Update README.md

README.md

@@ -1,11 +1,11 @@
 # (ALPHA) Turkish Instruct LLM Based On Facebook-XGLM
 
-This is one of our early experimental models based on Facebook's XGLM, fine-tuned using PEFT LORA on a Turkish Instruction dataset. 
+This is one of our early experimental models based on Facebook's XGLM, fine-tuned using PEFT LORA on a Turkish Instruction dataset. Our goal is to share a Turkish LLM model in the future.
 
 The model provided originates from a checkpoint at the 5700th step.
 
 ## Inference
-Please note, utilizing this model necessitates the download of substantial data files, necessitating a minimum of 22GB VRAM.
+Please note, utilizing this model necessitates the download of substantial data files, necessitating a minimum of 24GB VRAM.
 
 Given its experimental nature, it is very likely that this model can produce garbage (biased & wrong) output from time to time, so **use it with caution.** 
 
@@ -28,7 +28,9 @@ bnb_config = BitsAndBytesConfig(
 model = AutoModelForCausalLM.from_pretrained("facebook/xglm-7.5B", 
                                              quantization_config=bnb_config,
                                              device_map="auto")
+
 model = PeftModel.from_pretrained(model, "myzens/XGLM_TR_FineTune_alpha")
+
 tokenizer = AutoTokenizer.from_pretrained("facebook/xglm-7.5B")
 ```