search

PatrickHua / SimSiam

A pytorch implementation for paper 'Exploring Simple Siamese Representation Learning'
MIT License
814 stars 135 forks source link

bugfix and support ckpt saved by DDP #14

Closed youqingxiaozhua closed 3 years ago

youqingxiaozhua commented 3 years ago
  1. model defined in #50 will overwrite the input parameter model passed by main.py which makes --eval_from out of action.
  2. support checkpoint file saved by DDP
leodesigner commented 3 years ago

Number of classes is hardcoded again :( https://github.com/PatrickHua/SimSiam/blob/d22ce2abbfc230e37cd055f14f932afba9360a08/linear_eval.py#L68

youqingxiaozhua commented 3 years ago

Number of classes is hardcoded again :( https://github.com/PatrickHua/SimSiam/blob/d22ce2abbfc230e37cd055f14f932afba9360a08/linear_eval.py#L68

Sorry about that, but if you only modify the first defined classifier, the later one will overwrite it. Maybe we need another PR now