MCPcopy Create free account

hub / github.com/LeelaChessZero/lc0 / functions

Functions2,664 in github.com/LeelaChessZero/lc0

↓ 3 callersMethodGetDrawScore
Root is depth 0, i.e. even depth.
src/search/classic/search.cc:421
↓ 3 callersMethodGetEdgeToNode
src/search/classic/node.cc:225
↓ 3 callersMethodGetEstimatedRemainingTimeMs
src/search/classic/stoppers/timemgr.cc:40
↓ 3 callersFunctionGetFileList
src/utils/filesystem.posix.cc:44
↓ 3 callersFunctionGetGemmTensorDesc
src/neural/backends/dx/layers_dx.cc:116
↓ 3 callersMethodGetInt
src/utils/random.cc:40
↓ 3 callersMethodGetMUtility
Calculates the utility for favoring shorter wins and longer losses.
src/search/dag_classic/search.cc:110
↓ 3 callersMethodGetMUtility
Calculates the utility for favoring shorter wins and longer losses.
src/search/classic/search.cc:110
↓ 3 callersMethodGetMoves
src/selfplay/game.cc:309
↓ 3 callersMethodGetName
src/neural/wrapper.h:47
↓ 3 callersMethodGetName
(self)
scripts/compile_proto.py:492
↓ 3 callersMethodGetPliesSincePrevRepetition
How many half-moves since the same position appeared in the game before.
src/chess/position.h:58
↓ 3 callersMethodGetSearchNames
src/search/register.cc:43
↓ 3 callersMethodGetStickyEndgames
src/search/classic/params.h:99
↓ 3 callersMethodGetType
(self)
scripts/compile_proto.py:498
↓ 3 callersMethodGetU
Returns U = numerator * p / N. Passed numerator is expected to be equal to (cpuct * sqrt(N[parent])).
src/search/dag_classic/node.h:617
↓ 3 callersMethodGetUciOption
src/utils/optionsparser.h:53
↓ 3 callersMethodGetVariableCppType
(self)
scripts/compile_proto.py:208
↓ 3 callersMethodHasChildren
src/search/dag_classic/node.cc:148
↓ 3 callersMethodHasChildren
Returns whether a node has children.
src/search/classic/node.h:159
↓ 3 callersMethodIncrementNInFlight
src/search/dag_classic/node.cc:384
↓ 3 callersMethodIncrementNInFlight
When search decides to treat one visit as several (in case of collisions or visiting terminal nodes several times), it amplifies the visit by incremen
src/search/classic/node.h:219
↓ 3 callersMethodIndex
Index in parent's edges - useful for correlated ordering.
src/search/dag_classic/node.h:312
↓ 3 callersFunctionIsHectopliesFormat
src/neural/encoder.cc:117
↓ 3 callersFunctionLiteralOutInInOp
src/neural/xla/onnx2hlo.cc:124
↓ 3 callersFunctionLiteralOutInOp
src/neural/xla/onnx2hlo.cc:115
↓ 3 callersMethodMaximum
src/neural/xla/hlo_builder.cc:226
↓ 3 callersMethodOutputBestMove
src/chess/callbacks.h:187
↓ 3 callersMethodOutputThinkingInfo
src/chess/callbacks.h:188
↓ 3 callersMethodPlay
src/selfplay/game.cc:131
↓ 3 callersFunctionPopulateTimeManagementOptions
src/search/classic/stoppers/factory.cc:68
↓ 3 callersFunctionPositionToFen
src/chess/position.cc:151
↓ 3 callersFunctionPrintAligned
src/utils/histogram.cc:41
↓ 3 callersFunctionRELEASE
src/utils/mutex.h:56
↓ 3 callersFunctionREQUIRES
src/search/classic/search.cc:260
↓ 3 callersFunctionReadVarInt
src/utils/protomessage.cc:10
↓ 3 callersMethodReduceMean
src/neural/onnx/builder.cc:470
↓ 3 callersMethodRelu
src/neural/onnx/builder.cc:228
↓ 3 callersMethodResolveForwardDeclarations
(self, _)
scripts/compile_proto.py:504
↓ 3 callersMethodSetBounds
src/search/dag_classic/node.cc:255
↓ 3 callersMethodSetFilename
src/utils/logging.cc:57
↓ 3 callersMethodSetP
If the two assumed-on exponent bits (3<<28) are in fact off, the input is rounded up to the smallest value with them on. We accomplish this by subtrac
src/search/classic/node.cc:161
↓ 3 callersFunctionShowNetworkFormatInfo
src/tools/describenet.cc:67
↓ 3 callersFunctionShowNetworkOnnxInfo
src/tools/describenet.cc:226
↓ 3 callersMethodSortEdges
src/search/dag_classic/node.cc:195
↓ 3 callersMethodStartClock
src/search/classic/wrapper.cc:61
↓ 3 callersFunctionStrSplitAtWhitespace
src/utils/string.cc:47
↓ 3 callersMethodSub
src/neural/onnx/builder.cc:416
↓ 3 callersMethodToString
src/neural/backends/xla/pjrt.cc:196
↓ 3 callersFunctionTransposeBitsInBytes
Transpose across the diagonal connecting bit 7 to bit 56.
