🔧🚀 Updated app.py: BitNet Model Repo Swap

swapping out the dead “microsoft/bitnet‑b1.58‑2B‑4T” identifier for the live “1bitLLM/bitnet_b1_58-large” repo—which actually carries the needed configuration_bitnet.py (and its tokenizer/model code)

app.py

@@ -1,6 +1,6 @@
 import gradio as gr
 import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer, AutoConfig
+from transformers import AutoModelForCausalLM, AutoTokenizer
 
 # Singleton for model and tokenizer
 _model = None
@@ -9,18 +9,13 @@ _tokenizer = None
 def load_model():
     global _model, _tokenizer
     if _model is None or _tokenizer is None:
-        model_id = "microsoft/bitnet-b1.58-2B-4T"
+        model_id = "1bitLLM/bitnet_b1_58-large"  # patched from microsoft/bitnet‑b1.58‑2B‑4T
         _tokenizer = AutoTokenizer.from_pretrained(
             model_id,
             trust_remote_code=True
         )
-        config = AutoConfig.from_pretrained(
-            model_id,
-            trust_remote_code=True
-        )
         _model = AutoModelForCausalLM.from_pretrained(
             model_id,
-            config=config,
             torch_dtype=torch.bfloat16,
             trust_remote_code=True
         )
@@ -157,4 +152,4 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
 if __name__ == "__main__":
     # Preload model to avoid threading issues
     load_model()
-    demo.launch(ssr_mode=False, share=True)
+    demo.launch(ssr_mode=False, share=True)