2-layer-LSTM+Deep-Gradient-Compression
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
- Tsinghua University,Stanford University,NVIDIA
- Organisation type
- Academia,Academia,Industry
- Country
- China, United States of America
- Published
- 5 December 2017
- Authors
- Yujun Lin, Song Han, Huizi Mao, Yu Wang, William J. Dally
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Language
- Task
- Language modeling
- Numerical format
- FP32
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
- 6M
- Training data
- 929,000 tokens
- Epochs
- 40
Here is a summary of the calculations to determine the number of parameters in the 2-layer LSTM model with Deep Gradient Compression described in the paper: Model has 2 LSTM layers with 1500 hidden units each For an LSTM layer: W matrix: number of hidden units * input dimension U matrix: number of hidden units * number of hidden units 4 bias vectors with size = number of hidden units So params per LSTM layer = hidden_units * (input_dim + hidden_units + 4) For layer 1: Input dim is word embe…
" The warm-up period is 1 epoch out of 40 epochs"
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
- 1.3 × 10¹⁵ FLOP
- How it was established
- Operation counting
6 FLOP / parameter / token * 6020000 parameters ['Likely' confidence] * 929000 tokens * 40 epochs = 1.3422192e+15 FLOP
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
repo, but no code for language modeling: https://github.com/synxlin/deep-gradient-compression
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
- 1,627
- Benchmark data
- 2-layer-LSTM+Deep-Gradient-Compression
Sources
Where this record came from and when it was last checked.
- Reference
- Deep Gradient Compression: Reducing the Communication Bandwidth for Distributed Training
- Last updated
- 25 May 2026
What the numbers mean
Where it came from
2-layer-LSTM+Deep-Gradient-Compression was published by Tsinghua University,Stanford University,NVIDIA, in China, in December 2017. It comes out of academia,Academia,Industry.
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 1.3 × 10¹⁵ FLOP of arithmetic, which is a statement about the training budget rather than about inference.
It was trained on about 929,000 tokens of text.
Answers
2-layer-LSTM+Deep-Gradient-Compression — common questions
Is 2-layer-LSTM+Deep-Gradient-Compression open source?
No. 2-layer-LSTM+Deep-Gradient-Compression has not had its weights published, so it exists only as a service controlled by its owner.
How many parameters does 2-layer-LSTM+Deep-Gradient-Compression have?
2-layer-LSTM+Deep-Gradient-Compression has 6M parameters. Here is a summary of the calculations to determine the number of parameters in the 2-layer LSTM model with Deep Gradient Compression described in the paper: Model has 2 LSTM layers with 1500 hidden units each For an LSTM layer: W matrix: number of hidden units * input dimension U matrix: number of hidden units * number of hidden units 4 bias vectors with size = number of hidden units So params per LSTM layer = hidden_units * (input_dim + hidden_units + 4) For layer 1: Input dim is word embedding size (estimated 300) Params = 1500 * (300 + 1500 + 4) = 1,512,000 For layer 2: Input is previous layer output (1500) Params = 1500 * (1500 + 1500 + 4) = 4,506,000 Total Params = Layer 1 + Layer 2 = 1,512,000 + 4,506,000 = 6,018,000 The Deep Gradient Compression technique does not change number of parameters. So total parameters for the model is approximately 6 million. 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.
Who created 2-layer-LSTM+Deep-Gradient-Compression?
2-layer-LSTM+Deep-Gradient-Compression was published by Tsinghua University,Stanford University,NVIDIA, based in China, categorised as academia,Academia,Industry.
When was 2-layer-LSTM+Deep-Gradient-Compression released?
2-layer-LSTM+Deep-Gradient-Compression was published in December 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.
What is 2-layer-LSTM+Deep-Gradient-Compression used for?
2-layer-LSTM+Deep-Gradient-Compression works in Language, and is recorded as handling language modeling. 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.
How much compute was used to train 2-layer-LSTM+Deep-Gradient-Compression?
Around 1.3 × 10¹⁵ 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.
What GPU do I need to run 2-layer-LSTM+Deep-Gradient-Compression?
None. 2-layer-LSTM+Deep-Gradient-Compression 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.
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.