Timezone: »

Mimetic Initialization of Self-Attention Layers
Asher Trockman · Zico Kolter

Wed Jul 26 02:00 PM -- 03:30 PM (PDT) @ Exhibit Hall 1 #442

It is notoriously difficult to train Transformers on small datasets; typically, large pre-trained models are instead used as the starting point. We explore the weights of such pre-trained Transformers (particularly for vision) to attempt to find reasons for this discrepancy. Surprisingly, we find that simply initializing the weights of self-attention layers so that they "look" more like their pre-trained counterparts allows us to train vanilla Transformers faster and to higher final accuracies, particularly on vision tasks such as CIFAR-10 and ImageNet classification, where we see gains in accuracy of over 5% and 4%, respectively. Our initialization scheme is closed form, learning-free, and very simple: we set the product of the query and key weights to be approximately the identity, and the product of the value and projection weights to approximately the negative identity. As this mimics the patterns we saw in pre-trained Transformers, we call the technique "mimetic initialization".

Author Information

Asher Trockman (Carnegie Mellon University)
Zico Kolter (Carnegie Mellon University / Bosch Center for AI)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors