diff --git a/LTA_openwebtext_dualt/logs/lta_lm1b_classic_dirichlet_len256_gbs512_4gpu_10k_save1k_20260523_driver.log b/LTA_openwebtext_dualt/logs/lta_lm1b_classic_dirichlet_len256_gbs512_4gpu_10k_save1k_20260523_driver.log new file mode 100644 index 0000000000000000000000000000000000000000..4ee02f241cea1ba5120e2b8d7a5835717d2ddbdf --- /dev/null +++ b/LTA_openwebtext_dualt/logs/lta_lm1b_classic_dirichlet_len256_gbs512_4gpu_10k_save1k_20260523_driver.log @@ -0,0 +1,336 @@ +[launch] method=categorical_fullvocab_c1024_fullycoupled host=di-20260411014000-djqhq time=2026-05-23T13:51:35+00:00 +[launch] cwd=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt +[launch] run_name=lta_lm1b_classic_dirichlet_len256_gbs512_4gpu_10k_save1k_20260523 +[launch] save_dir=runs/lta_lm1b_classic_dirichlet_len256_gbs512_4gpu_10k_save1k_20260523 +[launch] log_file=logs/lta_lm1b_classic_dirichlet_len256_gbs512_4gpu_10k_save1k_20260523.log + +***************************************** +Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. +***************************************** +NCCL version 2.25.1+cuda12.8 +resumed_from=runs/lta_lm1b_classic_dirichlet_len256_gbs512_4gpu_10k_save1k_20260523/step_0005000.pt start_step=5001 +{ + "device": "cuda:0", + "rank": 0, + "world_size": 4, + "samples": "wrapped_stream", + "vocab_size": 30522, + "tokenizer_vocab_size": 30522, + "save_dir": "runs/lta_lm1b_classic_dirichlet_len256_gbs512_4gpu_10k_save1k_20260523", + "max_len": 256, + "effective_model_max_len": 256, + "batch_size": 32, + "grad_accum": 4, + "effective_batch_size": 512, + "global_batch_size": 512, + "lr_schedule": "constant_warmup", + "optimizer": "adamw", + "epochs": 0.0, + "steps_per_epoch": 0, + "total_steps": 20000, + "warmup_steps": 2500, + "warmup_epochs": -1.0, + "min_lr": 6e-05, + "weight_decay": 0.0, + "output_weight_decay": -1.0, + "adamw_param_groups": "nanogpt", + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "adam_eps": 1e-08, + "muon_impl": "legacy", + "muon_momentum": 0.95, + "muon_ns_steps": 5, + "muon_update_scale": 1.0, + "muon_nesterov": false, + "muon_width_scale": false, + "muon_grouping": "", + "muon_param_count": 0, + "muon_adam_param_count": 0, + "muon_param_names": [], + "muon_adam_param_names": [], + "muon_effective_nesterov": false, + "muon_effective_width_scale": false, + "muon_effective_weight_decay": 0.0, + "muon_adam_fallback_nesterov": false, + "muon_adam_fallback_weight_decay": 0.0, + "ema_decay": 0.0, + "ema_start_step": 0, + "model_type": "ddit", + "ddit_mlp_type": "gelu", + "block_anchor_every": 0, + "block_anchor_init_std": 0.02, + "bos_anchor_every": 0, + "bos_anchor_token_id": -1, + "bos_anchor_extra_len": 0, + "abs_pos_embed": false, + "abs_pos_init_std": 0.02, + "elf_num_time_tokens": 4, + "elf_num_model_mode_tokens": 0, + "qk_norm": true, + "output_bias": false, + "output_init_std": -1.0, + "norm_type": "rmsnorm", + "target_loss": "hard_ce", + "linear_soft_target_power": 1.0, + "linear_soft_target_min_conf": 0.0, + "linear_soft_target_max_conf": 1.0, + "t_sampling_mode": "uniform", + "t_sampling_power": 1.0, + "t_sampling_eps": 0.0001, + "t_sampling_logit_mean": -1.5, + "t_sampling_logit_std": 0.8, + "t_sampling_gumbel_loc": 2.2, + "t_sampling_gumbel_scale": 0.8, + "dual_t": true, + "corrupt_t_mode": "same", + "corrupt_min_t": 0.0, + "corrupt_max_t": 1.0, + "prefix_block_prob": 0.0, + "prefix_block_len": 128, + "block_ar_two_stream": false, + "block_ar_block_len": 128, + "mask_ratio_floor_schedule": "none", + "dirichlet_endpoint_mode": "categorical_dual_t", + "dirichlet_semantic_t_mode": "same", + "dirichlet_semantic_t_value": 0.0, + "dirichlet_semantic_t_curve": "linear", + "dirichlet_semantic_t_power": 1.0, + "dirichlet_support_t_curve": "linear", + "dirichlet_support_t_power": 1.0, + "endpoint_sequence_random_prob_alpha": 0.0, + "categorical_wrong_from_full_vocab": true, + "categorical_wrong_from_batch_valid_tokens": false, + "categorical_wrong_basin_token_ids": "", + "categorical_wrong_basin_prob": 0.0, + "categorical_wrong_unigram_prob": 0.0, + "categorical_wrong_uniform_prob": 0.0, + "categorical_wrong_prob_floor": 0.0, + "categorical_gold_prob_floor": 0.0, + "categorical_gold_prob_ceil": 1.0, + "categorical_wrong_corpus_unigram_path": "", + "categorical_wrong_corpus_unigram_alpha": 1.0, + "categorical_wrong_basin_shared_prob": 0.0, + "categorical_wrong_unigram_shared_prob": 0.0, + "mask_mixture_original_prob": 0.0, + "mask_mixture_lowk_prob": 0.0, + "mask_mixture_lowcorrupt_prob": 0.0, + "mask_mixture_block_prob": 0.0, + "mask_mixture_all_prob": 0.0, + "mask_mixture_lowk_clean_tokens": "1,2,4,8,16,32,64", + "mask_mixture_lowcorrupt_tokens": "1,2,4,8,16,32,64", + "mask_mixture_block_tokens": "64,128", + "simplex_bridge_sampler": "dirichlet", + "logistic_normal_sigma_min": 0.18, + "logistic_normal_sigma_max": 2.2, + "logistic_normal_tau_min": 0.65, + "logistic_normal_tau_max": 1.15, + "torch_compile": false, + "compile_mode": "max-autotune", + "state_format": "prob", + "meanflow_weight": 0.0, + "rollout_train_prob": 0.0, + "rollout_train_steps": 1, + "rollout_train_steps_min": -1, + "rollout_train_infer_steps": 64, + "rollout_train_time_mode": "fixed_steps", + "rollout_train_s_dist": "uniform", + "rollout_train_s_min_frac": 0.0, + "rollout_train_s_max_frac": 0.125, + "rollout_train_s_beta_alpha": 2.0, + "rollout_train_s_beta_beta": 6.0, + "rollout_train_temp": 1.0, + "rollout_train_max_gamma": 