MCPcopy Create free account

hub / github.com/HazyResearch/HipKittens / functions

Functions1,340 in github.com/HazyResearch/HipKittens

⨍ Functions 1,340◇ Types & classes 190↳ Endpoints 1
↓ 1 callersMethodgrid
analysis/bf16_gemm/mi350x/kernel.cpp:33
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/256_256_64_32_with32x16.cpp:29
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/256_256_64_32_with16x32.cpp:33
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/micro_04_2stage_12c4p.cpp:30
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/micro_09_async.cpp:31
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/micro_03_3stage_8c4p.cpp:31
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/micro_02_2stage_8c4p.cpp:31
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/timing/micro_01_syncthreads.cpp:23
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/archive/micro_06_2stage_8c4p_64x96.cpp:32
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/archive/micro_06_2stage_8c4p_96x64.cpp:31
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/archive/micro_07_2stage_8c4p_nblock8.cpp:29
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/archive/micro_08_2stage_4c4p.cpp:33
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/archive/micro_08_async.cpp:31
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/archive/micro_08_4stage_4c4p.cpp:32
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/32x16/archive/micro_05_2stage_16c2p.cpp:30
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/16x32/micro_04_2stage_12c4p.cpp:32
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/16x32/micro_03_3stage_8c4p.cpp:32
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/16x32/micro_02_2stage_8c4p.cpp:32
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/16x32/micro_05_async.cpp:31
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/16x32/archive/micro_06_async.cpp:31
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/16x32/archive/micro_03_3stage_8c4p_async.cpp:32
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/producer_consumer/16x32/archive/micro_07_async.cpp:31
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/hint_based/kernel.cpp:34
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/192x256/kernel.cpp:34
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/micros/192x256/kernelv2.cpp:34
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/archive/256_256_64_32/256_256_64_32.cpp:33
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/archive/256_256_64_16/256_256_64_16.cpp:33
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/archive/micros/test_accum_layout/kernel.cpp:21
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/archive/micros/64_16/kernel.cpp:22
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/archive/micros/32_16/kernel.cpp:22
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/archive/micros/256_64/kernel.cpp:32
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/archive/micros/test_load_to_lds/kernel.cpp:21
↓ 1 callersMethodgrid
kernels/gemm/bf16fp32/archive/micros/transpose_matmul/256_256_64_16.cpp:29
↓ 1 callersMethodgrid
kernels/layernorm/kernel.cpp:79
↓ 1 callersMethodgrid
Each warp handles 32 rows. 4 warps per block = 128 rows per block.
kernels/softmax/kernel.cpp:25
↓ 1 callersMethodgrid
kernels/attn/gqa_backwards/attn_bkwd_non_causal.cpp:42
↓ 1 callersMethodgrid
kernels/attn/gqa_backwards/attn_fwd_non_causal.cpp:75
↓ 1 callersMethodgrid
kernels/attn/gqa_backwards/archive/GQA_bkwd_asm.cpp:28
↓ 1 callersMethodgrid
kernels/attn/gqa_backwards/archive/GQA_bkwd_4warps.cpp:38
↓ 1 callersMethodgrid
kernels/attn/gqa_backwards/archive/GQA_bkwd_8warps.cpp:39
↓ 1 callersMethodgrid
kernels/attn/gqa_backwards/archive/GQA_fwd.cpp:49
↓ 1 callersMethodgrid
kernels/attn/gqa_causal_backwards/attn_bkwd_causal.cpp:43
↓ 1 callersMethodgrid
kernels/attn/gqa_causal_backwards/attn_fwd_causal.cpp:155
↓ 1 callersMethodgrid
kernels/attn/gqa/kernel.cpp:97
↓ 1 callersMethodgrid
kernels/attn/gqa/kernel_d64.cpp:97
↓ 1 callersMethodgrid
kernels/attn/gqa_causal/kernel.cpp:176
↓ 1 callersMethodgrid
kernels/attn/gqa_causal/kernel_d64.cpp:176
↓ 1 callersMethodgrid
kernels/rotary/kernel.cpp:39
↓ 1 callersMethodgrid
training/llama/csrc/attn_bkwd_causal_HBN.cpp:43
↓ 1 callersMethodgrid
training/llama/csrc/attn_fwd_causal.cpp:155
↓ 1 callersMethodgrid
training/llama/csrc/attn_bkwd_causal_HNB.cpp:43
↓ 1 callersFunctiongroup_parameters_for_optimizer
Set weight_decay=0.0 for parameters in model.no_weight_decay, for parameters with attribute _no_weight_decay==True, for bias parameters if bias_we
training/llama/train/optim/param_grouping.py:12
↓ 1 callersFunctioninfer_dataset_impl
(path)
training/llama/train/datamodules/datasets/indexed_dataset.py:35
