MCPcopy Create free account

hub / github.com/HazyResearch/HipKittens / functions

Functions1,340 in github.com/HazyResearch/HipKittens

↓ 1 callersFunctionrun_benchmark
(custom, args)
analysis/baselines/attn/triton_baseline_v02.py:1671
↓ 1 callersFunctionrun_profiling
Run the kernel with profiling and return conflict information.
analysis/paper_experiments/phases/ds_write_b64/bank_solver.py:17
↓ 1 callersFunctionrun_profiling
Run the kernel with profiling and return conflict information.
analysis/paper_experiments/phases/ds_write_b64/phase_solver.py:12
↓ 1 callersFunctionrun_profiling
Run the kernel with profiling and return conflict information.
analysis/paper_experiments/phases/ds_read_b128/bank_solver.py:17
↓ 1 callersFunctionrun_profiling
Run the kernel with profiling and return conflict information.
analysis/paper_experiments/phases/ds_read_b128/phase_solver.py:12
↓ 1 callersFunctionrun_profiling
Run the kernel with profiling and return conflict information.
analysis/paper_experiments/phases/ds_read_b64/bank_solver.py:17
↓ 1 callersFunctionrun_profiling
Run the kernel with profiling and return conflict information.
analysis/paper_experiments/phases/ds_read_b64/phase_solver.py:12
↓ 1 callersFunctionrun_profiling
Run the kernel with profiling and return conflict information.
analysis/paper_experiments/phases/ds_read_b96/bank_solver.py:17
↓ 1 callersFunctionrun_profiling
Run the kernel with profiling and return conflict information.
analysis/paper_experiments/phases/ds_read_b96/phase_solver.py:12
↓ 1 callersFunctionsave_swizzle_diagram
Save a swizzle diagram image. Args: filename (str): Output file path (PNG recommended). rows (int): Number of rows (default 16).
analysis/paper_experiments/swizzle_plot_scripts/swizzle_diagram.py:28
↓ 1 callersFunctionsave_value_table
( filename: str, values_path: str, rows: int = 16, cols: int = 32, delimiter: str = ",",
analysis/paper_experiments/swizzle_plot_scripts/table_values_16x32.py:22
↓ 1 callersFunctionsave_value_table
( filename: str, values_path: str, rows: int = 16, cols: int = 64, delimiter: str = ",",
analysis/paper_experiments/swizzle_plot_scripts/table_values_64x16.py:26
↓ 1 callersFunctionselection_test
()
analysis/baselines/gemm/triton_gemm_v03.py:18
↓ 1 callersMethodset_int8_params
(self, q_descale, k_descale, v_descale, p_scale, p_descale)
analysis/baselines/attn/triton_baseline_v02.py:68
↓ 1 callersFunctionsinusoidal_init_
tensor: (max_len, d_model)
training/llama/llama/models/seq_common.py:148
↓ 1 callersFunctionsolve_num_banks
Solve for the number of LDS banks by testing wraparound behavior.
analysis/paper_experiments/phases/ds_write_b64/bank_solver.py:111
↓ 1 callersFunctionsolve_num_banks
Solve for the number of LDS banks by testing wraparound behavior.
analysis/paper_experiments/phases/ds_read_b128/bank_solver.py:111
↓ 1 callersFunctionsolve_num_banks
Solve for the number of LDS banks by testing wraparound behavior.
analysis/paper_experiments/phases/ds_read_b64/bank_solver.py:111
↓ 1 callersFunctionsolve_num_banks
Solve for the number of LDS banks by testing wraparound behavior.
analysis/paper_experiments/phases/ds_read_b96/bank_solver.py:111
↓ 1 callersFunctionsolve_phases
Solve for the phase assignment of all threads by testing all pairs.
analysis/paper_experiments/phases/ds_write_b64/phase_solver.py:107
↓ 1 callersFunctionsolve_phases
Solve for the phase assignment of all threads by testing all pairs.
analysis/paper_experiments/phases/ds_read_b128/phase_solver.py:107
↓ 1 callersFunctionsolve_phases
Solve for the phase assignment of all threads by testing all pairs.
analysis/paper_experiments/phases/ds_read_b64/phase_solver.py:107
↓ 1 callersFunctionsolve_phases
Solve for the phase assignment of all threads by testing all pairs.
