[MUSA]: Add flashinfer sampling backend (#24978)

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: zhiguo.qin <zhiguo.qin@mthreads.com>
Co-authored-by: R0CKSTAR <yeahdongcn@gmail.com>
This commit is contained in:
ori
2026-05-14 20:23:15 -07:00
committed by GitHub
co-authored by github-actions[bot] zhiguo.qin R0CKSTAR
parent d9fa84b25b
commit 897587b03a
9 changed files with 229 additions and 4 deletions
+18
View File
@@ -78,3 +78,21 @@ void musa_top_k_top_p_sampling_from_probs(
double top_p_val,
bool deterministic,
std::optional<at::Generator> gen);
void min_p_sampling_from_probs(
at::Tensor probs,
at::Tensor output,
std::optional<at::Tensor> maybe_indices,
std::optional<at::Tensor> maybe_min_p_arr,
double min_p_val,
bool deterministic,
std::optional<at::Generator> gen);
void top_p_sampling_from_probs(
at::Tensor probs,
at::Tensor output,
std::optional<at::Tensor> maybe_indices,
std::optional<at::Tensor> maybe_top_p_arr,
double top_p_val,
bool deterministic,
std::optional<at::Generator> gen);