MCPcopy Create free account

hub / github.com/bytedance/ByteTransformer / functions

Functions317 in github.com/bytedance/ByteTransformer

⨍ Functions 317◇ Types & classes 134
↓ 20 callersFunctiondevice_malloc
unit_test/helper.h:35
↓ 18 callersFunctionload_weights
unit_test/helper.h:40
↓ 17 callersFunctionset_dtype
(ts: torch.Tensor, dtype: str)
unit_test/python_scripts/bert_transformer_test.py:50
↓ 15 callersMethodadd_tile_offset
Advances an iterator along logical dimensions of matrix in units of whole tiles
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:906
↓ 15 callersMethodclear
< Efficiently disables all accesses guarded by mask
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/predicated_tile_iterator.h:151
↓ 14 callersMethodload
Loads a fragment from memory
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/predicated_tile_iterator.h:598
↓ 13 callersMethodget
Returns a pointer
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:912
↓ 12 callersMethodvalid
Returns whether access is valid or not
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:279
↓ 11 callersFunctionwarpReduceMax
bytetransformer/include/reduce.h:90
↓ 11 callersFunctionwarpReduceSum
bytetransformer/include/reduce.h:33
↓ 10 callersMethodset_iteration_index
Overrides the internal iteration index
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/predicated_tile_iterator.h:632
↓ 8 callersMethodadd_pointer_offset
Adds a pointer offset in units of Element
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/predicated_tile_iterator.h:592
↓ 6 callersMethodclear_mask
< Efficiently disables all accesses guarded by mask
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/predicated_tile_iterator.h:658
↓ 6 callersMethodstore
Stores a fragment to memory
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/predicated_tile_iterator.h:615
↓ 5 callersMethodset_iteration_index
Overrides the internal iteration index
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:227
↓ 4 callersFunctiongetLaneId
bytetransformer/include/reduce.h:26
↓ 4 callersMethodinfer
bytetransformer/include/attention.h:124
↓ 4 callersMethodset_kgroup_index
cutlass_contrib/include/cutlass/contrib/args_pack_def.h:101
↓ 3 callersMethodclear_mask
Clears the predicate set efficiently
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:243
↓ 3 callersMethodenable
< CUTLASS_HOST_DEVICE enables all accesses guarded by mask
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/predicated_tile_iterator.h:159
↓ 3 callersMethodenable_mask
Clears the predicate set efficiently
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:252
↓ 3 callersMethodget_mask
Gets the mask
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:270
↓ 3 callersFunctionload_with_byte_offset
Loads a fragment from memory
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/predicated_tile_iterator.h:260
↓ 3 callersMethodset_mask
Sets the predicate mask, overriding value stored in predicate iterator
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:261
↓ 3 callersFunctionstore_with_byte_offset
Stores a fragment to memory
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/predicated_tile_iterator.h:313
↓ 3 callersFunctiontanht
bytetransformer/include/gemm_bias_act.h:295
↓ 3 callersFunctiontranspose_for_scores
(x, n_heads, head_size)
unit_test/python_scripts/bert_transformer_test.py:34
↓ 2 callersMethodadd_pointer_offset
Adds a pointer offset in units of Element
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_iterator.h:612
↓ 2 callersMethodcal_bufsize
bytetransformer/include/attention.h:109
↓ 2 callersMethodclear_mask
Clears the predicate set efficiently
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_iterator.h:641
↓ 2 callersMethodcompute_predicates_
Computes predicates based on internally tracked per-thread offset.
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:132
↓ 2 callersFunctioncopy_tiles_and_advance
cutlass_contrib/include/cutlass/contrib/gemm/threadblock/mma_multistage.h:98
↓ 2 callersMethodenable_mask
Clears the predicate set efficiently
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_iterator.h:647
↓ 2 callersFunctionfast_tanh
bytetransformer/include/gemm_bias_act.h:67
↓ 2 callersMethodforward
bytetransformer/th_op/ths_op.h:80
↓ 2 callersFunctionfused_infer
bytetransformer/include/variety_attention_fused.h:635
↓ 2 callersFunctionfused_long_infer
bytetransformer/include/variety_attention_fused.h:677
↓ 2 callersFunctionfused_long_rm_infer
bytetransformer/include/variety_attention_fused.h:739
↓ 2 callersFunctionfused_rm_infer
bytetransformer/include/variety_attention_fused.h:656
↓ 2 callersFunctiongelu
bytetransformer/include/gemm_bias_act.h:306
↓ 2 callersMethodget_mask
Gets the mask
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_iterator.h:659
↓ 2 callersMethodinitialize
bytetransformer/include/attention.h:105
↓ 2 callersMethodset_mask
Sets the predicate mask, overriding value stored in predicate iterator
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_iterator.h:653
↓ 1 callersMethodBertTransformer
bytetransformer/th_op/ths_op.h:31
↓ 1 callersFunction__half2add
bytetransformer/include/reduce.h:63
↓ 1 callersFunction__tanhf
Compute hyperbolic tangent for < sm_75. maxulperr = 108.82848, maxrelerr = 9.3450e-6
bytetransformer/include/gemm_bias_act.h:51
↓ 1 callersFunction_cudaGetErrorEnum
