3-ensemble of Self-ensembles on CIFAR-100

Closed weights Google DeepMind 60.2M parameters August 2024

No estimate

No hardware requirements for this model

The weights for this model have not been published, so it cannot be downloaded or run on your own hardware at any size. It is reachable only through its provider, and no graphics card changes that.

On record

Full specification

Everything on record for this model. Most of it describes how it was trained rather than how it runs — useful context for judging how much work went into it, and how it compares with models built at a different scale.

Origin

Who built this model, where, and when it was published.

Organisation
Google DeepMind
Organisation type
Industry
Country
United States of America
Published
8 August 2024
Authors
Stanislav Fort, Balaji Lakshminarayanan

What it does

The problem areas the model was built for. A model can carry several of each.

Domain
Vision
Task
Image classification
Base model
ResNet-152 (ImageNet)

Size

How large the model is and how much data it was trained on. Parameters are the figure that decides whether it fits on a given graphics card.

Parameters
60.2M

as in resnet-152

Training data
tokens

"We focused on the CIFAR-* datasets (Krizhevsky, 2009; Krizhevsky et al.) that comprise 50,000 32 × 32 × 3 images. We arbitrarily chose 𝑁 = 4 and the resolutions we used are 32 × 32, 16 × 16, 8 × 8, 4 × 4 (see Figure 3)." Assuming (!) each image (32×32+16×16+8×8+4×4)×3 = 4080 tokens then the entire dataset was 50,000*4080 = 204000000 tokens

Epochs
13

Training compute

The arithmetic performed to train the model, measured in floating-point operations. It is a measure of what the training run cost, not of how fast the finished model answers you.

How it was established
Operation counting,Hardware
Fine-tuning compute
9.6 × 10¹⁷ FLOP

6ND = 6 FLOP / token / parameter * 60200000 parameters* 204000000 tokens * 13 epochs = 9.579024e+17 FLOP "On an A100, the Colab runs for 20 min for the backbone training (6 epochs) but more epochs and especially lower learning rate training later helps. " 312000000000000 FLOP / GPU / sec [bf16 assumed] * 1 GPU * 20 minutes *60 sec / minute * 0.3 [assumed utilization] = 1.1232e+17 FLOP (an underestimation since the training took more epochs)

The training run

What it physically took to train: which chips, how many, for how long, and what that drew from the wall.

Training hardware
NVIDIA A100

Availability

Whether you can obtain the model and run it on your own hardware, which is what decides if any of the graphics-card figures on this page apply.

Weights
Closed — provider access only
Model access
Unreleased
Training code
Open source

MIT license https://github.com/stanislavfort/ensemble-everything-everywhere

How it is classified

Labels the source dataset applies when tracking notable models, and how confident it is in the entry.

Record confidence
Likely

Sources

Where this record came from and when it was last checked.

Reference
Ensemble everything everywhere: Multi-scale aggregation for adversarial robustness
Last updated
28 November 2025

What the numbers mean

Background

3-ensemble of Self-ensembles on CIFAR-100 was published by Google DeepMind, in United States of America, in August 2024. It comes out of industry.

It works in Vision, and is recorded as doing image classification.

Its starting point was ResNet-152 (ImageNet) — most models at this scale are adapted from an existing base rather than built from nothing.

Because the weights are not available, none of the hardware figures elsewhere on this site apply to it.

Answers

3-ensemble of Self-ensembles on CIFAR-100 — common questions

01

Who created 3-ensemble of Self-ensembles on CIFAR-100?

3-ensemble of Self-ensembles on CIFAR-100 was published by Google DeepMind, based in United States of America, categorised as industry.

02

When was 3-ensemble of Self-ensembles on CIFAR-100 released?

3-ensemble of Self-ensembles on CIFAR-100 was published in August 2024. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

03

What is 3-ensemble of Self-ensembles on CIFAR-100 used for?

3-ensemble of Self-ensembles on CIFAR-100 works in Vision, and is recorded as handling image classification. A model can carry several of each, so these are the areas it was built for rather than a limit on what it will attempt.

04

What GPU do I need to run 3-ensemble of Self-ensembles on CIFAR-100?

None. 3-ensemble of Self-ensembles on CIFAR-100 is a closed model — its weights were never published, so it cannot be downloaded or run on your own hardware at any price. It is reachable only through its provider.

05

Is 3-ensemble of Self-ensembles on CIFAR-100 open source?

No. 3-ensemble of Self-ensembles on CIFAR-100 has not had its weights published, so it exists only as a service controlled by its owner.

06

How many parameters does 3-ensemble of Self-ensembles on CIFAR-100 have?

3-ensemble of Self-ensembles on CIFAR-100 has 60.2M parameters. as in resnet-152. That figure is the total, and it is what decides how much memory the model needs — roughly half a gigabyte per billion at the compression most people use.

Source

Original publication

Record last updated 28 November 2025

The other direction

Looking at it from the other side?

This page starts from the model. If you already own a card and want to know everything it will run, start from the hardware instead.