Skip to main content

ClikaRT::ops::multinomial

function

multinomial()​

Tensor multinomial(
    Tensor probabilities,
    int64_t num_samples,
    bool replacement = false,
    StreamOrDevice s = {}
)

Categorical sampling: draws num_samples category indices per row of a weight tensor.

probabilities holds non-negative weights (they need not sum to 1), typically [batch, num_categories]; the output replaces the category axis with num_samples and is Int64. Without replacement each row samples distinct categories.

Parameters

  • probabilities: non-negative category weights [.., C].
  • num_samples: draws per row; without replacement must be <= C.
  • replacement: sample with replacement; default false.
  • s: stream or device; default = the input's.

Returns: Int64 category indices, [.., num_samples].

Throws

  • ClikaRT::Error: (INVALID_ARGUMENT) when a weight is negative or num_samples exceeds C without replacement.
auto next = ClikaRT::ops::multinomial(probs, 1);   // one token per row

Declared in ClikaRT/compute/ops.h, line 4035