Functions — github.com/HazyResearch/HipKittens

CodeHub 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

Functionmma_ABt_one

kernels/gemm/fp8fp32/FP8_4wave/utils.cpp:161

distributed-kernels/iris_py.cpp:40

Methodneed_dropout

(self, dropout_p, return_encoded_softmax)

analysis/baselines/attn/triton_baseline_v02.py:95

Methodnum_tokens

training/llama/train/datamodules/datasets/indexed_dataset.py:210

Methodon_before_optimizer_step

(self, trainer: Trainer, *args: Any, **kwargs: Any)

training/llama/train/callbacks/loss_scale_monitor.py:17

Methodon_before_optimizer_step

(self, trainer: Trainer, pl_module, *args: Any, **kwargs: Any)

training/llama/train/callbacks/norm_monitor.py:33

Methodon_exception

(self, trainer: "pl.Trainer", *_: Any, **__: Any)

training/llama/train/callbacks/model_checkpoint.py:14

Methodon_fit_start

(self, trainer: Trainer, pl_module: LightningModule)

training/llama/train/callbacks/params_log.py:23

Methodon_fit_start

(self, trainer: Trainer, pl_module: LightningModule)

training/llama/train/callbacks/flop_count.py:34

Methodon_keyboard_interrupt

(self, trainer, pl_module)

training/llama/train/callbacks/wandb_callbacks.py:119

Methodon_sanity_check_end

Start executing this callback only after all validation sanity checks end.

training/llama/train/callbacks/wandb_callbacks.py:151

Methodon_sanity_check_end

Start executing this callback only after all validation sanity checks end.

training/llama/train/callbacks/wandb_callbacks.py:209

Methodon_sanity_check_end

Start executing this callback only after all validation sanity checks end.

training/llama/train/callbacks/wandb_callbacks.py:273

Methodon_sanity_check_start

(self, trainer, pl_module)

training/llama/train/callbacks/wandb_callbacks.py:148

Methodon_sanity_check_start

(self, trainer, pl_module)

training/llama/train/callbacks/wandb_callbacks.py:206

Methodon_sanity_check_start

(self, trainer, pl_module)

training/llama/train/callbacks/wandb_callbacks.py:270

Methodon_save_checkpoint

(self, checkpoint)

training/llama/train/tasks/seq.py:167

Methodon_test_epoch_start

(self, trainer: "pl.Trainer", pl_module: "pl.LightningModule")

training/llama/train/callbacks/speed_monitor.py:38

Methodon_train_batch_end

( self, trainer: "pl.Trainer", pl_module: "pl.LightningModule", outputs: STEP_

training/llama/train/callbacks/speed_monitor.py:64

Methodon_train_batch_start

( self, trainer: "pl.Trainer", pl_module: "pl.LightningModule", batch: Any,

training/llama/train/callbacks/speed_monitor.py:42

Methodon_train_epoch_end

(self, trainer: "pl.Trainer", pl_module: "pl.LightningModule",)

training/llama/train/callbacks/speed_monitor.py:89

Methodon_train_epoch_start

(self, trainer: "pl.Trainer", pl_module: "pl.LightningModule")

training/llama/train/callbacks/speed_monitor.py:30

Methodon_train_start

(self, trainer: "pl.Trainer", pl_module: "pl.LightningModule")

training/llama/train/callbacks/speed_monitor.py:27

Methodon_train_start

(self, trainer, pl_module)

training/llama/train/callbacks/wandb_callbacks.py:45

Methodon_train_start

(self, trainer, pl_module)

training/llama/train/callbacks/wandb_callbacks.py:57

Methodon_train_start

(self, trainer, pl_module)

training/llama/train/callbacks/wandb_callbacks.py:79

Methodon_validation_batch_end

Gather data from single batch.

training/llama/train/callbacks/wandb_callbacks.py:155

Methodon_validation_batch_end

Gather data from single batch.

training/llama/train/callbacks/wandb_callbacks.py:213

Methodon_validation_epoch_end

Generate confusion matrix.

training/llama/train/callbacks/wandb_callbacks.py:163

Methodon_validation_epoch_end

Generate f1, precision and recall heatmap.

