Spaces:

Lordemarco
/

sentiment-fastapi

Sleeping

App Files Files Community

LorenzoBioinfo commited on Nov 3, 2025

Commit

a66d87f

1 Parent(s): 0ac2632

Train model also on youtube data and admin page

Browse files

Files changed (5) hide show

app_templates/admin.html +85 -0
app_templates/metrics.html +125 -0
src/app.py +43 -0
src/monitoring.py +24 -0
src/train_model.py +6 -2

app_templates/admin.html ADDED Viewed

	@@ -0,0 +1,85 @@

+<!DOCTYPE html>
+<html lang="it">
+<head>
+  <meta charset="UTF-8">
+  <title>⚙️ Pannello Admin - Sentiment App</title>
+  <style>
+    body {
+      font-family: "Segoe UI", sans-serif;
+      background-color: #f4f6fa;
+      margin: 0;
+      padding: 2rem;
+    }
+    .container {
+      max-width: 800px;
+      margin: 0 auto;
+      background: #fff;
+      padding: 2rem;
+      border-radius: 10px;
+      box-shadow: 0 3px 10px rgba(0,0,0,0.1);
+    }
+    h1 {
+      color: #0052cc;
+      text-align: center;
+    }
+    .section {
+      margin-top: 1.5rem;
+    }
+    button, a.button {
+      background-color: #0052cc;
+      color: white;
+      border: none;
+      padding: 10px 18px;
+      border-radius: 8px;
+      cursor: pointer;
+      text-decoration: none;
+      font-weight: 500;
+      margin-top: 10px;
+      display: inline-block;
+    }
+    button:hover, a.button:hover {
+      background-color: #003d99;
+    }
+    .metrics-link {
+      display: block;
+      text-align: center;
+      margin-top: 1.5rem;
+      font-weight: bold;
+    }
+    .back {
+      display: inline-block;
+      margin-top: 2rem;
+      text-decoration: none;
+      color: #0052cc;
+      text-align: center;
+      width: 100%;
+    }
+  </style>
+</head>
+<body>
+  <div class="container">
+    <h1>⚙️ Pannello di Amministrazione</h1>
+    <p style="text-align:center;">Gestisci il modello di analisi del sentiment, il training e il monitoring.</p>
+    <div class="section">
+      <h3>🎓 Training del Modello</h3>
+      <p>Avvia un nuovo training usando i dati <strong>TweetEval</strong>.</p>
+      <form action="/admin/train" method="post">
+        <button type="submit">Esegui Training</button>
+      </form>
+    </div>
+    <div class="section">
+      <h3>📊 Monitoring</h3>
+      <p>Analizza le performance del modello sui dataset disponibili.</p>
+      <form action="/admin/monitoring" method="post">
+        <button type="submit">Esegui Monitoring</button>
+      </form>
+      <a href="/admin/metrics" class="metrics-link button">📈 Visualizza Metriche</a>
+    </div>
+    <a class="back" href="/">← Torna alla Home</a>
+  </div>
+</body>
+</html>

