Re-assert offline mode inside the ZeroGPU worker itself

The previous fix (1cb4beb) patched huggingface_hub.constants.HF_HUB_OFFLINE
in the main process at startup, verified correct in isolation — but a live
diagnostic (direct HTTP call to /gradio_api/queue/join for MMAudio's
_run_mmaudio, bypassing the browser entirely to reproduce the "Connection
errored out" report) showed open_clip's CLIP loader still re-downloading its
~4GB checkpoint live inside the actual GPU call, confirming the worker
process ZeroGPU dispatches @spaces.GPU calls to doesn't reliably inherit
that patch from the main process.

Fix: re-set both os.environ["HF_HUB_OFFLINE"] and the huggingface_hub
constant at the top of _catch_oom's wrapper, which already wraps all six
@spaces.GPU functions as the innermost decorator — so it executes inside
whatever process actually runs the GPU call, at call time, regardless of
ZeroGPU's fork/spawn timing.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

app.py

@@ -744,9 +744,22 @@ def _catch_oom(fn):
     checkpoint re-downloaded, ~2 minutes) instead of just showing an error.
     Applied as the innermost decorator (below @spaces.GPU) so spaces.GPU's
     duration estimator still sees the original call signature via
-    functools.wraps."""
+    functools.wraps.
+
+    Also re-asserts offline mode here, at the top of the actual function body.
+    ZeroGPU runs @spaces.GPU functions in a separate worker process, and that
+    worker does not reliably inherit the main process's post-import os.environ
+    / huggingface_hub.constants mutations made at startup (confirmed live: the
+    same os.environ + constants patch that fixed offline resolution when
+    tested in-process still didn't stop MMAudio's open_clip CLIP loader from
+    re-downloading its ~4GB checkpoint from inside an actual GPU call). Setting
+    it again here runs inside whatever process actually executes the GPU
+    call, so it can't miss the worker regardless of ZeroGPU's fork/spawn
+    timing."""
     @functools.wraps(fn)
     def _wrapped(*args, **kwargs):
+        os.environ["HF_HUB_OFFLINE"] = "1"
+        _hf_constants.HF_HUB_OFFLINE = True
         try:
             return fn(*args, **kwargs)
         except torch.cuda.OutOfMemoryError as e: