install.md

Example conda environment setup

Step 1. Create a conda environment and activate it.

conda create -n intern_clean python=3.9 -y
conda activate intern_clean

Step 2. Install PyTorch (We use PyTorch 2.0.1 / CUDA 11.7)

pip3 install torch torchvision torchaudio

# Please use the following command to install PyTorch so you can replicate our results:
# pip install torch==2.0.1+cu117 torchvision==0.15.2+cu117 torchaudio==2.0.2 --index-url https://download.pytorch.org/whl/cu117

Step 3. Install require packages

pip install transformers==4.33.2 timm==0.4.12 sentencepiece==0.1.99 gradio==4.13.0 markdown2==2.4.10 xlsxwriter==3.1.2 einops

Step 4. Install flash-attention2 to save GPU memory

We strongly recommend installing flash-attention2 to save GPU memory, although you can run IXC models without it.

How to install flash-attention2: https://github.com/Dao-AILab/flash-attention?tab=readme-ov-file#installation-and-features

Optional: 4-bit inference

Please install the pypi package with pip install lmdeploy. By default, it depends on CUDA 12.x. For a CUDA 11.x environment, please refer to the installation guide.

Optional: Fine-tuning

Fine-turning requires deepspeed, peft (optional for LoRA fine-tuning)

# install deepspeed
pip install deepspeed==0.12.3

# install peft
pip install peft==0.8.2