app_templates/metrics.html ADDED Viewed

	@@ -0,0 +1,125 @@

+<!DOCTYPE html>
+<html lang="it">
+<head>
+  <meta charset="UTF-8">
+  <title>📈 Metriche del Modello</title>
+  <script src="https://cdn.jsdelivr.net/npm/chart.js"></script>
+  <style>
+    body {
+      font-family: "Segoe UI", sans-serif;
+      background-color: #f5f7fa;
+      margin: 0;
+      padding: 2rem;
+    }
+    .container {
+      max-width: 800px;
+      margin: 0 auto;
+      background: white;
+      padding: 2rem;
+      border-radius: 10px;
+      box-shadow: 0 3px 10px rgba(0,0,0,0.1);
+    }
+    h1 {
+      color: #0052cc;
+      text-align: center;
+    }
+    table {
+      width: 100%;
+      border-collapse: collapse;
+      margin-top: 1.5rem;
+    }
+    th, td {
+      text-align: left;
+      padding: 10px;
+      border-bottom: 1px solid #ddd;
+    }
+    th {
+      background-color: #0052cc;
+      color: white;
+    }
+    canvas {
+      margin-top: 30px;
+      width: 100%;
+      height: 300px;
+    }
+    .button {
+      background-color: #0052cc;
+      color: white;
+      border: none;
+      padding: 10px 18px;
+      border-radius: 8px;
+      cursor: pointer;
+      text-decoration: none;
+      font-weight: 500;
+      display: inline-block;
+      margin-top: 1rem;
+    }
+    .button:hover {
+      background-color: #003d99;
+    }
+    .back {
+      display: block;
+      text-align: center;
+      margin-top: 2rem;
+      text-decoration: none;
+      color: #0052cc;
+    }
+  </style>
+</head>
+<body>
+  <div class="container">
+    <h1>📊 Metriche del Modello</h1>
+    {% if metrics %}
+    <table>
+      <thead>
+        <tr>
+          <th>Metrica</th>
+          <th>Valore</th>
+        </tr>
+      </thead>
+      <tbody>
+        {% for key, value in metrics.items() %}
+          <tr>
+            <td>{{ key }}</td>
+            <td>{{ "%.3f"|format(value) }}</td>
+          </tr>
+        {% endfor %}
+      </tbody>
+    </table>
+    <canvas id="metricsChart"></canvas>
+    <script>
+      const ctx = document.getElementById('metricsChart').getContext('2d');
+      const labels = {{ metrics.keys() | list | tojson }};
+      const data = {{ metrics.values() | list | tojson }};
+      new Chart(ctx, {
+        type: 'bar',
+        data: {
+          labels: labels,
+          datasets: [{
+            label: 'Valori delle metriche',
+            data: data,
+            backgroundColor: 'rgba(0, 82, 204, 0.6)',
+            borderRadius: 6
+          }]
+        },
+        options: {
+          scales: {
+            y: { beginAtZero: true }
+          }
+        }
+      });
+    </script>
+    {% else %}
+      <p style="text-align:center;">Nessun dato disponibile. Esegui il monitoring per visualizzare le metriche.</p>
+    {% endif %}
+    <div style="text-align:center;">
+      <a class="button" href="/admin">← Torna all’Area Admin</a>
+    </div>
+  </div>
+</body>
+</html>

src/app.py CHANGED Viewed

@@ -8,6 +8,8 @@ from datasets import load_dataset, load_from_disk
 import torch
 import random
 import subprocess
 # Caricamento del modello e dei dati se già scaricati
 MODEL= "cardiffnlp/twitter-roberta-base-sentiment-latest"
@@ -123,6 +125,47 @@ def random_youtube_comment(request: Request):
     )
 if __name__=="__main__":
     import uvicorn
     uvicorn.run(app,host="0.0.0.0",port=8000)

 import torch
 import random
 import subprocess
+import json
+import os
 # Caricamento del modello e dei dati se già scaricati
 MODEL= "cardiffnlp/twitter-roberta-base-sentiment-latest"
     )
+@app.get("/admin", response_class=HTMLResponse)
+async def admin_dashboard(request: Request):
+    """Pagina principale dell'area admin."""
+    metrics = None
+    metrics_path = "reports/metrics.json"
+    if os.path.exists(metrics_path):
+        with open(metrics_path, "r") as f:
+            metrics = json.load(f)
+    return templates.TemplateResponse(
+        "admin.html",
+        {"request": request, "metrics": metrics}
+    )
+@app.post("/admin/train")
+async def retrain_model():
+    """Lancia lo script di training."""
+    subprocess.run(["python", "src/train.py"], check=True)
+    return {"status": "Training completato"}
+@app.post("/admin/monitor")
+async def run_monitoring():
+    """Esegue il monitoring e aggiorna metrics.json."""
+    subprocess.run(["python", "src/monitoring.py"], check=True)
+    return {"status": "Monitoring completato"}
+@app.get("/admin/metrics", response_class=HTMLResponse)
+def view_metrics(request: Request):
+    """Visualizza i risultati del monitoring in forma tabellare e grafica."""
+    metrics_path = "reports/metrics.json"
+    metrics = None
+    if os.path.exists(metrics_path):
+        with open(metrics_path, "r") as f:
+            metrics = json.load(f)
+    return templates.TemplateResponse(
+        "metrics.html",
+        {"request": request, "metrics": metrics}
+    )
 if __name__=="__main__":
     import uvicorn
     uvicorn.run(app,host="0.0.0.0",port=8000)

