Update app.py

app.py

@@ -1,21 +1,45 @@
 import torch
-from transformers import AutoTokenizer, AutoModel
+from transformers import AutoModelForSequenceClassification, Trainer, TrainingArguments, AutoTokenizer
+from datasets import load_dataset
 
-# Modell & Tokenizer laden
+# 1️⃣ Modell & Tokenizer laden
 model_name = "allenai/scibert_scivocab_uncased"
 tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModel.from_pretrained(model_name)
+model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=3)  # z.B. für 3 Kategorien
 
-# Explizit auf CPU setzen
-device = torch.device("cpu")
-model.to(device)
+# 2️⃣ Dataset laden (ersetze mit deinem Dataset)
+dataset = load_dataset("scientific_papers", "arxiv")  # Hugging Face Datasets
 
-# Beispiel-Text für den Test
-text = "This paper introduces a novel deep learning approach for cancer diagnosis."
+# 3️⃣ Tokenisierung der Texte
+def tokenize_function(examples):
+    return tokenizer(examples["abstract"], padding="max_length", truncation=True)
 
-# Tokenisierung & Modell-Durchlauf
-inputs = tokenizer(text, return_tensors="pt").to(device)
-outputs = model(**inputs)
+tokenized_datasets = dataset.map(tokenize_function, batched=True)
 
-print("Modell erfolgreich geladen!")
-print("Shape der Ausgabe:", outputs.last_hidden_state.shape)
+# 4️⃣ Trainingsparameter setzen
+training_args = TrainingArguments(
+    output_dir="./results",
+    evaluation_strategy="epoch",
+    save_strategy="epoch",
+    per_device_train_batch_size=8,
+    per_device_eval_batch_size=8,
+    num_train_epochs=3,
+    weight_decay=0.01,
+    logging_dir="./logs",
+)
+
+# 5️⃣ Training starten
+trainer = Trainer(
+    model=model,
+    args=training_args,
+    train_dataset=tokenized_datasets["train"],
+    eval_dataset=tokenized_datasets["validation"],
+)
+
+trainer.train()
+
+# 6️⃣ Speichern des Modells nach dem Training
+model.save_pretrained("./trained_model")
+tokenizer.save_pretrained("./trained_model")
+
+print("✅ Training abgeschlossen! Modell gespeichert.")