Monarch-GPT-2-Medium
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
- Stanford University,University at Buffalo,University of Michigan
- Organisation type
- Academia,Academia,Academia
- Country
- United States of America
- Published
- 1 April 2022
- Authors
- Tri Dao, Beidi Chen, Nimit Sohoni, Arjun Desai, Michael Poli, Jessica Grogan, Alexander Liu, Aniruddh Rao, Atri Rudra, Christopher Ré
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
- 165M
- Training data
- tokens
- Epochs
- 110
165M table 2 "For dense models, we use standard implementations of GPT-2 [86] from Huggingface transformers library and from Nvidia’s Megatron-LM repo. We follow the training recipe of the Megatron-LM repo. The Monarch version of these models simply swap out the dense weight matrices in the attention blocks (projection matrices) and in the FFN block (linear layers) with Monarch matrices."
"We use an effective batch size of 512" GPT-2-Medium: 10+100 epochs on wikitext-103 + 4k/400k iterations on openwebtext
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
- 4.4 × 10²⁰ FLOP
- How it was established
- Operation counting
"We use an effective batch size of 512" GPT-2-Medium: 10+100 epochs on wikitext-103 + 4k/400k iterations on openwebtext 6 FLOP / parameter / token * 165 * 10^6 parameters * (103000000 tokens * 110 epochs + 512 * 404000) = 1.142147952 × 10^19 FLOP
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
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
apache for code: https://github.com/HazyResearch/fly/tree/master https://github.com/HazyResearch/fly/blob/master/src/train.py
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Record confidence
- Confident
- Citations
- 130
- Benchmark data
- Monarch-GPT-2-Medium
Sources
Where this record came from and when it was last checked.
- Reference
- Monarch: Expressive Structured Matrices for Efficient and Accurate Training
- Last updated
- 25 May 2026
What the numbers mean
What this model is
Monarch-GPT-2-Medium was published by Stanford University,University at Buffalo,University of Michigan, in the country recorded as United States of America, during April 2022. The category the publisher falls under is academia,Academia,Academia.
It works in the domain of Language, and is recorded as performing the task of language modeling.
Because the weights are not available, none of the hardware figures elsewhere on this site apply to it.
Training and provenance
The training run consumed about 4.4 × 10²⁰ FLOP, on hardware recorded as NVIDIA V100. That figure measures what producing the model cost, and has no bearing on how fast it answers.
Answers
Monarch-GPT-2-Medium — common questions
Monarch-GPT-2-Medium— how many parameters does it have?
It has a parameter count of 165M. 165M table 2 "For dense models, we use standard implementations of GPT-2 [86] from Huggingface transformers library and from Nvidia’s Megatron-LM repo. We follow the training recipe of the Megatron-LM repo. The Monarch version of these models simply swap out the dense weight matrices in the attention blocks (projection matrices) and in the FFN block (linear layers) with Monarch matrices.". 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.
Monarch-GPT-2-Medium— who created it?
It was published by Stanford University,University at Buffalo,University of Michigan, based in United States of America, an organisation categorised as academia,Academia,Academia.
Monarch-GPT-2-Medium— when was it released?
It was published in April 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.
Monarch-GPT-2-Medium— what is it used for?
It works in the domain of Language, and is recorded as handling the task of 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.
Monarch-GPT-2-Medium— how much compute was used to train it?
Training consumed around 4.4 × 10²⁰ FLOP, on hardware recorded as 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.
Monarch-GPT-2-Medium— 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.
Monarch-GPT-2-Medium— is it open source?
No. Its weights have not been published, so it exists only as a service controlled by its owner.
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.