Progressive layer dropping during domain fine-tuning can halve an LLM's depth with small accuracy loss, yielding 2-5x throughput gains on consumer GPUs.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
TrimLLM: Progressive Layer Dropping for Domain-Specific LLMs
Progressive layer dropping during domain fine-tuning can halve an LLM's depth with small accuracy loss, yielding 2-5x throughput gains on consumer GPUs.