[Attention][DSA] Enable W4A16 DSA (#51724)
Signed-off-by: Shiyang Chen <shiychen@nvidia.com>
This commit is contained in:
@@ -672,6 +672,15 @@ void cp_gather_and_upconvert_fp8_kv_cache(
|
||||
int64_t batch_size,
|
||||
std::optional<torch::stable::Tensor> seq_starts = std::nullopt);
|
||||
|
||||
// Gather and upconvert an nvfp4_ds_mla KV cache to a BF16 workspace
|
||||
void cp_gather_and_upconvert_nvfp4_kv_cache(
|
||||
torch::stable::Tensor const& src_cache, // [NUM_BLOCKS, BLOCK_SIZE,
|
||||
// 352]
|
||||
torch::stable::Tensor const& dst, // [TOT_TOKENS, 576]
|
||||
torch::stable::Tensor const& block_table, // [BATCH, BLOCK_INDICES]
|
||||
torch::stable::Tensor const& workspace_starts, // [BATCH]
|
||||
int64_t batch_size);
|
||||
|
||||
// Indexer K quantization and cache function
|
||||
void indexer_k_quant_and_cache(
|
||||
torch::stable::Tensor& k, // [num_tokens, head_dim]
|
||||
|
||||
Reference in New Issue
Block a user