For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).
Mojo struct
Struct_topk_topp_masked_probs
struct Struct_topk_topp_masked_probs
Registers the sampler.topk_topp_masked_probs graph op.
Writes each row's top-k/top-p masked renormalized softmax, without sampling. Speculative decoding verification reads the target's masked probabilities and builds its rejection residual straight from this tensor, so nothing is sorted and no distribution is rebuilt in-graph.
Implemented traits
Methods
execute
static def execute[dtype: DType, target: StringSpan[ImmStaticOrigin], _trace_name: StringSpan[ImmStaticOrigin]](probs: ManagedTensorSlice[IOSpec[_, _].Output, static_spec=probs.static_spec], K: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=K.static_spec], temperature: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=temperature.static_spec], top_p: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=top_p.static_spec], input: ManagedTensorSlice[IOSpec[_, _].Input, static_spec=input.static_spec], ctx: DeviceContext)