Create app.py

app.py

@@ -0,0 +1,49 @@
+import streamlit as st
+import torch
+from transformers import DistilBertForSequenceClassification, DistilBertTokenizerFast
+
+
+target_to_ind = 
+{'cs': 0,
+ 'econ': 1,
+ 'eess': 2,
+ 'math': 3,
+ 'phys': 4,
+ 'q-bio': 5,
+ 'q-fin': 6,
+ 'stat': 7}
+
+ind_to_target = {ind: target for target, ind in target_to_ind.items()}
+
+@st.cache_resource
+def load_model_and_tokenizer():
+    model_name = 'distilbert/distilbert-base-cased'
+    tokenizer = AutoTokenizer.from_pretrained(model_name)
+    model = AutoModelForSequenceClassification.from_pretrained("./weights/model.safetensors", num_labels=len(target_to_ind))
+    
+    return model, tokenizer
+
+
+def get_predict(title: str, abstract: str) -> (str, float, dict):
+    tokenized_text = tokenizer(title + tokenizer.sep_token + abstact[:128], padding="max_length", truncation=True)
+
+    with torch.no_grad():
+        outputs = model(tokenized_text)
+        probs = torch.nn.functional.softmax(out.logits, dim=-1)
+
+        return list(sorted([(p, ind_to_target[i]) for i, p in enumerate(probs)], reversed=True))
+
+
+title = st.text_area("Title ", "", height=100)
+abstract = st.text_area("Abstract ", "", height=150)
+
+if st.button("Классифицировать", key="manual"):
+    if len(title_text) == 0:
+        st.error("Please, provide paper's title")
+    else:
+        with st.spinner("Be patient, I'm doing my best"):
+            predict = get_predict(title, abstract)
+        st.success(f"Предсказанный тэг: **{predict[0][1]}**")
+        
+        
+model, tokenizer = load_model_and_tokenizer()