update

app.py

@@ -337,19 +337,11 @@ def http_bot(
 
 
 title_markdown = """
-# 🌋 LLaVA: Large Language and Vision Assistant
+# 🗺️ Spatial-LLaVA
 [[Code]](https://github.com/xi-jiajun/Spatial-LLaVA) [[Model]](https://huggingface.co/rogerxi/Spatial-LLaVA-7B)
 
 ONLY WORKS WITH GPU!
 
-You can load the model with 4-bit or 8-bit quantization to make it fit in smaller hardwares. Setting the environment variable `bits` to control the quantization.
-*Note: 8-bit seems to be slower than both 4-bit/16-bit. Although it has enough VRAM to support 8-bit, until we figure out the inference speed issue, we recommend 4-bit for A10G for the best efficiency.*
-
-Recommended configurations:
-| Hardware          | T4-Small (16G)  | A10G-Small (24G) | A100-Large (40G) |
-|-------------------|-----------------|------------------|------------------|
-| **Bits**          | 4 (default)     | 4                | 16               |
-
 """
 
 tos_markdown = """
@@ -611,7 +603,7 @@ if __name__ == "__main__":
     logger.info(f"args: {args}")
 
     model_path = "rogerxi/Spatial-LLaVA-7B"
-    bits = int(os.getenv("bits", 8))
+    bits = int(os.getenv("bits", 16))
 
     controller_proc = start_controller()
     worker_proc = start_worker(model_path, bits=bits)