a

app.py

@@ -4,6 +4,11 @@ from huggingface_hub import InferenceClient
 """
 Copied from inference in colab notebook
 """
+import torch
+
+# Monkey-patch to avoid CUDA initialization issues
+torch.cuda.get_device_capability = lambda *args, **kwargs: (0, 0)
+
 from unsloth.chat_templates import get_chat_template
 from unsloth import FastLanguageModel
 

requirements.txt

@@ -1,7 +1,3 @@
 huggingface_hub==0.25.2
 
-torch==2.0.1
-torchvision==0.15.2
--f https://download.pytorch.org/whl/cpu
-
 unsloth