MCPcopy Create free account

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

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

↓ 1 callersFunction_ns_inner
(X: Tensor)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:331
↓ 1 callersFunction_ns_inner
(X: Tensor)
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:239
↓ 1 callersFunction_ns_inner
(X: Tensor)
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:267
↓ 1 callersFunction_ns_inner
(X: Tensor)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:330
↓ 1 callersFunction_ns_inner
(X: Tensor)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:313
↓ 1 callersMethod_outer_step
(self)
records/track_3_optimization/results/20260504_muloco_normuonh/train_gpt_simple_muloco_normuonh.py:298
↓ 1 callersMethod_parameter_offsets
(self, params)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:1075
↓ 1 callersFunction_peek_data_shard
(filename)
records/track_1_short/2024-10-09_SOAP/train_gpt2.py:586
↓ 1 callersFunction_peek_data_shard
(filename)
records/track_1_short/2024-12-04_ValueEmbed/train_gpt2.py:291
↓ 1 callersFunction_peek_data_shard
(filename)
records/track_1_short/2024-10-14_ModernArch/train_gpt2.py:242
↓ 1 callersFunction_peek_data_shard
(filename)
records/track_1_short/2024-10-10_Muon/train_gpt2.py:252
↓ 1 callersMethod_precondition_and_grafting
( self, state_lists: dict[str, Any], masked_filtered_grad_list: tuple[torch.Tensor, ..
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/distributed_shampoo.py:974
↓ 1 callersMethod_precondition_grads
(self)
records/track_3_optimization/results/20260505_newton_muon/train_gpt_simple_newton_muon.py:328
↓ 1 callersMethod_preconditioner_list_factory
(self)
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/preconditioner/tests/shampoo_preconditioner_list_test.py:131
↓ 1 callersMethod_refresh_preconditioner
(self)
records/track_3_optimization/results/20260505_newton_muon/train_gpt_simple_newton_muon.py:297
↓ 1 callersMethod_refresh_q
(self, state: dict)
records/track_3_optimization/results/20260518_soaph/train_gpt_simple_soaph.py:295
↓ 1 callersMethod_reset_momentum
(self, params)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:1069
↓ 1 callersFunction_scheduled_muon_history_blend
(step: int)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:418
↓ 1 callersFunction_scheduled_muon_history_blend
(step: int)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:417
↓ 1 callersFunction_scheduled_muon_history_blend
(step: int)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:400
↓ 1 callersFunction_soft_coefficients
(p: float)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:344
↓ 1 callersFunction_soft_coefficients
(p: float)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:399
↓ 1 callersFunction_soft_coefficients
(p: float)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:486
↓ 1 callersMethod_split_local_dist_buffers
Split distributed buffers for each local rank into views for each assigned block. Args: buffer_size_ranks (tuple[tuple[int, int],
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/distributor/shampoo_ddp_distributor.py:406
↓ 1 callersMethod_split_local_dist_buffers
Split distributed buffers for each local rank into views for each assigned block. Args: buffer_size_ranks (tuple[tuple[int, int],
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/distributor/shampoo_hsdp_distributor.py:409
↓ 1 callersMethod_split_local_dist_buffers
Split distributed buffers for each local rank into views for each assigned block. Args: buffer_size_ranks (tuple[tuple[int, int],
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/distributor/shampoo_hybrid_shard_distributor.py:422
↓ 1 callersMethod_step_param
(self, p: Tensor, group: dict)
records/track_3_optimization/results/20260518_soaph/train_gpt_simple_soaph.py:316
↓ 1 callersFunction_symmetrize
(matrix: Tensor)
records/track_3_optimization/results/20260518_soaph/train_gpt_simple_soaph.py:171
↓ 1 callersMethod_test_linear_forward_backward
( self, feature_vector: torch.Tensor, in_features: int, out_features: int,
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/utils/tests/shampoo_model_utils_test.py:42
↓ 1 callersMethod_unflatten_params
(self, params, flat_params)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:1062
↓ 1 callersMethod_update_factor_matrices
(self, masked_grad_list: tuple[Tensor, ...])
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/preconditioner/shampoo_preconditioner_list.py:1340
↓ 1 callersMethod_update_preconditioners
( self, state_lists: dict[str, Any], step: torch.Tensor, perform_amortized_com
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/distributed_shampoo.py:1033
↓ 1 callersMethodaccum_lookahead
Update nesterov's lookahead direction.
