|
InfiniOps
Operator Library for Accelerators
|
This is the complete list of members for infini::ops::FlashAttnWithKvcache, including all inherited members.
| active_implementation_indices(Device::Type dev_type) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| active_implementation_indices(Device::Type dev_type) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| alibi_slopes_dtype_ | infini::ops::FlashAttnWithKvcache | protected |
| alibi_slopes_shape_ | infini::ops::FlashAttnWithKvcache | protected |
| alibi_slopes_strides_ | infini::ops::FlashAttnWithKvcache | protected |
| batch_size_ | infini::ops::FlashAttnWithKvcache | protected |
| block_table_dtype_ | infini::ops::FlashAttnWithKvcache | protected |
| block_table_shape_ | infini::ops::FlashAttnWithKvcache | protected |
| block_table_strides_ | infini::ops::FlashAttnWithKvcache | protected |
| cache_batch_idx_dtype_ | infini::ops::FlashAttnWithKvcache | protected |
| cache_batch_idx_shape_ | infini::ops::FlashAttnWithKvcache | protected |
| cache_batch_idx_strides_ | infini::ops::FlashAttnWithKvcache | protected |
| cache_leftpad_dtype_ | infini::ops::FlashAttnWithKvcache | protected |
| cache_leftpad_shape_ | infini::ops::FlashAttnWithKvcache | protected |
| cache_leftpad_strides_ | infini::ops::FlashAttnWithKvcache | protected |
| cache_seqlens_dtype_ | infini::ops::FlashAttnWithKvcache | protected |
| cache_seqlens_shape_ | infini::ops::FlashAttnWithKvcache | protected |
| cache_seqlens_strides_ | infini::ops::FlashAttnWithKvcache | protected |
| Call(const Handle &handle, const Config &config, const Args &... args) | infini::ops::Operator< FlashAttnWithKvcache > | static |
| Call(const Tensor tensor, const Args &... args) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| Call(const TensorLike &tensor, const Args &... args) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| Call(const Handle &handle, const Config &config, const Args &... args) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| Call(const Tensor tensor, const Args &... args) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| Call(const TensorLike &tensor, const Args &... args) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| clear_cache() | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| clear_cache() | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| config_ptr_ | infini::ops::OperatorBase | protected |
| device_index_ | infini::ops::FlashAttnWithKvcache | protected |
| device_type_ | infini::ops::Operator< FlashAttnWithKvcache > | protectedstatic |
| FlashAttnWithKvcache(const Tensor q, Tensor k_cache, Tensor v_cache, Tensor out) | infini::ops::FlashAttnWithKvcache | inline |
| FlashAttnWithKvcache(const Tensor q, Tensor k_cache, Tensor v_cache, const std::optional< Tensor > k, const std::optional< Tensor > v, const std::optional< Tensor > rotary_cos, const std::optional< Tensor > rotary_sin, const int64_t cache_seqlens, const std::optional< Tensor > cache_batch_idx, const std::optional< Tensor > cache_leftpad, const std::optional< Tensor > block_table, const std::optional< Tensor > alibi_slopes, const std::optional< double > softmax_scale, const bool causal, const std::vector< int64_t > window_size, const double softcap, const bool rotary_interleaved, const int64_t num_splits, const bool return_softmax_lse, Tensor out, std::optional< Tensor > softmax_lse) | infini::ops::FlashAttnWithKvcache | inline |
| FlashAttnWithKvcache(const Tensor q, Tensor k_cache, Tensor v_cache, const std::optional< Tensor > k, const std::optional< Tensor > v, const std::optional< Tensor > rotary_cos, const std::optional< Tensor > rotary_sin, const std::optional< Tensor > cache_seqlens, const std::optional< Tensor > cache_batch_idx, const std::optional< Tensor > cache_leftpad, const std::optional< Tensor > block_table, const std::optional< Tensor > alibi_slopes, const std::optional< double > softmax_scale, const bool causal, const std::vector< int64_t > window_size, const double softcap, const bool rotary_interleaved, const int64_t num_splits, const bool return_softmax_lse, Tensor out, std::optional< Tensor > softmax_lse) | infini::ops::FlashAttnWithKvcache | inline |
| handle_ptr_ | infini::ops::OperatorBase | protected |
| has_alibi_slopes_ | infini::ops::FlashAttnWithKvcache | protected |
| has_block_table_ | infini::ops::FlashAttnWithKvcache | protected |
| has_cache_batch_idx_ | infini::ops::FlashAttnWithKvcache | protected |
| has_k_ | infini::ops::FlashAttnWithKvcache | protected |
| has_rotary_cos_ | infini::ops::FlashAttnWithKvcache | protected |
| has_rotary_sin_ | infini::ops::FlashAttnWithKvcache | protected |
| has_softmax_lse_ | infini::ops::FlashAttnWithKvcache | protected |
| has_v_ | infini::ops::FlashAttnWithKvcache | protected |
| head_size_ | infini::ops::FlashAttnWithKvcache | protected |
| implementation_index_ | infini::ops::Operator< FlashAttnWithKvcache > | protectedstatic |
| k_cache_dtype_ | infini::ops::FlashAttnWithKvcache | protected |
| k_cache_shape_ | infini::ops::FlashAttnWithKvcache | protected |
| k_cache_strides_ | infini::ops::FlashAttnWithKvcache | protected |
| k_dtype_ | infini::ops::FlashAttnWithKvcache | protected |
| k_shape_ | infini::ops::FlashAttnWithKvcache | protected |
| k_strides_ | infini::ops::FlashAttnWithKvcache | protected |
