MCPcopy Create free account

hub / github.com/KellerJordan/modded-nanogpt / functions

Functions2,320 in github.com/KellerJordan/modded-nanogpt

↓ 1 callersFunctionload_screening_checkpoint
(path: str)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:1034
↓ 1 callersMethodlookahead_step
Performs nesterov's lookahead.
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:700
↓ 1 callersMethodlookahead_step
Performs nesterov's lookahead.
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:551
↓ 1 callersMethodlookahead_step
Performs nesterov's lookahead.
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:962
↓ 1 callersMethodlookahead_step
Performs nesterov's lookahead.
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:595
↓ 1 callersMethodlookahead_step
(self)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:527
↓ 1 callersMethodlookahead_step
Performs nesterov's lookahead.
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:699
↓ 1 callersMethodlookahead_step
Performs nesterov's lookahead.
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:682
↓ 1 callersFunctionmacro_accum_in_scope
(name: str, p: Tensor)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1609
↓ 1 callersFunctionmacro_accum_kernel_weight
(block_step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1629
↓ 1 callersFunctionmacro_cem_group
(name: str, p: Tensor)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1638
↓ 1 callersFunctionmain
()
records/track_1_short/2026-05-07_XSAGatedLayers/analyze_sweep.py:85
↓ 1 callersFunctionmain
()
records/track_3_optimization/results/20260529_tail_phase_readout_2850/summarize_submission_stats.py:102
↓ 1 callersFunctionmain
(cfg: DictConfig)
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/examples/cifar10_example.py:67
↓ 1 callersFunctionmain
()
records/track_3_optimization/results/20260520_tail_refinterp_2900/summarize_submission_stats.py:102
↓ 1 callersFunctionmaybe_apply_final_readout
(step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1441
↓ 1 callersFunctionmaybe_apply_phase_readout
(step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1472
↓ 1 callersFunctionmaybe_apply_trail_delta
(step: int)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:1152
↓ 1 callersFunctionmaybe_apply_trail_delta
(step: int)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:1151
↓ 1 callersFunctionmaybe_apply_trail_delta
(step: int)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:1097
↓ 1 callersFunctionmaybe_capture_readout_anchor
(step: int)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:1145
↓ 1 callersFunctionmaybe_capture_readout_anchor
(step: int)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:1144
↓ 1 callersFunctionmaybe_capture_readout_anchor
(step: int)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:1090
↓ 1 callersFunctionmaybe_capture_ref_extrap_state
(step: int)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:1416
↓ 1 callersFunctionmaybe_capture_ref_extrap_state
(step: int)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:728
↓ 1 callersMethodmaybe_outer_step
(self)
records/track_3_optimization/results/20260504_muloco_normuonh/train_gpt_simple_muloco_normuonh.py:291
↓ 1 callersFunctionmaybe_update_tail_ema
(step: int)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:1130
↓ 1 callersFunctionmaybe_update_tail_ema
(step: int)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:1129
↓ 1 callersFunctionmaybe_update_target_uw_train_loss_gate
(step: int)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:815
↓ 1 callersFunctionmpe_sidi_extrapolate
(iterates, regularization=1e-6, damping=1.0)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:762
↓ 1 callersFunctionmuon_lr_mult_for_step
(step: int)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:767
↓ 1 callersFunctionmuon_sq_update
(grad, momentum, v, mu=0.95, beta2=0.95, eps=1e-10, nesterov=True)
records/track_3_optimization/results/20260501_muonsq/train_gpt_simple_muonsq.py:191
↓ 1 callersFunctionmuon_update
(grad, momentum, mu=0.95, nesterov=True)
records/track_3_optimization/train_gpt_simple.py:183
↓ 1 callersFunctionmuon_update
(update, second_moment, step, beta2=NOR_BETA2, use_contra=True, use_soft=True)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:630
↓ 1 callersFunctionmuon_update
(update)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:550
↓ 1 callersFunctionmuon_update
(grad, momentum, mu=0.95, nesterov=True)
records/track_3_optimization/results/20260519_dynmuon/train_gpt_simple_dynmuon.py:217
↓ 1 callersFunctionmuon_update
(update, second_moment, beta2=NOR_BETA2)
records/track_3_optimization/results/20260506_trustlight/train_gpt_simple_trustlight.py:384
