MCPcopy Create free account

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

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

↓ 2 callersFunctionget_bigram_hash
Computes bigram hash for each position using [prev_token, curr_token]. Multiply by arbitary large ints to get even spread over int32 range.
train_gpt.py:1597
↓ 2 callersFunctionget_bs
(step: int)
train_gpt_medium.py:1402
↓ 2 callersMethodget_state
(self)
train_gpt.py:1981
↓ 2 callersFunctionget_ws
(step: int)
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:1442
↓ 2 callersFunctionimpl
(x: Tensor, w: Tensor)
train_gpt_medium.py:43
↓ 2 callersFunctionimpl
(x: Tensor, w: Tensor)
train_gpt.py:67
↓ 2 callersFunctionimpl
(x: Tensor, w: Tensor)
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:47
↓ 2 callersFunctioninit_linear
(w: Tensor)
records/track_2_medium/2025-11-12_BlockMaskRedundantOp/train_gpt_medium.py:123
↓ 2 callersFunctionis_attn_param
(name: str)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:436
↓ 2 callersFunctionis_attn_param
(name: str)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:578
↓ 2 callersFunctionis_attn_proj_param
(name: str)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:433
↓ 2 callersFunctionis_attn_proj_param
(name: str)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:575
↓ 2 callersFunctionis_k_param
(name: str)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:450
↓ 2 callersFunctionis_q_param
(name: str)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:447
↓ 2 callersFunctionis_v_param
(name: str)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:293
↓ 2 callersMethodload_async
Returns getter function for async shard loading
train_gpt.py:1581
↓ 2 callersFunctionmacro_accum_enabled
()
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1292
↓ 2 callersFunctionmaybe_apply_broad_delta
(step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1373
↓ 2 callersFunctionmaybe_apply_macro_accum_step
(step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1777
↓ 2 callersFunctionmaybe_apply_trail_delta_step
(step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1657
↓ 2 callersFunctionmaybe_capture_broad_delta
(step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1361
↓ 2 callersFunctionmaybe_capture_final_readout_anchor
(step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1402
↓ 2 callersFunctionmaybe_capture_phase_readout_anchor
(step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1448
↓ 2 callersFunctionmaybe_save_checkpoint
(step: int)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1947
↓ 2 callersFunctionmaybe_save_checkpoint
(step: int)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:1491
↓ 2 callersFunctionmuon_update
(update, second_moment, step, beta2=NOR_BETA2, use_contra=True, use_soft=True)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:657
↓ 2 callersFunctionnext_multiple_of_n
(v: float | int, *, n: int)
train_gpt_medium.py:1028
↓ 2 callersFunctionnext_multiple_of_n
(v: float | int, *, n: int)
records/track_2_medium/2025-11-12_BlockMaskRedundantOp/train_gpt_medium.py:210
↓ 2 callersFunctionnext_multiple_of_n
(v: float | int, *, n: int)
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:1004
↓ 2 callersFunctionnorm_preserving_blend
(raw: Tensor, soap: Tensor, gate: Tensor, eps: float = 1e-8)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:557
↓ 2 callersFunctionnormal_cdf
(x: float)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/summarize_submission_stats.py:98
↓ 2 callersFunctionnormal_cdf
(x: float)
records/track_3_optimization/results/20260520_tail_refinterp_2900/summarize_submission_stats.py:98
↓ 2 callersFunctionparam_matches_spec
(name: str, spec: str)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:475
↓ 2 callersFunctionparam_matches_spec
(name: str, spec: str)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:502
↓ 2 callersFunctionparam_matches_spec
(name: str, spec: str)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:589
↓ 2 callersFunctionparse_logfile
(path)
records/track_3_optimization/results/20260507_pmuon/make_figures.py:88
↓ 2 callersFunctionpartition_param_list
Partitions a list of parameters into FSDP, HSDP, and other parameters. Args: original_params (Iterable[torch.Tensor]): Iterable o
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/distributor/shampoo_fsdp_utils.py:106
↓ 2 callersFunctionphase_readout_in_scope
(name: str, p: Tensor)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:1320
↓ 2 callersFunctionplot_results
(ax, plot_results, target_label_x, title_date)
records/track_3_optimization/make_figures.py:135
↓ 2 callersMethodprecond_flag_for_step
(self, step: int)
records/track_3_optimization/results/20260505_newton_muon/train_gpt_simple_newton_muon.py:293
↓ 2 callersMethodproject
Projects the gradient to the eigenbases of the preconditioner.