| Make(const Config &config, const Tensor tensor, Args &&... args) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| Make(const Tensor tensor, Args &&... args) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| Make(const Config &config, const std::vector< Tensor > tensors, Args &&... args) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| Make(const std::vector< Tensor > tensors, Args &&... args) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| Make(const Config &config, const Tensor tensor, Args &&... args) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| Make(const Tensor tensor, Args &&... args) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| Make(const Config &config, const std::vector< Tensor > tensors, Args &&... args) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| Make(const std::vector< Tensor > tensors, Args &&... args) | infini::ops::Operator< FlashAttnWithKvcache > | inlinestatic |
| operator()(const Tensor q, Tensor k_cache, Tensor v_cache, Tensor out) const | infini::ops::FlashAttnWithKvcache | inline |
| operator()(const Tensor q, Tensor k_cache, Tensor v_cache, const std::optional< Tensor > k, const std::optional< Tensor > v, const std::optional< Tensor > rotary_cos, const std::optional< Tensor > rotary_sin, const int64_t cache_seqlens, const std::optional< Tensor > cache_batch_idx, const std::optional< Tensor > cache_leftpad, const std::optional< Tensor > block_table, const std::optional< Tensor > alibi_slopes, const std::optional< double > softmax_scale, const bool causal, const std::vector< int64_t > window_size, const double softcap, const bool rotary_interleaved, const int64_t num_splits, const bool return_softmax_lse, Tensor out, std::optional< Tensor > softmax_lse) const =0 | infini::ops::FlashAttnWithKvcache | pure virtual |
| operator()(const Tensor q, Tensor k_cache, Tensor v_cache, const std::optional< Tensor > k, const std::optional< Tensor > v, const std::optional< Tensor > rotary_cos, const std::optional< Tensor > rotary_sin, const std::optional< Tensor > cache_seqlens, const std::optional< Tensor > cache_batch_idx, const std::optional< Tensor > cache_leftpad, const std::optional< Tensor > block_table, const std::optional< Tensor > alibi_slopes, const std::optional< double > softmax_scale, const bool causal, const std::vector< int64_t > window_size, const double softcap, const bool rotary_interleaved, const int64_t num_splits, const bool return_softmax_lse, Tensor out, std::optional< Tensor > softmax_lse) const =0 | infini::ops::FlashAttnWithKvcache | pure virtual |
| Operator< FlashAttnWithKvcache >::operator()(const Handle &handle, const Args &... args) | infini::ops::Operator< FlashAttnWithKvcache > | inline |
| Operator< FlashAttnWithKvcache >::operator()(const Args &... args) const | infini::ops::Operator< FlashAttnWithKvcache > | inline |
| out_dtype_ | infini::ops::FlashAttnWithKvcache | protected |
| out_shape_ | infini::ops::FlashAttnWithKvcache | protected |
| out_strides_ | infini::ops::FlashAttnWithKvcache | protected |
| q_dtype_ | infini::ops::FlashAttnWithKvcache | protected |
| q_shape_ | infini::ops::FlashAttnWithKvcache | protected |
| q_strides_ | infini::ops::FlashAttnWithKvcache | protected |
| rotary_cos_dtype_ | infini::ops::FlashAttnWithKvcache | protected |
| rotary_cos_shape_ | infini::ops::FlashAttnWithKvcache | protected |
| rotary_cos_strides_ | infini::ops::FlashAttnWithKvcache | protected |
| rotary_sin_dtype_ | infini::ops::FlashAttnWithKvcache | protected |
| rotary_sin_shape_ | infini::ops::FlashAttnWithKvcache | protected |
| rotary_sin_strides_ | infini::ops::FlashAttnWithKvcache | protected |
| set_config(const Config &config) | infini::ops::OperatorBase | inline |
| set_config(const Config &config) | infini::ops::OperatorBase | inline |
| set_handle(const Handle &handle) | infini::ops::OperatorBase | inline |
| set_handle(const Handle &handle) | infini::ops::OperatorBase | inline |
| set_stream(void *stream) | infini::ops::OperatorBase | inline |
| set_stream(void *stream) | infini::ops::OperatorBase | inline |
| set_workspace(void *workspace) | infini::ops::OperatorBase | inline |
| set_workspace(void *workspace) | infini::ops::OperatorBase | inline |
| set_workspace_size_in_bytes(std::size_t workspace_size_in_bytes) | infini::ops::OperatorBase | inline |
| set_workspace_size_in_bytes(std::size_t workspace_size_in_bytes) | infini::ops::OperatorBase | inline |
| softmax_lse_dtype_ | infini::ops::FlashAttnWithKvcache | protected |
| softmax_lse_shape_ | infini::ops::FlashAttnWithKvcache | protected |
| softmax_lse_strides_ | infini::ops::FlashAttnWithKvcache | protected |
| stream_ | infini::ops::OperatorBase | protected |
| v_cache_dtype_ | infini::ops::FlashAttnWithKvcache | protected |
| v_cache_shape_ | infini::ops::FlashAttnWithKvcache | protected |
| v_cache_strides_ | infini::ops::FlashAttnWithKvcache | protected |
| v_dtype_ | infini::ops::FlashAttnWithKvcache | protected |
| v_shape_ | infini::ops::FlashAttnWithKvcache | protected |
| v_strides_ | infini::ops::FlashAttnWithKvcache | protected |
| ValidateIndexVector(const std::optional< Tensor > &tensor) const | infini::ops::FlashAttnWithKvcache | inlineprotected |
| workspace_ | infini::ops::OperatorBase | protected |
| workspace_size_in_bytes() const | infini::ops::OperatorBase | inlinevirtual |
| workspace_size_in_bytes() const | infini::ops::OperatorBase | inlinevirtual |
| workspace_size_in_bytes_ | infini::ops::OperatorBase | protected |
| ~OperatorBase()=default | infini::ops::OperatorBase | virtual |
| ~OperatorBase()=default | infini::ops::OperatorBase | virtual |