implement the functions

README.md

@@ -11,4 +11,3 @@ license: mit
 short_description: Predict the probability of a chemical compound to be natural
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -1,15 +1,64 @@
 import gradio as gr
+import numpy as np
 import spaces
-import torch
+from transformers import AutoModelForSequenceClassification, AutoTokenizer, Trainer
 
-zero = torch.Tensor([0]).cuda()
-print(zero.device) # <-- 'cpu' 🤔
+# Global variables to store model and tokenizer
+model = None
+tokenizer = None
+
+def load_model(model_path):
+    """Load the fine-tuned model and tokenizer from Hugging Face"""
+    global model, tokenizer
+    model = AutoModelForSequenceClassification.from_pretrained(model_path)
+    tokenizer = AutoTokenizer.from_pretrained(model_path)
+    print(f"Model loaded from {model_path}")
 
 @spaces.GPU
-def greet(n):
-    print(zero.device) # <-- 'cuda:0' 🤗
-    return f"Hello {zero + n} Tensor"
+def predict(input_text):
+        """Make prediction on the input text directly without creating a dataset"""
+        if model is None or tokenizer is None:
+            return "Error: Model not loaded"
+        
+        model.to('cuda')
+        # Tokenize input directly
+        inputs = tokenizer(input_text, padding='max_length', truncation=True, max_length=512, return_tensors="pt")
+        
+        # Move input tensors to GPU
+        inputs = {k: v.to('cuda') for k, v in inputs.items()}
+
+        # Get model predictions
+        outputs = model(**inputs)
+        logits = outputs.logits.detach().cpu().numpy()
+        
+        # Stable softmax to get probabilities
+        exp_logits = np.exp(logits - np.max(logits, axis=1, keepdims=True))
+        probs = exp_logits / np.sum(exp_logits, axis=1, keepdims=True)
+        
+        # Get predicted label
+        pred_label = np.argmax(probs, axis=1)[0]
+        # Map prediction to label
+        label_map = {0: "Unnatural", 1: "Natural"}
+        pred_label_text = label_map[pred_label]
+        
+        # Format output
+        result = f"Prediction: {pred_label_text}\n"
+        natural_prob = probs[0][1] if pred_label == 1 else 1 - probs[0][0]
+        result += f"Natural Product Probability: {natural_prob:.4f}\n"
+        
+        return result
+
+# Load model on initialization
+load_model("shulik7/NP_SMILES_tokenized_PubChem_shard00_160k")  
 
-demo = gr.Interface(fn=greet, inputs=gr.Number(), outputs=gr.Text())
-demo.launch()
+# Create Gradio interface
+iface = gr.Interface(
+    fn=predict,
+    inputs=gr.Textbox(lines=5, placeholder="Enter the SMILES here..."),
+    outputs=gr.Textbox(label="Prediction Results"),
+    title="Naturalness Prediction",
+    description="Enter SMILES string to get the prediction from the fine-tuned ChemBERTa model."
+)
 
+if __name__ == "__main__":
+    iface.launch()

requirements.txt

@@ -0,0 +1,5 @@
+gradio
+spaces
+numpy
+transformers
+torch