records/track_1_short/2024-10-09_SOAP/train_gpt2.py:236
↓ 2 callersFunctionproject_from_klsoap_basis
(grad: Tensor, state)
records/track_3_optimization/results/20260511_klsoap_h_lr_power_decay/train_gpt_simple_klsoap_h_lr_power_decay.py:191
↓ 2 callersFunctionproject_to_klsoap_basis
(grad: Tensor, state)
records/track_3_optimization/results/20260511_klsoap_h_lr_power_decay/train_gpt_simple_klsoap_h_lr_power_decay.py:186
↓ 2 callersMethodquantize_
(self)
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/utils/shampoo_quantization.py:213
↓ 2 callersFunctionref_extrap_gamma_for_step
(step: int)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:1447
↓ 2 callersFunctionref_extrap_gamma_for_step
(step: int)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:740
↓ 2 callersFunctionrescale_to_radius
(param: Tensor, target_norm: Tensor, eps: float = 1e-8)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:588
↓ 2 callersMethodreset
(self)
records/track_2_medium/2024-12-31_Target350M/train_gpt.py:347
↓ 2 callersMethodreset
(self)
records/track_1_short/2024-12-04_ValueEmbed/train_gpt2.py:338
↓ 2 callersFunctionrotary
(x_BTHD: Tensor, cos: Tensor, sin: Tensor)
train_gpt_medium.py:901
↓ 2 callersFunctionrotary
(x_BTHD: Tensor, cos: Tensor, sin: Tensor)
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:881
↓ 2 callersFunctionsame_float
(left: float, right: float)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/summarize_submission_stats.py:26
↓ 2 callersFunctionsame_float
(left: float, right: float)
records/track_3_optimization/results/20260520_tail_refinterp_2900/summarize_submission_stats.py:26
↓ 2 callersFunctionscale_radial_update
(update: Tensor, param: Tensor, eps: float = 1e-12)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:563
↓ 2 callersFunctionsetup_environment
()
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/examples/utils.py:62
↓ 2 callersFunctionshift_legend_left_half_width
(ax, legend, anchor_x=0.95, anchor_y=1.01, width_fraction=0.9)
records/track_3_optimization/make_figures.py:124
↓ 2 callersFunctionshould_soap_param
(name: str)
records/track_3_optimization/results/20260506_trustlight/train_gpt_simple_trustlight.py:260
↓ 2 callersFunctionshould_soap_param
(name: str)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:286
↓ 2 callersFunctionsoap_eigenbasis
(mat: Tensor)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:565
↓ 2 callersFunctionsoap_eigenbasis
(mat: Tensor)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:503
↓ 2 callersFunctionsoap_eigenbasis
(mat: Tensor)
records/track_3_optimization/results/20260506_trustlight/train_gpt_simple_trustlight.py:327
↓ 2 callersFunctionsoap_eigenbasis
(mat: Tensor)
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:355
↓ 2 callersFunctionsoap_eigenbasis
(mat: Tensor)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:592
↓ 2 callersFunctionsoap_eigenbasis
(mat: Tensor)
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:383
↓ 2 callersFunctionsoap_eigenbasis
(mat: Tensor)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:709
↓ 2 callersFunctionsoap_eigenbasis
(mat: Tensor)
records/track_3_optimization/results/20260504_contra_muon_mlp_soapish/train_gpt_contra_normuon_soapish_mlp.py:221
↓ 2 callersFunctionsoap_eigenbasis
(mat: Tensor)
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:329
↓ 2 callersFunctionsoap_eigenbasis
(mat: Tensor)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:502
↓ 2 callersFunctionsoap_eigenbasis
(mat: Tensor)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:485
↓ 2 callersFunctionsoap_precondition_momentum
(update, state, beta2=SOAP_BETA2, eps=1e-8, blend=SOAP_BLEND, denom_floor_ratio
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:587
↓ 2 callersFunctionsoap_precondition_momentum
(update, state, beta2=SOAP_BETA2, eps=1e-8)
records/track_3_optimization/results/20260611_tailema_2720_submission/train_gpt_tailema_2720.py:525
↓ 2 callersFunctionsoap_precondition_momentum
(update, state, beta2=SOAP_BETA2, eps=1e-8, blend=SOAP_BLEND, denom_floor_ratio
records/track_3_optimization/results/20260506_trustlight/train_gpt_simple_trustlight.py:349