1.0, + "rollout_train_rule": "flowmap", + "rollout_train_corrupt_only": true, + "rollout_train_samplewise": false, + "rollout_train_compute_always": false, + "rollout_train_keep_grad": false, + "rollout_train_sync_t": false, + "rollout_train_state_mix_mode": "final", + "rollout_train_state_mix_alpha": 0.5, + "bridge_noise_init": "logistic_normal", + "noise_sigma": -1.0, + "allow_tf32": true, + "activation_checkpointing": false, + "activation_checkpoint_interval": 1, + "activation_checkpoint_scope": "block", + "ddp_static_graph": false, + "ddp_gradient_as_bucket_view": true, + "blocking_data_transfer": false, + "dataloader_prefetch_factor": 2, + "full_train_stats": false, + "tokenized_hf": false, + "tokenized_pad_token": "pad", + "elf_conditional_hf": false, + "record_pad_truncate": false, + "record_add_eos": false, + "record_add_special_tokens": false, + "record_pad_token": "pad", + "record_shuffle_buffer": 10000, + "wrap": true, + "wrap_mode": "stream", + "wrap_record_buffer_size": 200, + "owt_cached_chunks": false, + "owt_chunk_cache_dir": "", + "owt_chunk_cache_rebuild": false, + "owt_chunk_cache_write_batch": 4096, + "owt_exact_repeat_per_chunk": 0, + "online_chunk_shuffle": false, + "online_chunk_shuffle_buffer": 10000, + "openwebtext_split": "all", + "detokenizer": "auto", + "resolved_detokenizer": "lm1b", + "num_workers": 0, + "latest_every": 1000, + "resume_path": "runs/lta_lm1b_classic_dirichlet_len256_gbs512_4gpu_10k_save1k_20260523/step_0005000.pt" +} +step=5100 micro_steps=20400 elapsed=126.0s lr=3.000000e-04 loss=2.3515 loss_recon=2.3515 loss_meanflow=0.0000 mean_model_t=0.5045 mean_corrupt_t=0.5045 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7900 corrupt_frac=0.5523 acc_corrupt=0.6325 loss_corrupt=2.3515 wrong_frac=0.4952 init_acc_corrupt=0.4708 acc_corrupt_t_0p0_0p2=0.2505 corrupt_frac_t_0p0_0p2=0.1950 acc_corrupt_t_0p2_0p4=0.4625 corrupt_frac_t_0p2_0p4=0.1984 acc_corrupt_t_0p4_0p6=0.6689 corrupt_frac_t_0p4_0p6=0.2045 acc_corrupt_t_0p6_0p8=0.8185 corrupt_frac_t_0p6_0p8=0.1999 acc_corrupt_t_0p8_1p0=0.9438 corrupt_frac_t_0p8_1p0=0.2038 out_w_norm=88.5733 out_g_norm=0.2436 loss_all=1.4167 init_gold_top10=0.4359 init_gold_top100=0.4426 +step=5200 micro_steps=20800 elapsed=72.1s lr=3.000000e-04 loss=2.3775 loss_recon=2.3775 loss_meanflow=0.0000 mean_model_t=0.5022 mean_corrupt_t=0.5022 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7887 corrupt_frac=0.5522 acc_corrupt=0.6298 loss_corrupt=2.3775 wrong_frac=0.4978 init_acc_corrupt=0.4678 acc_corrupt_t_0p0_0p2=0.2513 corrupt_frac_t_0p0_0p2=0.2007 acc_corrupt_t_0p2_0p4=0.4576 corrupt_frac_t_0p2_0p4=0.1950 acc_corrupt_t_0p4_0p6=0.6709 corrupt_frac_t_0p4_0p6=0.2053 acc_corrupt_t_0p6_0p8=0.8207 corrupt_frac_t_0p6_0p8=0.1982 acc_corrupt_t_0p8_1p0=0.9435 corrupt_frac_t_0p8_1p0=0.2017 out_w_norm=89.5800 out_g_norm=0.2422 loss_all=1.5736 init_gold_top10=0.4234 init_gold_top100=0.4293 +step=5300 micro_steps=21200 elapsed=55.2s lr=3.000000e-04 loss=2.3725 loss_recon=2.3725 loss_meanflow=0.0000 mean_model_t=0.5013 mean_corrupt_t=0.5013 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7893 corrupt_frac=0.5498 acc_corrupt=0.6295 loss_corrupt=2.3725 wrong_frac=0.4989 init_acc_corrupt=0.4663 acc_corrupt_t_0p0_0p2=0.2519 corrupt_frac_t_0p0_0p2=0.2007 acc_corrupt_t_0p2_0p4=0.4610 corrupt_frac_t_0p2_0p4=0.1992 acc_corrupt_t_0p4_0p6=0.6698 corrupt_frac_t_0p4_0p6=0.1993 acc_corrupt_t_0p6_0p8=0.8183 corrupt_frac_t_0p6_0p8=0.1984 acc_corrupt_t_0p8_1p0=0.9443 corrupt_frac_t_0p8_1p0=0.2029 out_w_norm=90.5869 out_g_norm=0.2403 loss_all=1.6992 init_gold_top10=0.4632 init_gold_top100=0.4664 +step=5400 micro_steps=21600 elapsed=55.1s lr=3.000000e-04 loss=2.3809 loss_recon=2.3809 loss_meanflow=0.0000 mean_model_t=0.4956 mean_corrupt_t=0.4956 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7896 corrupt_frac=0.5459 acc_corrupt=0.6277 loss_corrupt=2.3809 wrong_frac=0.5035 init_acc_corrupt=0.4615 acc_corrupt_t_0p0_0p2=0.2547 corrupt_frac_t_0p0_0p2=0.2062 acc_corrupt_t_0p2_0p4=0.4618 corrupt_frac_t_0p2_0p4=0.1984 acc_corrupt_t_0p4_0p6=0.6720 corrupt_frac_t_0p4_0p6=0.1967 acc_corrupt_t_0p6_0p8=0.8193 corrupt_frac_t_0p6_0p8=0.2008 acc_corrupt_t_0p8_1p0=0.9442 corrupt_frac_t_0p8_1p0=0.1994 out_w_norm=91.5814 out_g_norm=0.2391 loss_all=1.1516 init_gold_top10=0.5024 init_gold_top100=0.5086 +step=5500 micro_steps=22000 elapsed=55.1s lr=3.000000e-04 loss=2.3760 loss_recon=2.3760 loss_meanflow=0.0000 mean_model_t=0.5002 mean_corrupt_t=0.5002 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7892 corrupt_frac=0.5499 acc_corrupt=0.6286 loss_corrupt=2.3760 wrong_frac=0.5019 init_acc_corrupt=0.4633 acc_corrupt_t_0p0_0p2=0.2500 corrupt_frac_t_0p0_0p2=0.2039 acc_corrupt_t_0p2_0p4=0.4666 corrupt_frac_t_0p2_0p4=0.1970 acc_corrupt_t_0p4_0p6=0.6700 corrupt_frac_t_0p4_0p6=0.2052 acc_corrupt_t_0p6_0p8=0.8211 corrupt_frac_t_0p6_0p8=0.1966 acc_corrupt_t_0p8_1p0=0.9449 corrupt_frac_t_0p8_1p0=0.1984 out_w_norm=92.5716 out_g_norm=0.2356 loss_all=1.3463 init_gold_top10=0.5272 init_gold_top100=0.5328 +step=5600 micro_steps=22400 elapsed=55.1s lr=3.000000e-04 loss=2.3794 loss_recon=2.3794 loss_meanflow=0.0000 mean_model_t=0.4962 mean_corrupt_t=0.4962 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7886 corrupt_frac=0.5504 acc_corrupt=0.6278 loss_corrupt=2.3794 wrong_frac=0.5025 init_acc_corrupt=0.4626 acc_corrupt_t_0p0_0p2=0.2528 corrupt_frac_t_0p0_0p2=0.2041 acc_corrupt_t_0p2_0p4=0.4620 corrupt_frac_t_0p2_0p4=0.2006 acc_corrupt_t_0p4_0p6=0.6715 corrupt_frac_t_0p4_0p6=0.2009 acc_corrupt_t_0p6_0p8=0.8213 corrupt_frac_t_0p6_0p8=0.1941 acc_corrupt_t_0p8_1p0=0.9450 corrupt_frac_t_0p8_1p0=0.2002 out_w_norm=93.5642 out_g_norm=0.2336 loss_all=1.1209 init_gold_top10=0.5413 init_gold_top100=0.5458 +step=5700 micro_steps=22800 elapsed=55.2s lr=3.000000e-04 loss=2.3515 loss_recon=2.3515 loss_meanflow=0.0000 mean_model_t=0.5019 mean_corrupt_t=0.5019 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7905 corrupt_frac=0.5520 acc_corrupt=0.6321 loss_corrupt=2.3515 wrong_frac=0.4981 init_acc_corrupt=0.4672 acc_corrupt_t_0p0_0p2=0.2540 corrupt_frac_t_0p0_0p2=0.1989 acc_corrupt_t_0p2_0p4=0.4609 corrupt_frac_t_0p2_0p4=0.2008 acc_corrupt_t_0p4_0p6=0.6731 corrupt_frac_t_0p4_0p6=0.1988 acc_corrupt_t_0p6_0p8=0.8239 corrupt_frac_t_0p6_0p8=0.1988 acc_corrupt_t_0p8_1p0=0.9440 corrupt_frac_t_0p8_1p0=0.2028 out_w_norm=94.5536 out_g_norm=0.2311 loss_all=1.2391 init_gold_top10=0.5221 init_gold_top100=0.5275 +step=5800 micro_steps=23200 elapsed=55.2s lr=3.000000e-04 loss=2.4007 loss_recon=2.4007 loss_meanflow=0.0000 mean_model_t=0.4973 mean_corrupt_t=0.4973 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7876 corrupt_frac=0.5487 acc_corrupt=0.6249 loss_corrupt=2.4007 wrong_frac=0.5057 init_acc_corrupt=0.4585 acc_corrupt_t_0p0_0p2=0.2515 corrupt_frac_t_0p0_0p2=0.2097 acc_corrupt_t_0p2_0p4=0.4589 corrupt_frac_t_0p2_0p4=0.1992 acc_corrupt_t_0p4_0p6=0.6741 corrupt_frac_t_0p4_0p6=0.1998 acc_corrupt_t_0p6_0p8=0.8213 corrupt_frac_t_0p6_0p8=0.1930 acc_corrupt_t_0p8_1p0=0.9444 corrupt_frac_t_0p8_1p0=0.1992 out_w_norm=95.5347 out_g_norm=0.2298 loss_all=1.3713 init_gold_top10=0.4254 init_gold_top100=0.4330 +step=5900 micro_steps=23600 elapsed=55.1s lr=3.000000e-04 loss=2.3311 loss_recon=2.3311 loss_meanflow=0.0000 mean_model_t=0.5005 mean_corrupt_t=0.5005 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7932 corrupt_frac=0.5480 acc_corrupt=0.6341 loss_corrupt=2.3311 wrong_frac=0.4986 init_acc_corrupt=0.4668 acc_corrupt_t_0p0_0p2=0.2584 corrupt_frac_t_0p0_0p2=0.1928 acc_corrupt_t_0p2_0p4=0.4617 corrupt_frac_t_0p2_0p4=0.2051 acc_corrupt_t_0p4_0p6=0.6716 corrupt_frac_t_0p4_0p6=0.2019 acc_corrupt_t_0p6_0p8=0.8240 corrupt_frac_t_0p6_0p8=0.1990 acc_corrupt_t_0p8_1p0=0.9444 corrupt_frac_t_0p8_1p0=0.2012 out_w_norm=96.5049 out_g_norm=0.2286 loss_all=1.1085 init_gold_top10=0.4756 init_gold_top100=0.4833 +step=6000 micro_steps=24000 elapsed=55.2s lr=3.000000e-04 loss=2.3443 loss_recon=2.3443 loss_meanflow=0.0000 mean_model_t=0.4990 mean_corrupt_t=0.4990 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7906 corrupt_frac=0.5530 acc_corrupt=0.6323 loss_corrupt=2.3443 wrong_frac=0.5001 init_acc_corrupt=0.4654 acc_corrupt_t_0p0_0p2=0.2580 corrupt_frac_t_0p0_0p2=0.2035 acc_corrupt_t_0p2_0p4=0.4661 corrupt_frac_t_0p2_0p4=0.2002 acc_corrupt_t_0p4_0p6=0.6733 corrupt_frac_t_0p4_0p6=0.1944 acc_corrupt_t_0p6_0p8=0.8225 corrupt_frac_t_0p6_0p8=0.1999 acc_corrupt_t_0p8_1p0=0.9451 corrupt_frac_t_0p8_1p0=0.2030 out_w_norm=97.4815 out_g_norm=0.2252 loss_all=1.5445 init_gold_top10=0.4857 init_gold_top100=0.4926 +step=6100 micro_steps=24400 elapsed=118.4s lr=3.000000e-04 loss=2.3494 loss_recon=2.3494 loss_meanflow=0.0000 mean_model_t=0.4971 mean_corrupt_t=0.4971 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7915 corrupt_frac=0.5479 acc_corrupt=0.6305 loss_corrupt=2.3494 wrong_frac=0.5038 init_acc_corrupt=0.4608 acc_corrupt_t_0p0_0p2=0.2587 corrupt_frac_t_0p0_0p2=0.2003 acc_corrupt_t_0p2_0p4=0.4651 corrupt_frac_t_0p2_0p4=0.2062 acc_corrupt_t_0p4_0p6=0.6750 corrupt_frac_t_0p4_0p6=0.1982 acc_corrupt_t_0p6_0p8=0.8211 corrupt_frac_t_0p6_0p8=0.2001 acc_corrupt_t_0p8_1p0=0.9449 corrupt_frac_t_0p8_1p0=0.1957 out_w_norm=98.4490 out_g_norm=0.2245 loss_all=1.3603 init_gold_top10=0.5530 init_gold_top100=0.5566 +step=6200 micro_steps=24800 elapsed=84.2s lr=3.000000e-04 loss=2.3486 loss_recon=2.3486 loss_meanflow=0.0000 mean_model_t=0.4975 mean_corrupt_t=0.4975 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7902 corrupt_frac=0.5527 acc_corrupt=0.6313 loss_corrupt=2.3486 wrong_frac=0.5033 init_acc_corrupt=0.4620 acc_corrupt_t_0p0_0p2=0.2596 corrupt_frac_t_0p0_0p2=0.2026 acc_corrupt_t_0p2_0p4=0.4676 corrupt_frac_t_0p2_0p4=0.2020 acc_corrupt_t_0p4_0p6=0.6744 corrupt_frac_t_0p4_0p6=0.2013 acc_corrupt_t_0p6_0p8=0.8219 corrupt_frac_t_0p6_0p8=0.1986 acc_corrupt_t_0p8_1p0=0.9444 corrupt_frac_t_0p8_1p0=0.1969 out_w_norm=99.4166 out_g_norm=0.2215 loss_all=1.3413 init_gold_top10=0.4376 init_gold_top100=0.4425 +step=6300 micro_steps=25200 elapsed=55.3s lr=3.000000e-04 loss=2.3106 loss_recon=2.3106 loss_meanflow=0.0000 mean_model_t=0.5031 mean_corrupt_t=0.5031 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7955 corrupt_frac=0.5466 acc_corrupt=0.6364 loss_corrupt=2.3106 wrong_frac=0.4975 init_acc_corrupt=0.4680 acc_corrupt_t_0p0_0p2=0.2596 corrupt_frac_t_0p0_0p2=0.1966 acc_corrupt_t_0p2_0p4=0.4672 corrupt_frac_t_0p2_0p4=0.1989 acc_corrupt_t_0p4_0p6=0.6767 corrupt_frac_t_0p4_0p6=0.2030 acc_corrupt_t_0p6_0p8=0.8245 corrupt_frac_t_0p6_0p8=0.2022 acc_corrupt_t_0p8_1p0=0.9437 corrupt_frac_t_0p8_1p0=0.2004 out_w_norm=100.3855 out_g_norm=0.2203 loss_all=1.0722 init_gold_top10=0.5374 init_gold_top100=0.5437 +step=6400 micro_steps=25600 elapsed=55.6s lr=3.000000e-04 loss=2.3505 loss_recon=2.3505 loss_meanflow=0.0000 mean_model_t=0.4966 mean_corrupt_t=0.4966 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7899 corrupt_frac=0.5523 acc_corrupt=0.6303 loss_corrupt=2.3505 wrong_frac=0.5048 init_acc_corrupt=0.4605 acc_corrupt_t_0p0_0p2=0.2585 corrupt_frac_t_0p0_0p2=0.2075 acc_corrupt_t_0p2_0p4=0.4654 corrupt_frac_t_0p2_0p4=0.1988 acc_corrupt_t_0p4_0p6=0.6780 corrupt_frac_t_0p4_0p6=0.1986 acc_corrupt_t_0p6_0p8=0.8247 corrupt_frac_t_0p6_0p8=0.2043 acc_corrupt_t_0p8_1p0=0.9467 corrupt_frac_t_0p8_1p0=0.1913 out_w_norm=101.3503 out_g_norm=0.2183 loss_all=0.9810 init_gold_top10=0.5506 init_gold_top100=0.5557 +step=6500 micro_steps=26000 elapsed=55.6s lr=3.000000e-04 loss=2.3192 loss_recon=2.3192 loss_meanflow=0.0000 mean_model_t=0.4977 mean_corrupt_t=0.4977 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7934 corrupt_frac=0.5499 acc_corrupt=0.6347 loss_corrupt=2.3192 wrong_frac=0.5009 init_acc_corrupt=0.4645 acc_corrupt_t_0p0_0p2=0.2624 corrupt_frac_t_0p0_0p2=0.2041 acc_corrupt_t_0p2_0p4=0.4689 corrupt_frac_t_0p2_0p4=0.2003 acc_corrupt_t_0p4_0p6=0.6765 corrupt_frac_t_0p4_0p6=0.1977 acc_corrupt_t_0p6_0p8=0.8249 corrupt_frac_t_0p6_0p8=0.1980 acc_corrupt_t_0p8_1p0=0.9443 corrupt_frac_t_0p8_1p0=0.2025 out_w_norm=102.3130 out_g_norm=0.2172 loss_all=1.3621 init_gold_top10=0.4290 init_gold_top100=0.4339 +step=6600 micro_steps=26400 elapsed=55.6s lr=3.000000e-04 loss=2.3402 loss_recon=2.3402 loss_meanflow=0.0000 mean_model_t=0.4968 mean_corrupt_t=0.4968 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7907 corrupt_frac=0.5527 acc_corrupt=0.6317 loss_corrupt=2.3402 wrong_frac=0.5035 init_acc_corrupt=0.4620 acc_corrupt_t_0p0_0p2=0.2570 corrupt_frac_t_0p0_0p2=0.2040 acc_corrupt_t_0p2_0p4=0.4667 corrupt_frac_t_0p2_0p4=0.1986 acc_corrupt_t_0p4_0p6=0.6771 corrupt_frac_t_0p4_0p6=0.2014 acc_corrupt_t_0p6_0p8=0.8253 corrupt_frac_t_0p6_0p8=0.2013 acc_corrupt_t_0p8_1p0=0.9457 corrupt_frac_t_0p8_1p0=0.1946 out_w_norm=103.2708 out_g_norm=0.2150 loss_all=1.4859 init_gold_top10=0.3859 init_gold_top100=0.3947 +step=6700 micro_steps=26800 elapsed=55.6s lr=3.000000e-04 loss=2.3093 loss_recon=2.3093 loss_meanflow=0.0000 mean_model_t=0.4990 mean_corrupt_t=0.4990 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7942 corrupt_frac=0.5508 acc_corrupt=0.6360 loss_corrupt=2.3093 wrong_frac=0.5005 init_acc_corrupt=0.4648 acc_corrupt_t_0p0_0p2=0.2627 corrupt_frac_t_0p0_0p2=0.1997 acc_corrupt_t_0p2_0p4=0.4675 corrupt_frac_t_0p2_0p4=0.1986 acc_corrupt_t_0p4_0p6=0.6768 corrupt_frac_t_0p4_0p6=0.2045 acc_corrupt_t_0p6_0p8=0.8244 corrupt_frac_t_0p6_0p8=0.1996 acc_corrupt_t_0p8_1p0=0.9465 corrupt_frac_t_0p8_1p0=0.2011 out_w_norm=104.2309 out_g_norm=0.2135 loss_all=1.7234 init_gold_top10=0.3940 init_gold_top100=0.4036 +step=6800 micro_steps=27200 elapsed=55.6s lr=3.000000e-04 loss=2.2579 loss_recon=2.2579 loss_meanflow=0.0000 mean_model_t=0.5066 mean_corrupt_t=0.5066 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7974 corrupt_frac=0.5527 acc_corrupt=0.6432 loss_corrupt=2.2579 wrong_frac=0.4926 init_acc_corrupt=0.4735 acc_corrupt_t_0p0_0p2=0.2647 corrupt_frac_t_0p0_0p2=0.1937 acc_corrupt_t_0p2_0p4=0.4705 corrupt_frac_t_0p2_0p4=0.1973 acc_corrupt_t_0p4_0p6=0.6764 corrupt_frac_t_0p4_0p6=0.1958 acc_corrupt_t_0p6_0p8=0.8268 corrupt_frac_t_0p6_0p8=0.2052 acc_corrupt_t_0p8_1p0=0.9464 corrupt_frac_t_0p8_1p0=0.2085 out_w_norm=105.1726 out_g_norm=0.2116 loss_all=0.9669 init_gold_top10=0.5611 init_gold_top100=0.5643 +step=6900 micro_steps=27600 elapsed=55.6s lr=3.000000e-04 loss=2.2972 loss_recon=2.2972 loss_meanflow=0.0000 mean_model_t=0.5009 mean_corrupt_t=0.5009 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7955 corrupt_frac=0.5492 acc_corrupt=0.6374 loss_corrupt=2.2972 wrong_frac=0.4997 init_acc_corrupt=0.4658 acc_corrupt_t_0p0_0p2=0.2612 corrupt_frac_t_0p0_0p2=0.2025 acc_corrupt_t_0p2_0p4=0.4741 corrupt_frac_t_0p2_0p4=0.1978 acc_corrupt_t_0p4_0p6=0.6790 corrupt_frac_t_0p4_0p6=0.1968 acc_corrupt_t_0p6_0p8=0.8260 corrupt_frac_t_0p6_0p8=0.2018 acc_corrupt_t_0p8_1p0=0.9453 corrupt_frac_t_0p8_1p0=0.2021 out_w_norm=106.1272 out_g_norm=0.2120 loss_all=1.2162 init_gold_top10=0.5118 init_gold_top100=0.5181 +step=7000 micro_steps=28000 elapsed=55.5s lr=3.000000e-04 loss=2.3056 loss_recon=2.3056 loss_meanflow=0.0000 mean_model_t=0.4964 mean_corrupt_t=0.4964 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7955 corrupt_frac=0.5461 acc_corrupt=0.6356 loss_corrupt=2.3056 wrong_frac=0.5025 init_acc_corrupt=0.4626 acc_corrupt_t_0p0_0p2=0.2619 corrupt_frac_t_0p0_0p2=0.2010 acc_corrupt_t_0p2_0p4=0.4736 corrupt_frac_t_0p2_0p4=0.2064 acc_corrupt_t_0p4_0p6=0.6807 corrupt_frac_t_0p4_0p6=0.1978 acc_corrupt_t_0p6_0p8=0.8273 corrupt_frac_t_0p6_0p8=0.1979 acc_corrupt_t_0p8_1p0=0.9475 corrupt_frac_t_0p8_1p0=0.1979 out_w_norm=107.0884 out_g_norm=0.2104 loss_all=1.5313 init_gold_top10=0.4293 init_gold_top100=0.4348 +step=7100 micro_steps=28400 elapsed=110.3s lr=3.000000e-04 loss=2.2829 loss_recon=2.2829 loss_meanflow=0.0000 mean_model_t=0.4989 mean_corrupt_t=0.4989 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7961 corrupt_frac=0.5486 acc_corrupt=0.6383 loss_corrupt=2.2829 wrong_frac=0.5004 init_acc_corrupt=0.4652 acc_corrupt_t_0p0_0p2=0.2658 corrupt_frac_t_0p0_0p2=0.2004 acc_corrupt_t_0p2_0p4=0.4735 corrupt_frac_t_0p2_0p4=0.1990 acc_corrupt_t_0p4_0p6=0.6794 corrupt_frac_t_0p4_0p6=0.2032 acc_corrupt_t_0p6_0p8=0.8269 corrupt_frac_t_0p6_0p8=0.2007 acc_corrupt_t_0p8_1p0=0.9464 corrupt_frac_t_0p8_1p0=0.1978 out_w_norm=108.0449 out_g_norm=0.2091 loss_all=0.9928 init_gold_top10=0.4920 init_gold_top100=0.4985 +step=7200 micro_steps=28800 elapsed=92.1s lr=3.000000e-04 loss=2.2894 loss_recon=2.2894 loss_meanflow=0.0000 mean_model_t=0.4986 mean_corrupt_t=0.4986 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7948 corrupt_frac=0.5515 acc_corrupt=0.6378 loss_corrupt=2.2894 wrong_frac=0.5007 init_acc_corrupt=0.4646 acc_corrupt_t_0p0_0p2=0.2632 corrupt_frac_t_0p0_0p2=0.1995 acc_corrupt_t_0p2_0p4=0.4709 corrupt_frac_t_0p2_0p4=0.2027 acc_corrupt_t_0p4_0p6=0.6815 corrupt_frac_t_0p4_0p6=0.1992 acc_corrupt_t_0p6_0p8=0.8285 corrupt_frac_t_0p6_0p8=0.1980 