dense_auto_fp8_or_bf16

metrale_model_layers::weight_map::model_a

Function dense_auto_fp8_or_bf16 

Source 
pub fn dense_auto_fp8_or_bf16(
    store: &WeightStore,
    prefix: &str,
    gpu: &dyn GpuBackend,
) -> Result<DenseWeight>
Expand description

2026-09-25: Load {prefix}.weight as BF16: a BF16 tensor is the store pointer, FP8E4M3 goes through dequant_fp8_blockscaled_to_bf16, and any other dtype is an error.