[Kernel/Model] Migrate mamba_ssm and causal_conv1d kernels to vLLM (#7651)

This commit is contained in:
Mor Zusman
2024-08-28 15:06:52 -07:00
committed by GitHub
parent 8c56e57def
commit fdd9daafa3
20 changed files with 2815 additions and 31 deletions
+22
View File
@@ -195,6 +195,28 @@ void moe_align_block_size(torch::Tensor topk_ids, int64_t num_experts,
torch::Tensor experts_ids,
torch::Tensor num_tokens_post_pad);
std::vector<torch::Tensor> selective_scan_fwd(
const torch::Tensor& u, const torch::Tensor& delta, const torch::Tensor& A,
const torch::Tensor& B, const torch::Tensor& C,
const c10::optional<torch::Tensor>& D_,
const c10::optional<torch::Tensor>& z_,
const c10::optional<torch::Tensor>& delta_bias_, bool delta_softplus,
const c10::optional<torch::Tensor>& index_,
const c10::optional<torch::Tensor>& x);
at::Tensor causal_conv1d_update(const at::Tensor& x,
const at::Tensor& conv_state,
const at::Tensor& weight,
const c10::optional<at::Tensor>& bias_,
bool silu_activation);
at::Tensor causal_conv1d_fwd(const at::Tensor& x, const at::Tensor& weight,
const c10::optional<at::Tensor>& bias_,
const c10::optional<at::Tensor>& seq_idx_,
const c10::optional<at::Tensor>& initial_states_,
const c10::optional<at::Tensor>& final_states_out_,
bool silu_activation);
#ifndef USE_ROCM
using fptr_t = int64_t;
fptr_t init_custom_ar(torch::Tensor& meta, torch::Tensor& rank_data,