Edit ‘softmax_bottleneck’

This commit is contained in:
osmarks 2024-09-17 20:00:24 +00:00 committed by wikimind
parent 0bda0e4b0e
commit aaa4bfd79c

View File

@ -1,4 +1,4 @@
Almost all modern [[large language model|LLMs]] map relatively low-dimensional hidden states to high-dimensional probability distributions over [[tokenizer|tokens]] using a single [[matrix]] and a [[softmax]] operation. The [[rank]] of this transformation is limited to the hidden size, so not all valid probability distributions can be represented. This has a number of [[consequences]]. Almost all modern [[large language model|LLMs]] map relatively low-dimensional hidden states to high-dimensional probability distributions over [[tokenizer|tokens]] using a single [[matrix]] and a [[softmax]] operation. The [[rank]] of this transformation is limited to the hidden size, so not all valid probability distributions can be represented. Some mixtures of tokens are not representable without introducing additional higher-probability tokens, particularly where a mixture of such would not be common in the training data. This has a number of [[consequences]].
References: References: