TaLK Convolution

Closed weights Carleton University 240M parameters February 2020

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
Carleton University
Organisation type
Academia
Country
Canada
Published
8 February 2020
Authors
Vasileios Lioutas, Yuhong Guo

What it does

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

Domain
Language
Task
Language modeling, Translation, Text summarization
Numerical format
FP16

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
240M

Table 5 "For the language model, we followed the same configuration as Baevski & Auli (2019). We used 17 decoding layers, each layer with a 1024 hidden size, a 4096 feed-forward hidden size and 8 heads. The adaptive input factor was set to 4."

Training data
103,000,000 tokens

" We replicated their setup and partition the training data into blocks of 512 contiguous tokens" " same setup as in Baevski & Auli (2019)" https://arxiv.org/abs/1809.10853 in that paper they trained for 286k steps in batches of 65,536 tokens. 286000*65536 / 103000000= 182 epochs (same as in as in Baevski & Auli (2019))

Epochs
182
Batch size
65,536

same setup as in Baevski & Auli (2019)

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
2.7 × 10¹⁹ FLOP

6 FLOP / parameter / token * 240000000 parameters * 286000 steps * 65536 tokens per batch = 2.6990346e+19 FLOP

How it was established
Operation counting

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 GeForce RTX 2080 Ti 11GB
Chips used
8
Power draw
4.1 kW

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
Unreleased

MIT, code and weights (though this repo is for translation not WT-103): https://github.com/lioutasb/TaLKConvolutions?tab=readme-ov-file

How it is classified

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

Why it is tracked
SOTA improvement

"[We] set a new state-of-the-art result on the IWSLT De-En and CNN-DailyMail datasets"

Record confidence
Likely
Citations
30
Benchmark data
TaLK Convolution

Sources

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

Reference
Time-aware Large Kernel Convolutions
Last updated
28 November 2025

What the numbers mean

About this model

TaLK Convolution was published by Carleton University, in Canada, in February 2020. The organisation is categorised as academia.

It works in Language, and is recorded as doing language modeling, Translation, Text summarization.

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

How it was trained

The training run consumed about 2.7 × 10¹⁹ FLOP, on NVIDIA GeForce RTX 2080 Ti 11GB. That figure describes the cost of creating it and has no bearing on how quickly it generates text.

Around 103,000,000 tokens went into training it.

Its inclusion criterion is sOTA improvement.

Answers

TaLK Convolution — common questions

01

How much compute was used to train TaLK Convolution?

Around 2.7 × 10¹⁹ FLOP, on NVIDIA GeForce RTX 2080 Ti 11GB. 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.

02

What GPU do I need to run TaLK Convolution?

None. TaLK Convolution 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.

03

Is TaLK Convolution open source?

No. TaLK Convolution has not had its weights published, so it exists only as a service controlled by its owner.

04

How many parameters does TaLK Convolution have?

TaLK Convolution has 240M parameters. Table 5 "For the language model, we followed the same configuration as Baevski & Auli (2019). We used 17 decoding layers, each layer with a 1024 hidden size, a 4096 feed-forward hidden size and 8 heads. The adaptive input factor was set to 4.". 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.

05

Who created TaLK Convolution?

TaLK Convolution was published by Carleton University, based in Canada, categorised as academia.

06

When was TaLK Convolution released?

TaLK Convolution was published in February 2020. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

07

What is TaLK Convolution used for?

TaLK Convolution works in Language, and is recorded as handling language modeling, Translation, Text summarization. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

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.