acc_corrupt_t_0p8_1p0=0.9471 corrupt_frac_t_0p8_1p0=0.2010 out_w_norm=108.9972 out_g_norm=0.2066 loss_all=1.2372 init_gold_top10=0.5186 init_gold_top100=0.5275 +step=7300 micro_steps=29200 elapsed=55.1s lr=3.000000e-04 loss=2.2544 loss_recon=2.2544 loss_meanflow=0.0000 mean_model_t=0.5031 mean_corrupt_t=0.5031 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7971 corrupt_frac=0.5535 acc_corrupt=0.6428 loss_corrupt=2.2544 wrong_frac=0.4959 init_acc_corrupt=0.4708 acc_corrupt_t_0p0_0p2=0.2643 corrupt_frac_t_0p0_0p2=0.1939 acc_corrupt_t_0p2_0p4=0.4727 corrupt_frac_t_0p2_0p4=0.1967 acc_corrupt_t_0p4_0p6=0.6803 corrupt_frac_t_0p4_0p6=0.2043 acc_corrupt_t_0p6_0p8=0.8285 corrupt_frac_t_0p6_0p8=0.2071 acc_corrupt_t_0p8_1p0=0.9475 corrupt_frac_t_0p8_1p0=0.1990 out_w_norm=109.9447 out_g_norm=0.2053 loss_all=1.0846 init_gold_top10=0.5578 init_gold_top100=0.5626 +step=7400 micro_steps=29600 elapsed=55.2s lr=3.000000e-04 loss=2.2395 loss_recon=2.2395 loss_meanflow=0.0000 mean_model_t=0.5065 mean_corrupt_t=0.5065 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7998 corrupt_frac=0.5503 acc_corrupt=0.6451 loss_corrupt=2.2395 wrong_frac=0.4925 init_acc_corrupt=0.4730 acc_corrupt_t_0p0_0p2=0.2639 corrupt_frac_t_0p0_0p2=0.1898 acc_corrupt_t_0p2_0p4=0.4744 corrupt_frac_t_0p2_0p4=0.2118 acc_corrupt_t_0p4_0p6=0.6819 corrupt_frac_t_0p4_0p6=0.1838 acc_corrupt_t_0p6_0p8=0.8289 corrupt_frac_t_0p6_0p8=0.2028 acc_corrupt_t_0p8_1p0=0.9481 corrupt_frac_t_0p8_1p0=0.2122 out_w_norm=110.8972 out_g_norm=0.2041 loss_all=1.1188 init_gold_top10=0.5453 init_gold_top100=0.5469 +step=7500 micro_steps=30000 elapsed=55.2s lr=3.000000e-04 loss=2.2359 loss_recon=2.2359 loss_meanflow=0.0000 mean_model_t=0.5037 mean_corrupt_t=0.5037 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8010 corrupt_frac=0.5458 acc_corrupt=0.6447 loss_corrupt=2.2359 wrong_frac=0.4946 init_acc_corrupt=0.4711 acc_corrupt_t_0p0_0p2=0.2691 corrupt_frac_t_0p0_0p2=0.1960 acc_corrupt_t_0p2_0p4=0.4756 corrupt_frac_t_0p2_0p4=0.1972 acc_corrupt_t_0p4_0p6=0.6811 corrupt_frac_t_0p4_0p6=0.1969 acc_corrupt_t_0p6_0p8=0.8285 corrupt_frac_t_0p6_0p8=0.2026 acc_corrupt_t_0p8_1p0=0.9467 corrupt_frac_t_0p8_1p0=0.2077 out_w_norm=111.8448 out_g_norm=0.2043 loss_all=1.2759 init_gold_top10=0.4327 init_gold_top100=0.4401 +step=7600 micro_steps=30400 elapsed=55.2s lr=3.000000e-04 loss=2.2620 loss_recon=2.2620 loss_meanflow=0.0000 mean_model_t=0.5006 mean_corrupt_t=0.5006 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7987 corrupt_frac=0.5471 acc_corrupt=0.6414 loss_corrupt=2.2620 wrong_frac=0.4985 init_acc_corrupt=0.4672 acc_corrupt_t_0p0_0p2=0.2643 corrupt_frac_t_0p0_0p2=0.1989 acc_corrupt_t_0p2_0p4=0.4768 corrupt_frac_t_0p2_0p4=0.2003 acc_corrupt_t_0p4_0p6=0.6847 corrupt_frac_t_0p4_0p6=0.1952 acc_corrupt_t_0p6_0p8=0.8279 corrupt_frac_t_0p6_0p8=0.2059 acc_corrupt_t_0p8_1p0=0.9476 corrupt_frac_t_0p8_1p0=0.1998 out_w_norm=112.7959 out_g_norm=0.2019 loss_all=1.3598 init_gold_top10=0.4540 init_gold_top100=0.4601 +step=7700 micro_steps=30800 elapsed=55.2s lr=3.000000e-04 loss=2.2534 loss_recon=2.2534 loss_meanflow=0.0000 mean_model_t=0.5024 mean_corrupt_t=0.5024 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7982 corrupt_frac=0.5498 acc_corrupt=0.6424 loss_corrupt=2.2534 wrong_frac=0.4979 init_acc_corrupt=0.4684 acc_corrupt_t_0p0_0p2=0.2654 corrupt_frac_t_0p0_0p2=0.1979 acc_corrupt_t_0p2_0p4=0.4785 corrupt_frac_t_0p2_0p4=0.1966 acc_corrupt_t_0p4_0p6=0.6815 corrupt_frac_t_0p4_0p6=0.2029 acc_corrupt_t_0p6_0p8=0.8274 corrupt_frac_t_0p6_0p8=0.2016 acc_corrupt_t_0p8_1p0=0.9485 corrupt_frac_t_0p8_1p0=0.2021 out_w_norm=113.7400 out_g_norm=0.2011 loss_all=1.4613 init_gold_top10=0.4488 init_gold_top100=0.4568 +step=7800 micro_steps=31200 elapsed=55.2s lr=3.000000e-04 loss=2.2673 loss_recon=2.2673 loss_meanflow=0.0000 mean_model_t=0.5009 mean_corrupt_t=0.5009 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7980 corrupt_frac=0.5474 acc_corrupt=0.6406 loss_corrupt=2.2673 wrong_frac=0.5005 init_acc_corrupt=0.4648 acc_corrupt_t_0p0_0p2=0.2678 corrupt_frac_t_0p0_0p2=0.2028 acc_corrupt_t_0p2_0p4=0.4747 corrupt_frac_t_0p2_0p4=0.1978 acc_corrupt_t_0p4_0p6=0.6826 corrupt_frac_t_0p4_0p6=0.1965 acc_corrupt_t_0p6_0p8=0.8316 corrupt_frac_t_0p6_0p8=0.2057 acc_corrupt_t_0p8_1p0=0.9489 corrupt_frac_t_0p8_1p0=0.1977 out_w_norm=114.6713 out_g_norm=0.1996 loss_all=1.2077 init_gold_top10=0.4915 init_gold_top100=0.4986 +step=7900 micro_steps=31600 elapsed=55.2s lr=3.000000e-04 loss=2.2700 loss_recon=2.2700 loss_meanflow=0.0000 mean_model_t=0.4990 mean_corrupt_t=0.4990 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7974 corrupt_frac=0.5483 acc_corrupt=0.6396 loss_corrupt=2.2700 wrong_frac=0.5024 init_acc_corrupt=0.4624 acc_corrupt_t_0p0_0p2=0.2675 corrupt_frac_t_0p0_0p2=0.2026 acc_corrupt_t_0p2_0p4=0.4795 corrupt_frac_t_0p2_0p4=0.2026 acc_corrupt_t_0p4_0p6=0.6856 corrupt_frac_t_0p4_0p6=0.2037 acc_corrupt_t_0p6_0p8=0.8311 corrupt_frac_t_0p6_0p8=0.1936 acc_corrupt_t_0p8_1p0=0.9479 corrupt_frac_t_0p8_1p0=0.1985 out_w_norm=115.6093 out_g_norm=0.1992 loss_all=1.2517 init_gold_top10=0.5114 init_gold_top100=0.5191 +step=8000 micro_steps=32000 elapsed=55.2s lr=3.000000e-04 loss=2.2735 loss_recon=2.2735 loss_meanflow=0.0000 mean_model_t=0.4974 mean_corrupt_t=0.4974 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7969 corrupt_frac=0.5483 acc_corrupt=0.6397 loss_corrupt=2.2735 wrong_frac=0.5022 init_acc_corrupt=0.4636 acc_corrupt_t_0p0_0p2=0.2660 corrupt_frac_t_0p0_0p2=0.2059 acc_corrupt_t_0p2_0p4=0.4802 corrupt_frac_t_0p2_0p4=0.1965 acc_corrupt_t_0p4_0p6=0.6847 