karpathy
Who has trained a model as good as the pretrained checkpoint model ?
There are too many hyper-parameters on train.lua and it is hard to train the model as good as the pretrained checkpoint model.
The model I trained only achieve 62%~84% of the pretrained checkpoint model.
Can you share your hyper-parameters if you achieve a better result? I use the mscoco datasets.
How many iterations have you run? could you please share your hyper-parameters?
There are too many hyper-parameters on train.lua and it is hard to train the model as good as the pretrained checkpoint model.
The model I trained only achieve 62%~84% of the pretrained checkpoint model.
Can you share your hyper-parameters if you achieve a better result? I use the mscoco datasets.
What did you do finally? Could you share your experiences? And could you share an accurate pretrained checkpoint model if you have any?
