flashinfer.cake_lm_head_nondet_train.cake_lm_head_nondet_train_support_logprob

flashinfer.cake_lm_head_nondet_train.cake_lm_head_nondet_train_support_logprob(X: Tensor, W: Tensor, support_ids: Tensor, support_offsets: Tensor, labels: Tensor | None = None, *, k_cap: int, temperature: None | float | Tensor = 1.0, chunk_size: int | None = None, check_inputs: bool = False, backend: str = 'cake') → Tensor

Differentiable log-probabilities over a ragged candidate support per token, SM100 / SM103 / SM107 only.

See flashinfer.cake_lm_head_det_train.cake_lm_head_det_train_support_logprob() for the computation and the arguments. The entry makes no run-to-run bitwise promise for the gradients: its backward raises PyTorch’s alert under torch’s deterministic mode before a dense dX of more than one K slice (with warn_only=True it warns and runs on); there is no fallback to the deterministic entry.