supports it!**
# 5. Install Correct CUDA and cuDNN Versions
TensorFlow 2.10 **requires CUDA 11.2 and cuDNN 8.1**. Install them inside Conda:
conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1
# 6. Verify Installation
Run this in Python:
import tensorflow as tf
print("TensorFlow version:", tf.__version__)
print("GPUs available:", tf.config.list_physical_devices('GPU'))
Expected Output:
TensorFlow version: 2.10.0
GPUs available: [PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]
If the **GPU list is empty (**`[]`**)**, TensorFlow is running on the **CPU**. Try restarting your terminal and running again.
# 7. (Optional) Set Up TensorFlow in PyCharm
If you're using **PyCharm**, you need to manually add the Conda environment:
1. Go to **File > Settings > Project: > Python Interpreter**.
2. Click **Add Interpreter > Add Local Interpreter**.
3. Select **Existing Environment** and browse to: C:\\Users\\\\miniconda3\\envs\\tf-2.10\\python.exe
4. Select your **Environment** : tf-2.10
5. Click **OK**.
# Done!