1

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+epoch_2.pt filter=lfs diff=lfs merge=lfs -text

app.py

@@ -0,0 +1,64 @@
+import streamlit as st
+import torch
+from transformers import DistilBertTokenizer, DistilBertForSequenceClassification
+import pickle
+
+# Скачать и загрузить токенизатор
+tokenizer = DistilBertTokenizer.from_pretrained('distilbert-base-cased')
+
+# Загружаем квантизированную модель
+def load_quantized_model(model_path, num_labels):
+    model = DistilBertForSequenceClassification.from_pretrained('distilbert-base-cased', num_labels=num_labels)
+    model.load_state_dict(torch.load(model_path, map_location=torch.device('cpu'))['model_state_dict'])
+    return model
+
+model_path = "epoch_2.pt"
+num_labels = 126  # или другое количество меток в зависимости от вашей задачи
+model = load_quantized_model(model_path, num_labels)
+model.eval()  # Перевод модели в режим оценки
+
+# Установите CUDA, если доступно, иначе используйте CPU
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+model.to(device)
+
+# Streamlit интерфейс
+st.title('Arxiv tag classification')
+
+title = st.text_input('Title:', '')
+summary = st.text_area('Summary:', '')
+
+if st.button('Predict'):
+    # Подготовка данных
+    combined_text = f"{title} {summary}"
+    inputs = tokenizer(
+        combined_text,
+        padding=True,
+        truncation=True,
+        max_length=512,
+        return_tensors='pt'
+    ).to(device)
+
+    with torch.no_grad():
+        outputs = model(**inputs)
+        logits = outputs.logits
+        probs = torch.nn.functional.softmax(logits, dim=-1).squeeze()
+
+    # Отсортируем вероятности и классы
+    sorted_indices = torch.argsort(probs, descending=True)
+    cumulative_probs = 0.0
+    selected_indices = []
+
+    # Выбираем классы пока сумма вероятностей не станет >= 0.95
+    for idx in sorted_indices:
+        cumulative_probs += probs[idx].item()
+        selected_indices.append((idx.item(), probs[idx].item()))
+        if cumulative_probs >= 0.95:
+            break
+
+    with open('label_encoder.pkl', 'rb') as f:
+        label_encoder = pickle.load(f)
+
+    # Отображение результата
+    st.write("Predicted classes by probability up to 95%:")
+    for class_id, probability in selected_indices:
+        st.write(f'Class : {label_encoder.inverse_transform([class_id])[0]}, Probability: {probability * 100:.2f}%')

epoch_2.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4288752cd10f875ea22c7aafe7c55082b42745bf9ea80d29e745d37539593d1c
+size 790669426

label_encoder.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a070ca9de881130ec4ccda071fdec6842f604051a7c47507857c0d2343c76eca
+size 9312