Update README.md

README.md

@@ -19,10 +19,31 @@ We carry out byte-pair encoding (BPE) tokenization on our dataset, tailored for
 
 ## **Usage**
 ```python
-```
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM
 
-### **Example Generation Pipeline**
-```python
+# Load the tokenizer and model
+tokenizer = AutoTokenizer.from_pretrained("metagene-ai/METAGENE-1")
+model = AutoModelForCausalLM.from_pretrained("metagene-ai/METAGENE-1", torch_dtype=torch.bfloat16)
+
+# Example input: Hexamita inflata 5.8S ribosomal RNA gene sequence
+input_sequence = (
+    "TCACCGTTCTACAATCCCAAGCTGGAGTCAAGCTCAACAGGGTCTTCTTGCCCCGCTGAGGGTTACACTCGCCCGTTCCCGAGTCTGTGGTTTCGCGAAGATATGACCAGGGACAGTAAGAACC"
+)
+
+# Tokenize the input sequence and truncate to the first 12 tokens
+input_tokens = tokenizer.encode(input_sequence, return_tensors="pt", add_special_tokens=False)[..., :12]
+
+# Generate output from the model with a max sequence length of 32 tokens
+generated_tokens = model.generate(input_tokens, max_length=32)
+
+# Decode the generated output and clean up the result
+generated_sequence = tokenizer.decode(generated_tokens[0], skip_special_tokens=True)
+generated_sequence = generated_sequence.replace(" ", "").replace("_", "")
+
+# Print the original input and the model's output
+print(f"📄 Input Sequence:\n{input_sequence}")
+print(f"🔬 Generated Sequence:\n{generated_sequence}")
 ```
 
 ## **Benchmark Performance**