↓ 1 callersFunctionmuon_update
(update)
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:402
↓ 1 callersFunctionmuon_update
(grad, momentum, second_moment, mu=0.95, beta2=0.95, nesterov=True)
records/track_3_optimization/results/20260501_contra_muon/train_gpt_simple_contra_muon_2.py:224
↓ 1 callersFunctionmuon_update
(grad, momentum, mu=0.95, nesterov=True)
records/track_3_optimization/results/20260430_muonh/train_gpt_simple_muonh.py:190
↓ 1 callersFunctionmuon_update
(update)
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:430
↓ 1 callersFunctionmuon_update
(grad, momentum, mu=0.95, nesterov=True)
records/track_3_optimization/results/20260611_muonh_tuned_aux_3250/train_gpt_simple_muonh_tuned_aux_3250.py:190
↓ 1 callersFunctionmuon_update
( update, second_moment, step, beta2=NOR_BETA2, use_contra=True, use_soft=True, us
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:925
↓ 1 callersFunctionmuon_update
(grad, momentum, mu=0.95, nesterov=True)
records/track_3_optimization/results/20260505_newton_muon/train_gpt_simple_newton_muon.py:222
↓ 1 callersFunctionmuon_update
(grad, momentum, second_moment, mu=0.95, beta2=0.95, nesterov=True)
records/track_3_optimization/results/20260509_contra_muon_split_cooldown/train_gpt_simple_contra_muon_split_cooldown.py:224
↓ 1 callersFunctionmuon_update
(update, second_moment, beta2=0.95)
records/track_3_optimization/results/20260504_contra_muon_mlp_soapish/train_gpt_contra_normuon_soapish_mlp.py:274
↓ 1 callersFunctionmuon_update
(update, second_moment, step, beta2=NOR_BETA2)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:396
↓ 1 callersFunctionmuon_update
(update)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:549
↓ 1 callersFunctionmuon_update
(update)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:532
↓ 1 callersMethodnesterov_step
(self)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:738
↓ 1 callersMethodnesterov_step
(self)
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:589
↓ 1 callersMethodnesterov_step
(self)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1000
↓ 1 callersMethodnesterov_step
(self)
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:633
↓ 1 callersMethodnesterov_step
(self)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:553
↓ 1 callersMethodnesterov_step
(self)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:737
↓ 1 callersMethodnesterov_step
(self)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:720
↓ 1 callersMethodnext_batch
(self, num_tokens_local: int, max_seq_len: int)
train_gpt_medium.py:1271
↓ 1 callersMethodnext_batch
(self, num_tokens_local: int, max_seq_len: int)
train_gpt.py:1555
↓ 1 callersMethodnext_batch
(self, num_tokens_local: int, max_seq_len: int)
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:1185
↓ 1 callersFunctionnorm_preserving_blend
(raw: Tensor, soap: Tensor, gate: Tensor, eps: float = 1e-8)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:530
↓ 1 callersFunctionnorm_preserving_blend
(raw: Tensor, soap: Tensor, gate: Tensor, eps: float = 1e-8)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:412
↓ 1 callersFunctionnorm_preserving_blend
(raw: Tensor, soap: Tensor, gate: Tensor, eps: float = 1e-8)
records/track_3_optimization/results/20260506_trustlight/train_gpt_simple_trustlight.py:321
↓ 1 callersFunctionnorm_preserving_blend
(raw: Tensor, soap: Tensor, gate: Tensor, eps: float = 1e-8)
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:320
↓ 1 callersFunctionnorm_preserving_blend
(raw: Tensor, soap: Tensor, gate: Tensor, eps: float = 1e-8)
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:348
↓ 1 callersFunctionnorm_preserving_blend
(raw: Tensor, soap: Tensor, gate: Tensor, eps: float = 1e-8)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:644
↓ 1 callersFunctionnorm_preserving_blend
(raw: Tensor, soap: Tensor, gate: Tensor, eps: float = 1e-8)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:411
↓ 1 callersFunctionnorm_preserving_blend
(raw: Tensor, soap: Tensor, gate: Tensor, eps: float = 1e-8)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:394
↓ 1 callersFunctionnorm_preserving_soap_blend
(raw: Tensor, soap: Tensor, eps: float = 1e-8)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:382
↓ 1 callersFunctionnormuon_update
NorMuon direction: NS-orthogonalised gradient followed by Adafactor-style variance preconditioning along the SHORT axis (per https://arxiv.org/pdf
records/track_3_optimization/results/20260504_muloco_normuonh/train_gpt_simple_muloco_normuonh.py:196
↓ 1 callersMethodnum_bytes
(self)
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/preconditioner/preconditioner_list.py:101
↓ 1 callersFunctionnvidia_smi
()
train_gpt_medium.py:1682
