second model ensemble

app.py

@@ -3,12 +3,18 @@ from transformers import AutoTokenizer, AutoModelForSequenceClassification
 import torch
 
 model_path = "modernbert.bin"
+huggingface_model_url = "https://huggingface.co/mihalykiss/modernbert_2/resolve/main/Model_groups_3class_seed12"
 device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
 tokenizer = AutoTokenizer.from_pretrained("answerdotai/ModernBERT-base")
-model = AutoModelForSequenceClassification.from_pretrained("answerdotai/ModernBERT-base", num_labels=41)
-model.load_state_dict(torch.load(model_path, map_location=device))
-model.to(device).eval()
+
+model_1 = AutoModelForSequenceClassification.from_pretrained("answerdotai/ModernBERT-base", num_labels=41)
+model_1.load_state_dict(torch.load(model_path, map_location=device))
+model_1.to(device).eval()
+
+model_2 = AutoModelForSequenceClassification.from_pretrained("answerdotai/ModernBERT-base", num_labels=41)
+model_2.load_state_dict(torch.hub.load_state_dict_from_url(huggingface_model_url, map_location=device))
+model_2.to(device).eval()
 
 label_mapping = {
     0: '13B', 1: '30B', 2: '65B', 3: '7B', 4: 'GLM130B', 5: 'bloom_7b',
@@ -30,7 +36,11 @@ def classify_text(text):
     inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True).to(device)
 
     with torch.no_grad():
-        probabilities = torch.softmax(model(**inputs).logits, dim=1)[0]
+        logits_1 = model_1(**inputs).logits
+        logits_2 = model_2(**inputs).logits
+
+        avg_logits = (logits_1 + logits_2) / 2
+        probabilities = torch.softmax(avg_logits, dim=1)[0]
 
     ai_probs = probabilities.clone()
     ai_probs[24] = 0
@@ -53,7 +63,6 @@ def classify_text(text):
     return result_message
 
 
-
 title = "AI Text Detector"
 
 description = """