Update app.py

app.py

@@ -9,17 +9,18 @@ import torch
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
 # === Tokenizery i modele ABSA ===
-aspect_tokenizer = AutoTokenizer.from_pretrained("EfektMotyla/bert-aspect-ner")
-aspect_model = AutoModelForTokenClassification.from_pretrained("EfektMotyla/bert-aspect-ner").to(device)
+with gr.StatusTracker("Ładowanie modeli..."):
+    aspect_tokenizer = AutoTokenizer.from_pretrained("EfektMotyla/bert-aspect-ner")
+    aspect_model = AutoModelForTokenClassification.from_pretrained("EfektMotyla/bert-aspect-ner").to(device)
 
-sentiment_tokenizer = AutoTokenizer.from_pretrained("EfektMotyla/absa-roberta")
-sentiment_model = AutoModelForSequenceClassification.from_pretrained("EfektMotyla/absa-roberta").to(device)
+    sentiment_tokenizer = AutoTokenizer.from_pretrained("EfektMotyla/absa-roberta")
+    sentiment_model = AutoModelForSequenceClassification.from_pretrained("EfektMotyla/absa-roberta").to(device)
 
-en_to_pl_tokenizer = MarianTokenizer.from_pretrained("gsarti/opus-mt-tc-en-pl")
-en_to_pl_model = MarianMTModel.from_pretrained("gsarti/opus-mt-tc-en-pl").to(device)
+    en_to_pl_tokenizer = MarianTokenizer.from_pretrained("gsarti/opus-mt-tc-en-pl")
+    en_to_pl_model = MarianMTModel.from_pretrained("gsarti/opus-mt-tc-en-pl").to(device)
 
-pl_to_en_tokenizer = MarianTokenizer.from_pretrained("Helsinki-NLP/opus-mt-pl-en")
-pl_to_en_model = MarianMTModel.from_pretrained("Helsinki-NLP/opus-mt-pl-en").to(device)
+    pl_to_en_tokenizer = MarianTokenizer.from_pretrained("Helsinki-NLP/opus-mt-pl-en")
+    pl_to_en_model = MarianMTModel.from_pretrained("Helsinki-NLP/opus-mt-pl-en").to(device)
 
 def translate(texts, tokenizer, model):
     inputs = tokenizer(texts, return_tensors="pt", padding=True, truncation=True).to(device)
@@ -55,15 +56,19 @@ def extract_aspects(text):
         aspects.append(aspect_tokenizer.convert_tokens_to_string(current_tokens).strip())
     return list(set(aspects))
 
-def analyze(text_pl):
+def analyze(text_pl, progress=gr.Progress()):
     try:
+        progress(0, desc="Tłumaczenie na angielski...")
         text_en = translate_pl_to_en([text_pl])[0]
+
+        progress(0.3, desc="Wykrywanie aspektów...")
         aspects_en = extract_aspects(text_en)
         if not aspects_en:
             return "Nie wykryto żadnych aspektów."
-        
+
         results = []
-        for asp in aspects_en:
+        for i, asp in enumerate(aspects_en):
+            progress(0.4 + i/len(aspects_en)*0.6, desc=f"Analiza aspektu: {asp}")
             input_text = f"{text_en} [SEP] {asp}"
             inputs = sentiment_tokenizer(input_text, return_tensors="pt", truncation=True, padding=True).to(device)
             with torch.no_grad():
@@ -74,7 +79,7 @@ def analyze(text_pl):
                 results.append(f"{asp_pl.capitalize()} -> **{sentiment_label}**")
         return "\n".join(results)
     except Exception as e:
-        return f"Błąd: {str(e)}"
+        return f"Błąd podczas analizy: {e}"
 
 # === Gradio UI ===
 demo = gr.Interface(
@@ -82,7 +87,8 @@ demo = gr.Interface(
     inputs=gr.Textbox(label="Komentarz po polsku", placeholder="Np. Pizza była pyszna, ale kelner był nieuprzejmy."),
     outputs=gr.Markdown(label="Wyniki analizy"),
     title="ABSA – Analiza komentarzy restauracyjnych",
-    description="Wykrywa aspekty i przypisuje im sentymenty (pozytywny / negatywny / neutralny / konfliktowy)."
+    description="Wykrywa aspekty i przypisuje im sentymenty (pozytywny / negatywny / neutralny / konfliktowy).",
+    allow_flagging="never"
 )
 
-demo.launch(share=True)
+demo.launch()