Update app.py

app.py

@@ -4,22 +4,22 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 import torch
 
 # Load the model and config when the script starts
-config = PeftConfig.from_pretrained("phearion/bigbrain-v0.0.1")
-
-model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-v0.1")
-model = PeftModel.from_pretrained(model,"phearion/bigbrain-v0.0.1")
-
-# Load the tokenizer
-tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-v0.1")
+peft_model_id = "phearion/bigbrain-v0.0.1"
+config = PeftConfig.from_pretrained(peft_model_id)
+model = AutoModelForCausalLM.from_pretrained(config.base_model_name_or_path, return_dict=True)
+tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)
 
+# Load the Lora model
+model = PeftModel.from_pretrained(model, peft_model_id)
 
 
 def greet(text):
-    batch = tokenizer(f"'{text}' ->: ", return_tensors='pt')
+    batch = tokenizer(f"\"{text}\" ->: ", return_tensors='pt')
     
     # Use torch.no_grad to disable gradient calculation
     with torch.no_grad():
-        output_tokens = model.generate(**batch, do_sample=True, max_new_tokens=15)
+        output_tokens = model.generate(**batch, do_sample=True, max_new_tokens=15
+)
 
     return tokenizer.decode(output_tokens[0], skip_special_tokens=True)