MCPcopy Create free account

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

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

↓ 1 callersFunctionearly_trust_floor_for_step
(step: int)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:503
↓ 1 callersFunctionearly_trust_floor_for_step
(step: int)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:396
↓ 1 callersFunctionearly_trust_floor_for_step
(step: int)
records/track_3_optimization/results/20260506_trustlight/train_gpt_simple_trustlight.py:305
↓ 1 callersFunctionearly_trust_floor_for_step
(step: int)
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:304
↓ 1 callersFunctionearly_trust_floor_for_step
(step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:530
↓ 1 callersFunctionearly_trust_floor_for_step
(step: int)
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:332
↓ 1 callersFunctionearly_trust_floor_for_step
(step: int)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:617
↓ 1 callersFunctionearly_trust_floor_for_step
(step: int)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:395
↓ 1 callersFunctionearly_trust_floor_for_step
(step: int)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:378
↓ 1 callersFunctioneffective_target_uw_for_update
(step: int, cur_uw: Tensor, scheduled_target_uw: float)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:847
↓ 1 callersFunctionema_nesterov_suffix
()
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:774
↓ 1 callersFunctionenable_precond_hooks
(model: GPT)
records/track_3_optimization/results/20260505_newton_muon/train_gpt_simple_newton_muon.py:193
↓ 1 callersMethodexit
(self)
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/utils/tests/shampoo_utils_test.py:266
↓ 1 callersFunctionextrap_layer_key
(name)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:1030
↓ 1 callersFunctionextrap_param_matches
(name: str)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:456
↓ 1 callersFunctionfast_spectral
( G: Tensor, p: float, epsilon: float = 1e-7, order: int = 2, )
records/track_3_optimization/results/20260519_dynmuon/train_gpt_simple_dynmuon.py:226
↓ 1 callersFunctionfinal_readout_in_scope
(name: str, p: Tensor)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1295
↓ 1 callersFunctionfinalize_sequence
(inputs: list[int], targets: list[int], tasks: list[HellaswagTask], seq_len: int)
evals/hellaswag.py:50
↓ 1 callersFunctionformat_evidence
(evidence, steps_to_target)
records/track_3_optimization/make_figures.py:64
↓ 1 callersMethodfrom_kronecker_factors_state
Constructs a RootInvShampooKroneckerFactorsUnwrapped object from the given Kronecker factors state. This method converts the wrapped
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/preconditioner/shampoo_preconditioner_list.py:426
↓ 1 callersMethodgenerate_custom_param_groups
Implementation requires that a single GPU does not receive both attn and mlp params when a param group is split across GPUs.
train_gpt_medium.py:524
↓ 1 callersMethodgenerate_custom_param_groups
Implementation requires that a single GPU does not receive both attn and mlp params when a param group is split across GPUs.
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:533
↓ 1 callersFunctiongenerate_param_shapes
Generate parameter shapes for testing. For N parameters, we generate the following shapes: [(1, 2), (2, 3), (3, 4), ..., (N, N + 1)].
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/utils/gpu_tests/shampoo_utils_test.py:34
↓ 1 callersMethodgenerate_standard_param_groups
Use this method if running on less than 8 GPU or experimenting with additional attn or mlp modules. Creates one param group per modul
train_gpt_medium.py:508
↓ 1 callersMethodgenerate_standard_param_groups
Use this method if running on less than 8 GPU or experimenting with additional attn or mlp modules. Creates one param group per modul
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:517
↓ 1 callersMethodget
(self)
train_gpt_medium.py:1318
↓ 1 callersMethodget
(self)
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:1232
↓ 1 callersFunctionget_adam_lr_scale
(step: int)
records/track_3_optimization/results/20260527_psgd/train_psgd.py:351
↓ 1 callersFunctionget_all_unique_subclasses
Gets all unique subclasses of a given class recursively.
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/utils/commons.py:34
↓ 1 callersFunctionget_bs
(step: int)
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:1453
↓ 1 callersFunctionget_data_loader_and_sampler
( data_path: Path | str, world_size: int, rank: int, batch_size: int )
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/examples/utils.py:136
↓ 1 callersFunctionget_dynmuon_p
(step, total_steps, p_max=0.25, p_min=-0.25, tau_ratio=0.04, width_ratio=0.04)
records/track_3_optimization/results/20260519_dynmuon/train_gpt_simple_dynmuon.py:468
↓ 1 callersFunctionget_logfile_paths
(logfile)
records/track_3_optimization/make_figures.py:163
↓ 1 callersFunctionget_logfile_paths
(logfile)
records/track_3_optimization/results/20260507_pmuon/make_figures.py:81
↓ 1 callersFunctionget_lr
(step: int)
train_gpt_medium.py:1420
↓ 1 callersFunctionget_lr
(step: int)
records/track_2_medium/2025-11-12_BlockMaskRedundantOp/train_gpt_medium.py:449
↓ 1 callersFunctionget_lr
(step: int)
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:1427
↓ 1 callersMethodget_lr
(self, step: int)
train_gpt.py:1781
↓ 1 callersMethodget_lr_lambda
(self)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:692
↓ 1 callersMethodget_lr_lambda
(self)
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:543
↓ 1 callersMethodget_lr_lambda
(self)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:954
↓ 1 callersMethodget_lr_lambda
(self)
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:587
↓ 1 callersMethodget_lr_lambda
(self)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:522
↓ 1 callersMethodget_lr_lambda
(self)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:691
↓ 1 callersMethodget_lr_lambda
(self)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:674
↓ 1 callersFunctionget_muon_momentum
(step: int, muon_warmup_steps=300, muon_cooldown_steps=50, momentum_min=0.85, momentum_max=0.95)
train_gpt_medium.py:1437
↓ 1 callersFunctionget_muon_momentum
(step: int, muon_warmup_steps=300, muon_cooldown_steps=50, momentum_min=0.85, momentum_max=0.95)
train_gpt.py:1808
↓ 1 callersFunctionget_muon_momentum
(step: int, muon_warmup_steps=300, muon_cooldown_steps=50, momentum_min=0.85, momentum_max=0.95)
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:1461
↓ 1 callersMethodget_orthogonal_matrix
Computes the eigenbases of the preconditioner using torch.linalg.eigh decomposition.
records/track_1_short/2024-10-09_SOAP/train_gpt2.py:327
↓ 1 callersMethodget_orthogonal_matrix_QR
Computes the eigenbases of the preconditioner using one round of power iteration followed by torch.linalg.qr decomposition.
records/track_1_short/2024-10-09_SOAP/train_gpt2.py:363
↓ 1 callersFunctionget_psgd_lr
(step: int)
records/track_3_optimization/results/20260527_psgd/train_psgd.py:344
↓ 1 callersFunctionget_results
(logfiles)
records/track_3_optimization/make_figures.py:33
↓ 1 callersFunctionget_sequences_for_current_rank
(seq_len: int)
evals/hellaswag.py:144
↓ 1 callersMethodget_state
(self)
train_gpt_medium.py:1604
↓ 1 callersMethodget_transition_steps
(self)
train_gpt_medium.py:1533
↓ 1 callersMethodget_transition_steps
(self)
train_gpt.py:1921
↓ 1 callersFunctionget_window_size_blocks_helper
(window_size: int)
records/track_2_medium/2025-11-12_BlockMaskRedundantOp/train_gpt_medium.py:459
↓ 1 callersFunctionheadmean_antithetic_pair
(shape, *, device, layer_idx: int, seed: int, head_dim: int = HEAD_DIM)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:903
↓ 1 callersFunctionheadmean_antithetic_pair
(shape, *, device, layer_idx: int, seed: int, head_dim: int = HEAD_DIM)
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:713
↓ 1 callersFunctionheadmean_antithetic_pair
(shape, *, device, layer_idx: int, seed: int, head_dim: int = HEAD_DIM)
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:757
↓ 1 callersFunctionheadmean_antithetic_pair
(shape, *, device, layer_idx: int, seed: int, head_dim: int = HEAD_DIM)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:902
↓ 1 callersFunctionheadmean_antithetic_pair
(shape, *, device, layer_idx: int, seed: int, head_dim: int = HEAD_DIM)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:879
↓ 1 callersFunctioninit_2d_klsoap_state_
(state, grad: Tensor, shampoo_beta: float, init_factor: float)
