---
# IMPORTANT: Choose the correct license identifier from https://hf.co/docs/hub/repositories-licenses
license: apache-2.0 # Or cc-by-sa-4.0, mit, etc. - CHOOSE THE CORRECT ONE

# IMPORTANT: Choose the most accurate pipeline tag for your model's task.
# See: https://huggingface.co/docs/hub/models-widgets#pipeline-types
# Examples for genomics:
# token-classification: If predicting labels for each base/token (e.g., is this base part of a TATA box?)
# text-classification: If classifying the whole sequence (e.g., promoter vs. non-promoter)
pipeline_tag: text-classification # <-- EDIT THIS BASED ON YOUR MODEL'S TASK

tags:
- pytorch
- genomics
- dna
- promoter-prediction

---

# GenomeOcean-100M-finetuned-prom_300_tata

## Model Description

This repository contains the `GenomeOcean-100M-finetuned-prom_300_tata` model.
It is a transformer model fine-tuned

You can use this model with the following Python code. Make sure to use the AutoModelFor... class that matches your pipeline_tag (e.g., AutoModelForTokenClassification, AutoModelForSequenceClassification).
```
from transformers import AutoTokenizer, AutoModelForTokenClassification # <-- CHANGE AutoModel class if pipeline_tag is different

model_id = "magicslabnu/GenomeOcean-100M-finetuned-prom_300_tata"

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_id)

# Load model (Ensure the AutoModel class matches your task)
model = AutoModelForTokenClassification.from_pretrained(model_id)

# --- Inference Example ---
# Prepare your DNA sequence(s)
# Ensure sequence format matches what the tokenizer expects (e.g., spaces between bases if needed)
dna_sequence = "[Your example DNA sequence here, e.g., 'A C G T A C G T']"

# Tokenize the input
inputs = tokenizer(dna_sequence, return_tensors="pt") # "pt" for PyTorch

# Perform inference
# For Token Classification:
outputs = model(**inputs)
predictions = outputs.logits.argmax(dim=-1)
# You might need to map prediction IDs back to labels
print("Token predictions:", predictions)

# For Sequence Classification:
# outputs = model(**inputs)
# predictions = outputs.logits.softmax(dim=-1)
# print("Sequence probabilities:", predictions)
# -------------------------

# [Add code here to interpret the predictions based on your specific task
# e.g., mapping token IDs to labels like 'Promoter', 'Non-Promoter', 'TATA-box']
````