Unsloth aur Docker ke saath LLMs ko Fine-Tune Karein

Tutorial • Docker Fine-Tuning

Unsloth aur Docker ke saath LLMs ko Fine-Tune Kaise Karein 🐳

Local training aksar dependency issues ya environment break hone ki vajah se complex ho sakti hai. Unsloth ki Docker image in sab pareshaniyon se bachne ka ek shandar tarika hai. Koi setup ki zaroorat nahi: bas image ko pull karo, run karo, aur training shuru karo.

Aap ab hamari main Docker image unsloth/unsloth ko Blackwell aur 50-series GPUs ke liye bhi use kar sakte hain - alag se image ki zaroorat nahi hai.

⚡ Step-by-Step Tutorial

Step 1: Docker aur NVIDIA Container Toolkit Install Karein

Pehle Docker install karein Linux ya Desktop ke liye. Phir, NVIDIA Container Toolkit install karein:

export NVIDIA_CONTAINER_TOOLKIT_VERSION=1.17.8-1
sudo apt-get update && sudo apt-get install -y \
  nvidia-container-toolkit=${NVIDIA_CONTAINER_TOOLKIT_VERSION} \
  nvidia-container-toolkit-base=${NVIDIA_CONTAINER_TOOLKIT_VERSION} \
  libnvidia-container-tools=${NVIDIA_CONTAINER_TOOLKIT_VERSION} \
  libnvidia-container1=${NVIDIA_CONTAINER_TOOLKIT_VERSION}

Step 2: Container ko Run Karein

Unsloth ki official Docker image unsloth/unsloth hai. Isse run karne ke liye neeche di gayi command use karein. Yeh command aapke current directory ke `work` folder ko container ke andar map kar degi.

docker run -d -e JUPYTER_PASSWORD="mypassword" \
  -p 8888:8888 -p 2222:22 \
  -v $(pwd)/work:/workspace/work \
  --gpus all \
  unsloth/unsloth

Step 3: Jupyter Lab Access Karein

Apne browser me http://localhost:8888 par jayein. Aapko Jupyter Lab ka interface dikhega. Yahan aapko `unsloth-notebooks` folder milega jismein fine-tuning ke liye ready-to-use notebooks hain.

Step 4: Unsloth ke saath Training Shuru Karein

Ab aap kisi bhi notebook ko open karke training shuru kar sakte hain. Agar aap naye hain, to hamare Fine-tuning Guide ko follow karein.

⚙️ Advanced Settings

Aap environment variables ka use karke container ko configure kar sakte hain:

Variable Description Default
JUPYTER_PASSWORD Jupyter Lab access karne ke liye password set karein. unsloth
JUPYTER_PORT Container ke andar Jupyter Lab jis port par chalega. 8888
SSH_KEY SSH access ke liye apni public key yahan add karein. None
USER_PASSWORD Container ke andar `sudo` commands ke liye 'unsloth' user ka password. unsloth

Important: Apne kaam ko save karne ke liye hamesha volume mounts (-v) ka use karein, warna container restart hone par data delete ho jayega.

← Back to Fine-Tuning Projects