README.md

---
license: mit
pipeline_tag: text-generation
tags:
- biology
- genomics
- long-context
library_name: transformers
---
# GENERator-v2-prokaryote-3b-base model

## **Important Notice**
If you are using **GENERator** for sequence generation, please ensure that the length of each input sequence is a multiple of **6**. This can be achieved by either:  
1. Padding the sequence on the left with `'A'` (**left padding**);  
2. Truncating the sequence from the left (**left truncation**).  

This requirement arises because **GENERator** employs a 6-mer tokenizer. If the input sequence length is not a multiple of **6**, the tokenizer will append an `'<oov>'` (out-of-vocabulary) token to the end of the token sequence. This can result in uninformative subsequent generations, such as repeated `'AAAAAA'`.  

We apologize for any inconvenience this may cause and recommend adhering to the above guidelines to ensure accurate and meaningful generation results.

## Abouts
In this repository, we present GENERator-v2, a generative genomic foundation with enhanced performance in prokaryotic domain. More technical details are provided in the GENERator-v2 [technical report](https://www.biorxiv.org/content/10.64898/2026.01.27.702015v1).

Python scripts for downstream analysis are available on Github: [https://github.com/GenerTeam/GENERator](https://github.com/GenerTeam/GENERator).

## How to use

### Simple example1:  generation

```python

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

# Load the tokenizer and model.
tokenizer = AutoTokenizer.from_pretrained("GenerTeam/GENERator-v2-prokaryote-3b-base", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("GenerTeam/GENERator-v2-prokaryote-3b-base")
config = model.config

max_length = config.max_position_embeddings

# Define input sequences.
sequences = [
    "ATGAGGTGGCAAGAAATGGGCTAC",
    "GAATTCCATGAGGCTATAGAATAATCTAAGAGAAAT"
]

def left_padding(sequence, padding_char='A', multiple=6):
    remainder = len(sequence) % multiple
    if remainder != 0:
        padding_length = multiple - remainder
        return padding_char * padding_length + sequence
    return sequence

def left_truncation(sequence, multiple=6):
    remainder = len(sequence) % multiple
    if remainder != 0:
        return sequence[remainder:]
    return sequence

# Apply left_padding to all sequences
# padded_sequences = [left_padding(seq) for seq in sequences]

# Apply left_truncation to all sequences
truncated_sequences = [left_truncation(seq) for seq in sequences]

# Process the sequences
sequences = [tokenizer.bos_token + sequence for sequence in truncated_sequences]

# Tokenize the sequences
tokenizer.padding_side = "left"
inputs = tokenizer(
    sequences,
    add_special_tokens=False,
    return_tensors="pt",
    padding=True,
    truncation=True,
    max_length=max_length
)

# Generate the sequences
with torch.inference_mode():
    outputs = model.generate(**inputs, max_new_tokens=32, temperature=0.00001, top_k=1)

# Decode the generated sequences
decoded_sequences = tokenizer.batch_decode(outputs, skip_special_tokens=True)

# Print the decoded sequences
print(decoded_sequences)

# It is expected to observe non-sense decoded sequences (e.g., 'AAAAAA')
# The input sequences are too short to provide sufficient context.
```

### Simple example2: embedding

```python

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

# Load the tokenizer and model
tokenizer = AutoTokenizer.from_pretrained("GENERator-v2-prokaryote-3b-base", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("GENERator-v2-prokaryote-3b-base")

# Get model configuration
config = model.config
max_length = config.max_position_embeddings

# Define input sequences
sequences = [
    "ATGAGGTGGCAAGAAATGGGCTAC",
    "GAATTCCATGAGGCTATAGAATAATCTAAGAGAAAT"
]

# Truncate each sequence to the nearest multiple of 6
processed_sequences = [tokenizer.bos_token + seq[:len(seq)//6*6] for seq in sequences]

# Tokenization
tokenizer.padding_side = "right"
inputs = tokenizer(
    processed_sequences,
    add_special_tokens=True,
    return_tensors="pt",
    padding=True,
    truncation=True,
    max_length=max_length
)

# Model Inference
with torch.inference_mode():
    outputs = model(**inputs, output_hidden_states=True)

hidden_states = outputs.hidden_states[-1]
attention_mask = inputs["attention_mask"]

# Option 1: Last token (EOS) embedding
last_token_indices = attention_mask.sum(dim=1) - 1
eos_embeddings = hidden_states[torch.arange(hidden_states.size(0)), last_token_indices, :]

# Option 2: Mean pooling over all tokens
expanded_mask = attention_mask.unsqueeze(-1).expand(hidden_states.size()).to(torch.float32)
sum_embeddings = torch.sum(hidden_states * expanded_mask, dim=1)
mean_embeddings = sum_embeddings / expanded_mask.sum(dim=1)

# Output
print("EOS (Last Token) Embeddings:", eos_embeddings)
print("Mean Pooling Embeddings:", mean_embeddings)

# ============================================================================
# Additional notes:
# - The preprocessing step ensures sequences are multiples of 6 for 6-mer tokenizer
# - For causal LM, the last token embedding (EOS) is commonly used
# - Mean pooling considers all tokens including BOS and content tokens
# - The choice depends on your downstream task requirements
# - Both methods handle variable sequence lengths via attention mask
# ============================================================================

```

## Citation
```
@article {li2026generator2,
	author = {Li, Qiuyi and Zhan, Zhihao and Feng, Shikun and Zhu, Yiheng and He, Yuan and Wu, Wei and Shi, Zhenghang and Wang, Shengjie and Hu, Zongyong and Yang, Zhao and Li, Jiaoyang and Tang, Jian and Liu, Haiguang and Qin, Tao},
	title = {Functional In-Context Learning in Genomic Language Models with Nucleotide-Level Supervision and Genome Compression},
	elocation-id = {2026.01.27.702015},
	year = {2026},
	doi = {10.64898/2026.01.27.702015},
	publisher = {Cold Spring Harbor Laboratory},
	URL = {https://www.biorxiv.org/content/early/2026/01/29/2026.01.27.702015},
	journal = {bioRxiv}
}

@article{wu2025generator,
  title={GENERator: a long-context generative genomic foundation model},
  author={Wu, Wei and Li, Qiuyi and Li, Mingyang and Fu, Kun and Feng, Fuli and Ye, Jieping and Xiong, Hui and Wang, Zheng},
  journal={arXiv preprint arXiv:2502.07272},
  year={2025}
}

```