requirements: remove invalid stdlib entries (pickle, asyncio.selector_events)

Fix infinite loop: build test_dl explicitly with num_workers=0; suppress progress bar

Fix: num_workers=0 prevents DataLoader deadlock in container; ssr_mode=False

Rewrite: minimal clean app with targeted asyncio fix and compact plum shims

Silence asyncio GC teardown ValueError (invalid fd -1) with Gradio

Gradio: assign demo variable; disable experimental SSR

Remove debug logging; keep plum unpickle shims

Harvest plum Function impls from _f/_resolved/_resolver; NotFoundLookupError uses MethodList()

Compat1xFunction: init _pending for plum parity

Compat1xFunction: implement _resolve_method_with_cache for fasttransform

Plum MethodList: use list subclass shim for unpickle append

Log _BUILD_TAG in startup for deploy verification

Force Space rebuild: plum setstate tuple fix + _Compat1xFunction tuple state

Sync app, requirements, and model from fastai_model_cursor (plum pickle compat)

Fix: custom Unpickler shim for plum 1.x Function; revert plum-dispatch pin

Fix: pin plum-dispatch<2.0 to restore plum.function required by exported .pkl

Fix: bypass load_learner, use torch.load directly; add probes for torch.load compat

Debug: probe torch.load pickle_module compat + robust patch for PyTorch 2.11

Fix: patch torch.load for PyTorch 2.6+ weights_only default change

Remove debug instrumentation, clean up app.py

Upload app.py with huggingface_hub

Upload README.md with huggingface_hub

Upload requirements.txt with huggingface_hub

Upload app.py with huggingface_hub

Upload folder using huggingface_hub

initial commit