revert unquantized

app.py

@@ -1,15 +1,22 @@
 import gradio as gr
-from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
+from transformers import AutoTokenizer, pipeline
+from peft import AutoPeftModelForCausalLM
+import torch
 
-checkpoint = 'matteosz/phi3-mini-4k-sft-dpo-quant'
+checkpoint = 'ernestoBocini/Phi3-mini-DPO-Tuned'
 
 tokenizer = AutoTokenizer.from_pretrained(checkpoint, trust_remote_code=True)
 tokenizer.model_max_length = 256
-model = AutoModelForCausalLM.from_pretrained(
+tokenizer.pad_token = tokenizer.unk_token 
+tokenizer.pad_token_id = tokenizer.convert_tokens_to_ids(tokenizer.pad_token)
+tokenizer.padding_side = 'left'
+
+model = AutoPeftModelForCausalLM.from_pretrained(
     checkpoint,
     trust_remote_code=True,
     device_map='auto',
-).eval()
+    torch_dtype=torch.bfloat16
+).merge_and_unload().to(torch.float16).eval()
 
 pipe = pipeline(
     "text-generation",