Update app.py

app.py

@@ -22,18 +22,16 @@ MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 HF_TOKEN = os.getenv("HF_TOKEN")
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
-model_name = "tiiuae/Falcon3-10B-Instruct"
+model_name = "ehristoforu/FalconThink-10B-IT"
 
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
-    torch_dtype="auto",
+    torch_dtype="fp16",
     device_map="auto"
 )
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 
-peft_model = AutoPeftModelForCausalLM.from_pretrained("ehristoforu/falconthink-10b-lora")
-merged_model = peft_model.merge_and_unload()
-merged_model.save_pretrained("./falconthink")
+model.save_pretrained("./falconthink")
 
 from huggingface_hub import HfApi
 
@@ -43,7 +41,7 @@ api = HfApi()
 
 api.upload_folder(
     folder_path="./falconthink",
-    repo_id="ehristoforu/FalconThink-10B-IT",
+    repo_id="ehristoforu/FT-10B-fp16",
     repo_type="model",
     token=HF_TOKEN,
 )