Update src/streamlit_app.py

src/streamlit_app.py

@@ -1,40 +1,103 @@
-import altair as alt
-import numpy as np
-import pandas as pd
 import streamlit as st
+import torch
+import numpy as np
+from huggingface_hub import hf_hub_download
+import pickle
+from transformers import AutoTokenizer, AutoModelForSequenceClassification
+
+st.set_page_config(
+    page_title="Anime Genre Classifier",
+)
+
+bert_model = "NickolayFM/ml2_hw_anime_space"
+
+@st.cache_resource
+def load_model():
+    tokenizer = AutoTokenizer.from_pretrained(bert_model)
+    model = AutoModelForSequenceClassification.from_pretrained(bert_model)
+    model.eval()
+
+    mlb_path = hf_hub_download(repo_id=bert_model, filename="mlb.pkl")
+    with open(mlb_path, "rb") as f:
+        mlb = pickle.load(f)
+
+    return tokenizer, model, mlb
+
+
+def predict_genres(synopsis, tokenizer, model, mlb):
+    encoding = tokenizer(
+        synopsis,
+        truncation=True,
+        max_length=256,
+        padding="max_length",
+        return_tensors="pt"
+    )
+
+    with torch.no_grad():
+        outputs = model(
+            input_ids=encoding["input_ids"],
+            attention_mask=encoding["attention_mask"]
+        )
+
+    probs = torch.sigmoid(outputs.logits).cpu().numpy()[0]
+    total = probs.sum()
+    if total == 0:
+        return []
+    probs_norm = probs / total
+    sorted_idx = np.argsort(probs_norm)[::-1]
+    result = []
+    cumsum = 0.0
+    for idx in sorted_idx:
+        result.append((mlb.classes_[idx], float(probs[idx])))
+        cumsum += probs_norm[idx]
+        if cumsum >= 0.95:
+            break
+
+    return result
+
+
+st.title("Anime Genre Classifier")
+st.write("Введи описание аниме — модель предскажет жанры.")
+
+synopsis = st.text_area(
+    label="Описание аниме (на АНГЛИЙСКОМ)",
+    placeholder="Примерчик: Determined to put his life back on track, Keyaru decided to unleash a powerful.",
+    height=100
+)
+
+predict_button = st.button("Показать жанры", type="primary")
+
+if predict_button:
+    if not synopsis.strip():
+        st.warning("Ошибочка: пустое описание")
+    elif len(synopsis.strip().split()) < 5:
+        st.warning("Напиши хоть пару предложений")
+    else:
+        with st.spinner("Работаем"):
+            tokenizer, model, mlb = load_model()
+            genres = predict_genres(synopsis, tokenizer, model, mlb)
+            if not genres:
+                st.error("Ошибочка: не удалось определить жанры")
+            else:
+                st.subheader("Предсказанные жанры:")
+                for genre, prob in genres:
+                    st.write(f"**{genre}**")
+                    st.progress(float(prob))
+                    st.caption(f"Вероятность: {prob:.4%}")
+
+st.divider()
+st.subheader("Примеры для теста:")
+
+
+for title, text in {"Dragon Ball": "Goku is a young boy with a monkey tail and incredible strength. He embarks on a journey to collect the seven Dragon Balls, meeting friends and fighting powerful enemies along the way.", "Spirited Away": "A young girl named Chihiro wanders into a spirit world. Her parents are turned into pigs, and she must work in a bathhouse to save them and return to the human world.", "Death Note": "A high school student finds a supernatural notebook that allows him to kill anyone whose name he writes in it. He decides to use it to rid the world of criminals."}.items():
+    if st.button(f"📺 {title}"):
+        st.session_state["example_text"] = text
+        st.rerun()
 
-"""
-# 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])),
-    ))
+if "example_text" in st.session_state:
+    st.text_area(
+        "Описание (из примера):",
+        value=st.session_state["example_text"],
+        height=100
+    )
+    del st.session_state["example_text"]