src/utils/bititer.h:81
↓ 3 callersFunctionTrim
src/utils/string.cc:89
↓ 3 callersMethodUpdateEstimatedNps
src/search/classic/stoppers/timemgr.cc:52
↓ 3 callersMethodUpdateEstimatedRemainingPlayouts
src/search/classic/stoppers/timemgr.cc:44
↓ 3 callersFunctionWDLRescale
WDL conversion formula based on random walk model.
src/search/dag_classic/search.cc:229
↓ 3 callersFunctionWDLRescale
WDL conversion formula based on random walk model.
src/search/classic/search.cc:223
↓ 3 callersMethodWait
src/engine_test.cc:97
↓ 3 callersFunctionaddBias_NCHW
src/neural/backends/sycl/common_kernels.dp.cpp:416
↓ 3 callersMethodclear
Not thread safe.
src/utils/atomic_vector.h:68
↓ 3 callersMethodconstructor_name
(self)
scripts/pybind/__init__.py:124
↓ 3 callersFunctiondtz_before_zeroing
DTZ tables don't store valid scores for moves that reset the rule50 counter like captures and pawn moves but we can easily recover the correct dtz of
src/syzygy/syzygy.cc:177
↓ 3 callersMethodedge
src/search/dag_classic/node.h:578
↓ 3 callersFunctionencode
src/syzygy/syzygy.cc:490
↓ 3 callersFunctionexpandPlanes_Fp16_NCHW
src/neural/backends/sycl/common_kernels.dp.cpp:716
↓ 3 callersFunctionexpandPlanes_Fp32_NCHW
src/neural/backends/sycl/common_kernels.dp.cpp:601
↓ 3 callersFunctionfill_squares
p[i] is to contain the square 0-63 (A1-H8) for a piece of type pc[i] ^ flip, where 1 = white pawn, ..., 14 = black king and pc ^ flip flips between wh
src/syzygy/syzygy.cc:851
↓ 3 callersFunctionfirst_ei
src/syzygy/syzygy.cc:274
↓ 3 callersMethodgetCommandList
src/neural/backends/dx/network_dx.h:177
↓ 3 callersMethodis_en_passant
src/chess/types.h:171
↓ 3 callersMethodname_at_caller
(self)
scripts/pybind/parameters.py:123
↓ 3 callersMethodroot_probe_wdl
Use the WDL tables to rank root moves. This is a fallback for the case that some or all DTZ tables are missing. A return value false indicates that n
src/syzygy/syzygy.cc:1691
↓ 3 callersMethodset_they_can_00
src/chess/board.h:134
↓ 3 callersMethodset_they_can_000
src/chess/board.h:135
↓ 3 callersMethodset_we_can_00
src/chess/board.h:132
↓ 3 callersMethodset_we_can_000
src/chess/board.h:133
↓ 2 callersMethodAddAliasDict
src/utils/optionsdict.cc:63
↓ 2 callersMethodAddBackend
src/neural/register.h:40
↓ 2 callersMethodAddVectors
src/neural/backends/dx/shader_wrapper.cc:413
↓ 2 callersMethodAdjustForTerminal
src/search/dag_classic/node.cc:346
↓ 2 callersFunctionApplyBias
src/neural/backends/blas/fully_connected_layer.cc:28
↓ 2 callersFunctionAsNnueString
src/trainingdata/rescorer.cc:438
↓ 2 callersMethodAsPtr
src/neural/backend.h:65
↓ 2 callersMethodAwait
src/neural/backends/xla/pjrt.cc:251
↓ 2 callersMethodAwaitAndReleaseBuffer
src/neural/backends/xla/pjrt.cc:303
↓ 2 callersMethodBackoff
src/utils/spinhelper.h:40
↓ 2 callersFunctionBoardToFen
src/chess/board.cc:1227
↓ 2 callersFunctionBoolValue
src/neural/xla/print_hlo.cc:72
↓ 2 callersFunctionBuildAttacksTable
Builds rook or bishop attacks table.
src/chess/board.cc:263
↓ 2 callersFunctionCachedValueToEvalResult
src/neural/memcache.cc:53
↓ 2 callersFunctionCanConvertConstant
src/neural/xla/onnx2hlo.cc:47
↓ 2 callersMethodCancelScoreUpdate
src/search/dag_classic/node.cc:327
↓ 2 callersMethodCancelScoreUpdate
src/search/classic/node.cc:354
↓ 2 callersFunctionChangeInputFormat
src/trainingdata/rescorer.cc:342
↓ 2 callersFunctionCheckOutOfBounds
src/utils/protomessage.cc:23
↓ 2 callersFunctionChooseTransform
src/neural/encoder.cc:50
↓ 2 callersFunctionClippy
src/tools/backendbench.cc:54
↓ 2 callersMethodCompare
src/neural/xla/hlo_builder.cc:367
↓ 2 callersFunctionComputeCpuct
src/search/dag_classic/search.cc:455
↓ 2 callersFunctionComputeEstimatedMovesToGo
src/search/classic/stoppers/legacy.cc:37
↓ 2 callersFunctionComputeEvalPositionHash
TODO For now it uses the hash of the current position, ignoring repetitions and history. We'll likely need to have configurable hash function that we'
src/neural/memcache.cc:41
↓ 2 callersMethodConv
src/neural/onnx/builder.cc:142
← previousnext →401–500 of 2,664, ranked by callers