Closed fabianlim closed 1 week ago
Right now the fused ops do not support dropout, but perhaps it can be quite trivally supported as this is the implementation of the dropout in QuantLinear in both peft.tuners.lora.bnb and peft.tuners.lora.gptq
QuantLinear
peft.tuners.lora.bnb
peft.tuners.lora.gptq
output = lora_B(lora_A(dropout(x)))
Right now the fused ops do not support dropout, but perhaps it can be quite trivally supported as this is the implementation of the dropout in
QuantLinear
in bothpeft.tuners.lora.bnb
andpeft.tuners.lora.gptq