AWD-LSTM-MoS + dynamic evaluation (WT2, 2017)

Closed weights Carnegie Mellon University (CMU) 35M parameters November 2017

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
Carnegie Mellon University (CMU)
Organisation type
Academia
Country
United States of America
Published
10 November 2017
Authors
Zhilin Yang, Zihang Dai, Ruslan Salakhutdinov, William W. Cohen

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
35M

35M (Table 2)

Training data
2,000,000 tokens

max sequence length from https://github.com/zihangdai/mos/blob/master/main.py is 110 tokens batch size 15 (table 8) epochs: 8000 (from https://github.com/zihangdai/mos/blob/master/main.py)

Epochs
8,000

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
3.4 × 10¹⁸ FLOP

6 FLOP / parameter / token * 35000000 parameters * 2000000 tokens * 8000 epochs = 3.36e+18 FLOP _____________ in the algorithmic progress paper the estimation was 4.37 × 10^17 FLOP based on 1000 epochs assumption

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 source

MIT code: https://github.com/zihangdai/mos

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

"Experimental results confirm that the proposed method significantly improves state-of-the-art language models, achieving a perplexity of 55.31 and 62.89 on the test set of Penn Treebank and WikiText-2"

Record confidence
Likely
Citations
416
Benchmark data
AWD-LSTM-MoS + dynamic evaluation (WT2, 2017)

Sources

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

Reference
Breaking the Softmax Bottleneck: A High-Rank RNN Language Model
Last updated
25 May 2026

What the numbers mean

What this model is

AWD-LSTM-MoS + dynamic evaluation (WT2, 2017) was published by Carnegie Mellon University (CMU), in United States of America, in November 2017. The organisation is categorised as academia.

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

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

Training and provenance

Producing it required around 3.4 × 10¹⁸ FLOP of arithmetic, which is a statement about the training budget rather than about inference.

It was trained on about 2,000,000 tokens of text.

The reason it appears in this catalogue at all is sOTA improvement.

Answers

AWD-LSTM-MoS + dynamic evaluation (WT2, 2017) — common questions

01

How much compute was used to train AWD-LSTM-MoS + dynamic evaluation (WT2, 2017)?

Around 3.4 × 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.

02

What GPU do I need to run AWD-LSTM-MoS + dynamic evaluation (WT2, 2017)?

None. AWD-LSTM-MoS + dynamic evaluation (WT2, 2017) 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 AWD-LSTM-MoS + dynamic evaluation (WT2, 2017) open source?

No. AWD-LSTM-MoS + dynamic evaluation (WT2, 2017) has not had its weights published, so it exists only as a service controlled by its owner.

04

How many parameters does AWD-LSTM-MoS + dynamic evaluation (WT2, 2017) have?

AWD-LSTM-MoS + dynamic evaluation (WT2, 2017) has 35M parameters. 35M (Table 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.

05

Who created AWD-LSTM-MoS + dynamic evaluation (WT2, 2017)?

AWD-LSTM-MoS + dynamic evaluation (WT2, 2017) was published by Carnegie Mellon University (CMU), based in United States of America, categorised as academia.

06

When was AWD-LSTM-MoS + dynamic evaluation (WT2, 2017) released?

AWD-LSTM-MoS + dynamic evaluation (WT2, 2017) was published in November 2017. 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 AWD-LSTM-MoS + dynamic evaluation (WT2, 2017) used for?

AWD-LSTM-MoS + dynamic evaluation (WT2, 2017) 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.

Source

Original publication

Record last updated 25 May 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.