Implementation:Princeton nlp SimPO Conda Environment Create

Knowledge Sources	SimPO Conda Docs
Domains	MLOps, Environment_Management
Last Updated	2026-02-08 04:30 GMT

Overview

External CLI tool for creating the SimPO training environment from a Conda specification file.

Description

The environment.yml file defines a complete Conda environment named simpo with Python 3.10.14, PyTorch 2.2.2 (CUDA 12.1), and all training dependencies. The environment is created using the conda env create command, followed by an editable install of the project package.

Usage

Use this before any training or inference script. The environment must be created once per machine or container. Activate with conda activate simpo before running training commands.

Code Reference

Source Location

• Repository: SimPO
• File: environment.yml (Lines 1-260)

Signature

# Step 1: Create the conda environment
conda env create -f environment.yml

# Step 2: Activate the environment
conda activate simpo

# Step 3: Install the project in editable mode
pip install -e .

Import

# No Python import — this is a CLI tool
# After environment creation, activate with:
conda activate simpo

I/O Contract

Inputs

Name	Type	Required	Description
environment.yml	File	Yes	Conda environment specification with all dependencies

Outputs

Name	Type	Description
simpo environment	Conda env	Fully configured Python 3.10 environment with all training libraries

Usage Examples

Full Environment Setup

# Clone the repository
git clone https://github.com/princeton-nlp/SimPO.git
cd SimPO

# Create the conda environment from the specification
conda env create -f environment.yml

# Activate the environment
conda activate simpo

# Install the project in editable mode
pip install -e .

# Verify key packages
python -c "import torch; print(torch.__version__)"        # 2.2.2
python -c "import transformers; print(transformers.__version__)"  # 4.44.2
python -c "import trl; print(trl.__version__)"             # 0.9.6

Key Dependencies

python==3.10.14
torch==2.2.2 (with CUDA 12.1)
transformers==4.44.2
trl==0.9.6
accelerate==0.29.2
peft==0.7.1
deepspeed==0.12.2
flash-attn==2.5.7
bitsandbytes==0.41.2.post2
datasets==2.18.0
wandb==0.13.11

Related Pages

Implements Principle

• Principle:Princeton_nlp_SimPO_Environment_Setup