Fold2Seq
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
- IBM,Texas A&M
- Organisation type
- Industry,Academia
- Country
- United States of America
- Published
- 24 June 2021
- Authors
- Yue Cao, Payel Das, Vijil Chenthamarakshan, Pin-Yu Chen, Igor Melnyk, Yang Shen
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Biology
- Task
- Proteins, Protein generation, Protein inverse folding, Protein fold classification
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
- 12.4M
- Training data
- 45,995 tokens
- Epochs
- 200
- Batch size
- 128
Three components, a sequence encoder, fold encoder, and sequence decoder. (1) and (3) have similar architectures, a sequence embedder + transformer block (the encoder uses a transformer encoder, the decoder uses a decoder block). The fold encoder adds 6 residual conv blocks and has a 3d positional encoder rather than sequence embedder. Each residual block has two 3D-convolutional layers (3×3×3) and batch normalization layers. Each transformer block has 4 layers and d = 256 latent dimensions. …
"Training set includes 45995 proteins belonging to a total of 971 folds"
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
- 1.4 × 10¹⁷ FLOP
- How it was established
- Operation counting
"We train our model on 2 Tesla K80 GPUs, with batch size 128. In every training stage we train up to 200 epochs with an early stopping strategy based on the validation loss" See calculations here: https://docs.google.com/document/d/1luTCTQLmfaBfmnjbsBpWuG51x_IwuiN7mqNoOXtoi8Y/edit?usp=sharing Block 1: CNN 1: 2*20*20*20*3*3*3*4*8=13824000 CNN 2: 2*20*20*20*3*3*3*8*8=27648000 Block 2: CNN 1: 2*20*20*20*3*3*3*8*16=55296000 CNN 2: 2*20*20*20*3*3*3*16*16=110592000 Block 3: CNN 1: 2*20*20*20*3*3*3*16…
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 Tesla K80
- Chips used
- 2
- Power draw
- 1.2 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
- Open source
Apache 2.0 https://github.com/IBM/fold2seq
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
- Record confidence
- Speculative
- Citations
- 56
"On test sets with single, high-resolution and complete structure inputs for individual folds, our experiments demonstrate improved or comparable performance of Fold2Seq in terms of speed, coverage, and reliability for sequence design, when compared to existing state-of-the-art methods that include data-driven deep generative models and physics-based RosettaDesign." [Abstract]
Sources
Where this record came from and when it was last checked.
- Reference
- Fold2Seq: A Joint Sequence(1D)-Fold(3D) Embedding-based Generative Model for Protein Design
- Last updated
- 25 May 2026
What the numbers mean
About this model
Fold2Seq was published by IBM,Texas A&M, in the country recorded as United States of America, during June 2021. The publishing organisation is categorised as industry,Academia.
It works in the domain of Biology, and is recorded as performing the task of proteins, Protein generation, Protein inverse folding, Protein fold classification.
Its weights were never published, so it can only be reached through its provider. No graphics card changes that.
What went into building it
The training run consumed about 1.4 × 10¹⁷ FLOP, on hardware recorded as NVIDIA Tesla K80. That figure measures what producing the model cost, and has no bearing on how fast it answers.
It was trained on a corpus of about 45,995 tokens of text.
The reason it appears in this catalogue at all: sOTA improvement.
Answers
Fold2Seq — common questions
Fold2Seq— who created it?
It was published by IBM,Texas A&M, based in United States of America, an organisation categorised as industry,Academia.
Fold2Seq— when was it released?
It was published in June 2021. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.
Fold2Seq— what is it used for?
It works in the domain of Biology, and is recorded as handling the task of proteins, Protein generation, Protein inverse folding, Protein fold classification. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.
Fold2Seq— how much compute was used to train it?
Training consumed around 1.4 × 10¹⁷ FLOP, on hardware recorded as NVIDIA Tesla K80. 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.
Fold2Seq— 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.
Fold2Seq— is it open source?
No. Its weights have not been published, so it exists only as a service controlled by its owner.
Fold2Seq— how many parameters does it have?
It has a parameter count of 12.4M. Three components, a sequence encoder, fold encoder, and sequence decoder. (1) and (3) have similar architectures, a sequence embedder + transformer block (the encoder uses a transformer encoder, the decoder uses a decoder block). The fold encoder adds 6 residual conv blocks and has a 3d positional encoder rather than sequence embedder. Each residual block has two 3D-convolutional layers (3×3×3) and batch normalization layers. Each transformer block has 4 layers and d = 256 latent dimensions. Calculations here: https://docs.google.com/document/d/1luTCTQLmfaBfmnjbsBpWuG51x_IwuiN7mqNoOXtoi8Y/edit?usp=sharing. 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.
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.