Skip to content

Latest commit

 

History

History
42 lines (32 loc) · 2.42 KB

README.md

File metadata and controls

42 lines (32 loc) · 2.42 KB
Raw

PromptFolio

This repository contains the codebase for the paper: [NeurIPS 2024] Federated Learning from Vision-Language Foundation Models: Theoretical Analysis and Method.

Installation

Environment Settings

conda create -n PromptFolio python=3.8 yacs tqdm tabulate ftfy regex tensorboard
conda activate PromptFolio
pip install torch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0 --index-url https://download.pytorch.org/whl/cu121
pip install timm gdown prettytable scikit-learn einops 

git clone https://github.com/PanBikang/PromptFolio.git
cd PromptFolio

Our CUDA version is 12.3, so ensure your Torch version corresponds to this CUDA version.

Additionally, set your dataset directory path, for example, /storage/data, in the argparse --root option within federated_main.py.

For dataset configuration details, refer to DATASETS.md.

Running the Code

Example: Run PromptFolio with Caltech101

python federated_main.py --trainer "configs/datasets/caltech101.yaml" --model fedavg --trainer PromptFolio --frac_p 0.4 --num_users 10 --beta 0.3 --round 10

Example: Run the baselines with DTD

#python federated_main.py --dataset-config-file "configs/datasets/dtd.yaml" --num_users 10 --beta 0.3 --round 10 --model local --trainer PromptFL # CoOp
#python federated_main.py --dataset-config-file "configs/datasets/dtd.yaml" --num_users 10 --beta 0.3 --round 10 --model fedavg --trainer PromptFL # PromptFL
#python federated_main.py --dataset-config-file "configs/datasets/dtd.yaml" --num_users 10 --beta 0.3 --round 10 --model fedavg --trainer PromptFLFT # PromptFL+FT
#python federated_main.py --dataset-config-file "configs/datasets/dtd.yaml" --num_users 10 --beta 0.3 --round 10 --model fedavg --trainer PromptFL --fedprox_mu 1.0 # PromptFL+FedProx (e.g., with mu=1.0)
#python federated_main.py --dataset-config-file "configs/datasets/dtd.yaml" --num_users 10 --beta 0.3 --round 10 --model fedavg --trainer PromptFLFedPer
#python federated_main.py --dataset-config-file "configs/datasets/dtd.yaml" --num_users 10 --beta 0.3 --round 10 --model fedavg --trainer PromptFLFedAMP
#python federated_main.py --dataset-config-file "configs/datasets/dtd.yaml" --num_users 10 --beta 0.3 --round 10 --model fedavg --trainer FedTPG
#python federated_main.py --dataset-config-file "configs/datasets/dtd.yaml" --num_users 10 --beta 0.3 --round 10 --model fedavg --trainer pFedPrompt

=