4 ms·
With ELMo, the pretrained weights are frozen. Only the scalars for ELMo layers are tuned (as well as the additional top-level model, of course).
by phowon 8y ago
With ELMo, the pretrained weights are frozen. Only the scalars for ELMo layers are tuned (as well as the additional top-level model, of course).