records/track_3_optimization/results/20260511_klsoap_h_lr_power_decay/train_gpt_simple_klsoap_h_lr_power_decay.py:196
↓ 1 callersMethodinit_attn
(self, model_dim, head_dim, num_heads, num_layers, max_seq_len)
train_gpt.py:1180
↓ 1 callersMethodinit_from_dequantized_tensor
( cls, dequantized_values: Tensor, quantized_dtype: torch.dtype, block_info: B
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/utils/shampoo_quantization.py:56
↓ 1 callersMethodinit_misc
(self, model_dim, num_layers)
train_gpt.py:1244
↓ 1 callersMethodinit_mlp
(self, model_dim)
train_gpt.py:1230
↓ 1 callersMethodinit_mudd
Multiway Dynamic Dense Connections @lishengping. https://arxiv.org/abs/2502.12170 Expressive and efficient mechanism for data depende
train_gpt.py:1281
↓ 1 callersMethodinit_preconditioner
Initializes the preconditioner matrices (L and R in the paper).
records/track_1_short/2024-10-09_SOAP/train_gpt2.py:210
↓ 1 callersFunctioninit_psgd
(grad: Tensor)
records/track_3_optimization/results/20260527_psgd/train_psgd.py:165
↓ 1 callersMethodinitialize_buffers
(self)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:682
↓ 1 callersMethodinitialize_buffers
(self)
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:533
↓ 1 callersMethodinitialize_buffers
(self)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:944
↓ 1 callersMethodinitialize_buffers
(self)
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:577
↓ 1 callersMethodinitialize_buffers
(self)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:513
↓ 1 callersMethodinitialize_buffers
(self)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:681
↓ 1 callersMethodinitialize_buffers
(self)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:664
↓ 1 callersFunctioninstantiate_parallelism_strategy
(cfg: DictConfig)
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/examples/utils.py:117
↓ 1 callersFunctionis_attn_param
(name: str)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:369
↓ 1 callersFunctionis_attn_param
(name: str)
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:277
↓ 1 callersFunctionis_attn_param
(name: str)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:491
↓ 1 callersFunctionis_attn_param
(name: str)
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:305
↓ 1 callersFunctionis_attn_param
(name: str)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:296
↓ 1 callersFunctionis_attn_param
(name: str)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:368
↓ 1 callersFunctionis_attn_param
(name: str)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:351
↓ 1 callersFunctionis_attn_proj_param
(name: str)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:366
↓ 1 callersFunctionis_attn_proj_param
(name: str)
records/track_3_optimization/results/20260506_trustlight/train_gpt_simple_trustlight.py:283
↓ 1 callersFunctionis_attn_proj_param
(name: str)
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:274
↓ 1 callersFunctionis_attn_proj_param
(name: str)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:488
↓ 1 callersFunctionis_attn_proj_param
(name: str)
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:302
↓ 1 callersFunctionis_attn_proj_param
(name: str)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:365
↓ 1 callersFunctionis_attn_proj_param
(name: str)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:348
↓ 1 callersFunctionis_mlp_param
(name: str)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:453
↓ 1 callersFunctionis_v_param
(name: str)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:499
↓ 1 callersFunctionis_v_param
(name: str)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:586
↓ 1 callersFunctionklsoap_direction
(state, grad: Tensor, beta1: float, beta2: float, eps: float)
records/track_3_optimization/results/20260511_klsoap_h_lr_power_decay/train_gpt_simple_klsoap_h_lr_power_decay.py:261
↓ 1 callersFunctionload_checkpoint_state
(path: str)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1963
↓ 1 callersFunctionload_checkpoint_state
(path: str)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:1505
↓ 1 callersFunctionload_screening_checkpoint
(path: str)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:1060
↓ 1 callersFunctionload_screening_checkpoint
(path: str)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:1059
← previousnext →701–800 of 2,320, ranked by callers