extract emojis

app.py

@@ -18,21 +18,31 @@ tokenizer = AutoTokenizer.from_pretrained("sshleifer/tiny-gpt2")
 model = AutoModelForCausalLM.from_pretrained("sshleifer/tiny-gpt2")
 
 # conversion method
-def text_to_emoji(text):
-    # remove special characters
-    text_cleaned = re.sub(r"[.,!?;:]", "", text)
-    prompt = f"Convert the following sentence into an emoji-sequence which conveys a similar meaning and return only the emojis, no explanation:\n\n\"{text_cleaned}\""
+def text_to_emoji(input_text):
+    # Eingabetext bereinigen (optional)
+    cleaned_text = re.sub(r"[.,!?;:]", "", input_text)
 
-    # Tokenisieren
+    # Prompt vorbereiten
+    prompt = f'Convert the following sentence into an emoji-sequence which conveys a similar meaning and return only the emojis, no explanation:\n\n"{cleaned_text}"\n\n'
+
+    # Tokenisierung und Generation
     inputs = tokenizer(prompt, return_tensors="pt")
-    
-    # Antwort generieren
-    outputs = model.generate(**inputs, max_new_tokens=25, do_sample=True, temperature=0.7)
-    
-    # Antwort decodieren
-    result = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    
-    return result
+    outputs = model.generate(
+        **inputs,
+        max_new_tokens=30,
+        do_sample=True,
+        temperature=0.8,
+        top_k=50
+    )
+
+    # Decodieren
+    generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+    # Nur den Teil nach dem Prompt extrahieren
+    emoji_part = generated_text[len(prompt):].strip()
+
+    return emoji_part
+
 
 # Gradio UI
 iface = gr.Interface(