Awesome

eP-ALM: Efficient Perceptual Augmentation of Language Models

Official implementation of the paper:

• eP-ALM: Efficient Perceptual Augmentation of Language Models

In this repo, you will find the pretrained models and code to train and evaluate eP-ALM on Image/Video/Audio-Text tasks.

News

• [July-2023] A project page and a demo are now available.
• [July-2023] The paper is accepted at ICCV23 !
• [June-2023] A new version of the paper is released on arXiv:
  • We re-evaluate the models with greedy decoding.
  • We add comparison with SoTA.
  • We add new experiments, including pretraining on CC3M and evaluation in zero-shot (check Appendix).
• [May-2023] The code is optimized to train and evaluate with float16 mixed precision, using the accelerate library 🤗.
• [May-2023] We found greedy decoding with beam search is significantly better than multinomial/random sampling.
• [20-March-2023] The paper is released on arXiv.
• [March-2023] The paper is submitted for publication and currently under review.

Summary:

• Introduction
• Download
• Installation
• Evaluation
• Accelerated Training 🤗
• Training
• Citation
• Acknowledgment

Introduction

Large Language Models (LLMs) have so far impressed the world, with unprecedented capabilities that emerge in models at large scales. On the vision side, transformer models (i.e., ViT) are following the same trend, achieving the best performance on challenging benchmarks. With the abundance of such unimodal models, a natural question arises; do we need also to follow this trend to tackle multimodal tasks? In this work, we propose to rather direct effort to efficient adaptations of existing models, and propose to augment Language Models with perception. Existing approaches for adapting pretrained models for vision-language tasks still rely on several key components that hinder their efficiency. In particular, they still train a large number of parameters, rely on large multimodal pretraining, use encoders (e.g., CLIP) trained on huge image-text datasets, and add significant inference overhead. In addition, most of these approaches have focused on Zero-Shot and In Context Learning, with little to no effort on direct finetuning. We investigate the minimal computational effort needed to adapt unimodal models for multimodal tasks and propose a new challenging setup, alongside different approaches, that efficiently adapts unimodal pretrained models. We show that by freezing more than 99% of total parameters, training only one linear projection layer, and prepending only one trainable token, our approach (dubbed eP-ALM) significantly outperforms other baselines on VQA and Captioning across Image, Video, and Audio modalities, following the proposed setup.

Results

Comparison of eP-ALM with OPT-6.7B that train 0.06% (~4M) of parameters (greedy decoding with number of beams 3) to SoTA that train large number of parameters on large datasets. $^*$: test-std set. PT: pretrained, FT: finetuned, ZS: zero-shot.

Qualitative results of eP-ALM: the model is able to generate accurate answers and coherent descriptions of the image. Ground truth answers are highlighted in green (with multinomial sampling and OPT350M).

Download

OPT Model

First you need to download OPT models and tokenizers. You can use the following (for OPT-2.7B) to automatically download them:

from transformers import AutoTokenizer, OPTModel
tokenizer = AutoTokenizer.from_pretrained("facebook/opt-2.7b")
model = OPTModel.from_pretrained("facebook/opt-2.7b")

Pretrain Encoders

You also need to download the different encoders from the following:

• Audio AST encoder from this repo, specifically this model.
• Video TimeSformer encoder from this repo, specifically this model.
• Image Encoder should be downloaded automatically from with the timm library, specifically we use the vit_base_patch16_224 model.

Pretrained Models

We provide only the adaptation parameters (linear connection and Soft Prompt). You can download the following models:

For eP-ALM_pt-L (with OPT-6.7B and ViT-L), trained with float16 mixed precision (accelerate_training):

• VQA v2: ePALM_pt_L
• COCO Caption: ePALM_pt_L
• GQA: ePALM_pt_L

In the following, we provide smaller models that used to obtain the main results in the paper. Note that these models are trained with float32:

• VQA v2: ePALM
• COCO Caption: ePALM
• GQA: ePALM
• MSR-VTT Video Caption: ePALM
• MSRVTT-QA Video QA: ePALM
• MSVD-QA: ePALM
• AudioCaps Audio Captioning: ePALM

Data

More details on the download and the organization of datasets can be found here

Installation

Main requirements:

python >= 3.8+
torch >= 1.12+
transformers >= 4.24+
accelerate >= 0.11.0

More details can be found here.

Evaluation

To evaluate the trained models, you can use the same scripts in run_scripts/, and just pass the best checkpoint path to the --evaluate arguments.

To visualize the results and test on your own images, you can use this notebook ePALM.ipynb.

You should use the same script used for training to evaluate the model (e.g., run_scripts/accelerate_training for models trained with accelerate).

Note that you can evaluate the models trained with float32 with run_scripts/accelerate_training, but the you might obtain slightly different results (e.g., for caption we obtain 102 CIDEr instead of 97 as reported in the paper).

Accelerated Training 🤗

We optimized the code code based on the accelerate librairy. Mainly we train with mixed precision and keep the precision of the LM in float16, this significantly reduces the memory consumption (/2) and accelerates (x2) the training.

For example, after specifying the path to config file, data_dir and output_dir, to launch a training of eP-ALM_pt-L on VQA v2:

sh run_scripts/accelerate/image/ePALM_pt_L_vqa_acc.sh

To resume training, specify the initialization checkpoint to the --resume argument.

Training

Previous models are trained with float32 precision. You can launch the training/evaluation of eP-ALM using the different scripts in run_scripts/float32. For example you can launch a training on VQA v2 from the following script:

sh run_scripts/float32/image/ePALM_vqa.sh

Citation

@InProceedings{Shukor_2023_ICCV,
    author    = {Shukor, Mustafa and Dancette, Corentin and Cord, Matthieu},
    title     = {eP-ALM: Efficient Perceptual Augmentation of Language Models},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2023},
    pages     = {22056-22069}
}

Acknowledgment

Some code was borrowed from timm, transformers, TimeSformer, and VL-Adapter.