implementation detail · filed under optimization
FP32 attention-output retention
The attention output is kept in FP32 during training.
- source
- 1
- model
- 1
- lab adopt it
- 1
- strongest
- used
How sources treat it
One count per evidence span, weakest treatment to strongest.
used 1
Documented in
Evidence
1 span quoted from the sources, strongest treatment first.
we adopt the method of [98] and keep the attention output in FP32 during training.
usedsoftware implementationin Kimi K3Moonshot AI
Filed alongside
Other methods under optimization :: training precision.
NVFP4BF16NVFP4 pre-trainingFP8 mixed-precision trainingFP4+FP8 mixed precisionFP8-precision reinforcement learningMXFP8BF16 gradient reductionBF16 mixed-precision trainingBlock-wise FP8 activation quantization with offloadE2M1FP32 gradient reductionFP8 storage for the residual stateHigh-precision final network layersMixed-FP8 quantizationMixed-precision trainingNVFP4 fine-grained micro-block scalingTwo-dimensional block quantization