Skip to content
Other

gt-embed-minimal-v1

Lightweight Human DNA Embeddings

2 min readupdated

Created by Bell Eapen, this compact model turns protein-coding DNA sequences (CDS) into 256-dimensional vector representations (embeddings). Under the hood is a 2-layer Transformer Encoder with ~2M parameters, trained via contrastive learning on ~17,000 human transcripts from GENCODE v50 (GRCh38 assembly). Its standout feature is codon-level tokenization (triplets of nucleotides that code for amino acids) via Genomic Tokenizer instead of single-character parsing, allowing the model to capture structure at the protein synthesis level. It is well-suited for bioinformaticians and developers looking for rapid prototyping and pipeline debugging.

Key features

  • 26-token codon vocabulary: 20 tokens for amino acid codons and 6 special tokens ([CLS], [SEP], [BOS], [MASK], [PAD], [UNK]).
  • Ultra-lightweight architecture: just ~2M parameters (2 layers, 4 attention heads, 256 hidden size), making it convenient for testing and running even on a CPU.
  • Normalized output: produces L2-normalized 256-dimensional vectors for sequences up to 2,048 tokens (~6,144 nucleotides) long.
  • Part of a model lineup: this is the minimal development release—the family also includes planned Small (~10M parameters) and Regular (~50M) variants for higher-precision tasks.

What you can do

  • Gene prioritization: plug it into research pipelines when studying rare hereditary disorders and searching for disease-associated genes.
  • Variant annotation: score and cluster novel genetic variants in protein-coding regions.
  • Knowledge graphs: build vector links between genes and diseases in bioinformatics graphs.

Example

import torch
from genomic_tokenizer import GenomicTokenizer

tokenizer = GenomicTokenizer(model_max_length=2048)
cds_sequence = "ATGGCCGCCTACAAGCAGCAGTCTGGGGAAGACGGGCTACCTTGCCCCTGA"
tokens = tokenizer(cds_sequence)
input_ids = torch.tensor([tokens['input_ids']])
with torch.no_grad():
    embedding = model(input_ids)  # shape: [1, 256], L2-normalized
Read moreDetails
Training: contrastive learning on ~17,000 human protein-coding transcripts (GENCODE v50, GRCh38.p14 genome).
Hardware & requirements: ~2M parameters, runs on CPU with PyTorch 2.5.1 and the genomic-tokenizer / gtembed packages.
Limitations: human genome only (excludes non-coding and regulatory regions), max length of 2,048 tokens, not clinically validated and not intended for medical diagnosis.
License: Apache 2.0.
Discuss Telegram 1 40

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

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.

    # beapen/gt-embed-minimal-v1
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    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 beapen/gt-embed-minimal-v1 --local-dir ~/models/gt-embed-minimal-v1
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to gt-embed-minimal-v1
  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 gt-embed-minimal-v1 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

beapen/gt-embed-minimal-v1
Author
beapen
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

401

Ask AI about this model

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

Comments