Skip to content
Other

EpiClock

transformer models for DNA age estimation

2 min readupdated

EpiClock is a family of five transformer models that predict human age from DNA methylation using Illumina 450K or EPIC data. The models are trained on a filtered set of healthy donor profiles and tested on independent studies. Developed for research purposes only, this tool is not a medical device. It is useful for data scientists and researchers in bioinformatics and biogerontology.

Key features

  • Includes five variants of different sizes: ranging from EpiClock-T (2.1M parameters) to EpiClock-XL (48.0M parameters).
  • Each CpG site is processed as a token, taking into account genomic context and beta values.
  • Pre-training involves masking and reconstructing 25% of CpG sites without using age labels.
  • Fine-tuning combines linear clocks with a correction learned by the transformer.

What you can do

  • Run age predictions on a DNA beta-value matrix using the ready-to-use epiclock.py script.
  • Compare the accuracy of the model family against classical methods like Elastic net or Horvath V1/V2.
  • Integrate the model into your own research pipelines for methylome analysis.

Example

from epiclock import epiclock_predict

# rows = samples, columns = CpG ids (cg...), values = beta in [0, 1]; missing values are allowed
res = epiclock_predict("EpiClock-B/model.pt", betas_dataframe)
ages = res["epigenetic_age"]
Read moreDetails
Benchmarks: the mean absolute error of the models on the test set is about 4.78–4.89 years.
Hardware requirements: standard PyTorch checkpoints, model sizes from 2.1M to 48M parameters.
Limitations: works only with a fixed panel of 9,098 CpG sites, primarily trained on blood and adult data.
License: CC BY-SA 4.0.
Examples
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.2 GB · ready files0.7 GB
  • FP8 / INT8recommendedweights 0.1 GB0.6 GB
  • 4-bitweights 0.1 GB0.6 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # dhillonv10/epiclock
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download dhillonv10/epiclock --local-dir ~/models/epiclock
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to epiclock
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of epiclock (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

dhillonv10/epiclock
Author
dhillonv10
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments

Other

WVY-3D-1M

WVY-3D-1M