corrupt_frac_t_0p4_0p6=0.2023 acc_corrupt_t_0p6_0p8=0.8308 corrupt_frac_t_0p6_0p8=0.1940 acc_corrupt_t_0p8_1p0=0.9481 corrupt_frac_t_0p8_1p0=0.2013 out_w_norm=116.5355 out_g_norm=0.1975 loss_all=1.1563 init_gold_top10=0.5354 init_gold_top100=0.5416 +step=8100 micro_steps=32400 elapsed=97.7s lr=3.000000e-04 loss=2.2584 loss_recon=2.2584 loss_meanflow=0.0000 mean_model_t=0.5001 mean_corrupt_t=0.5001 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7966 corrupt_frac=0.5529 acc_corrupt=0.6416 loss_corrupt=2.2584 wrong_frac=0.4993 init_acc_corrupt=0.4664 acc_corrupt_t_0p0_0p2=0.2681 corrupt_frac_t_0p0_0p2=0.2003 acc_corrupt_t_0p2_0p4=0.4782 corrupt_frac_t_0p2_0p4=0.1999 acc_corrupt_t_0p4_0p6=0.6834 corrupt_frac_t_0p4_0p6=0.1969 acc_corrupt_t_0p6_0p8=0.8290 corrupt_frac_t_0p6_0p8=0.2029 acc_corrupt_t_0p8_1p0=0.9475 corrupt_frac_t_0p8_1p0=0.2005 out_w_norm=117.4534 out_g_norm=0.1952 loss_all=1.6276 init_gold_top10=0.4510 init_gold_top100=0.4640 +step=8200 micro_steps=32800 elapsed=104.9s lr=3.000000e-04 loss=2.2617 loss_recon=2.2617 loss_meanflow=0.0000 mean_model_t=0.5004 mean_corrupt_t=0.5004 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7971 corrupt_frac=0.5507 acc_corrupt=0.6408 loss_corrupt=2.2617 wrong_frac=0.5010 init_acc_corrupt=0.4645 acc_corrupt_t_0p0_0p2=0.2681 corrupt_frac_t_0p0_0p2=0.2012 acc_corrupt_t_0p2_0p4=0.4801 corrupt_frac_t_0p2_0p4=0.2008 acc_corrupt_t_0p4_0p6=0.6831 corrupt_frac_t_0p4_0p6=0.2024 acc_corrupt_t_0p6_0p8=0.8305 corrupt_frac_t_0p6_0p8=0.1992 acc_corrupt_t_0p8_1p0=0.9484 corrupt_frac_t_0p8_1p0=0.1985 out_w_norm=118.3682 out_g_norm=0.1938 loss_all=1.0356 init_gold_top10=0.5087 init_gold_top100=0.5128 +step=8300 micro_steps=33200 elapsed=55.2s lr=3.000000e-04 loss=2.2511 loss_recon=2.2511 loss_meanflow=0.0000 mean_model_t=0.5004 mean_corrupt_t=0.5004 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7990 corrupt_frac=0.5474 acc_corrupt=0.6422 loss_corrupt=2.2511 wrong_frac=0.5007 init_acc_corrupt=0.4648 acc_corrupt_t_0p0_0p2=0.2690 corrupt_frac_t_0p0_0p2=0.2012 acc_corrupt_t_0p2_0p4=0.4803 corrupt_frac_t_0p2_0p4=0.1992 acc_corrupt_t_0p4_0p6=0.6844 corrupt_frac_t_0p4_0p6=0.1993 acc_corrupt_t_0p6_0p8=0.8302 corrupt_frac_t_0p6_0p8=0.1993 acc_corrupt_t_0p8_1p0=0.9481 corrupt_frac_t_0p8_1p0=0.2010 out_w_norm=119.2832 out_g_norm=0.1944 loss_all=1.4781 init_gold_top10=0.5541 init_gold_top100=0.5596 +step=8400 micro_steps=33600 elapsed=55.2s lr=3.000000e-04 loss=2.2411 loss_recon=2.2411 loss_meanflow=0.0000 mean_model_t=0.4999 mean_corrupt_t=0.4999 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7983 corrupt_frac=0.5518 acc_corrupt=0.6438 loss_corrupt=2.2411 wrong_frac=0.4980 init_acc_corrupt=0.4675 acc_corrupt_t_0p0_0p2=0.2700 corrupt_frac_t_0p0_0p2=0.2024 acc_corrupt_t_0p2_0p4=0.4805 corrupt_frac_t_0p2_0p4=0.1945 acc_corrupt_t_0p4_0p6=0.6819 corrupt_frac_t_0p4_0p6=0.2011 acc_corrupt_t_0p6_0p8=0.8327 corrupt_frac_t_0p6_0p8=0.1969 acc_corrupt_t_0p8_1p0=0.9475 corrupt_frac_t_0p8_1p0=0.2056 out_w_norm=120.1928 out_g_norm=0.1908 loss_all=1.5874 init_gold_top10=0.4601 init_gold_top100=0.4657 +step=8500 micro_steps=34000 elapsed=55.2s lr=3.000000e-04 loss=2.2281 loss_recon=2.2281 loss_meanflow=0.0000 mean_model_t=0.5015 mean_corrupt_t=0.5015 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7987 corrupt_frac=0.5527 acc_corrupt=0.6451 loss_corrupt=2.2281 wrong_frac=0.4974 init_acc_corrupt=0.4686 acc_corrupt_t_0p0_0p2=0.2717 corrupt_frac_t_0p0_0p2=0.1978 acc_corrupt_t_0p2_0p4=0.4783 corrupt_frac_t_0p2_0p4=0.2013 acc_corrupt_t_0p4_0p6=0.6892 corrupt_frac_t_0p4_0p6=0.1941 acc_corrupt_t_0p6_0p8=0.8278 corrupt_frac_t_0p6_0p8=0.2035 acc_corrupt_t_0p8_1p0=0.9475 corrupt_frac_t_0p8_1p0=0.2038 out_w_norm=121.0937 out_g_norm=0.1904 loss_all=1.0633 init_gold_top10=0.5148 init_gold_top100=0.5218 +step=8600 micro_steps=34400 elapsed=55.8s lr=3.000000e-04 loss=2.2317 loss_recon=2.2317 loss_meanflow=0.0000 mean_model_t=0.4998 mean_corrupt_t=0.4998 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8016 corrupt_frac=0.5444 acc_corrupt=0.6448 loss_corrupt=2.2317 wrong_frac=0.4993 init_acc_corrupt=0.4661 acc_corrupt_t_0p0_0p2=0.2720 corrupt_frac_t_0p0_0p2=0.1986 acc_corrupt_t_0p2_0p4=0.4816 corrupt_frac_t_0p2_0p4=0.2018 acc_corrupt_t_0p4_0p6=0.6873 corrupt_frac_t_0p4_0p6=0.1990 acc_corrupt_t_0p6_0p8=0.8331 corrupt_frac_t_0p6_0p8=0.2002 acc_corrupt_t_0p8_1p0=0.9479 corrupt_frac_t_0p8_1p0=0.2004 out_w_norm=121.9897 out_g_norm=0.1897 loss_all=1.0934 init_gold_top10=0.4684 init_gold_top100=0.4757 +step=8700 micro_steps=34800 elapsed=55.8s lr=3.000000e-04 loss=2.2557 loss_recon=2.2557 loss_meanflow=0.0000 mean_model_t=0.4960 mean_corrupt_t=0.4960 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7960 corrupt_frac=0.5547 acc_corrupt=0.6411 loss_corrupt=2.2557 wrong_frac=0.5025 init_acc_corrupt=0.4626 acc_corrupt_t_0p0_0p2=0.2727 corrupt_frac_t_0p0_0p2=0.2044 acc_corrupt_t_0p2_0p4=0.4815 corrupt_frac_t_0p2_0p4=0.2008 acc_corrupt_t_0p4_0p6=0.6833 corrupt_frac_t_0p4_0p6=0.1974 acc_corrupt_t_0p6_0p8=0.8314 corrupt_frac_t_0p6_0p8=0.1994 acc_corrupt_t_0p8_1p0=0.9494 corrupt_frac_t_0p8_1p0=0.1989 out_w_norm=122.8828 out_g_norm=0.1896 loss_all=0.9112 init_gold_top10=0.5567 init_gold_top100=0.5591 +step=8800 micro_steps=35200 elapsed=55.7s lr=3.000000e-04 loss=2.2512 loss_recon=2.2512 loss_meanflow=0.0000 mean_model_t=0.5025 mean_corrupt_t=0.5025 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7979 corrupt_frac=0.5518 acc_corrupt=0.6424 loss_corrupt=2.2512 wrong_frac=0.5004 init_acc_corrupt=0.4649 acc_corrupt_t_0p0_0p2=0.2670 corrupt_frac_t_0p0_0p2=0.1958 acc_corrupt_t_0p2_0p4=0.4766 