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:720
↓ 1 callersMethodaccum_lookahead
Update nesterov's lookahead direction.
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:571
↓ 1 callersMethodaccum_lookahead
Update nesterov's lookahead direction.
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:982
↓ 1 callersMethodaccum_lookahead
Update nesterov's lookahead direction.
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:615
↓ 1 callersMethodaccum_lookahead
(self)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:540
↓ 1 callersMethodaccum_lookahead
Update nesterov's lookahead direction.
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:719
↓ 1 callersMethodaccum_lookahead
Update nesterov's lookahead direction.
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:702
↓ 1 callersFunctionaccum_xtx_
(x: Tensor, accum: Tensor, count: Tensor)
records/track_3_optimization/results/20260505_newton_muon/train_gpt_simple_newton_muon.py:56
↓ 1 callersFunctionadam_lr_mult_for_step
(step: int)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:773
↓ 1 callersFunctionadd_legend
(ax, legend_entries)
records/track_3_optimization/make_figures.py:69
↓ 1 callersMethodadvance
(self)
records/track_1_short/2024-10-09_SOAP/train_gpt2.py:640
↓ 1 callersMethodadvance
(self)
records/track_1_short/2024-10-14_ModernArch/train_gpt2.py:296
↓ 1 callersMethodadvance
(self)
records/track_1_short/2024-10-10_Muon/train_gpt2.py:306
↓ 1 callersFunctionalign_train_loader_to_step
(step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:2001
↓ 1 callersMethodallocate_ones_tensor
( size: tuple[int, ...], dtype: torch.dtype | None = None, device: torch.d
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/distributor/shampoo_block_info.py:110
↓ 1 callersMethodamortized_computation
Performs amortized computation for Shampoo preconditioners. This method orchestrates the execution of the computationally expensive matrix op
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/preconditioner/shampoo_preconditioner_list.py:210
↓ 1 callersMethodapply
(self, old_window: int, new_window: int, alpha: int=1, beta: int=32)
train_gpt_medium.py:890
↓ 1 callersFunctionapply_agreement_shrink
(update: Tensor, state: dict, step: int, eps: float = 1e-12)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:689
↓ 1 callersFunctionapply_aurora_relax
(aurora: Tensor, plain: Tensor, step: int | None)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:436
↓ 1 callersMethodapply_final_ws_ext
(self)
train_gpt_medium.py:1520
↓ 1 callersMethodapply_final_ws_ext
(self)
train_gpt.py:1906
↓ 1 callersFunctionapply_muon_history
(update: Tensor, state: dict, step: int)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:427
↓ 1 callersFunctionapply_muon_history
(update: Tensor, state: dict, step: int)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:426
↓ 1 callersFunctionapply_muon_history
(update: Tensor, state: dict, step: int)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:409
↓ 1 callersFunctionapply_muon_weight_scale
(factor: float)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:1380
↓ 1 callersFunctionapply_normuon_variance_reduction
NorMuon variance reduction. Algebraically fuses the normalization steps to minimize memory ops.
train_gpt_medium.py:437
↓ 1 callersFunctionapply_normuon_variance_reduction
NorMuon variance reduction. Algebraically fuses the normalization steps to minimize memory ops.
