🚀 EasyTrain: Minimalist LLM Training & Inference Toolkit

EasyTrain is a lightweight, cross-platform CLI tool that makes training and running large language models (LLMs) as simple as a single command. Designed for developers who value simplicity over complexity.

---

🔑 Key Features

• ✅ Minimalist Design
  Just python3 EasyTrain.py runlm [...] or python3 EasyTrain.py trainlm [...]

• 🧠 LLM Support

  • Run: LLaMA, Mistral, Phi, Falcon, etc.
  • Train: LoRA/PEFT support for efficient fine-tuning
  • CUDA acceleration (when available)

• 🌍 Cross-Platform

  • Works on Linux, macOS, and Windows (NVIDIA GPUs only)
  • AMD GPU support on Linux only (via ROCm)

• 🧪 Smart Defaults

  • Automatic dataset format detection
  • Built-in token streaming
  • PEFT/LoRA support for efficient training

• 🔧 Modular Architecture

  • Clean Python modules for extension
  • Importable components:

    from text import textgenerationmodel
    from text.train import trainllm

---

🧰 System Requirements

Component	Minimum Requirement
OS	Linux/macOS/Windows (Linux preferred)
Python	3.8+
GPU (Optional)	CUDA-capable NVIDIA GPU (for speed), or AMD ROCm-capable GPU
Disk Space	20GB+ for models (varies by model)

Note: AMD ROCm support only available on Linux.

---

🛠️ Installation

🐧 Linux / macOS

# Make installer executable
chmod +x autoinstall.sh

# Run installer
./autoinstall.sh

🪟 Windows

Currently no AMD GPU support. For NVIDIA GPUs:

# Install CUDA dependencies manually
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

# Install EasyTrain dependencies
pip install transformers datasets accelerate peft

🚀 Quick Start Run a Model

python3 EasyTrain.py runlm \
  --model meta-llama/Llama-2-7b-hf \
  --prompt "Explain quantum computing in simple terms" \
  --stream

Train with LoRA

python3 EasyTrain.py trainlm \
  --model meta-llama/Llama-2-7b-hf \
  --dataset training_data.jsonl \
  --use-peft \
  --lora-rank 64 \
  --lora-alpha 128 \
  --output-dir ./fine_tuned_model

📦 CLI Arguments runlm Mode (Inference) Argument Default Description --model REQUIRED Model name/path --prompt REQUIRED Input text --system-prompt "You are an AI assistant" Context prompt --temperature 0.7 Sampling temperature --stream False Stream tokens in real-time trainlm Mode (Training) Argument Default Description --model REQUIRED Base model name/path --dataset REQUIRED Dataset path (JSON/JSONL/CSV/TXT) --use-peft False Enable PEFT/LoRA training --lora-rank 8 LoRA attention dimension --lora-alpha 16 LoRA scaling parameter --training-steps 400 Total training steps (min 400) 📁 Project Structure

EasyTrain/ ├── EasyTrain.py # Main CLI interface ├── text/ │ ├── textgenerationmodel.py # Inference model │ └── train/ │ └── trainllm.py # Training module ├── autoinstall.sh # Environment setup script └── README.md # This file

📚 Credits

Built with 🧠 HuggingFace Transformers Optimized with 🔧 PEFT (LoRA) Streamed with 📡 TextStreamer

📜 License

MIT License - See LICENSE for details 🔭 Contributing

Contributions are welcome! Whether you want to:

Add Windows AMD support when available Improve dataset format detection Add new training features Optimize inference speed

COMMON ISSUES

1. When running autoinstall.sh, the script might sometimes falsely detect a gpu (at least on my laptop) which I have not found a way around yet but here is the fix, if you run cpu only: For apt:

sudo apt install python3 python3-pip
pip3 install -r requirements.txt && pip3 install torch (--break-system-requirements flag if needed)

dnf:

sudo dnf install python3 python3-pip
pip3 install -r requirements.txt && pip3 install torch (--break-system-requirements flag if needed)

yum:

sudo yum install python3 python3-pip
pip3 install -r requirements.txt && pip3 install torch (--break-system-requirements flag if needed)

pacman:

sudo pacman -Syu install python3 python3-pip
pip3 install -r requirements.txt && pip3 install torch (--break-system-requirements flag if needed)

