Create app.py

app.py

@@ -0,0 +1,41 @@
+import gradio as gr
+import torch
+import numpy as np
+import pickle
+from transformers import AutoTokenizer, AutoModel
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+tokenizer = AutoTokenizer.from_pretrained("facebook/esm2_t12_35M_UR50D")
+model = AutoModel.from_pretrained("facebook/esm2_t12_35M_UR50D")
+model = model.to(device)
+model.eval()
+
+with open("plastic_classifier.pkl", "rb") as f:
+    clf = pickle.load(f)
+
+def get_embedding(sequence, max_length=512):
+    inputs = tokenizer(sequence, return_tensors="pt", truncation=True, max_length=max_length)
+    inputs = {k: v.to(device) for k, v in inputs.items()}
+    with torch.no_grad():
+        outputs = model(**inputs)
+    return outputs.last_hidden_state.mean(dim=1).squeeze().cpu().numpy()
+
+def predict(sequence):
+    sequence = sequence.strip().upper()
+    if not sequence:
+        return "enter a sequence"
+    emb = get_embedding(sequence)
+    proba = clf.predict_proba([emb])[0][1]
+    label = clf.predict([emb])[0]
+    result = "plastic" if label == 1 else "not plastic"
+    return f"{result} ({proba:.2%})"
+
+demo = gr.Interface(
+    fn=predict,
+    inputs=gr.Textbox(label="protein sequence", placeholder="MNFPRASRLMQAAVLGGLMAVSAAATA..."),
+    outputs=gr.Textbox(label="result"),
+    title="protein plastic classifier",
+    description="predicts whether a protein can degrade plastic (PET)"
+)
+
+demo.launch()