corrupt_frac_t_0p2_0p4=0.2074 acc_corrupt_t_0p4_0p6=0.6878 corrupt_frac_t_0p4_0p6=0.1946 acc_corrupt_t_0p6_0p8=0.8315 corrupt_frac_t_0p6_0p8=0.2075 acc_corrupt_t_0p8_1p0=0.9479 corrupt_frac_t_0p8_1p0=0.1957 out_w_norm=123.7751 out_g_norm=0.1875 loss_all=1.3859 init_gold_top10=0.4596 init_gold_top100=0.4640 +step=8900 micro_steps=35600 elapsed=55.7s lr=3.000000e-04 loss=2.2581 loss_recon=2.2581 loss_meanflow=0.0000 mean_model_t=0.4994 mean_corrupt_t=0.4994 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7974 corrupt_frac=0.5512 acc_corrupt=0.6414 loss_corrupt=2.2581 wrong_frac=0.5014 init_acc_corrupt=0.4631 acc_corrupt_t_0p0_0p2=0.2696 corrupt_frac_t_0p0_0p2=0.1982 acc_corrupt_t_0p2_0p4=0.4752 corrupt_frac_t_0p2_0p4=0.2063 acc_corrupt_t_0p4_0p6=0.6891 corrupt_frac_t_0p4_0p6=0.1972 acc_corrupt_t_0p6_0p8=0.8312 corrupt_frac_t_0p6_0p8=0.2045 acc_corrupt_t_0p8_1p0=0.9498 corrupt_frac_t_0p8_1p0=0.1939 out_w_norm=124.6735 out_g_norm=0.1864 loss_all=0.9142 init_gold_top10=0.5545 init_gold_top100=0.5573 +step=9000 micro_steps=36000 elapsed=55.7s lr=3.000000e-04 loss=2.2359 loss_recon=2.2359 loss_meanflow=0.0000 mean_model_t=0.5010 mean_corrupt_t=0.5010 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7990 corrupt_frac=0.5514 acc_corrupt=0.6444 loss_corrupt=2.2359 wrong_frac=0.4991 init_acc_corrupt=0.4661 acc_corrupt_t_0p0_0p2=0.2702 corrupt_frac_t_0p0_0p2=0.2004 acc_corrupt_t_0p2_0p4=0.4786 corrupt_frac_t_0p2_0p4=0.1991 acc_corrupt_t_0p4_0p6=0.6879 corrupt_frac_t_0p4_0p6=0.1981 acc_corrupt_t_0p6_0p8=0.8305 corrupt_frac_t_0p6_0p8=0.2023 acc_corrupt_t_0p8_1p0=0.9502 corrupt_frac_t_0p8_1p0=0.2011 out_w_norm=125.5700 out_g_norm=0.1853 loss_all=1.5726 init_gold_top10=0.4751 init_gold_top100=0.4798 +step=9100 micro_steps=36400 elapsed=88.6s lr=3.000000e-04 loss=2.2463 loss_recon=2.2463 loss_meanflow=0.0000 mean_model_t=0.5006 mean_corrupt_t=0.5006 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7984 corrupt_frac=0.5503 acc_corrupt=0.6427 loss_corrupt=2.2463 wrong_frac=0.5016 init_acc_corrupt=0.4638 acc_corrupt_t_0p0_0p2=0.2695 corrupt_frac_t_0p0_0p2=0.2021 acc_corrupt_t_0p2_0p4=0.4796 corrupt_frac_t_0p2_0p4=0.2029 acc_corrupt_t_0p4_0p6=0.6888 corrupt_frac_t_0p4_0p6=0.1965 acc_corrupt_t_0p6_0p8=0.8341 corrupt_frac_t_0p6_0p8=0.1974 acc_corrupt_t_0p8_1p0=0.9483 corrupt_frac_t_0p8_1p0=0.2026 out_w_norm=126.4580 out_g_norm=0.1850 loss_all=1.0020 init_gold_top10=0.5376 init_gold_top100=0.5423 +step=9200 micro_steps=36800 elapsed=114.0s lr=3.000000e-04 loss=2.2255 loss_recon=2.2255 loss_meanflow=0.0000 mean_model_t=0.4991 mean_corrupt_t=0.4991 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8011 corrupt_frac=0.5471 acc_corrupt=0.6449 loss_corrupt=2.2255 wrong_frac=0.5002 init_acc_corrupt=0.4648 acc_corrupt_t_0p0_0p2=0.2734 corrupt_frac_t_0p0_0p2=0.1988 acc_corrupt_t_0p2_0p4=0.4856 corrupt_frac_t_0p2_0p4=0.2024 acc_corrupt_t_0p4_0p6=0.6864 corrupt_frac_t_0p4_0p6=0.2014 acc_corrupt_t_0p6_0p8=0.8319 corrupt_frac_t_0p6_0p8=0.1985 acc_corrupt_t_0p8_1p0=0.9494 corrupt_frac_t_0p8_1p0=0.1990 out_w_norm=127.3393 out_g_norm=0.1846 loss_all=1.4781 init_gold_top10=0.4442 init_gold_top100=0.4488 +step=9300 micro_steps=37200 elapsed=55.3s lr=3.000000e-04 loss=2.2602 loss_recon=2.2602 loss_meanflow=0.0000 mean_model_t=0.4943 mean_corrupt_t=0.4943 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7962 corrupt_frac=0.5523 acc_corrupt=0.6401 loss_corrupt=2.2602 wrong_frac=0.5062 init_acc_corrupt=0.4587 acc_corrupt_t_0p0_0p2=0.2703 corrupt_frac_t_0p0_0p2=0.2049 acc_corrupt_t_0p2_0p4=0.4804 corrupt_frac_t_0p2_0p4=0.2004 acc_corrupt_t_0p4_0p6=0.6889 corrupt_frac_t_0p4_0p6=0.2019 acc_corrupt_t_0p6_0p8=0.8330 corrupt_frac_t_0p6_0p8=0.2043 acc_corrupt_t_0p8_1p0=0.9490 corrupt_frac_t_0p8_1p0=0.1895 out_w_norm=128.2191 out_g_norm=0.1826 loss_all=1.2276 init_gold_top10=0.5235 init_gold_top100=0.5282 +step=9400 micro_steps=37600 elapsed=55.1s lr=3.000000e-04 loss=2.2091 loss_recon=2.2091 loss_meanflow=0.0000 mean_model_t=0.5030 mean_corrupt_t=0.5030 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8007 corrupt_frac=0.5514 acc_corrupt=0.6474 loss_corrupt=2.2091 wrong_frac=0.4978 init_acc_corrupt=0.4687 acc_corrupt_t_0p0_0p2=0.2694 corrupt_frac_t_0p0_0p2=0.1976 acc_corrupt_t_0p2_0p4=0.4845 corrupt_frac_t_0p2_0p4=0.1963 acc_corrupt_t_0p4_0p6=0.6885 corrupt_frac_t_0p4_0p6=0.2071 acc_corrupt_t_0p6_0p8=0.8363 corrupt_frac_t_0p6_0p8=0.2010 acc_corrupt_t_0p8_1p0=0.9490 corrupt_frac_t_0p8_1p0=0.1992 out_w_norm=129.0950 out_g_norm=0.1813 loss_all=1.4108 init_gold_top10=0.5643 init_gold_top100=0.5681 +step=9500 micro_steps=38000 elapsed=55.1s lr=3.000000e-04 loss=2.2348 loss_recon=2.2348 loss_meanflow=0.0000 mean_model_t=0.4969 mean_corrupt_t=0.4969 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7995 corrupt_frac=0.5486 acc_corrupt=0.6434 loss_corrupt=2.2348 wrong_frac=0.5023 init_acc_corrupt=0.4632 acc_corrupt_t_0p0_0p2=0.2755 corrupt_frac_t_0p0_0p2=0.1967 acc_corrupt_t_0p2_0p4=0.4797 corrupt_frac_t_0p2_0p4=0.2091 acc_corrupt_t_0p4_0p6=0.6828 corrupt_frac_t_0p4_0p6=0.2025 acc_corrupt_t_0p6_0p8=0.8350 corrupt_frac_t_0p6_0p8=0.1919 acc_corrupt_t_0p8_1p0=0.9489 corrupt_frac_t_0p8_1p0=0.2013 out_w_norm=129.9627 out_g_norm=0.1812 loss_all=1.4871 init_gold_top10=0.4609 init_gold_top100=0.4647 +step=9600 micro_steps=38400 elapsed=55.1s lr=3.000000e-04 loss=2.2399 loss_recon=2.2399 loss_meanflow=0.0000 mean_model_t=0.4987 mean_corrupt_t=0.4987 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7986 corrupt_frac=0.5491 acc_corrupt=0.6426 loss_corrupt=2.2399 wrong_frac=0.5029 init_acc_corrupt=0.4620 