<div id="readability-page-1" class="page"><div>


 

 <article>
 <ul>
 <li><a href="#1-install-pytorch">1. Install Pytorch</a></li>
 <li><a href="#2-tensorflow">2. Tensorflow</a></li>
 <li><a href="#3-vllm">3. vLLM</a></li>
 <li><a href="#stable-diffusion-webui">Stable Diffusion WebUI</a></li>
</ul>

 <h2 id="1-install-pytorch">1. Install Pytorch</h2>

It’s better to install Pytorch preview (nightly) version. 
I use CUDA Version 13.1

go to https://pytorch.org and find your matched version.
<ul>
 <li>PyTorch Build: Preview (Nightly)</li>
 <li>CUDA 13.0</li>
</ul>

<div><pre><code>$ pip3 install --pre torch torchvision --index-url https://download.pytorch.org/whl/nightly/cu130

# check pytorch version
$ python -c "import torch; print(torch.__version__)"
</code></pre></div>

here you can test it like this

<div><pre><code>import time
import os

# Suppress TF logs for cleaner output
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3' 

def test_pytorch():
 print("--- PyTorch Check ---")
 import torch
 
 if torch.cuda.is_available():
 device = torch.device("cuda")
 gpu_name = torch.cuda.get_device_name(0)
 vram = torch.cuda.get_device_properties(0).total_memory / 1e9
 print(f"✅ GPU Detected: {gpu_name} ({vram:.2f} GB VRAM)")
 
 # Simple Compute Test
 x = torch.randn(5000, 5000, device=device)
 y = torch.randn(5000, 5000, device=device)
 start = time.time()
 result = torch.matmul(x, y)
 torch.cuda.synchronize() # Wait for compute to finish
 print(f"✅ Matrix Mul (5k x 5k) Time: {time.time() - start:.4f}s")
 else:
 print("❌ PyTorch cannot see the GPU.")

test_pytorch()
</code></pre></div>

<h2 id="2-tensorflow">2. Tensorflow</h2>

<ul>
 <li>you need to install latest version of tensorflow “tf-nightly” with cuda option.</li>
</ul>

firstly, modify ~/.bashrc

<div><pre><code># CUDA for Tensorflow
NVIDIA_HOME="$HOME/.pyenv/versions/3.12.10/lib/python3.12/site-packages/nvidia"

# 2. 필요한 라이브러리 경로들을 LD_LIBRARY_PATH에 추가
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${NVIDIA_HOME}/cudnn/lib
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${NVIDIA_HOME}/cublas/lib
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${NVIDIA_HOME}/cufft/lib
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${NVIDIA_HOME}/curand/lib
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${NVIDIA_HOME}/cusolver/lib
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${NVIDIA_HOME}/cusparse/lib
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${NVIDIA_HOME}/nccl/lib
</code></pre></div>

<div><pre><code>$ pip install "tf-nightly[and-cuda]"

# 테스트
$ python -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
</code></pre></div>

if we use [and-cuda], it automatically downloads cudnn, cublas, … nccl. 
However if you want to install all of them manually, you can do it like this.

<div><pre><code>pip install nvidia-cudnn-cu12 \
 nvidia-cublas-cu12 \
 nvidia-cufft-cu12 \
 nvidia-curand-cu12 \
 nvidia-cusolver-cu12 \
 nvidia-cusparse-cu12 \
 nvidia-nccl-cu12
</code></pre></div>

here’s a bit more complex test.

<div><pre><code>import time
import os

def test_tensorflow():
 print("\n--- TensorFlow Check ---")
 import tensorflow as tf
 
 gpus = tf.config.list_physical_devices('GPU')
 if gpus:
 print(f"✅ GPU Detected: {len(gpus)} device(s)")
 for gpu in gpus:
 print(f" - {gpu.device_type}: {gpu.name}")
 
 # Simple Compute Test
 with tf.device('/GPU:0'):
 a = tf.random.normal([5000, 5000])
 b = tf.random.normal([5000, 5000])
 start = time.time()
 c = tf.matmul(a, b)
 _ = c.numpy() # Force execution
 print(f"✅ Matrix Mul (5k x 5k) Time: {time.time() - start:.4f}s")
 else:
 print("❌ TensorFlow cannot see the GPU.")


test_tensorflow()
</code></pre></div>

<h2 id="3-vllm">3. vLLM</h2>

<ul>
 <li>here, we need to install nightly version</li>
</ul>

<div><pre><code>$ pip install vllm --pre --extra-index-url https://wheels.vllm.ai/nightly
</code></pre></div>

<div><pre><code>python -m vllm.entrypoints.openai.api_server \
 --model moonshotai/Kimi-K2.5 \
 --tensor-parallel-size 1 \
 --gpu-memory-utilization 0.95 \
 --trust-remote-code \
 --served-model-name kimi-k2.5 \
 --max-model-len 32768
 
