improve

app.py

@@ -5,10 +5,13 @@ from time import time
 import gradio as gr
 import psutil
 
+# load like this - use tne variable everywhere 
+model_path=os.getenv("MODEL_PATH")
+# show warning, when empty and briefs description of how to set it 
+
 # Initing things                
-print("debug: init model")
+print(f"debug: init model: {model_path}")
 #llm = Llama(model_path="./model.bin")                             # LLaMa model
-llama_model_name = "TheBloke/dolphin-2.2.1-AshhLimaRP-Mistral-7B-GGUF" 
 print("! INITING DONE !")
 
 # Preparing things to work
@@ -45,13 +48,14 @@ MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 
 #download model here
 # check localstorage, if no there, load, else use existing.
+# check gradio - how does it dl? is there a function we can use?
 
 if torch.cuda.is_available():
     model_id = "mistralai/Mistral-7B-Instruct-v0.1"
     model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16, device_map="auto")
     tokenizer = AutoTokenizer.from_pretrained(model_id)
 
-
+# we need to make sure we only run one thread or we probably run out of ram
 def generate(
     message: str,
     chat_history: list[tuple[str, str]],
@@ -133,6 +137,7 @@ chat_interface = gr.ChatInterface(
         ),
     ],
     stop_btn=None,
+    # add more eval examples, like a long list taken from teknium and others maybe group by type
     examples=[
         ["Hello there! How are you doing?"],
         ["Can you explain briefly to me what is the Python programming language?"],
@@ -149,6 +154,7 @@ with gr.Blocks(css="style.css") as demo:
         value="Duplicate Space for private use",
         elem_id="duplicate-button",
         visible=os.getenv("SHOW_DUPLICATE_BUTTON") == "1",
+        # add 
     )
     chat_interface.render()