TY  - EJOUR
AU  - Finkbeiner, Jan Robert
AU  - Neftci, Emre
TI  - On-Chip Learning via Transformer In-Context Learning
M1  - FZJ-2025-01042
PY  - 2024
AB  - Autoregressive decoder-only transformers have become key components for scalable sequence processing and generation models. However, the transformer's self-attention mechanism requires transferring prior token projections from the main memory at each time step (token), thus severely limiting their performance on conventional processors. Self-attention can be viewed as a dynamic feed-forward layer, whose matrix is input sequence-dependent similarly to the result of local synaptic plasticity. Using this insight, we present a neuromorphic decoder-only transformer model that utilizes an on-chip plasticity processor to compute self-attention. Interestingly, the training of transformers enables them to ``learn'' the input context during inference. We demonstrate this in-context learning ability of transformers on the Loihi 2 processor by solving a few-shot classification problem. With this we emphasize the importance of pretrained models especially their ability to find simple, local, backpropagation free, learning rules enabling on-chip learning and adaptation in a hardware friendly manner.
LB  - PUB:(DE-HGF)25
DO  - DOI:10.34734/FZJ-2025-01042
UR  - https://juser.fz-juelich.de/record/1037904
ER  -