Update app.py

app.py

@@ -6,6 +6,7 @@ import gradio as gr
 import spaces
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
+from peft import AutoPeftModelForCausalLM
 
 DESCRIPTION = """\
 # Llama 3.2 3B Instruct
@@ -18,20 +19,36 @@ For more details, please check [our post](https://huggingface.co/blog/llama32).
 MAX_MAX_NEW_TOKENS = 2048
 DEFAULT_MAX_NEW_TOKENS = 1024
 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_id = "ehristoforu/HermesX2"
-tokenizer = AutoTokenizer.from_pretrained(model_id)
+model_name = "tiiuae/Falcon3-10B-Instruct"
+
 model = AutoModelForCausalLM.from_pretrained(
-    model_id,
-    device_map="auto",
-    torch_dtype=torch.bfloat16,
+    model_name,
+    torch_dtype="auto",
+    device_map="auto"
 )
-model.eval()
+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")
+
+from huggingface_hub import HfApi
 
+api = HfApi()
+
+
+
+api.upload_folder(
+    folder_path="./falconthink",
+    repo_id="ehristoforu/FalconThink-10B-IT",
+    repo_type="model",
+    token=HF_TOKEN,
+)
 
-@spaces.GPU(duration=90)
+@spaces.GPU(duration=240)
 def generate(
     message: str,
     chat_history: list[tuple[str, str]],