Update cuda().half()

.flake8

@@ -0,0 +1,21 @@
+[flake8]
+ignore =
+  # E203 whitespace before ':'
+  E203
+  D203,
+  # line too long
+  E501
+per-file-ignores =
+  # imported but unused
+  # __init__.py: F401
+  test_*.py: F401
+exclude =
+  .git,
+  __pycache__,
+  docs/source/conf.py,
+  old,
+  build,
+  dist,
+  .venv
+  pad*.py
+max-complexity = 25

.gitignore

@@ -0,0 +1 @@
+call-activate.bat

app.py

@@ -35,7 +35,7 @@ has_cuda = torch.cuda.is_available()
 if has_cuda:
     model = AutoModel.from_pretrained(
         model_name, trust_remote_code=True
-    ).cuda()  # 3.92G
+    ).cuda().half()  # 3.92G
 else:
     model = AutoModel.from_pretrained(
         model_name, trust_remote_code=True