Pin torch 2.9.1 + torchvision 0.24.1 (Qwen3VLVideoProcessor needs torchvision)

requirements.txt

@@ -1,11 +1,13 @@
-# torch is deliberately NOT pinned here: the ZeroGPU image provides a CUDA build of torch, and the
-# ZeroGPU docs list "almost all versions from 2.8.0 to latest" as supported (2.8.0, 2.9.1, 2.10.0,
-# 2.11.0, 2.12.1, 2.13.0). ASSUMPTION, not yet checked against a real Space build: the image ships
-# one of those. Everything in this release was validated on torch 2.9.1+cu128, which is on that
-# list. Gate the first Space build on its log: if it resolves a torch outside 2.8-2.13, add
-# `torch==2.9.1` here and record the resolved version in README.md before going public.
-# gradio is not pinned here either: `sdk_version` in README.md owns the Gradio version on Spaces
+# torch/torchvision are pinned as a matching pair. The first Space build left them to the resolver,
+# which (a) picked torch 2.14.0 - outside the versions ZeroGPU supports - and (b) never installed
+# torchvision at all, because nothing in this list depends on it. transformers 5's Qwen3VLProcessor
+# always builds a Qwen3VLVideoProcessor, which hard-requires torchvision, so QwenImage21Pipeline
+# .from_pretrained died with "Qwen3VLVideoProcessor requires the Torchvision library". 2.9.1/0.24.1
+# is the pair this release was validated on and 2.9.1 is on the ZeroGPU supported list.
+# gradio is not pinned here: `sdk_version` in README.md owns the Gradio version on Spaces
 # (5.50.0, the version app.py's component/argument use was checked against).
+torch==2.9.1
+torchvision==0.24.1
 
 # QwenImage21Pipeline is not in any released diffusers (0.40.0 on PyPI has no qwenimage21 files),
 # so install from upstream main at the commit this model was validated against.