</code></pre></div>

<h2 id="stable-diffusion-webui">Stable Diffusion WebUI</h2>

<ul>
 <li>in venv, you need to install pytorch nightly version.</li>
</ul>

<div><pre><code>$ export STABLE_DIFFUSION_REPO=https://github.com/joypaul162/Stability-AI-stablediffusion.git
$ ./webui.sh
</code></pre></div>

 </article>

 
 

 
 
 
 
</div></div>

Incredible.AI

1. Install Pytorch
It’s better to install Pytorch preview (nightly) version.
go to https://pytorch.org and find your matched version.
PyTorch Build: Preview (Nightly)
CUDA 13.0
$ pip3 install --pre torch torchvision --index-url https://download.pytorch.org/whl/nightly/cu130
# check pytorch version
$ python -c "import torch; print(torch.__version__)"
here you can test it like this
import time
import os
# Suppress TF logs for cleaner output
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3' 
def test_pytorch():
    print("--- PyTorch Check ---")
    import torch
    
    if torch.cuda.is_available():
        device = torch.device("cuda")
        gpu_name = torch.cuda.get_device_name(0)
        vram = torch.cuda.get_device_properties(0).total_memory / 1e9
        print(f"✅ GPU Detected: {gpu_name} ({vram:.2f} GB VRAM)")
        
        # Simple Compute Test
        x = torch.randn(5000, 5000, device=device)
        y = torch.randn(5000, 5000, device=device)
        start = time.time()
        result = torch.matmul(x, y)
        torch.cuda.synchronize() # Wait for compute to finish
        print(f"✅ Matrix Mul (5k x 5k) Time: {time.time() - start:.4f}s")
    else:
        print("❌ PyTorch cannot see the GPU.")
test_pytorch()
2. Tensorflow
you need to install latest version of tensorflow “tf-nightly” with cuda option.
firstly, modify ~/.bashrc
# CUDA for Tensorflow
NVIDIA_HOME="$HOME/.pyenv/versions/3.12.10/lib/python3.12/site-packages/nvidia"
# 2. 필요한 라이브러리 경로들을 LD_LIBRARY_PATH에 추가
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${NVIDIA_HOME}/cudnn/lib
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${NVIDIA_HOME}/cublas/lib
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${NVIDIA_HOME}/cufft/lib
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${NVIDIA_HOME}/curand/lib
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${NVIDIA_HOME}/cusolver/lib
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${NVIDIA_HOME}/cusparse/lib
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:${NVIDIA_HOME}/nccl/lib
$ pip install "tf-nightly[and-cuda]"
# 테스트
$ python -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
if we use [and-cuda], it automatically downloads cudnn, cublas, … nccl.
pip install nvidia-cudnn-cu12 \
            nvidia-cublas-cu12 \
            nvidia-cufft-cu12 \
            nvidia-curand-cu12 \
            nvidia-cusolver-cu12 \
            nvidia-cusparse-cu12 \
            nvidia-nccl-cu12
here’s a bit more complex test.
import time
import os
def test_tensorflow():
    print("\n--- TensorFlow Check ---")
    import tensorflow as tf
    
    gpus = tf.config.list_physical_devices('GPU')
    if gpus:
        print(f"✅ GPU Detected: {len(gpus)} device(s)")
        for gpu in gpus:
            print(f"   - {gpu.device_type}: {gpu.name}")
            
        # Simple Compute Test
        with tf.device('/GPU:0'):
            a = tf.random.normal([5000, 5000])
            b = tf.random.normal([5000, 5000])
            start = time.time()
            c = tf.matmul(a, b)
            _ = c.numpy() # Force execution
            print(f"✅ Matrix Mul (5k x 5k) Time: {time.time() - start:.4f}s")
    else:
        print("❌ TensorFlow cannot see the GPU.")
test_tensorflow()
3. vLLM
here, we need to install nightly version
$ pip install vllm --pre --extra-index-url https://wheels.vllm.ai/nightly
python -m vllm.entrypoints.openai.api_server \
    --model moonshotai/Kimi-K2.5 \
    --tensor-parallel-size 1 \
    --gpu-memory-utilization 0.95 \
    --trust-remote-code \
    --served-model-name kimi-k2.5 \
    --max-model-len 32768
    
Stable Diffusion WebUI
in venv, you need to install pytorch nightly version.
$ export STABLE_DIFFUSION_REPO=https://github.com/joypaul162/Stability-AI-stablediffusion.git
$ ./webui.sh

Nvidia RTX 6000 Pro Blackwell Workstation 600w Settings

댓글