training/llama/train/callbacks/wandb_callbacks.py:221

Methodon_validation_epoch_end

(self, trainer, pl_module)

training/llama/train/callbacks/wandb_callbacks.py:277

Methodon_validation_epoch_start

(self, trainer: "pl.Trainer", pl_module: "pl.LightningModule")

training/llama/train/callbacks/speed_monitor.py:35

Methodoptimizer_zero_grad

(self, epoch, batch_idx, optimizer, optimizer_idx)

training/llama/train/tasks/seq.py:159

Functionpatch_triton_compiler

analysis/baselines/gemm/utils/utils.py:98

Functionpersistent_gemm

( A, B, C, bias_ptr, P, locks, M, N, K, stride_am, stride_ak,

analysis/baselines/gemm/persistent_gemm.py:6

Functionprecompute_addresses

kernels/gemm/fp8fp32/FP8_4wave/utils.cpp:24

(self, indices)

training/llama/train/datamodules/datasets/indexed_dataset.py:237

Functionprefill_swizzled_offsets_old

kernels/gemm/bf16fp32/archive/utils.cpp:279

Functionprint_config

Prints content of DictConfig using Rich library and its tree structure. Args: config (DictConfig): Configuration composed by Hydra.

training/llama/train/utils/utils.py:89

Functionprint_gpu

(prefix, val=None)

analysis/baselines/attn/triton_baseline_v02.py:190

Functionrandom_init

kernels/gemm/fp8fp32/profile_utils.cpp:36

Functionrandom_init

kernels/torch_scaled/profile_utils.cpp:36

Functionreference_forward

GQA Reference implementation using BHND layout (batch, heads, seq, dim)

analysis/baselines/attn/attn_bwd_baselines.py:69

Functionreference_forward

GQA Reference implementation using BHND layout (batch, heads, seq, dim)

analysis/attn/bkwd/benchmark/test_python.py:83

Functionreference_forward

GQA Reference implementation using BHND layout (batch, heads, seq, dim)

kernels/attn/gqa_backwards/test_python.py:82

Functionreference_forward

GQA Reference implementation using BHND layout (batch, heads, seq, dim)

kernels/attn/gqa_backwards/archive/test_python.py:82

Functionreference_forward

GQA Reference implementation using BHND layout (batch, heads, seq, dim)

kernels/attn/gqa_causal_backwards/test_python.py:82

Functionreference_forward

GQA Reference implementation using BHND layout (batch, heads, seq, dim)

training/llama/csrc/test.py:82

training/llama/train/metrics/num_tokens.py:33

Functionrms_norm_ref

( x, weight, bias, residual=None, x1=None, weight1=None, bias1=None, eps=1e-6,

training/llama/llama/ops/triton/layer_norm.py:74

Functionrobustness_check

analysis/baselines/attn/attn_bwd_baselines.py:43

Functionrobustness_check

analysis/paper_experiments/phases/ds_write_b64/test_python.py:16

Functionrobustness_check

analysis/paper_experiments/phases/ds_read_b96/test_python.py:16

Functionrotary_kernel

( OUT, # Pointers to matrices X, COS, SIN, CU_SEQLENS, SEQLEN_OFFSETS, # this could

training/llama/llama/ops/triton/rotary.py:21

Functionrun_bash_command_wrapper

(commandstring, capture=True)

