fix: replace mmengine set_random_seed import with local impl (avoids PyTorch JIT LaTeX bug)

fix: set num_workers=0 for ZeroGPU daemon process compatibility

fix: from_pretrained must be a .pt file, not model/ dir

fix: disable apex FusedLayerNorm (use nn.LayerNorm)

fix: resolve model.from_pretrained placeholder

fix: move inference in-process to avoid ZeroGPU SIGSEGV (-11)

Fix numpy version conflict: numpy<2.0 + pickle compat shim

init