↓ 2 callersFunctionsoap_precondition_momentum
(update, state, beta2=SOAP_BETA2, eps=1e-8)
records/track_3_optimization/results/20260609_soap_f1_auxb2_clean/train_gpt_clean_SOTA.py:377
↓ 2 callersFunctionsoap_precondition_momentum
(update, state, beta2=SOAP_BETA2, eps=1e-8)
records/track_3_optimization/results/20260619_cwd_rowfloor_tailema/train_gpt_cwd_SOTA.py:405
↓ 2 callersFunctionsoap_precondition_momentum
(update, state, beta2=SOAP_BETA2, eps=1e-8, blend=SOAP_BLEND, denom_floor_ratio
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:731
↓ 2 callersFunctionsoap_precondition_momentum
(update, state, beta2=SOAP_BETA2, eps=1e-8, blend=SOAP_BLEND, denom_floor_ratio
records/track_3_optimization/results/20260525_aurora_ema_ref/train_gpt_simple_aurora_ema_ref.py:351
↓ 2 callersFunctionsoap_precondition_momentum
(update, state, beta2=SOAP_BETA2, eps=1e-8)
records/track_3_optimization/results/20260611_tailema_2730_submission/train_gpt_tailema_2730.py:524
↓ 2 callersFunctionsoap_precondition_momentum
(update, state, beta2=SOAP_BETA2, eps=1e-8)
records/track_3_optimization/results/20260611_tailema_2730_submission/ablation/combo_2740/train_gpt_combo_2740.py:507
↓ 2 callersFunctionsoap_update_preconditioner
(grad, state, shampoo_beta=SOAP_BETA2, precondition_frequency=SOAP_PRECONDITION_FREQUENCY)
records/track_3_optimization/results/20260520_rre_extrapolation_pr300_2925/train_gpt_simple_rre_pr300_2925.py:605
↓ 2 callersFunctionsoap_update_preconditioner
(grad, state, shampoo_beta=SOAP_BETA2, precondition_frequency=SOAP_PRECONDITION_FREQUENCY)
records/track_3_optimization/results/20260506_trustlight/train_gpt_simple_trustlight.py:367
↓ 2 callersFunctionsoap_update_preconditioner
(grad, state, shampoo_beta=SOAP_BETA2, precondition_frequency=SOAP_PRECONDITION_FREQUENCY)
records/track_3_optimization/results/20260520_tail_refinterp_2900/train_gpt_tail_refinterp_2900.py:749
↓ 2 callersMethodsparse_index_share
(self, step)
train_gpt.py:1998
↓ 2 callersMethodsparse_index_update
(self, step, bigram_indexes)
train_gpt.py:1984
↓ 2 callersFunctionspectral_norm
(A: Tensor)
records/track_3_optimization/results/20260527_psgd/train_psgd.py:174
↓ 2 callersMethodstart
(self)
train_gpt_medium.py:1313
↓ 2 callersMethodstart
(self)
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:1174
↓ 2 callersMethodstart
(self)
records/track_1_short/2025-12-11_NorMuonOptimsAndFixes/profiler-example-traces/train_gpt-profiler-example.py:1227
↓ 2 callersMethodstep
(self)
records/track_2_medium/2024-12-31_Target350M/train_gpt.py:86
↓ 2 callersMethodstep
(self)
records/track_2_medium/2025-11-12_BlockMaskRedundantOp/train_gpt_medium.py:94
↓ 2 callersMethodstep
Performs a single optimization step. Arguments: closure (`Callable`, *optional*): A closure that reevaluates the model a
records/track_1_short/2024-10-09_SOAP/train_gpt2.py:109
↓ 2 callersMethodstep
(self)
records/track_1_short/2024-12-04_ValueEmbed/train_gpt2.py:87
↓ 2 callersMethodstep
(self)
records/track_1_short/2024-10-14_ModernArch/train_gpt2.py:81
↓ 2 callersMethodstep
(self)
records/track_1_short/2024-10-10_Muon/train_gpt2.py:80
↓ 2 callersMethodstep_optimizers
(self, step: int)
train_gpt_medium.py:1558
↓ 2 callersMethodstep_optimizers
(self, step: int)
train_gpt.py:1943
↓ 2 callersFunctionsummarize
(label, rows)
records/track_1_short/2026-05-07_XSAGatedLayers/analyze_sweep.py:69
↓ 2 callersFunctiont_lower_p
One-sided P(T <= t) via regularized incomplete beta (no scipy).
records/track_1_short/2026-05-07_XSAGatedLayers/analyze_sweep.py:19
↓ 2 callersFunctiontarget_radius_after_update
(param: Tensor, update: Tensor, lr: float, eps: float = 1e-8)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:579
↓ 2 callersFunctiontrust_gate
(raw: Tensor, soap: Tensor, grad: Tensor, eps: float = 1e-8)
records/track_3_optimization/results/20260529_tail_phase_readout_2850/train_gpt_tail_phase_readout_2850.py:515
↓ 2 callersMethodupdate_global_metrics
(self)
records/track_3_optimization/results/20260513_shampoo_1_4_power/distributed_shampoo/examples/loss_metrics.py:74
← previousnext →401–500 of 2,320, ranked by callers