MemSizer (language modeling)

Closed weights Meta AI,Chinese University of Hong Kong (CUHK) 357M parameters March 2022

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
Meta AI,Chinese University of Hong Kong (CUHK)
Organisation type
Industry,Academia
Country
United States of America, Hong Kong
Published
23 March 2022
Authors
Yizhe Zhang, Deng Cai

What it does

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

Domain
Language
Task
Language modeling

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
357M

357M Table 3 "Following Kasai et al. (2021), we choose similar hyperparameters to prior work (Baevski and Auli, 2019; Fan et al., 2020): 32 layers, 8 heads, 128 head dimensions, 1024 model dimensions, 4096 fully connected dimensions and dropout (Srivastava et al., 2014) and layer dropout rates of 0.2. "

Training data
103,000,000 tokens

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.

Training compute
7.3 × 10¹⁸ FLOP

6 FLOP / token / parameter * 357000000 parameters * 103000000 tokens * 1 epoch [assumed for lower bound] = 2.20626e+17 FLOP also to consider: 7.3 × 10^18 FLOP SOURCE: Impute based on Baevski and Auli 2019 "We generally follow the optimization method from Baevski and Auli (2019), with a slight modification for some hyperparameters including learning rate (we use 10−4), which shows better convergence. " NOTES: Probably correct within an order of 2 or so

How it was established
Operation counting

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 (non-commercial)

code, no license: https://github.com/jcyk/memsizer wt103 train: https://github.com/jcyk/memsizer/blob/main/lm_wikitext-103.sh

How it is classified

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

Record confidence
Likely
Citations
6
Benchmark data
MemSizer

Sources

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

Reference
Linearizing Transformer with Key-Value Memory
Last updated
11 February 2026

What the numbers mean

What this model is

MemSizer (language modeling) was published by Meta AI,Chinese University of Hong Kong (CUHK), in the country recorded as United States of America, during March 2022. It comes out of an organisation categorised as industry,Academia.

It works in the domain of Language, and is recorded as performing the task of language modeling.

This is a closed model: the trained values stayed with whoever produced them, and there is no local version to run.

What went into building it

Producing it required arithmetic totalling around 7.3 × 10¹⁸ FLOP. That figure measures what producing the model cost, and has no bearing on how fast it answers.

Training consumed a corpus of around 103,000,000 tokens of text.

Answers

MemSizer (language modeling) — common questions

01

MemSizer (language modeling)— what is it used for?

It works in the domain of Language, and is recorded as handling the task of language modeling. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

02

MemSizer (language modeling)— how much compute was used to train it?

Training consumed around 7.3 × 10¹⁸ FLOP. That measures what producing the model cost and says nothing about how quickly it answers once trained — inference speed comes from memory bandwidth, not from the training budget.

03

MemSizer (language modeling)— what GPU do I need to run it?

None. This 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.

04

MemSizer (language modeling)— is it open source?

No. Its weights have not been published, so it exists only as a service controlled by its owner.

05

MemSizer (language modeling)— how many parameters does it have?

It has a parameter count of 357M. 357M Table 3 "Following Kasai et al. (2021), we choose similar hyperparameters to prior work (Baevski and Auli, 2019; Fan et al., 2020): 32 layers, 8 heads, 128 head dimensions, 1024 model dimensions, 4096 fully connected dimensions and dropout (Srivastava et al., 2014) and layer dropout rates of 0.2. ". 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.

06

MemSizer (language modeling)— who created it?

It was published by Meta AI,Chinese University of Hong Kong (CUHK), based in United States of America, an organisation categorised as industry,Academia.

07

MemSizer (language modeling)— when was it released?

It was published in March 2022. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

Source

Original publication

Record last updated 11 February 2026

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.