analysis/paper_experiments/phases/ds_read_b96/phase_solver.py:107
↓ 1 callersMethodstate_dict
(self)
training/llama/train/datamodules/fault_tolerant_sampler.py:26
↓ 1 callersMethodstep
(self, epoch=None)
training/llama/train/optim/timm_lr_scheduler.py:18
↓ 1 callersMethodstep
(self, batch: Any, is_train=True)
training/llama/train/tasks/seq.py:179
↓ 1 callersFunctionstore_acc16
* @brief Write a 16x16 wave-32 WMMA accumulator tile out to global bf16. * * Per the gfx1250 WMMA bf16 ISA layout, the 16x16 D output is held by 32
kernels/gemm/bf16fp32/gfx1250/common.h:65
↓ 1 callersFunctionsupported_layouts
()
analysis/baselines/attn/triton_baseline_v02.py:1785
↓ 1 callersFunctiontest_bank_conflict
Test if two threads have a bank conflict at the given offsets.
analysis/paper_experiments/phases/ds_write_b64/bank_solver.py:85
↓ 1 callersFunctiontest_bank_conflict
Test if two threads have a bank conflict at the given offsets.
analysis/paper_experiments/phases/ds_read_b128/bank_solver.py:85
↓ 1 callersFunctiontest_bank_conflict
Test if two threads have a bank conflict at the given offsets.
analysis/paper_experiments/phases/ds_read_b64/bank_solver.py:85
↓ 1 callersFunctiontest_bank_conflict
Test if two threads have a bank conflict at the given offsets.
analysis/paper_experiments/phases/ds_read_b96/bank_solver.py:85
↓ 1 callersFunctiontest_phase
Test if two threads are in the same phase by checking for bank conflicts.
analysis/paper_experiments/phases/ds_write_b64/phase_solver.py:81
↓ 1 callersFunctiontest_phase
Test if two threads are in the same phase by checking for bank conflicts.
analysis/paper_experiments/phases/ds_read_b128/phase_solver.py:81
↓ 1 callersFunctiontest_phase
Test if two threads are in the same phase by checking for bank conflicts.
analysis/paper_experiments/phases/ds_read_b64/phase_solver.py:81
↓ 1 callersFunctiontest_phase
Test if two threads are in the same phase by checking for bank conflicts.
analysis/paper_experiments/phases/ds_read_b96/phase_solver.py:81
↓ 1 callersMethodtie_weights
(self)
training/llama/llama/models/gpt.py:234
↓ 1 callersMethodtie_weights
(self)
training/llama/llama/models/gpt.py:309
↓ 1 callersFunctiontrain
Contains training pipeline. Instantiates all PyTorch Lightning objects from config. Args: config (DictConfig): Configuration composed
training/llama/train/training.py:103
↓ 1 callersFunctiontrain_epoch
( model: nn.Module, data_loader: DataLoader, loss_fn, optimizer, device, scheduler,
training/bert/tasks.py:217
↓ 1 callersMethodval_dataloader
(self, **kwargs)
training/llama/train/datamodules/base.py:223
↓ 1 callersMethodwarmstart
(self)
training/llama/train/tasks/seq.py:87
↓ 1 callersMethodworld_size
distributed-kernels/iris_py.cpp:69
↓ 1 callersMethodwriter
(cls, path, dtype)
training/llama/train/datamodules/datasets/indexed_dataset.py:356
MethodIrisInstance
distributed-kernels/iris_py.cpp:53
MethodIrisTensor
distributed-kernels/iris_py.cpp:19
FunctionPYBIND11_MODULE
distributed-kernels/iris_py.cpp:109
FunctionPYBIND11_MODULE
distributed-kernels/bf16_gemm/kernel.cpp:253
FunctionPYBIND11_MODULE
analysis/paper_experiments/phases/ds_write_b64/kernel.cpp:53
FunctionPYBIND11_MODULE
analysis/paper_experiments/phases/ds_read_b128/kernel.cpp:54
FunctionPYBIND11_MODULE
analysis/paper_experiments/phases/ds_read_b64/kernel.cpp:54
FunctionPYBIND11_MODULE
analysis/paper_experiments/phases/ds_read_b96/kernel.cpp:54
FunctionPYBIND11_MODULE
analysis/paper_experiments/grid_micro/kernel_8192_w0.cpp:147
FunctionPYBIND11_MODULE
analysis/paper_experiments/grid_micro/kernel_8192_w32.cpp:165
FunctionPYBIND11_MODULE
analysis/paper_experiments/grid_micro/kernel_8192_w2.cpp:165
FunctionPYBIND11_MODULE
analysis/paper_experiments/grid_micro/kernel_8192_w8.cpp:165
FunctionPYBIND11_MODULE
