Fix Docker build: correct Python version, add missing json file, use CPU torch

- Change python:3.13.5-slim (nonexistent) to python:3.10-slim
- Copy label_to_topic.json into image (was missing, causing runtime error)
- Install CPU-only torch to avoid ~2GB CUDA download and speed up build

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.13.5-slim
+FROM python:3.10-slim
 
 WORKDIR /app
 
@@ -10,8 +10,10 @@ RUN apt-get update && apt-get install -y \
 
 COPY requirements.txt ./
 COPY src/ ./src/
+COPY label_to_topic.json ./
 
-RUN pip3 install -r requirements.txt
+RUN pip3 install torch --index-url https://download.pytorch.org/whl/cpu && \
+    pip3 install -r requirements.txt
 
 EXPOSE 8501
 

label_to_topic.json

@@ -0,0 +1 @@
+{"0": "Artificial Intelligence", "1": "Computation and Language", "2": "Computer Vision and Pattern Recognition", "3": "Neural and Evolutionary Computing", "4": "Machine Learning", "5": "Machine Learning", "6": "Physics and Society", "7": "Applications", "8": "Robotics", "9": "Software Engineering", "10": "Multiagent Systems", "11": "Optimization and Control", "12": "Information Retrieval", "13": "Disordered Systems and Neural Networks", "14": "Methodology", "15": "Distributed, Parallel, and Cluster Computing", "16": "Computation", "17": "Neurons and Cognition", "18": "Computer Science and Game Theory", "19": "Multimedia", "20": "Cryptography and Security", "21": "Human-Computer Interaction", "22": "Sound", "23": "Graphics", "24": "Numerical Analysis", "25": "Computers and Society", "26": "Data Analysis, Statistics and Probability", "27": "Statistics Theory", "28": "Statistics Theory", "29": "Information Theory", "30": "Information Theory", "31": "Quantum Physics", "32": "Social and Information Networks", "33": "Databases", "34": "Logic in Computer Science", "35": "Adaptation and Self-Organizing Systems", "36": "Systems and Control", "37": "Computational Complexity", "38": "Quantitative Methods", "39": "Networking and Internet Architecture", "40": "Data Structures and Algorithms", "41": "Numerical Analysis", "42": "Discrete Mathematics", "43": "Probability", "44": "Computational Engineering, Finance, and Science", "45": "Programming Languages", "46": "Digital Libraries"}

requirements.txt

@@ -1,3 +1,4 @@
-altair
 pandas
-streamlit
+streamlit
+transformers
+torch

src/streamlit_app.py

@@ -1,40 +1,65 @@
-import altair as alt
 import numpy as np
-import pandas as pd
 import streamlit as st
+from typing import Dict, List
+import torch
+import json
+from transformers import AutoModelForSequenceClassification, AutoTokenizer
 
-"""
-# Welcome to Streamlit!
-
-Edit `/streamlit_app.py` to customize this app to your heart's desire :heart:.
-If you have any questions, checkout our [documentation](https://docs.streamlit.io) and [community
-forums](https://discuss.streamlit.io).
-
-In the meantime, below is an example of what you can do with just a few lines of code:
-"""
-
-num_points = st.slider("Number of points in spiral", 1, 10000, 1100)
-num_turns = st.slider("Number of turns in spiral", 1, 300, 31)
-
-indices = np.linspace(0, 1, num_points)
-theta = 2 * np.pi * num_turns * indices
-radius = indices
-
-x = radius * np.cos(theta)
-y = radius * np.sin(theta)
-
-df = pd.DataFrame({
-    "x": x,
-    "y": y,
-    "idx": indices,
-    "rand": np.random.randn(num_points),
-})
-
-st.altair_chart(alt.Chart(df, height=700, width=700)
-    .mark_point(filled=True)
-    .encode(
-        x=alt.X("x", axis=None),
-        y=alt.Y("y", axis=None),
-        color=alt.Color("idx", legend=None, scale=alt.Scale()),
-        size=alt.Size("rand", legend=None, scale=alt.Scale(range=[1, 150])),
-    ))
+@st.cache_resource
+def load_model():
+    model = AutoModelForSequenceClassification.from_pretrained("Kulebyaka-kokosik/articles-classifier-model")
+    tokenizer = AutoTokenizer.from_pretrained("distilbert/distilbert-base-cased")
+    model.eval()
+    return model, tokenizer
+
+
+@st.cache_resource
+def load_label_to_topic():
+    with open("label_to_topic.json") as file:
+        label_to_topic = json.load(file)
+    return label_to_topic
+
+
+def inference(
+        model: AutoModelForSequenceClassification,
+        tokenizer: AutoTokenizer,
+        title: str,
+        summary: str
+) -> torch.Tensor:
+    tokenized = tokenizer(title, summary, padding='max_length', truncation=True, return_tensors="pt")
+
+    with torch.no_grad():
+        logits = model(**tokenized).logits
+    return logits
+
+
+def predict_topics(
+        model: AutoModelForSequenceClassification,
+        tokenizer: AutoTokenizer,
+        title: str,
+        summary: str,
+        label_to_topic: Dict[str, str],
+        top_k: int = 5,
+) -> List[str]:
+    logits = inference(model, tokenizer, title, summary)
+
+    probs = torch.sigmoid(logits).squeeze(0)
+    labels = np.argsort(-probs)[:top_k].tolist()
+    topics = [label_to_topic[str(label)] for label in labels]
+    return topics
+
+
+
+def main():
+    model, tokenizer = load_model()
+    label_to_topic = load_label_to_topic()
+
+    st.title("Articles classifier")
+
+    title = st.text_input("Title")
+    abstract = st.text_area("Abstract")
+
+    topics = predict_topics(model, tokenizer, title, abstract, label_to_topic=label_to_topic)
+    st.write(topics)
+
+main()