Update ui/gpu.py

ui/gpu.py

@@ -20,7 +20,6 @@ Off ZeroGPU (local, or Spaces "CPU basic") @spaces.GPU is a transparent
 passthrough and ON_ZEROGPU is False, so this module quietly degrades to plain
 CPU work and nothing else in the app changes.
 """
-
 import os
 
 import spaces
@@ -36,9 +35,8 @@ GPU_DEVICE = "cuda" if ON_ZEROGPU else "cpu"
 # itself. A long paper on a cold cache is the worst case.
 _VECTORIZE_SECONDS = 120
 
+
 @spaces.GPU(duration=_VECTORIZE_SECONDS)
-zero_gpu_flag = os.environ.get("SPACES_ZERO_GPU")
-print("[GPU FN] pid=" + str(os.getpid()) + ", in ZeroGPU ctx: " + str(zero_gpu_flag))
 def vectorize_on_gpu(pdf_path: str) -> None:
     """Build and persist this paper's vectorstore with the embedder on GPU.
 
@@ -51,5 +49,15 @@ def vectorize_on_gpu(pdf_path: str) -> None:
     The caller then re-opens it on CPU, which costs nothing: build_vectorstore
     short-circuits to a plain load as soon as the persist dir exists.
     """
+    # Runs fresh on every call, inside the GPU worker — unlike a module-level
+    # print, this actually tells you what's true for *this* invocation.
+    zero_gpu_flag = os.environ.get("SPACES_ZERO_GPU")
+    print(
+        "[GPU FN] pid=" + str(os.getpid())
+        + ", in ZeroGPU ctx: " + str(zero_gpu_flag)
+        + ", device=" + GPU_DEVICE
+    )
+
     from vectorizeer import build_vectorstore
-    build_vectorstore(pdf_path, device=GPU_DEVICE)
+
+    build_vectorstore(pdf_path, device=GPU_DEVICE)