src/monitoring.py CHANGED Viewed

@@ -5,7 +5,9 @@ import torch
 import numpy as np
 import json
 import os
 MODEL_PATH = "models/sentiment_model"
 TWEET_PATH = "data/processed/tweet_eval_tokenized"
 YT_PATH = "data/processed/youtube_comments"
@@ -31,6 +33,26 @@ def evaluate_model(model, tokenizer, dataset, dataset_name, sample_size=300):
     print(f"{dataset_name} — Accuracy: {acc:.3f}, F1: {f1:.3f}")
     return {"dataset": dataset_name, "accuracy": acc, "f1": f1, "confusion_matrix": cm}
 def main():
     print("Caricamento del modello")
     model = AutoModelForSequenceClassification.from_pretrained(MODEL_PATH)
@@ -52,5 +74,7 @@ def main():
     print(f"Risultati salvati in: {metrics_path}")
 if __name__ == "__main__":
     main()

 import numpy as np
 import json
 import os
+from src.train_model import train_model
+ACCURACY_THRESHOLD = 0.75
 MODEL_PATH = "models/sentiment_model"
 TWEET_PATH = "data/processed/tweet_eval_tokenized"
 YT_PATH = "data/processed/youtube_comments"
     print(f"{dataset_name} — Accuracy: {acc:.3f}, F1: {f1:.3f}")
     return {"dataset": dataset_name, "accuracy": acc, "f1": f1, "confusion_matrix": cm}
+def retrain_on_youtube_sample():
+    from datasets import load_from_disk
+    youtube_data = load_from_disk(YT_PROCESSED_PATH)["train"]
+    youtube_sample = youtube_data.shuffle(seed=42).select(range(500))
+    train_model(additional_data=youtube_sample, output_dir=MODEL_OUTPUT_PATH)
+def monitor_model():
+    metrics = evaluate_model_on_youtube()
+    print(f"Accuracy su YouTube: {metrics['accuracy']:.3f}")
+    if metrics["accuracy"] < ACCURACY_THRESHOLD:
+        print("Performance sotto la soglia. Avvio retraining parziale...")
+        retrain_on_youtube_sample()
+    return metrics
 def main():
     print("Caricamento del modello")
     model = AutoModelForSequenceClassification.from_pretrained(MODEL_PATH)
     print(f"Risultati salvati in: {metrics_path}")
 if __name__ == "__main__":
     main()

src/train_model.py CHANGED Viewed

@@ -5,7 +5,7 @@ from transformers import (
     TrainingArguments,
     AutoTokenizer
 )
-from datasets import load_from_disk
 import evaluate
 import numpy as np
 import os
@@ -24,9 +24,13 @@ def compute_metrics(eval_pred):
     f1 = metric_f1.compute(predictions=predictions, references=labels, average="weighted")
     return {"accuracy": acc["accuracy"], "f1": f1["f1"]}
-def train_model(sample_train_size=1000, sample_eval_size=300):
     print("Caricamento dataset Tweet eval preprocessato")
     dataset = load_from_disk(DATA_PATH)
     #
     print(f"Riduzione dataset: {sample_train_size} per il train, {sample_eval_size} per la validazione.")

     TrainingArguments,
     AutoTokenizer
 )
+from datasets import load_from_disk,concatenate_datasets
 import evaluate
 import numpy as np
 import os
     f1 = metric_f1.compute(predictions=predictions, references=labels, average="weighted")
     return {"accuracy": acc["accuracy"], "f1": f1["f1"]}
+def train_model(additional_data=None,sample_train_size=1000, sample_eval_size=300):
     print("Caricamento dataset Tweet eval preprocessato")
     dataset = load_from_disk(DATA_PATH)
+    if additional_data is not None:
+        print("Aggiungo dati YouTube al training set...")
+        dataset["train"] = concatenate_datasets([dataset["train"], additional_data])
     #
     print(f"Riduzione dataset: {sample_train_size} per il train, {sample_eval_size} per la validazione.")