Salience-DETR
Salience-DETR copied to clipboard
pos_embed算子输入的shape不匹配
Question
跑训练过程遇到 pos_embed算子输入的shape不匹配,请教大概是什么原因呢?我是pt2.3,其他以来版本是requirments.txt中内容
[rank5]: File "/torch/venv3/pytorch/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl [rank5]: return forward_call(*args, **kwargs) [rank5]: File "/gpfs/xj/Salience-DETR_mlu/models/bricks/salience_transformer.py", line 565, in forward [rank5]: query_with_pos = key_with_pos = self.with_pos_embed(query, query_pos) [rank5]: File "/gpfs/xj/Salience-DETR_mlu/models/bricks/salience_transformer.py", line 545, in with_pos_embed [rank5]: return tensor if pos is None else tensor + pos [rank5]: RuntimeError: The size of tensor a (1092) must match the size of tensor b (192) at non-singleton dimension 1
补充信息
No response