TransformerXL + spectrum control

Closed weights University of California Los Angeles (UCLA),JD.com 151M parameters March 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
University of California Los Angeles (UCLA),JD.com
Organisation type
Academia,Industry
Country
United States of America, China
Published
11 March 2020
Authors
Lingxiao Wang, Jing Huang, Kevin Huang, Ziniu Hu, Guangtao Wang, Quanquan Gu

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
151M

151M (Table 2) " On the large WikiText-103 dataset, we implement our method based on the state-of-the-art Transformer-XL based models (Dai et al., 2019). We follow the same settings reported in (Dai et al., 2019), and our implementation is based on the official code for Transformer-XL."

Training data
103,000,000 tokens

"For WikiText-103 dataset, we use four NVIDIA Tesla V100 GPU and set the batch size to be 40." unknown amount of epochs and training steps but they say they follow original Transformer-XL setup (400K steps + 16K warmup steps, 128 batch size, 384 sequence length) 400000*4*40*384 / 103000000 = 248 epochs

Epochs
250
Batch size
61,440

4*40*384

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
2.6 × 10¹⁹ FLOP

6 FLOP / parameter / token * 151000000 parameters * 103000000 tokens * 250 epochs = 2.33295e+19 FLOP 31330000000000 FLOP / sec / GPU [fp16] * 4 GPUs * 3152 sec per epoch * 250 epochs * 0.3 [assumed precision] = 2.9625648e+19 FLOP sqrt(2.33295e+19*2.9625648e+19) = 2.6289761e+19 FLOP 'speculative' confidence due to the assumption of the amount of epochs

How it was established
Operation counting,Hardware

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 V100
Chips used
4
Wall-clock time
219 hours (9.1 days)

3152 sec per epoch (Table 7) * 250 epochs / 3600 sec/hour = 219 hours

Power draw
2.5 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

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 demonstrate that our spectrum control method outperforms the state-of-the-art Transformer-XL modeling for language model" " Our spectrum control method outperforms the latest state-of-the-art TransformerXL model on WikiText-103 dataset for language modeling; and obtains close to 1.5 BLEU improvement on IWSLT 2014 German-English translation task compared to the Transformer baseline model."

Record confidence
Speculative
Citations
90
Benchmark data
TransformerXL + spectrum control

Sources

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

Reference
Improving Neural Language Generation with Spectrum Control
Last updated
1 December 2025

What the numbers mean

About this model

TransformerXL + spectrum control was published by University of California Los Angeles (UCLA),JD.com, in United States of America, in March 2020. It comes out of academia,Industry.

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

Its weights were never published, so it can only be reached through its provider. No graphics card changes that.

How it was trained

The training run consumed about 2.6 × 10¹⁹ FLOP, on NVIDIA V100. 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.

It is tracked in the underlying dataset for one reason in particular: sOTA improvement.

Answers

TransformerXL + spectrum control — common questions

01

Who created TransformerXL + spectrum control?

TransformerXL + spectrum control was published by University of California Los Angeles (UCLA),JD.com, based in United States of America, categorised as academia,Industry.

02

When was TransformerXL + spectrum control released?

TransformerXL + spectrum control was published in March 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.

03

What is TransformerXL + spectrum control used for?

TransformerXL + spectrum control works in Language, and is recorded as handling language modeling. These are the areas it was designed around; they describe intent rather than a hard boundary.

04

How much compute was used to train TransformerXL + spectrum control?

Around 2.6 × 10¹⁹ FLOP, on NVIDIA V100. 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.

05

What GPU do I need to run TransformerXL + spectrum control?

None. TransformerXL + spectrum control 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.

06

Is TransformerXL + spectrum control open source?

No. TransformerXL + spectrum control has not had its weights published, so it exists only as a service controlled by its owner.

07

How many parameters does TransformerXL + spectrum control have?

TransformerXL + spectrum control has 151M parameters. 151M (Table 2) " On the large WikiText-103 dataset, we implement our method based on the state-of-the-art Transformer-XL based models (Dai et al., 2019). We follow the same settings reported in (Dai et al., 2019), and our implementation is based on the official code for Transformer-XL.". 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 1 December 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.