↓ 1 callersFunctionnvidia_smi
()
train_gpt.py:2040
↓ 1 callersFunctionnvidia_smi
()
records/track_2_medium/2025-11-12_BlockMaskRedundantOp/train_gpt_medium.py:405
↓ 1 callersFunctionnvidia_smi
()
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:1383
↓ 1 callersFunctionopt_params
(opt: torch.optim.Optimizer)
records/track_2_medium/2025-11-12_BlockMaskRedundantOp/train_gpt_medium.py:441
↓ 1 callersFunctionpack_tasks
Pack `raw_tasks` into a list of sequences. The tasks are packed into a sequence like this: <ctx1><end11>EOT_ID<ctx1><end12>EOT_ID<ctx1><end13
evals/hellaswag.py:77
↓ 1 callersFunctionparse_filename_gamma
(path: Path, eval_kind: str)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/summarize_submission_stats.py:30
↓ 1 callersFunctionparse_filename_gamma
(path: Path, eval_kind: str)
records/track_3_optimization/results/20260520_tail_refinterp_2900/summarize_submission_stats.py:30
↓ 1 callersFunctionparse_fsdp_params
Splits parameters into FSDP, HSDP, and the rest of parameters. This is useful for parsing the parameters when FSDP and HSDP are wrapping a subset
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/distributor/shampoo_fsdp_utils.py:190
↓ 1 callersFunctionparse_fully_shard_params
Splits parameters into fully shard, hybrid shard, and other parameters. This is useful for parsing the parameters when fully shard or hybrid shar
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/distributor/shampoo_fsdp_utils.py:224
↓ 1 callersFunctionparse_int_list
(raw: str)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:177
↓ 1 callersFunctionparse_logfile
(path)
records/track_3_optimization/make_figures.py:170
↓ 1 callersFunctionparse_loss
(path: Path, step: int | None, eval_kind: str, gamma: float | None)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/summarize_submission_stats.py:42
↓ 1 callersFunctionparse_loss
(path: Path, step: int | None, eval_kind: str, gamma: float | None)
records/track_3_optimization/results/20260520_tail_refinterp_2900/summarize_submission_stats.py:42
↓ 1 callersFunctionparse_value
( value: dict[str, Any] | torch.Tensor | OptimizerModule, )
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/utils/shampoo_state_dict_utils.py:63
↓ 1 callersFunctionpolar_express
Polar Express Sign Method: https://arxiv.org/pdf/2505.16932 by Noah Amsel, David Persson, Christopher Musco, Robert M. Gower.
train_gpt_medium.py:378
↓ 1 callersFunctionpolar_express
Fused Nesterov momentum + Polar Express Sign Method. Nesterov momentum is applied in FP32, then the result is cast to BF16 for polar express
train_gpt.py:170
↓ 1 callersFunctionpolar_express
Polar Express Sign Method: https://arxiv.org/pdf/2505.16932 by Noah Amsel, David Persson, Christopher Musco, Robert M. Gower.
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:382
↓ 1 callersMethodprecondition
(self, masked_grad_list: tuple[Tensor, ...])
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/preconditioner/spectral_descent_preconditioner_list.py:55
↓ 1 callersMethodproject_back
Projects the gradient back to the original space.
records/track_1_short/2024-10-09_SOAP/train_gpt2.py:299
↓ 1 callersFunctionpsgd_update_precond
(Q0: Tensor, Q1: Tensor, U: Tensor, v: Tensor, precond_lr=1.0)
records/track_3_optimization/results/20260527_psgd/train_psgd.py:187
↓ 1 callersFunctionquadratic_extrapolate
(iterates, regularization=1e-6, damping=1.0)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:867
↓ 1 callersMethodquantize
(self, tensor_list: tuple[Tensor, ...])
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/utils/shampoo_quantization.py:193
↓ 1 callersFunctionradial_outward_scale_for_step
(step: int)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:876
↓ 1 callersFunctionrecursive_compare
(d1: dict[str, Any], d2: dict[str, Any])
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/distributor/gpu_tests/shampoo_checkpoint_test.py:94
↓ 1 callersFunctionref_extrap_suffix
(step: int)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:777
↓ 1 callersMethodrefresh_assigned_full_params
Refresh the cached full parameter tensors from current model parameters. This method should be called after loading a checkpoint to ensure th
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/distributor/_shampoo_fully_shard_lossless_distributor.py:234
↓ 1 callersFunctionrefresh_klsoap_basis_
(state)
records/track_3_optimization/results/20260511_klsoap_h_lr_power_decay/train_gpt_simple_klsoap_h_lr_power_decay.py:251
↓ 1 callersMethodregister_backward_hooks
(self)
train_gpt_medium.py:743
↓ 1 callersMethodregister_backward_hooks
(self)
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:753
← previousnext →801–900 of 2,320, ranked by callers