Update README.md

README.md

@@ -27,6 +27,42 @@ Currently working to implement PICARD (Parsing Incrementally for Constrained Aut
 
 Results are currently being evaluated and will be posted here soon.
 
+## Usage
+
+```
+import torch
+from transformers import AutoTokenizer, T5ForConditionalGeneration
+
+# Load the tokenizer and model
+model_path = 'text2sql_model_path'
+tokenizer = AutoTokenizer.from_pretrained(model_path)
+model = T5ForConditionalGeneration.from_pretrained(model_path)
+
+# Function to generate SQL queries
+def generate_sql(prompt, schema):
+    input_text = "translate English to SQL: " + prompt + " " + schema
+    inputs = tokenizer(input_text, return_tensors="pt", max_length=512, truncation=True, padding="max_length")
+
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    inputs = {key: value.to(device) for key, value in inputs.items()}
+
+    max_output_length = 1024
+    outputs = model.generate(**inputs, max_length=max_output_length)
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+# Interactive loop
+print("Enter 'quit' to exit.")
+while True:
+    prompt = input("Insert prompt: ")
+    schema = input("Insert schema: ")
+    if prompt.lower() == 'quit':
+        break
+
+    sql_query = generate_sql(prompt, schema)
+    print(f"Generated SQL query: {sql_query}")
+    print()
+```
+
 ## Files
 
 - `optimizer.pt`: State of the optimizer.