GenomeOcean-100M-AWQ

Model Overview

This is a AWQ quantized version of GenomeOcean-100M, designed for high-efficiency DNA sequence modeling.

• Architecture: Mistral-based Genomic LLM
• Quantization: AWQ (4-bit)
• Primary Use: DNA sequence scoring, generation, and genomic feature analysis.

Benchmark Results (Local Evaluation)

Evaluation conducted on genomic sequences (max length 512) using TP=2.

Metric	FP16 (Original)	AWQ (4-bit)	Change
VRAM Footprint	~2x Model Size	2.2 GB	Optimized
Model Size	228.3 MB	68.5 MB	-70.0%
NLL Loss	6.2110	6.2790	+7.03% Drift
Perplexity (PPL)	498.1917	533.2338	+7.03% Drift
Generation Time	92.2s	42.2s	Optimized

Usage

Using vLLM

from vllm import LLM, SamplingParams

# Load the model
llm = LLM(model="ThomasYn/GenomeOcean-100M-AWQ", quantization="awq")

# Generate sequences
prompts = ["ATG", "GCA"]
sampling_params = SamplingParams(temperature=0.7, top_p=0.95, max_tokens=100)
outputs = llm.generate(prompts, sampling_params)

for output in outputs:
    print(f"Generated: {output.outputs[0].text}")

Using go-infer (GenomeOcean CLI)

# Scoring sequences
python -m genomeocean.cli score --model_dir ThomasYn/GenomeOcean-100M-AWQ --sequence_file data.txt

# Generating sequences
python -m genomeocean.cli generate --model_dir ThomasYn/GenomeOcean-100M-AWQ --num 10 --max_seq_len 512

Model Repository Structure

This repository contains the necessary configuration files and model weights for AWQ inference.

• model.safetensors: Quantized weights
• config.json: Model configuration
• modeling_mistral.py: Architecture implementation
• tokenizer.json & tokenizer_config.json: Genomic tokenizer files

Citation

If you use this model in your research, please cite:

@article{genomeocean2026,
  title={GenomeOcean: A Large-scale Foundation Model for Ocean Genomics},
  author={Thomas Yn, et al.},
  journal={bioRxiv},
  year={2026}
}