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:441
↓ 1 callersFunctionapply_proj_weight_scale
(factor: float)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:1395
↓ 1 callersFunctionapply_ref_extrap_weights
(gamma: float)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:744
↓ 1 callersFunctionapply_tempered_polar
(prepolar: Tensor, polar: Tensor, polar_norm: Tensor, step: int)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:913
↓ 1 callersFunctionattach_precond_stats
(model: GPT)
records/track_3_optimization/results/20260505_newton_muon/train_gpt_simple_newton_muon.py:178
↓ 1 callersMethodattach_preconditioner
(self)
records/track_3_optimization/results/20260505_newton_muon/train_gpt_simple_newton_muon.py:256
↓ 1 callersFunctionattention_soap_blend_for_step
(step: int)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:519
↓ 1 callersFunctionattention_soap_blend_for_step
(step: int)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:633
↓ 1 callersFunctionaurora_beta_for_step
(step: int | None)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:406
↓ 1 callersFunctionaurora_relax_lambda_for_step
(step: int | None)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:429
↓ 1 callersFunctionaverage_runs
(runs)
records/track_3_optimization/make_figures.py:189
↓ 1 callersFunctionba_plus_cAA
Launch Triton kernel to compute C = alpha * A @ A.T + beta * A
train_gpt_medium.py:336
↓ 1 callersFunctionba_plus_cAA
Launch Triton kernel to compute C = alpha * A @ A.T + beta * A
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:340
↓ 1 callersFunctionbounded_trust_gate
(gate: Tensor, step: int)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:514
↓ 1 callersFunctionbounded_trust_gate
(gate: Tensor, step: int)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:407
↓ 1 callersFunctionbounded_trust_gate
(gate: Tensor, step: int)
records/track_3_optimization/results/20260506_trustlight/train_gpt_simple_trustlight.py:316
↓ 1 callersFunctionbounded_trust_gate
(gate: Tensor, step: int)
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:315
↓ 1 callersFunctionbounded_trust_gate
(gate: Tensor, step: int)
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:343
↓ 1 callersFunctionbounded_trust_gate
(gate: Tensor, step: int)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:628
↓ 1 callersFunctionbounded_trust_gate
(gate: Tensor, step: int)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:406
↓ 1 callersFunctionbounded_trust_gate
(gate: Tensor, step: int)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:389
↓ 1 callersFunctionbroad_delta_in_scope
(name: str, p: Tensor)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1340
↓ 1 callersFunctionbuild_checkpoint_state
(step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1917
↓ 1 callersFunctionbuild_checkpoint_state
(step: int)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:1472
↓ 1 callersFunctioncautious_wd_and_update_inplace
Cautious weight decay + parameter update. wd_tensor and lr_tensor are 0-D CPU tensors.
train_gpt_medium.py:428
↓ 1 callersFunctioncautious_wd_and_update_inplace
Cautious weight decay + parameter update. wd_tensor and lr_tensor are 0-D CPU tensors.
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:432
↓ 1 callersFunctionce_fwd_bwd
( logits: torch.Tensor, targets: torch.Tensor, mtp_weights: torch.Tensor, losses: torch.Tensor
triton_kernels.py:937
↓ 1 callersFunctioncheckpoint_path_for_step
(step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1914
↓ 1 callersFunctioncheckpoint_path_for_step
(step: int)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:1455
↓ 1 callersFunctioncircuit_muon
Circuit-Muon for one OV pair (attn.v, attn.proj). Approximates each head's Gram root as isotropic C = c*I with c^2 = ||W_h||^2/hd, so the par
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:687
↓ 1 callersMethodcompress
(self, selector: tuple[bool, ...])
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/utils/shampoo_quantization.py:239
↓ 1 callersFunctioncompute_eigenvalue_threshold
Compute threshold for filtering eigenvalues based on numerical rank. Determines which eigenvalues should be considered numerically zero based
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/preconditioner/matrix_functions.py:200
↓ 1 callersFunctioncompute_validation_loss
()
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:766
↓ 1 callersFunctioncontra_coeff_for_step
(step: int)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:623
↓ 1 callersFunctioncontra_coeff_for_step
(step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:650
↓ 1 callersFunctioncontra_coeff_for_step
(step: int)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:886
↓ 1 callersFunctioncontra_coeff_for_step
(step: int)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:392
↓ 1 callersMethodcopy_lm_state_to_embed
Copy the optimizer state from the lm_head to the embed at the untie point. This requires an all-gather + reshard because of different
train_gpt.py:650
↓ 1 callersMethodcopy_lm_to_embed
(self)
train_gpt_medium.py:769
↓ 1 callersMethodcreate_blockmasks
(self, input_seq: Tensor, sliding_window_num_blocks: Tensor)
records/track_2_medium/2025-11-12_BlockMaskRedundantOp/train_gpt_medium.py:232
↓ 1 callersFunctioncreate_device_mesh
( dp_replicate_degree: int, world_size: int )
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/examples/utils.py:109
↓ 1 callersFunctiondynmuon_update
(grad, momentum, p: float, lam: float = 0.0, mu=0.95, nesterov=True)
records/track_3_optimization/results/20260519_dynmuon/train_gpt_simple_dynmuon.py:278
← previousnext →601–700 of 2,320, ranked by callers