Open Aspirinkb opened 2 years ago
The exported pytorch models are both ok.
Hi @Aspirinkb, I'm facing the same issue. Did you find a way to resolve it?
change
fx_t = fx.view(fx.shape[:2], self.feat_sz_s, self.feat_sz_s).contiguous()
to
fx_t = fx.view(1, fx.shape[1], self.feat_sz_s, self.feat_sz_s).contiguous()
the error is removed.
When convert stark_lighting model for the search region branch by
python3 tracking/ORT_lightning_X_trt_complete.py
, a shape miss-match error occur as following, for the template branch is ok: