MCPcopy Create free account

hub / github.com/KarelPeeters/Kyanite / functions

Functions1,306 in github.com/KarelPeeters/Kyanite

⨍ Functions 1,306◇ Types & classes 523
↓ 3 callersFunctionassert_tensors_match
(expected: &[DTensor], actual: &[DTensor], print_match: bool)
kn-cuda-eval/src/tester.rs:32
↓ 3 callersMethodcall
(&mut self, graph: &Graph, inputs: &[Value])
kn-graph/src/graph.rs:1319
↓ 3 callersMethodclone_map_inputs
(&self, mut f: impl FnMut(Value) -> Value)
kn-graph/src/graph.rs:253
↓ 3 callersMethodcomponents_factor_exp
(self)
kn-graph/src/shape.rs:270
↓ 3 callersMethodcopy_from
(&self, other: &DeviceTensor)
kn-cuda-eval/src/device_tensor.rs:102
↓ 3 callersFunctioncublasMigrateComputeType
kn-cuda-sys/doc_headers/cuda_include/cublas_api.h:5540
↓ 3 callersFunctioncudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags
kn-cuda-sys/doc_headers/cuda_include/cuda_device_runtime_api.h:104
↓ 3 callersFunctioncudaOccupancyMaxPotentialBlockSizeVariableSMemWithFlags
kn-cuda-sys/doc_headers/cuda_include/cuda_runtime.h:1749
↓ 3 callersMethoddefine
(&mut self, key: &'a str, value: T)
kn-graph/src/onnx/store.rs:14
↓ 3 callersFunctiondefine_tensor_data
( graph: &mut Graph, name: &str, tensor: &TensorProto, external: &mut dyn ExternalDataLoader,
kn-graph/src/onnx/load.rs:1230
↓ 3 callersMethoddescriptor
(&self, dtype: cudnnDataType_t)
kn-cuda-eval/src/shape.rs:241
↓ 3 callersMethoddiv_rem
(self, rhs: impl Into<Size>)
kn-graph/src/shape.rs:303
↓ 3 callersMethodfinish
(self, old_input: Value)
kn-graph/src/optimizer/affine.rs:151
↓ 3 callersFunctionfuse_affine_list
(channels: usize, operations: &[AffineOperation])
kn-graph/src/optimizer/affine.rs:352
↓ 3 callersMethodhas_simple_strides
(&self)
kn-cuda-eval/src/shape.rs:65
↓ 3 callersFunctionheap_recurse
Trick to avoid stackoverflow on deep graphs: We don't recuse at all, and if we would have wanted to to recurse we return `Err` instead. This function
kn-graph/src/optimizer/recurse.rs:6
↓ 3 callersMethodinner
(&self)
kn-cuda-sys/src/wrapper/event.rs:31
↓ 3 callersMethodinsert_mapping
(&mut self, value: Value, result: PlanTensor)
kn-cuda-eval/src/planner.rs:656
↓ 3 callersMethodis_const_zero
(&self, value: Value)
kn-graph/src/graph.rs:481
↓ 3 callersMethodkeep
Build a new shape with the shape at `axis` kept and all other axes replaced by `rest`.
kn-graph/src/shape.rs:165
↓ 3 callersMethodlayernorm
(&mut self, input: Value, axis: usize, eps: f32)
kn-graph/src/graph.rs:1136
↓ 3 callersMethodload
(self)
kn-graph/src/onnx/loader.rs:102
↓ 3 callersMethodload_operand_y
(&mut self, operand: &PlanTensor)
kn-cuda-eval/src/planner.rs:1055
↓ 3 callersFunctionmanual_tensor
(shape: I, data: Vec<T>)
kn-cuda-eval/tests/root/tensor_utils.rs:13
↓ 3 callersMethodmul
(self, rhs: Self)
kn-graph/src/onnx/typed_value.rs:249
↓ 3 callersMethodpermute
(&mut self, input: Value, permutation: Vec<usize>)
kn-graph/src/graph.rs:723
↓ 3 callersMethodplan_copy_tensor
TODO participate in scalar fusion
kn-cuda-eval/src/planner.rs:940
↓ 3 callersMethodpow
(&mut self, left: Value, right: Value)
kn-graph/src/graph.rs:1242
↓ 3 callersMethodprepare
(&self, graph: Graph, batch_size: usize)
kn-runtime/src/lib.rs:85
↓ 3 callersFunctionresolve_dtype
(data_type: DataType, node: &str)
kn-graph/src/onnx/load.rs:1364
↓ 3 callersMethodrun
(&self, handles: &Handles)
kn-cuda-eval/src/step.rs:342
↓ 3 callersFunctionrun_tensor_op
Runs `output = op(alpha_1 * input_1, alpha_2 * input_2) + b * output`
kn-cuda-sys/src/wrapper/operation.rs:268
↓ 3 callersFunctionsimple_strides
(shape: &[usize])
kn-cuda-eval/src/shape.rs:294
↓ 3 callersFunctionslice_info
( rank: usize, axis: usize, start: isize, end: Option<isize>, step: isize, )
kn-graph/src/cpu.rs:481
↓ 3 callersMethodsum
(mut iter: I)
kn-graph/src/shape.rs:490
↓ 3 callersFunctiontensor_to_image
(tensor: &Tensor<f32>)
kn-runtime/examples/stable_diffusion.rs:238
↓ 3 callersFunctiontest_all_exact_graph
TODO remove references to this
kn-cuda-eval/tests/root/runner.rs:29
↓ 3 callersMethodto_mat_mul_arg
TODO move this somewhere else, this is pretty random
kn-cuda-eval/src/offset_tensor.rs:118
↓ 3 callersFunctionunwrap_1
(slice: &[i64])
kn-graph/src/onnx/load.rs:1402
↓ 3 callersMethodunwrap_1
(&self)
kn-graph/src/shape.rs:103
↓ 3 callersMethodunwrap_3
(&self)
kn-graph/src/shape.rs:115
↓ 3 callersMethodvisit_ensure_simple_strides
(&mut self, value: Value)
kn-cuda-eval/src/planner.rs:575
↓ 2 callersFunction__bfloat162char_rz
kn-cuda-sys/doc_headers/cuda_include/cuda_bf16.hpp:1309
↓ 2 callersFunction__bfloat162int_rz
kn-cuda-sys/doc_headers/cuda_include/cuda_bf16.hpp:1249
↓ 2 callersFunction__bfloat162ll_rz
defined(__CUDACC__) || defined(_NVHPC_CUDA) */
kn-cuda-sys/doc_headers/cuda_include/cuda_bf16.hpp:1950
↓ 2 callersFunction__bfloat162uchar_rz
kn-cuda-sys/doc_headers/cuda_include/cuda_bf16.hpp:1343
↓ 2 callersFunction__bfloat162uint_rz
kn-cuda-sys/doc_headers/cuda_include/cuda_bf16.hpp:1571
↓ 2 callersFunction__bfloat162ull_rz
defined(__CUDACC__) || defined(_NVHPC_CUDA) */
kn-cuda-sys/doc_headers/cuda_include/cuda_bf16.hpp:1810
↓ 2 callersFunction__double2half
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:1001
↓ 2 callersFunction__half2char_rz
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:1335
↓ 2 callersFunction__half2int_rz
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:1450
↓ 2 callersFunction__half2ll_rz
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:1504
↓ 2 callersFunction__half2uchar_rz
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:1366
↓ 2 callersFunction__half2uint_rz
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:1477
↓ 2 callersFunction__half2ull_rz
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:1531
↓ 2 callersFunction__high2half
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:2135
↓ 2 callersFunction__hmin_nan
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:3900
↓ 2 callersFunction__internal_sin_cos_kernel
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:3467
↓ 2 callersFunction__internal_trig_reduction_kernel
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:3457
↓ 2 callersFunction__short2half_rn
defined(__CUDACC__) || defined(_NVHPC_CUDA) */
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:1677
↓ 2 callersFunction__ushort2half_rn
defined(__CUDACC__) || defined(_NVHPC_CUDA) */
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:1809
↓ 2 callersMethodalloc_buffer_shared
TODO add skip for zero-sized buffers?
kn-cuda-eval/src/planner.rs:983
↓ 2 callersMethodalloc_y
(&mut self)
kn-cuda-eval/src/planner.rs:1031
↓ 2 callersMethodapply
(self, graph: &mut Graph, input: Value)
kn-graph/src/optimizer/affine.rs:331
↓ 2 callersMethodas_fixed
(&self)
kn-graph/src/shape.rs:80
↓ 2 callersMethodassert_in_bounds
(self, size: usize)
kn-graph/src/graph.rs:1585
↓ 2 callersMethodassert_valid
(self)
kn-graph/src/graph.rs:1568
↓ 2 callersMethodattribute
(self, attribute: cudaDeviceAttr)
kn-cuda-sys/src/wrapper/handle.rs:78
↓ 2 callersFunctionbinary_op_str
(op: BinaryOp, a: &str, b: &str)
kn-cuda-eval/src/planner.rs:1139
↓ 2 callersMethodbroadcast
(&self, new_shape: Vec<usize>)
kn-cuda-eval/src/shape.rs:106
↓ 2 callersFunctionbroadcast_tensors_symmetric
( left: &'l Tensor<L>, right: &'r Tensor<R>, )
kn-graph/src/graph.rs:1450
↓ 2 callersFunctioncalculate_auto_padding
(graph: &Graph, conv_rank: usize, input: Value, filter: Value, strides: &[i64], dilations: &[i64], up: bool)
kn-graph/src/onnx/load.rs:1201
↓ 2 callersFunctioncheck_tensor_match_approx
( i: usize, expected_output: &Tensor<f64>, output: &Tensor<f64>, counts: &mut Counts, firs
kn-cuda-eval/src/tester.rs:251
↓ 2 callersMethodclamp_dyn
(&mut self, input: Value, min: DScalar, max: DScalar)
kn-graph/src/graph.rs:1193
↓ 2 callersFunctioncomplex_multiply
(new_axis: bool)
kn-cuda-eval/tests/root/graphs.rs:971
↓ 2 callersMethodcompute_capability
(self)
kn-cuda-sys/src/wrapper/handle.rs:86
↓ 2 callersFunctionconcatenate
Wrapper around [ndarray::concatenate()] that can handle an empty input list.
kn-graph/src/cpu.rs:336
↓ 2 callersMethodcopy_linear_from_device
(&self, other: &DevicePtr, len_bytes: usize)
kn-cuda-sys/src/wrapper/mem/device.rs:155
↓ 2 callersMethodcopy_simple_from_host
(&self, buffer: &[u8])
kn-cuda-eval/src/device_tensor.rs:39
↓ 2 callersMethodcopy_simple_to_host
(&self, buffer: &mut [u8])
kn-cuda-eval/src/device_tensor.rs:55
↓ 2 callersFunctioncpu_slice
(input: &Tensor<T>, axis: usize, range: SliceRange)
kn-graph/src/cpu.rs:286
↓ 2 callersFunctioncudaFuncGetAttributes
kn-cuda-sys/doc_headers/cuda_include/cuda_device_runtime_api.h:84
↓ 2 callersFunctioncudaGetDevice
kn-cuda-sys/doc_headers/cuda_include/cuda_device_runtime_api.h:94
↓ 2 callersFunctioncudaMalloc
kn-cuda-sys/doc_headers/cuda_include/cuda_device_runtime_api.h:79
↓ 2 callersFunctioncudaMemcpyAsync
kn-cuda-sys/doc_headers/cuda_include/cuda_device_runtime_api.h:474
↓ 2 callersFunctioncuda_device_count
()
kn-cuda-sys/src/wrapper/handle.rs:104
↓ 2 callersMethoddefine_y_simple
Define a new y value, computed by `expression`. The expression is not allowed to do any memory accesses, it should only use existing `y` values.
kn-cuda-eval/src/planner.rs:1070
↓ 2 callersMethoddevice
(&self)
kn-cuda-sys/src/wrapper/handle.rs:141
↓ 2 callersMethodeval
(&mut self, inputs: Vec<&PyUntypedArray>, py: Python<'py>)
kn-python/src/lib.rs:72
↓ 2 callersFunctioneval_cudnn
(graph: &Graph, batch_size: usize, inputs: &[DTensor], print_executor: bool)
kn-cuda-eval/src/tester.rs:20
↓ 2 callersFunctionfloat
* Type cast to \p float operator. */
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:338
↓ 2 callersFunctionfmt_shape_impl
(f: &mut Formatter, dims: &[impl Display])
kn-graph/src/shape.rs:554
↓ 2 callersMethodfollow_if
( &self, start: Value, mut next: impl FnMut(&Graph, Value, &Operation) -> VisitResult<
kn-graph/src/optimizer/core.rs:301
↓ 2 callersMethodfor_each_continuous_group
(&self, mut f: impl FnMut(usize, isize))
kn-cuda-eval/src/shape.rs:189
↓ 2 callersMethodfrom_bytes
(bytes: &[u8])
kn-python/src/lib.rs:35
↓ 2 callersMethodget_function_by_lower_name
It's probably easier to use [CompileResult::get_function_by_name] if possible.
kn-cuda-sys/src/wrapper/rtc/core.rs:198
↓ 2 callersMethodhas_stride
(&self)
kn-graph/src/graph.rs:345
↓ 2 callersFunctionhceil
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:1996
↓ 2 callersFunctionhfloor
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:2002
↓ 2 callersFunctionhrint
kn-cuda-sys/doc_headers/cuda_include/cuda_fp16.hpp:2008
← previousnext →201–300 of 1,306, ranked by callers