[Admin maintenance] Support new ZeroGPU hardware

README.md

@@ -4,7 +4,8 @@ emoji: 🪄
 colorFrom: yellow
 colorTo: purple
 sdk: gradio
-sdk_version: 4.44.1
+sdk_version: 5.49.1
+python_version: "3.12"
 app_file: app.py
 pinned: false
 license: apache-2.0

app.py

@@ -11,10 +11,21 @@ import shlex
 import string
 import subprocess
 import threading
-import spaces
 
-subprocess.run(shlex.split('pip install flash-attn --no-build-isolation'),
-               env=os.environ | {'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"})
+# Python 3.12 removed `importlib.find_loader`; scepter still calls it. Shim it
+# back via `importlib.util.find_spec` so the scepter import chain succeeds.
+import importlib
+if not hasattr(importlib, "find_loader"):
+    import importlib.util
+    def _find_loader(name, path=None):
+        try:
+            spec = importlib.util.find_spec(name)
+        except (ImportError, ValueError):
+            return None
+        return spec.loader if spec is not None else None
+    importlib.find_loader = _find_loader
+
+import spaces
 
 import cv2
 import gradio as gr
@@ -201,7 +212,6 @@ class ChatBotUI(object):
                         self.chatbot = gr.Chatbot(
                             height=600,
                             value=[],
-                            bubble_full_width=False,
                             show_copy_button=True,
                             container=False,
                             placeholder='<strong>Chat Box</strong>')

requirements.txt

@@ -2,7 +2,7 @@ huggingface_hub
 diffusers
 transformers
 gradio_imageslider
-torch==2.4.0
-torchvision
-gradio==4.44.1
-git+https://github.com/modelscope/scepter.git@v1.3.0_dev#egg=scepter
+torch==2.10.0
+torchvision==0.25.0
+https://github.com/Dao-AILab/flash-attention/releases/download/v2.8.1/flash_attn-2.8.1+cu12torch2.10cxx11abiTRUE-cp312-cp312-linux_x86_64.whl
+git+https://github.com/modelscope/scepter.git@v1.3.0_dev#egg=scepter