Update app.py

app.py

@@ -1,28 +1,20 @@
-from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig, pipeline
+from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
 import torch
 import gradio as gr
 
 model_id = "eduard76/Llama3-8b-good-new"
 
-#quant_config = BitsAndBytesConfig(
-#    load_in_4bit=True,
-#    bnb_4bit_compute_dtype=torch.float16,
-#    bnb_4bit_use_double_quant=True,
-#    bnb_4bit_quant_type="nf4"
-#)
-
 tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
+
 model = AutoModelForCausalLM.from_pretrained(
     model_id,
-    device_map="auto",
-    #torch_dtype=torch.float16,
-    quantization_config=quant_config,
+    device_map="auto",  # poate fi "cuda:0" sau "cpu" dacă ai eroare
+    torch_dtype=torch.float16,  # sau .bfloat16 dacă vrei
     trust_remote_code=True
 )
 
 pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
 
-# 🔐 Lista de topicuri din dataset (poți ajusta manual dacă vrei):
 covered_topics = {
     "ospf", "bgp", "eigrp", "vxlan", "evpn", "network design", "acl", "routing",
     "spine", "leaf", "underlay", "overlay", "mpls", "qos", "firewall",
@@ -32,7 +24,7 @@ covered_topics = {
 def chat(user_input):
     prompt = f"""You are a Cisco-certified network architect trained in OSPF, BGP, EIGRP, VLAN, STP, RSTP design principles.
 If the user's question is unclear, clarify first.
-If the topic is outside  OSPF, BGP, EIGRP, VLAN, STP, RSTP, respond with: "I'm not trained on that topic."
+If the topic is outside OSPF, BGP, EIGRP, VLAN, STP, RSTP, respond with: "I'm not trained on that topic."
 Give short, clear, non-repetitive answers.
 
 User: {user_input}
@@ -51,7 +43,6 @@ AI:"""
 
     return response[len(prompt):].strip()
 
-
 iface = gr.Interface(fn=chat, inputs="text", outputs="text", title="Eduard's 1st virtual Architect")
 
 if __name__ == "__main__":