Update app.py

app.py

@@ -1,73 +1,51 @@
-import os
-from threading import Thread
-from typing import Iterator
-
 import gradio as gr
-import spaces
-from transformers import AutoModelForCausalLM, AutoTokenizer
-
-# Konfigurationsparameter
-MAX_MAX_NEW_TOKENS = 100
-DEFAULT_MAX_NEW_TOKENS = 20
-MAX_INPUT_TOKEN_LENGTH = 400 # Begrenzung auf 400 Tokens
-
-# Modell und Tokenizer laden
-model_id = "Loewolf/GPT_1"
-model = AutoModelForCausalLM.from_pretrained(model_id)
-tokenizer = AutoTokenizer.from_pretrained(model_id)
-
-# Gradio Chat Interface Funktion
-def generate(
- message: str,
- chat_history: list[tuple[str, str]],
- system_prompt: str,
- max_new_tokens: int = DEFAULT_MAX_NEW_TOKENS,
- temperature: float = 0.6,
- top_p: float = 0.9,
- top_k: int = 50,
- repetition_penalty: float = 1.2,
-) -> str:
- conversation = []
- if system_prompt:
- conversation.append({"role": "system", "content": system_prompt})
- for user, assistant in chat_history:
- conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
- conversation.append({"role": "user", "content": message})
-
- input_ids = tokenizer(conversation, return_tensors="pt", truncation=True, max_length=MAX_INPUT_TOKEN_LENGTH)
- generate_kwargs = dict(
- input_ids=input_ids["input_ids"],
- max_length=input_ids["input_ids"].shape[1] + max_new_tokens,
- temperature=temperature,
- top_p=top_p,
- top_k=top_k,
- repetition_penalty=repetition_penalty,
- pad_token_id=tokenizer.eos_token_id
+from transformers import GPT2LMHeadModel, GPT2Tokenizer
+import torch
+
+# Initialisierung des Modells und des Tokenizers
+tokenizer = GPT2Tokenizer.from_pretrained("Loewolf/GPT_1")
+model = GPT2LMHeadModel.from_pretrained("Loewolf/GPT_1")
+
+def generate_text(prompt):
+ input_ids = tokenizer.encode(prompt, return_tensors="pt")
+
+ # Erstelle eine Attention-Mask, die überall '1' ist
+ attention_mask = torch.ones(input_ids.shape, dtype=torch.bool)
+
+ # Bestimmung der maximalen Länge
+ max_length = model.config.n_positions if len(input_ids[0]) > model.config.n_positions else len(input_ids[0]) + 20
+
+ # Erzeugen von Text mit spezifischen Parametern
+ beam_output = model.generate(
+ input_ids,
+ attention_mask=attention_mask,
+ max_length=max_length,
+ min_length=4, # Mindestlänge der Antwort
+ num_beams=5,
+ no_repeat_ngram_size=2,
+ early_stopping=True,
+ temperature=0.9,
+ top_p=0.90,
+ top_k=50,
+ length_penalty=2.0,
+ do_sample=True,
+ eos_token_id=tokenizer.eos_token_id, # EOS Token setzen
+ pad_token_id=tokenizer.eos_token_id 
  )
 
- outputs = model.generate(**generate_kwargs)
- return tokenizer.decode(outputs[0], skip_special_tokens=True)
+ text = tokenizer.decode(beam_output[0], skip_special_tokens=True)
+ return text
 
-# Gradio Interface
-chat_interface = gr.Interface(
- fn=generate,
- inputs=[
- gr.Textbox(label="Message"),
- gr.JSON(label="Chat History"),
- gr.Textbox(label="System Prompt", lines=2),
- gr.Slider(label="Max new tokens", minimum=1, maximum=MAX_MAX_NEW_TOKENS, step=1, value=DEFAULT_MAX_NEW_TOKENS),
- gr.Slider(label="Temperature", minimum=0.1, maximum=1.0, step=0.1, value=0.6),
- gr.Slider(label="Top-p (nucleus sampling)", minimum=0.05, maximum=1.0, step=0.05, value=0.9),
- gr.Slider(label="Top-k", minimum=1, maximum=1000, step=1, value=50),
- gr.Slider(label="Repetition penalty", minimum=1.0, maximum=2.0, step=0.05, value=1.2)
- ],
+# Erstellung des Chatbot-Interface mit dem Titel "Löwolf Chat"
+iface = gr.Interface(
+ fn=generate_text,
+ inputs=gr.Textbox(label="Schreibe hier...", placeholder="Stelle deine Frage..."),
  outputs="text",
- live=True
+ title="Löwolf Chat",
+ description="Willkommen beim Löwolf Chat. Stelle deine Fragen und erhalte Antworten vom KI-Chatbot."
 )
 
-# Starten des Gradio-Servers
-if __name__ == "__main__":
- chat_interface.launch()
-
+# Starten des Chatbot-Interfaces
+iface.launch()