analysis/paper_experiments/grid_micro/kernel_8192_w4.cpp:165
FunctionPYBIND11_MODULE
analysis/layernorm/mi350x/kernel_4096.cpp:137
FunctionPYBIND11_MODULE
analysis/layernorm/mi350x/kernel_1024.cpp:137
FunctionPYBIND11_MODULE
analysis/layernorm/mi350x/kernel_2048.cpp:137
FunctionPYBIND11_MODULE
analysis/layernorm/mi350x/kernel_16384.cpp:137
FunctionPYBIND11_MODULE
analysis/layernorm/mi350x/kernel_8192.cpp:137
FunctionPYBIND11_MODULE
analysis/attn/bkwd/benchmark/attn_bkwd_causal_HBN.cpp:3362
FunctionPYBIND11_MODULE
analysis/attn/bkwd/benchmark/attn_bkwd_non_causal_8_warps.cpp:852
FunctionPYBIND11_MODULE
analysis/attn/bkwd/benchmark/attn_bkwd_prep.cpp:226
FunctionPYBIND11_MODULE
analysis/attn/bkwd/benchmark/attn_fwd_causal.cpp:632
FunctionPYBIND11_MODULE
analysis/attn/bkwd/benchmark/attn_bkwd_causal_HNB.cpp:3362
FunctionPYBIND11_MODULE
analysis/attn/bkwd/benchmark/attn_fwd_non_causal.cpp:517
FunctionPYBIND11_MODULE
analysis/attn/bkwd/benchmark/attn_bkwd_non_causal_4_warps.cpp:1007
FunctionPYBIND11_MODULE
analysis/attn/bkwd/benchmark/attn_bkwd_non_causal_HBN.cpp:3171
FunctionPYBIND11_MODULE
analysis/attn/bkwd/benchmark/attn_bkwd_non_causal_HNB.cpp:3171
FunctionPYBIND11_MODULE
analysis/attn/fwd/benchmark/attn_fwd_causal.cpp:702
FunctionPYBIND11_MODULE
analysis/attn/fwd/benchmark/attn_fwd_causal_d64.cpp:702
FunctionPYBIND11_MODULE
analysis/attn/fwd/benchmark/attn_fwd_non_causal.cpp:582
FunctionPYBIND11_MODULE
analysis/attn/fwd/benchmark/attn_fwd_non_causal_d64.cpp:583
FunctionPYBIND11_MODULE
analysis/rotary/mi350x/kernel_4096.cpp:83
FunctionPYBIND11_MODULE
analysis/rotary/mi350x/kernel_1024.cpp:83
FunctionPYBIND11_MODULE
analysis/rotary/mi350x/kernel_2048.cpp:83
FunctionPYBIND11_MODULE
analysis/rotary/mi350x/kernel_16384.cpp:83
FunctionPYBIND11_MODULE
analysis/rotary/mi350x/kernel_8192.cpp:83
FunctionPYBIND11_MODULE
analysis/bf16_gemm/mi325x/kernel_4096.cpp:233
FunctionPYBIND11_MODULE
analysis/bf16_gemm/mi325x/kernel_1024.cpp:233
FunctionPYBIND11_MODULE
analysis/bf16_gemm/mi325x/kernel_2048.cpp:233
FunctionPYBIND11_MODULE
analysis/bf16_gemm/mi325x/kernel_16384.cpp:233
FunctionPYBIND11_MODULE
analysis/bf16_gemm/mi325x/kernel_8192.cpp:233
FunctionPYBIND11_MODULE
analysis/bf16_gemm/mi350x/kernel.cpp:335
FunctionPYBIND11_MODULE
kernels/gemm/bf16fp32/256_256_64_32_with32x16.cpp:160
FunctionPYBIND11_MODULE
kernels/gemm/bf16fp32/256_256_64_32_with16x32.cpp:335
FunctionPYBIND11_MODULE
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/micro_04_2stage_12c4p.cpp:167
FunctionPYBIND11_MODULE
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/micro_09_async.cpp:219
FunctionPYBIND11_MODULE
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/micro_03_3stage_8c4p.cpp:187
FunctionPYBIND11_MODULE
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/micro_02_2stage_8c4p.cpp:162
FunctionPYBIND11_MODULE
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/timing/micro_01_syncthreads.cpp:57
FunctionPYBIND11_MODULE
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/archive/micro_06_2stage_8c4p_64x96.cpp:149
FunctionPYBIND11_MODULE
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/archive/micro_06_2stage_8c4p_96x64.cpp:149
FunctionPYBIND11_MODULE
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/archive/micro_07_2stage_8c4p_nblock8.cpp:163
FunctionPYBIND11_MODULE
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/archive/micro_08_2stage_4c4p.cpp:190
FunctionPYBIND11_MODULE
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/archive/micro_08_async.cpp:188
FunctionPYBIND11_MODULE
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/archive/micro_08_4stage_4c4p.cpp:183
FunctionPYBIND11_MODULE
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/archive/micro_05_2stage_16c2p.cpp:179
← previousnext →601–700 of 1,340, ranked by callers