Runs On
What it takes to serve videomae-base (94M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.
SAVRN's Notes on videomae-base
Video classification with a self-supervised backbone is the job, and this checkpoint is the pretraining stage: VideoMAEForPreTraining, 94M parameters, 12 layers, pretrained on Kinetics-400 for 1600 epochs. At 16-bit the weights and the memory needed are both 0.2 GB, so the cheapest setup on the Index, one MI300X with 192 GB at $1.85 per hour on-demand, has room for hundreds of copies. Treat the $1.85 as the price of a whole video pipeline, not of this model.
The license is where this one stops. CC BY-NC 4.0 permits sharing and adapting with credit for non-commercial purposes only; commercial use needs separate permission from the rights holder, so get that permission before the first fine-tune, not after. Also confirm you want the pretraining head rather than a classifier, note the float32 stored precision behind the 6 files at roughly 754 MB, and read the two papers, arXiv:2203.12602 and arXiv:2111.06377.
SAVRN Research, 2026-09-18
Model Card
VideoMAE model pre-trained on Kinetics-400 for 1600 epochs in a self-supervised way. It was introduced in the paper VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training by Tong et al. and first released in this repository. Disclaimer: The team releasing VideoMAE did not write a model card for this model so this model card has been written by the Hugging Face team. VideoMAE is an extension of Masked Autoencoders (MAE) to video. The architecture of the model is very similar to that of a standard Vision Transformer (ViT), with a decoder on top for predicting pixel values for masked patches. Videos are presented to the model as a sequence of…
Excerpt from the card by Multimedia Computing Group-Nanjing University, licensed cc-by-nc-4.0.
Configuration
- Architecture
- VideoMAEForPreTraining
- Layers
- 12
- Hidden size
- 768
- Feed-forward size
- 3,072
- Attention heads
- 12
- Stored precision
- float32
- Model type
- videomae
Identity and Version
- Repository
- MCG-NJU/videomae-base
- Publisher
- Multimedia Computing Group-Nanjing University
- Task
- Video classification
- Modality
- Video
- Library
- transformers
- Parameters
- 94M parameters
- Languages
- Not stated by the source
- Revision
- dc740ceda42fce44faed2ea03c6d447db72f6af9
- First published
- 2022-08-03
- Last updated
- 2024-03-29
Files and Weights
6 files, 753.8 MB in total. The weights are 2 files totalling 753.8 MB in bin, safetensors.
Weights2 files · 753.8 MB
Configuration2 files · 996 B
Documentation1 file · 3.8 KB
Repository1 file · 1.4 KB
Every file
License and Download
- License
- cc-by-nc-4.0
- Access
- Open weights, no gate
- Download size
- 753.8 MB
Memory Requirements
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About videomae-base
How much GPU memory does videomae-base need?
About 0.2 GB at 16-bit and 0.1 GB at 4-bit: the weights (94M parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run videomae-base on?
At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use videomae-base commercially?
Not without separate permission. videomae-base is released under Creative Commons Attribution-NonCommercial 4.0. CC BY-NC 4.0 permits sharing and adapting with credit for non-commercial purposes only. Commercial use needs separate permission from the rights holder.
Similar Models
VideoMAE model pre-trained on Something-Something-v2 for 2400 epochs in a self-supervised way. It was introduced in the paper VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training by Tong et al. and first released in this repository. Disclaimer: The team releasing VideoMAE did not write a model card for this model so this model card has been written by the Hugging Face team. VideoMAE is an extension of Masked Autoencoders (MAE) to video. The architecture of the model is very similar to that of a standard Vision Transformer (ViT), with a decoder on top for predicting pixel values for masked patches. Videos are presented to the model as a sequence of…
Open weights
cc-by-nc-4.0
94M parameters
transformers
VideoMAE model pre-trained on Kinetics-400 for 800 epochs in a self-supervised way. It was introduced in the paper VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training by Tong et al. and first released in this repository. Disclaimer: The team releasing VideoMAE did not write a model card for this model so this model card has been written by the Hugging Face team. VideoMAE is an extension of Masked Autoencoders (MAE) to video. The architecture of the model is very similar to that of a standard Vision Transformer (ViT), with a decoder on top for predicting pixel values for masked patches. Videos are presented to the model as a sequence of fixed-size…
Open weights
cc-by-nc-4.0
94M parameters
transformers
VideoMAE model pre-trained for 1600 epochs in a self-supervised way and fine-tuned in a supervised way on Kinetics-400. It was introduced in the paper VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training by Tong et al. and first released in this repository. Disclaimer: The team releasing VideoMAE did not write a model card for this model so this model card has been written by the Hugging Face team. VideoMAE is an extension of Masked Autoencoders (MAE) to video. The architecture of the model is very similar to that of a standard Vision Transformer (ViT), with a decoder on top for predicting pixel values for masked patches. Videos are presented to…
Open weights
cc-by-nc-4.0
87M parameters
transformers
This repository houses a fine-tuned VideoMAE (Base) model optimized for multi-class Vietnamese Sign Language Recognition (VSLR). The model architecture adapts self-supervised video representations to accurately classify short video clips of sign gestures into distinct Vietnamese text labels. The model processes short video sequences by partitioning them into spatiotemporal patches, mapping sequential gestures (such as "Ăn", "Bệnh viện", "Xin lỗi") to their corresponding semantic classes. The training routine was monitored closely across key evaluation metrics to prevent overfitting while maximizing classification accuracy on the validation split. The plot below illustrates the progression…
Open weights
mit
86M parameters
transformers
This model is a fine-tuned version of MCG-NJU/videomae-base on an unknown dataset. It achieves the following results on the evaluation set: The following hyperparameters were used during training: - learningrate: 5e-05 - trainbatchsize: 8 - evalbatchsize: 8 - lrschedulertype: linear - trainingsteps: 370 - Transformers 5.16.1 - Pytorch 2.14.0+cu126 - Datasets 5.0.1 - Tokenizers 0.23.2
Open weights
cc-by-nc-4.0
86M parameters
transformers
This model is a fine-tuned version of MCG-NJU/videomae-base on an unknown dataset. It achieves the following results on the evaluation set: The following hyperparameters were used during training: - learningrate: 5e-05 - trainbatchsize: 8 - evalbatchsize: 8 - lrschedulertype: linear - trainingsteps: 370 - Transformers 5.16.1 - Pytorch 2.11.0+cu128 - Datasets 5.0.1 - Tokenizers 0.23.1
Open weights
cc-by-nc-4.0
86M parameters
transformers
