Update app.py

app.py

@@ -22,18 +22,34 @@ 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 = "ehristoforu/rufalcon3-3b-it"
+model_name = "Qwen/Qwen2.5-3B-Instruct"
 
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
-    torch_dtype=torch.float16,
+    torch_dtype="fp16",,
     device_map="auto"
 )
-tokenizer = AutoTokenizer.from_pretrained("ehristoforu/rufalcon3-3b-it")
+tokenizer = AutoTokenizer.from_pretrained(model_name)
 
+peft_model = AutoPeftModelForCausalLM.from_pretrained("ehristoforu/CoolQwen2.5-3b-it")
+merged_model = peft_model.merge_and_unload()
+merged_model.save_pretrained("./coolqwen")
+tokenizer.save_pretraiend("./coolqwen")
 
+from huggingface_hub import HfApi
 
-@spaces.GPU(duration=60)
+api = HfApi()
+
+
+
+api.upload_folder(
+    folder_path="./coolqwen",
+    repo_id="ehristoforu/CoolQwen2.5-3B-IT-fp16",
+    repo_type="model",
+    token=HF_TOKEN,
+)
+
+@spaces.GPU(duration=240)
 def generate(
     message: str,
     chat_history: list[tuple[str, str]],
@@ -136,4 +152,4 @@ with gr.Blocks(css="style.css", fill_height=True) as demo:
     chat_interface.render()
 
 if __name__ == "__main__":
-    demo.queue(max_size=20).launch()
+    demo.queue(max_size=20).launch()