Switch to smaller SmolLM2 model variant

- Change model from SmolLM2-1.7B-Instruct to SmolLM2-135M-Instruct
- Update comments to improve clarity and readability
- Maintain existing cache directory and authentication configuration

app.py

@@ -10,7 +10,7 @@ import os
 from dotenv import load_dotenv
 load_dotenv()
 
-# Configurar directorio de caché en un lugar con permisos
+# Configure cache directory
 os.environ["TRANSFORMERS_CACHE"] = "/tmp/transformers_cache"
 os.environ["HF_HOME"] = "/tmp/hf_home"
 
@@ -19,20 +19,20 @@ HUGGINGFACE_TOKEN = os.environ.get("HUGGINGFACE_TOKEN", os.getenv("HUGGINGFACE_T
 print(f"Token HuggingFace: {HUGGINGFACE_TOKEN}")
 
 # Model to use
-MODEL_NAME = "HuggingFaceTB/SmolLM2-1.7B-Instruct"
+MODEL_NAME = "HuggingFaceTB/SmolLM2-135M-Instruct"
 
 # Initialize the model and tokenizer locally with authentication
 print(f"Loading model {MODEL_NAME} locally...")
 tokenizer = AutoTokenizer.from_pretrained(
     MODEL_NAME,
     token=HUGGINGFACE_TOKEN,  # Add token for authentication
-    cache_dir="/tmp/transformers_cache"  # Especificar directorio de caché
+    cache_dir="/tmp/transformers_cache"  # Specify cache directory
 )
 model = AutoModelForCausalLM.from_pretrained(
     MODEL_NAME, 
     device_map="auto",
     token=HUGGINGFACE_TOKEN,  # Add token for authentication
-    cache_dir="/tmp/transformers_cache"  # Especificar directorio de caché
+    cache_dir="/tmp/transformers_cache"  # Specify cache directory
 )
 
 # Create a pipeline to facilitate generation