Update app.py

app.py

@@ -12,39 +12,18 @@ import torch
 from langchain_community.vectorstores import FAISS
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from huggingface_hub import login
-from transformers import BitsAndBytesConfig
 
 # Autenticación en Hugging Face
 huggingface_token = st.secrets["HUGGINGFACE_TOKEN"]
 login(huggingface_token)
 
 # Configurar modelo y tokenizador
-model_name = 'mistralai/Mistral-7B-Instruct-v0.3'
+model_name = 'Fawazzx/SaulLm_Finetuned_10k'
 model_config = AutoConfig.from_pretrained(model_name)
 tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
 tokenizer.pad_token = tokenizer.eos_token
 tokenizer.padding_side = "right"
 
-# Configuración de bitsandbytes para cargar el modelo en 4-bit
-use_4bit = True
-bnb_4bit_compute_dtype = "float16"
-bnb_4bit_quant_type = "nf4"
-use_nested_quant = False
-
-compute_dtype = getattr(torch, bnb_4bit_compute_dtype)
-
-bnb_config = BitsAndBytesConfig(
-    load_in_4bit=use_4bit,
-    bnb_4bit_quant_type=bnb_4bit_quant_type,
-    bnb_4bit_compute_dtype=compute_dtype,
-    bnb_4bit_use_double_quant=use_nested_quant,
-)
-
-model = AutoModelForCausalLM.from_pretrained(
-    model_name,
-    quantization_config=bnb_config,
-)
-
 text_generation_pipeline = pipeline(
     model=model,
     tokenizer=tokenizer,
@@ -52,7 +31,7 @@ text_generation_pipeline = pipeline(
     temperature=0.2,
     repetition_penalty=1.1,
     return_full_text=True,
-    max_new_tokens=300,
+    max_new_tokens=1000,
 )
 
 prompt_template = """