Installation¶

Prerequisites¶

Python 3.10+
PyTorch 2.1+ (with CUDA recommended)
OS: Linux, macOS, or Windows

Install from Source (Recommended)¶

# 1. Clone
git clone https://github.com/Yaselley/deepfense-framework
cd deepfense-framework

# 2. Create environment
conda create -n deepfense python=3.10
conda activate deepfense

# 3. Install
pip install -e .

This installs DeepFense in editable mode with all dependencies (torch, transformers, wandb, etc.).

Verify¶

python -c "import deepfense; print('OK')"
deepfense list   # shows all registered components

Frontend Checkpoints¶

DeepFense supports two ways to load pretrained SSL models:

Method	Config	What you need
HuggingFace	`source: "huggingface"`	Just an internet connection (downloads automatically)
Local .pt file	`source: "fairseq"`	A downloaded `.pt` checkpoint file

No separate fairseq installation is needed. DeepFense includes a built-in loader for fairseq-format .pt checkpoints.

Example -- HuggingFace (easiest):

frontend:
  type: "wav2vec2"
  args:
    source: "huggingface"
    ckpt_path: "facebook/wav2vec2-xls-r-300m"

Example -- local .pt file:

frontend:
  type: "wav2vec2"
  args:
    source: "fairseq"
    ckpt_path: "/path/to/xlsr2_300m.pt"

Troubleshooting¶

Problem	Fix
`ModuleNotFoundError: torch`	Install PyTorch first: https://pytorch.org/get-started/locally/
`CUDA not available`	`python -c "import torch; print(torch.cuda.is_available())"` -- install CUDA-enabled PyTorch
`deepfense: command not found`	Run `pip install -e .` again from the repo root

Next: Quick Start