Add load 8bit back

app.py

@@ -3,17 +3,15 @@ import transformers as t
 import torch
 import peft
 
-checkpoint = 1200
-
 # Load your fine-tuned model and tokenizer
 tokenizer = t.AutoTokenizer.from_pretrained("NousResearch/Llama-2-7b-hf")
-model = t.AutoModelForCausalLM.from_pretrained("NousResearch/Llama-2-7b-hf")
+model = t.AutoModelForCausalLM.from_pretrained("NousResearch/Llama-2-7b-hf",load_in_8bit=True, torch_dtype=torch.float16)
 tokenizer.pad_token_id = 0
 
 config = peft.LoraConfig(r=8, lora_alpha=16, target_modules=["q_proj", "v_proj"], lora_dropout=0.005, bias="none", task_type="CAUSAL_LM")
 model = peft.get_peft_model(model, config)
 
-peft.set_peft_model_state_dict(model, torch.load(f"./output/checkpoint-{checkpoint}/adapter_model.bin"))
+peft.set_peft_model_state_dict(model, torch.load(f".weights/adapter_model.bin"))
 
 # Define a prediction function
 def generate_article(title):