Masked Vision and Language Modeling for Multi-modal Representation Learning

Authors: Gukyeong Kwon, Zhaowei Cai, Avinash Ravichandran, Erhan Bas, Rahul Bhotika, Stefano Soatto

ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments on various V+L tasks show that the proposed method, along with common V+L alignment losses, achieves state-of-the-art performance in the regime of millions of pre-training data.
Researcher Affiliation Industry Gukyeong Kwon, Zhaowei Cai, Avinash Ravichandran, Erhan Bas, Rahul Bhotika, Stefano Soatto AWS AI Labs {gukyeong,zhaoweic,soattos}@amazon.com
Pseudocode No No explicit pseudocode or algorithm blocks are provided in the paper.
Open Source Code No No explicit statement about releasing source code for the described methodology or a direct link to a code repository is found in the paper.
Open Datasets Yes We use the union of four datasets for pre-training... These datasets are Conceptual Captions (CC) (Sharma et al., 2018), SBU Captions (Ordonez et al., 2011), Visual Genome (VG) (Krishna et al., 2017), and COCO Captions (Lin et al., 2014).
Dataset Splits Yes To be specific, we follow data splits proposed in (Karpathy & Fei-Fei, 2015) and an average recall over image and text retrieval is used to find the best model in the validation set.
Hardware Specification Yes A batch size of 512 is used with 16 NVIDIA Tesla V100 GPUs.
Software Dependencies No We used the Image Net pretrained Vi T (vit base patch16 224) from (Wightman, 2019) and the pre-trained Ro BERTa (roberta-base) from Hugging Face (Wolf et al., 2020).
Experiment Setup Yes We pre-train the model for 50 epochs when the 4M dataset is used and 30 epochs for all other experiments. A batch size of 512 is used with 16 NVIDIA Tesla V100 GPUs. All parameters are optimized using Adam W (Loshchilov & Hutter, 2017) with a weight decay of 0.05. Following (Xie et al., 2021), we use the image masking ratio of 60%. While 15% masking ratio is used for text in language models (Devlin et al., 2018; Liu et al., 2019), we use 30% since the paired image can provide additional information for text reconstruction. During pre-training, the learning rate is warmed up to 3 × 10−4 in the first 5 epochs and decayed to 3 × 10−5 using a cosine scheduler. The learning rates for the image encoder and the text encoder are set to 10−5, which is less than that of the cross-modality encoders. An image size of 224 × 224 and Rand Augment (Cubuk et al., 2020) are used.