MacOS:

brew install python
pip3 install -r requirements.txt && pip3 install torch (--break-system-requirements flag if needed)

2. Sometimes AMD ROCm fails to install due to bad requests and stuff. It may also happen if you are using a lesser known distro like Oracle Linux or Azure Linux. To install fresh ROCm:

Ubuntu 24.04:

wget https://repo.radeon.com/amdgpu-install/6.4.1/ubuntu/noble/amdgpu-install_6.4.60401-1_all.deb
sudo apt install ./amdgpu-install_6.4.60401-1_all.deb
sudo apt update
sudo apt install python3-setuptools python3-wheel
sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
sudo apt install rocm

Ubuntu 22.04:

wget https://repo.radeon.com/amdgpu-install/6.4.1/ubuntu/jammy/amdgpu-install_6.4.60401-1_all.deb
sudo apt install ./amdgpu-install_6.4.60401-1_all.deb
sudo apt update
sudo apt install python3-setuptools python3-wheel
sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
sudo apt install rocm

Debian 12:

wget https://repo.radeon.com/amdgpu-install/6.4.1/ubuntu/jammy/amdgpu-install_6.4.60401-1_all.deb
sudo apt install ./amdgpu-install_6.4.60401-1_all.deb
sudo apt update
sudo apt install -y python3-setuptools python3-wheel
sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
sudo apt install rocm

Red Hat 9.6:

sudo dnf install https://repo.radeon.com/amdgpu-install/6.4.1/rhel/9.6/amdgpu-install-6.4.60401-1.el9.noarch.rpm
sudo dnf clean all
wget https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm
sudo rpm -ivh epel-release-latest-9.noarch.rpm
sudo dnf install dnf-plugin-config-manager
sudo crb enable
sudo dnf install python3-setuptools python3-wheel
sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
sudo dnf install rocm

Red Hat 9.5:

sudo dnf install https://repo.radeon.com/amdgpu-install/6.4.1/rhel/9.5/amdgpu-install-6.4.60401-1.el9.noarch.rpm
sudo dnf clean all
sudo dnf install "kernel-headers-$(uname -r)" "kernel-devel-$(uname -r)" "kernel-devel-matched-$(uname -r)"
sudo dnf install amdgpu-dkms

Red Hat 9.4:

sudo dnf install https://repo.radeon.com/amdgpu-install/6.4.1/rhel/9.4/amdgpu-install-6.4.60401-1.el9.noarch.rpm
sudo dnf clean all
sudo dnf install "kernel-headers-$(uname -r)" "kernel-devel-$(uname -r)" "kernel-devel-matched-$(uname -r)"
sudo dnf install amdgpu-dkms

Red Hat 8.10

sudo dnf install https://repo.radeon.com/amdgpu-install/6.4.1/rhel/8.10/amdgpu-install-6.4.60401-1.el8.noarch.rpm
sudo dnf clean all
sudo dnf install "kernel-headers-$(uname -r)" "kernel-devel-$(uname -r)"
sudo dnf install amdgpu-dkms

Oracle Linux

sudo dnf install https://repo.radeon.com/amdgpu-install/6.4.1/el/9.5/amdgpu-install-6.4.60401-1.el9.noarch.rpm
sudo dnf clean all
sudo dnf install "kernel-uek-devel-$(uname -r)"
sudo dnf install amdgpu-dkms

SUSE

sudo SUSEConnect -p sle-module-desktop-applications/15.6/x86_64
sudo SUSEConnect -p sle-module-development-tools/15.6/x86_64
sudo SUSEConnect -p PackageHub/15.6/x86_64
sudo zypper install zypper
sudo zypper --no-gpg-checks install https://repo.radeon.com/amdgpu-install/6.4.1/sle/15.6/amdgpu-install-6.4.60401-1.noarch.rpm
sudo zypper --gpg-auto-import-keys refresh
sudo zypper install kernel-default-devel
sudo zypper install amdgpu-dkms

Azure Linux

sudo tdnf install "kernel-headers-$(uname -r)" "kernel-devel-$(uname -r)"
sudo tdnf install azurelinux-repos-amd
sudo tdnf repolist --refresh
sudo tdnf install amdgpu

Then install PyTorch for ROCm:

 pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/rocm6.4/

Just fork and submit a PR! 📬 Feedback

Have ideas to make this simpler? Found a bug? Open an issue or PR on GitHub!