analysis/baselines/gemm/utils/utils.py:42

Functionsched_barrier_exp_pairs

analysis/attn/bkwd/benchmark/attn_fwd_causal.cpp:49

Functionsched_barrier_exp_pairs

analysis/attn/bkwd/benchmark/attn_fwd_non_causal.cpp:48

Functionsched_barrier_exp_pairs

analysis/attn/fwd/benchmark/attn_fwd_causal.cpp:52

Functionsched_barrier_exp_pairs

analysis/attn/fwd/benchmark/attn_fwd_causal_d64.cpp:52

Functionsched_barrier_exp_pairs

analysis/attn/fwd/benchmark/attn_fwd_non_causal.cpp:50

Functionsched_barrier_exp_pairs

analysis/attn/fwd/benchmark/attn_fwd_non_causal_d64.cpp:51

Functionsched_barrier_exp_pairs

kernels/attn/gqa_backwards/attn_fwd_non_causal.cpp:48

Functionsched_barrier_exp_pairs

kernels/attn/gqa_backwards/archive/GQA_fwd.cpp:34

Functionsched_barrier_exp_pairs

kernels/attn/gqa_causal_backwards/attn_fwd_causal.cpp:49

Functionsched_barrier_exp_pairs

kernels/attn/gqa/kernel.cpp:51

Functionsched_barrier_exp_pairs

kernels/attn/gqa/kernel_d64.cpp:51

Functionsched_barrier_exp_pairs

kernels/attn/gqa_causal/kernel.cpp:52

Functionsched_barrier_exp_pairs

kernels/attn/gqa_causal/kernel_d64.cpp:52

Functionsched_barrier_exp_pairs

training/llama/csrc/attn_fwd_causal.cpp:49

Functionsched_barrier_pairs

analysis/attn/bkwd/benchmark/attn_fwd_causal.cpp:42

Functionsched_barrier_pairs

analysis/attn/bkwd/benchmark/attn_fwd_non_causal.cpp:41

Functionsched_barrier_pairs

analysis/attn/fwd/benchmark/attn_fwd_causal.cpp:45

Functionsched_barrier_pairs

analysis/attn/fwd/benchmark/attn_fwd_causal_d64.cpp:45

Functionsched_barrier_pairs

analysis/attn/fwd/benchmark/attn_fwd_non_causal.cpp:43

Functionsched_barrier_pairs

analysis/attn/fwd/benchmark/attn_fwd_non_causal_d64.cpp:44

Functionsched_barrier_pairs

kernels/attn/gqa_backwards/attn_fwd_non_causal.cpp:41

Functionsched_barrier_pairs

kernels/attn/gqa_backwards/archive/GQA_fwd.cpp:27

Functionsched_barrier_pairs

kernels/attn/gqa_causal_backwards/attn_fwd_causal.cpp:42

Functionsched_barrier_pairs

kernels/attn/gqa/kernel.cpp:44

Functionsched_barrier_pairs

kernels/attn/gqa/kernel_d64.cpp:44

Functionsched_barrier_pairs

kernels/attn/gqa_causal/kernel.cpp:45

Functionsched_barrier_pairs

kernels/attn/gqa_causal/kernel_d64.cpp:45

Functionsched_barrier_pairs

training/llama/csrc/attn_fwd_causal.cpp:42

Functionsegfault_handler

Handle segmentation faults and memory access violations.

training/bert/models/hipkittens.py:14

Methodset_debug

analysis/baselines/gemm/utils/gemm_wrapper.py:17

Methodset_doc_idx

(self, doc_idx_)

training/llama/train/datamodules/datasets/indexed_dataset.py:557

Methodset_varlen_params

(self, cu_seqlens_q, cu_seqlens_k)

analysis/baselines/attn/triton_baseline_v02.py:52

(self, stage=None)

training/llama/train/datamodules/language_modeling_hf.py:86

Methodshared_step

(self, batch: Any, batch_idx: int, phase='train')

training/llama/train/tasks/seq.py:190

Functionsimple_flash_backward

GQA version that should match PyTorch exactly

analysis/baselines/attn/attn_bwd_baselines.py:87

Functionsimple_flash_backward

GQA version that should match PyTorch exactly

analysis/attn/bkwd/benchmark/test_python.py:101

Functionsimple_flash_backward

GQA version that should match PyTorch exactly

kernels/attn/gqa_backwards/test_python.py:100

Functionsimple_flash_backward

GQA version that should match PyTorch exactly

kernels/attn/gqa_backwards/archive/test_python.py:100

Functionsimple_flash_backward

GQA version that should match PyTorch exactly

kernels/attn/gqa_causal_backwards/test_python.py:100

Functionsimple_flash_backward

GQA version that should match PyTorch exactly

training/llama/csrc/test.py:100

training/llama/train/datamodules/datasets/indexed_dataset.py:458

training/llama/train/datamodules/datasets/indexed_dataset.py:547

Functionsoftmax_tk

kernels/softmax/kernel.cpp:31

Methodsplit_train_val

Randomly split self.dataset_train into a new (self.dataset_train, self.dataset_val) pair.

training/llama/train/datamodules/base.py:202

Methodstate_dict

training/llama/train/datamodules/fault_tolerant_sampler.py:72

← previous next →1,201–1,300 of 1,340, ranked by callers