bytetransformer/include/common.h:91
↓ 1 callersFunctionadd_pointer_offset
Adds a pointer offset in units of Element
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:443
↓ 1 callersMethodadd_pointer_offset
Adds a pointer offset in units of Element
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:899
↓ 1 callersMethodadvance
cutlass_contrib/include/cutlass/contrib/gemm/kernel/gemm_grouped.h:222
↓ 1 callersFunctioncopysignf_pos
bytetransformer/include/gemm_bias_act.h:46
↓ 1 callersFunctiondiffTime
unit_test/helper.h:30
↓ 1 callersMethodget_pickle_info
bytetransformer/th_op/ths_op.h:99
↓ 1 callersMethodis_source_needed
Returns true if source is needed
bytetransformer/include/gemm_bias_act.h:172
↓ 1 callersMethodload_with_byte_offset
Loads a fragment from memory
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_iterator.h:671
↓ 1 callersMethodload_with_pointer_offset
Loads a fragment from memory
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_iterator.h:665
↓ 1 callersFunctionmax_
bytetransformer/include/reduce.h:85
↓ 1 callersMethodnext_tile
Returns true if there is a tile to compute
cutlass_contrib/include/cutlass/contrib/gemm/kernel/gemm_grouped.h:152
↓ 1 callersMethodoperand_A_ref
Returns a TensorRef to the A operand
cutlass_contrib/include/cutlass/contrib/gemm/threadblock/mma_base.h:196
↓ 1 callersMethodoperator()
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/epilogue.h:194
↓ 1 callersMethodoperator++
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:936
↓ 1 callersFunctionparse_args
()
unit_test/python_scripts/bert_transformer_test.py:58
↓ 1 callersMethodproblem_index
Gets the index of the problem
cutlass_contrib/include/cutlass/contrib/gemm/kernel/gemm_grouped.h:205
↓ 1 callersFunctionresult_check
unit_test/helper.h:59
↓ 1 callersFunctionseqlen_to_mask
(lengths, max_len)
unit_test/python_scripts/bert_transformer_test.py:41
↓ 1 callersFunctionset_iteration_index
Overrides the internal iteration index
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:437
↓ 1 callersMethodset_predicates
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:181
↓ 1 callersMethodstore_with_byte_offset
Store a fragment to memory
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_iterator.h:689
↓ 1 callersMethodstore_with_pointer_offset
Store a fragment to memory
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_iterator.h:683
↓ 1 callersMethodthread_offset
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_iterator.h:305
↓ 1 callersMethodthreadblock_index
cutlass_contrib/include/cutlass/contrib/gemm/kernel/gemm_grouped.h:218
↓ 1 callersMethodtotal
bytetransformer/include/cutlass_attention.h:60
↓ 1 callersFunctionwarpPrefixSum
bytetransformer/include/remove_padding.h:27
MethodAttention
bytetransformer/include/attention.h:78
MethodAttentionBatchCountGeneratorOp
bytetransformer/include/cutlass_attention_operators.h:76
MethodAttentionParam
bytetransformer/include/attention.h:37
MethodBTEncoder
bytetransformer/th_op/bert_transformer_ext.h:43
MethodBertTransformer
bytetransformer/include/bert_transformer.h:120
MethodBertTransformerParam
bytetransformer/include/bert_transformer.h:58
FunctionCUTLASS_HOST_DEVICE PredicatedTileAccessIterator
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:309
MethodCutlassAttention
bytetransformer/include/cutlass_attention.h:66
FunctionFragmentOutput operator()
bytetransformer/include/cutlass_attention_operators.h:273
MethodGemmGrouped
Constructs the GEMM.
cutlass_contrib/include/cutlass/contrib/gemm/device/gemm_grouped.h:70
MethodGemmGrouped
cutlass_contrib/include/cutlass/contrib/gemm/kernel/gemm_grouped.h:412
MethodIdentityPrologueOperator
cutlass_contrib/include/cutlass/contrib/args_pack_def.h:91
MethodInterleavedConvPredicatedTileIterator
Constructor
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/predicated_tile_iterator.h:829
MethodInterleavedPredicatedTileIterator
Constructor
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/predicated_tile_iterator.h:563
MethodLayoutA
cutlass_contrib/include/cutlass/contrib/gemm/threadblock/mma_base.h:185
MethodLayoutB
cutlass_contrib/include/cutlass/contrib/gemm/threadblock/mma_base.h:191
MethodLinearCombinationGELUFAST
Constructs the function object, possibly loading from pointers in host memory
bytetransformer/include/gemm_bias_act.h:165
MethodMask
Mask
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/predicated_tile_iterator.h:145
MethodMask
Mask
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/predicated_tile_iterator.h:504
MethodMask
Mask
cutlass_contrib/include/cutlass/contrib/epilogue/threadblock/predicated_tile_iterator.h:757
MethodParams
Ctor
cutlass_contrib/include/cutlass/contrib/gemm/kernel/gemm_grouped.h:72
MethodParams
cutlass_contrib/include/cutlass/contrib/gemm/kernel/gemm_grouped.h:367
MethodParams
Default ctor
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:348
MethodParams
Default ctor
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:650
MethodParams
Default ctor
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_access_iterator.h:839
MethodParams
Construct the Params object given a pitch-linear tensor's layout
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_iterator.h:116
MethodParams
cutlass_contrib/include/cutlass/contrib/transform/threadblock/predicated_tile_iterator.h:366
next →1–100 of 317, ranked by callers