↓ 1 callersMethodinstantiate_datamodule
(self)
training/llama/train/tasks/seq.py:38
↓ 1 callersMethodinstantiate_loss
(self)
training/llama/train/tasks/seq.py:66
↓ 1 callersMethodinstantiate_metrics
(self)
training/llama/train/tasks/seq.py:74
↓ 1 callersMethodinstantiate_model
(self)
training/llama/train/tasks/seq.py:47
↓ 1 callersFunctionis_cdna
()
analysis/baselines/attn/triton_baseline_v02.py:380
↓ 1 callersFunctionis_hip_mi200
()
analysis/baselines/gemm/triton_gemm_v01.py:162
↓ 1 callersFunctionis_rdna
()
analysis/baselines/attn/triton_baseline_v02.py:385
↓ 1 callersFunctionkey_mapping
(key)
training/llama/train/utils/checkpoint.py:25
↓ 1 callersFunctionleaky_relu
(x)
analysis/baselines/gemm/triton_gemm_v01.py:321
↓ 1 callersFunctionload_checkpoint
(path, device='cpu')
training/llama/train/eval.py:28
↓ 1 callersFunctionload_checkpoint
(path, device='cpu')
training/llama/train/utils/checkpoint.py:8
↓ 1 callersFunctionload_config_hf
(model_name)
training/llama/llama/models/utils/hf.py:9
↓ 1 callersMethodload_state_dict
(self, state_dict, strict=True)
training/llama/llama/models/gpt.py:350
↓ 1 callersFunctionload_state_dict_hf
(model_name, device=None, dtype=None)
training/llama/llama/models/utils/hf.py:13
↓ 1 callersFunctionlog_predictions
One issue is that this can only be run with trainer.devices=1 right now. So restart with that after training completes. Add resume=Tr
training/llama/train/training.py:28
↓ 1 callersFunctionmain
()
analysis/baselines/attn/triton_baseline_v02.py:1831
↓ 1 callersFunctionmain
()
analysis/paper_experiments/swizzle_plot_scripts/table_values_16x32.py:210
↓ 1 callersFunctionmain
()
analysis/paper_experiments/swizzle_plot_scripts/table_values_64x16.py:212
↓ 1 callersFunctionmain
()
analysis/paper_experiments/swizzle_plot_scripts/swizzle_diagram.py:193
↓ 1 callersFunctionmain
()
analysis/paper_experiments/compile_time/measure_compile_time.py:113
↓ 1 callersFunctionmain
()
analysis/bf16_gemm/mi350x/analyze_prof.py:88
↓ 1 callersFunctionmain
(config: DictConfig)
training/llama/train/run.py:34
↓ 1 callersFunctionmain
()
docs/profiling/analyze_pmc_counter_output.py:88
↓ 1 callersFunctionmeasure_compile_time
Measure compile time for a single kernel. Returns dict with timing info and status.
analysis/paper_experiments/compile_time/measure_compile_time.py:39
↓ 1 callersMethodneed_bias
(self, bias, batch, nheads, seqlen_q, seqlen_k)
analysis/baselines/attn/triton_baseline_v02.py:78
↓ 1 callersFunctionnonvarlen_benchmark_configs
()
analysis/baselines/attn/triton_baseline_v02.py:1653
↓ 1 callersMethodon_train_end
(self, trainer, pl_module)
training/llama/train/callbacks/wandb_callbacks.py:123
↓ 1 callersFunctionparse_args
()
analysis/baselines/attn/triton_baseline_v02.py:1794
↓ 1 callersFunctionparse_args
()
analysis/paper_experiments/swizzle_plot_scripts/table_values_16x32.py:180
↓ 1 callersFunctionparse_args
()
analysis/paper_experiments/swizzle_plot_scripts/table_values_64x16.py:184
↓ 1 callersFunctionparse_args
()
analysis/paper_experiments/swizzle_plot_scripts/swizzle_diagram.py:170
↓ 1 callersFunctionparse_makefile_targets
(makefile_path)
kernels/gemm/bf16fp32/micros/hint_based/test_python.py:36
↓ 1 callersFunctionparse_makefile_targets
(makefile_path)
kernels/gemm/bf16fp32/archive/256_256_64_32/test_python.py:26
↓ 1 callersFunctionparse_makefile_targets
(makefile_path)
kernels/gemm/bf16fp32/archive/256_256_64_16/test_python.py:26
↓ 1 callersFunctionparse_makefile_targets
(makefile_path)
kernels/gemm/bf16fp32/archive/micros/transpose_matmul/test_python.py:26
↓ 1 callersFunctionpick_inputs
()
analysis/bf16_gemm/mi350x/analyze_prof.py:40
↓ 1 callersFunctionpick_inputs
()
docs/profiling/analyze_pmc_counter_output.py:40
↓ 1 callersFunctionquantize_input
(q, k, v, input_metadata: MetaData, quantize_p=False, int8_kv=False)
analysis/baselines/attn/triton_baseline_v02.py:1282
↓ 1 callersMethodrank
distributed-kernels/iris_py.cpp:68
↓ 1 callersMethodread_index
(self, path)
training/llama/train/datamodules/datasets/indexed_dataset.py:147
↓ 1 callersFunctionread_yaml_file
(file_path)
analysis/baselines/gemm/utils/solution_selection.py:7
↓ 1 callersFunctionremove_model_prefix
(state_dict)
training/llama/train/utils/checkpoint.py:68
↓ 1 callersFunctionremove_prefix
(text: str, prefix: str)
training/llama/train/eval.py:22
↓ 1 callersMethodreset_parameters
(self)
training/llama/llama/ops/triton/layer_norm.py:940
↓ 1 callersFunctionrms_norm_fn
( x, weight, bias, residual=None, x1=None, weight1=None, bias1=None, eps=1e-6,
training/llama/llama/ops/triton/layer_norm.py:893
↓ 1 callersFunctionrobustness_check
(ref, pred)
analysis/baselines/attn/attn_fwd_baselines.py:36
↓ 1 callersFunctionrotate_half
(x, interleaved=False)
training/llama/llama/models/rotary.py:14
← previousnext →501–600 of 1,340, ranked by callers