acc_corrupt_t_0p0_0p2=0.2707 corrupt_frac_t_0p0_0p2=0.2054 acc_corrupt_t_0p2_0p4=0.4828 corrupt_frac_t_0p2_0p4=0.2011 acc_corrupt_t_0p4_0p6=0.6909 corrupt_frac_t_0p4_0p6=0.1976 acc_corrupt_t_0p6_0p8=0.8336 corrupt_frac_t_0p6_0p8=0.1972 acc_corrupt_t_0p8_1p0=0.9502 corrupt_frac_t_0p8_1p0=0.1992 out_w_norm=130.8293 out_g_norm=0.1805 loss_all=1.6742 init_gold_top10=0.4245 init_gold_top100=0.4328 +step=9700 micro_steps=38800 elapsed=55.1s lr=3.000000e-04 loss=2.2033 loss_recon=2.2033 loss_meanflow=0.0000 mean_model_t=0.5015 mean_corrupt_t=0.5015 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8019 corrupt_frac=0.5489 acc_corrupt=0.6479 loss_corrupt=2.2033 wrong_frac=0.4979 init_acc_corrupt=0.4684 acc_corrupt_t_0p0_0p2=0.2699 corrupt_frac_t_0p0_0p2=0.1956 acc_corrupt_t_0p2_0p4=0.4855 corrupt_frac_t_0p2_0p4=0.1979 acc_corrupt_t_0p4_0p6=0.6876 corrupt_frac_t_0p4_0p6=0.2033 acc_corrupt_t_0p6_0p8=0.8347 corrupt_frac_t_0p6_0p8=0.2057 acc_corrupt_t_0p8_1p0=0.9494 corrupt_frac_t_0p8_1p0=0.1976 out_w_norm=131.6922 out_g_norm=0.1788 loss_all=1.3464 init_gold_top10=0.4513 init_gold_top100=0.4585 +step=9800 micro_steps=39200 elapsed=55.1s lr=3.000000e-04 loss=2.2432 loss_recon=2.2432 loss_meanflow=0.0000 mean_model_t=0.4972 mean_corrupt_t=0.4972 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7981 corrupt_frac=0.5506 acc_corrupt=0.6421 loss_corrupt=2.2432 wrong_frac=0.5046 init_acc_corrupt=0.4605 acc_corrupt_t_0p0_0p2=0.2721 corrupt_frac_t_0p0_0p2=0.2012 acc_corrupt_t_0p2_0p4=0.4831 corrupt_frac_t_0p2_0p4=0.2045 acc_corrupt_t_0p4_0p6=0.6898 corrupt_frac_t_0p4_0p6=0.2026 acc_corrupt_t_0p6_0p8=0.8334 corrupt_frac_t_0p6_0p8=0.1991 acc_corrupt_t_0p8_1p0=0.9497 corrupt_frac_t_0p8_1p0=0.1926 out_w_norm=132.5535 out_g_norm=0.1786 loss_all=1.1403 init_gold_top10=0.5167 init_gold_top100=0.5187 +step=9900 micro_steps=39600 elapsed=55.1s lr=3.000000e-04 loss=2.2405 loss_recon=2.2405 loss_meanflow=0.0000 mean_model_t=0.5000 mean_corrupt_t=0.5000 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7976 corrupt_frac=0.5537 acc_corrupt=0.6432 loss_corrupt=2.2405 wrong_frac=0.5015 init_acc_corrupt=0.4636 acc_corrupt_t_0p0_0p2=0.2701 corrupt_frac_t_0p0_0p2=0.2054 acc_corrupt_t_0p2_0p4=0.4791 corrupt_frac_t_0p2_0p4=0.1986 acc_corrupt_t_0p4_0p6=0.6904 corrupt_frac_t_0p4_0p6=0.2000 acc_corrupt_t_0p6_0p8=0.8351 corrupt_frac_t_0p6_0p8=0.1951 acc_corrupt_t_0p8_1p0=0.9493 corrupt_frac_t_0p8_1p0=0.2029 out_w_norm=133.4086 out_g_norm=0.1760 loss_all=1.4809 init_gold_top10=0.4857 init_gold_top100=0.4941 +step=10000 micro_steps=40000 elapsed=55.1s lr=3.000000e-04 loss=2.2283 loss_recon=2.2283 loss_meanflow=0.0000 mean_model_t=0.5005 mean_corrupt_t=0.5005 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7980 corrupt_frac=0.5550 acc_corrupt=0.6448 loss_corrupt=2.2283 wrong_frac=0.5007 init_acc_corrupt=0.4644 acc_corrupt_t_0p0_0p2=0.2677 corrupt_frac_t_0p0_0p2=0.1973 acc_corrupt_t_0p2_0p4=0.4843 corrupt_frac_t_0p2_0p4=0.2078 acc_corrupt_t_0p4_0p6=0.6898 corrupt_frac_t_0p4_0p6=0.1990 acc_corrupt_t_0p6_0p8=0.8353 corrupt_frac_t_0p6_0p8=0.1967 acc_corrupt_t_0p8_1p0=0.9496 corrupt_frac_t_0p8_1p0=0.2002 out_w_norm=134.2528 out_g_norm=0.1752 loss_all=1.0706 init_gold_top10=0.5461 init_gold_top100=0.5496 +step=10100 micro_steps=40400 elapsed=112.7s lr=3.000000e-04 loss=2.2620 loss_recon=2.2620 loss_meanflow=0.0000 mean_model_t=0.4958 mean_corrupt_t=0.4958 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7987 corrupt_frac=0.5465 acc_corrupt=0.6409 loss_corrupt=2.2620 wrong_frac=0.5037 init_acc_corrupt=0.4613 acc_corrupt_t_0p0_0p2=0.2729 corrupt_frac_t_0p0_0p2=0.2040 acc_corrupt_t_0p2_0p4=0.4781 corrupt_frac_t_0p2_0p4=0.2006 acc_corrupt_t_0p4_0p6=0.6865 corrupt_frac_t_0p4_0p6=0.2036 acc_corrupt_t_0p6_0p8=0.8316 corrupt_frac_t_0p6_0p8=0.1957 acc_corrupt_t_0p8_1p0=0.9495 corrupt_frac_t_0p8_1p0=0.1971 out_w_norm=135.0591 out_g_norm=0.1728 loss_all=1.2958 init_gold_top10=0.4689 init_gold_top100=0.4778 +step=10200 micro_steps=40800 elapsed=89.8s lr=3.000000e-04 loss=2.2387 loss_recon=2.2387 loss_meanflow=0.0000 mean_model_t=0.4981 mean_corrupt_t=0.4981 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7989 corrupt_frac=0.5514 acc_corrupt=0.6446 loss_corrupt=2.2387 wrong_frac=0.4993 init_acc_corrupt=0.4661 acc_corrupt_t_0p0_0p2=0.2704 corrupt_frac_t_0p0_0p2=0.2029 acc_corrupt_t_0p2_0p4=0.4824 corrupt_frac_t_0p2_0p4=0.1957 acc_corrupt_t_0p4_0p6=0.6879 corrupt_frac_t_0p4_0p6=0.1978 acc_corrupt_t_0p6_0p8=0.8328 corrupt_frac_t_0p6_0p8=0.2049 acc_corrupt_t_0p8_1p0=0.9485 corrupt_frac_t_0p8_1p0=0.1996 out_w_norm=135.8204 out_g_norm=0.1699 loss_all=1.5501 init_gold_top10=0.4729 init_gold_top100=0.4788 +step=10300 micro_steps=41200 elapsed=55.5s lr=3.000000e-04 loss=2.2466 loss_recon=2.2466 loss_meanflow=0.0000 mean_model_t=0.4969 mean_corrupt_t=0.4969 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7989 corrupt_frac=0.5503 acc_corrupt=0.6433 loss_corrupt=2.2466 wrong_frac=0.5011 init_acc_corrupt=0.4645 acc_corrupt_t_0p0_0p2=0.2734 corrupt_frac_t_0p0_0p2=0.2033 acc_corrupt_t_0p2_0p4=0.4794 corrupt_frac_t_0p2_0p4=0.1975 acc_corrupt_t_0p4_0p6=0.6891 corrupt_frac_t_0p4_0p6=0.2029 acc_corrupt_t_0p6_0p8=0.8327 corrupt_frac_t_0p6_0p8=0.1986 acc_corrupt_t_0p8_1p0=0.9489 corrupt_frac_t_0p8_1p0=0.1991 out_w_norm=136.5715 out_g_norm=0.1690 loss_all=1.4086 init_gold_top10=0.3860 init_gold_top100=0.3935 +step=10400 micro_steps=41600 elapsed=55.5s lr=3.000000e-04 loss=2.2302 loss_recon=2.2302 loss_meanflow=0.0000 mean_model_t=0.4998 mean_corrupt_t=0.4998 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8008 corrupt_frac=0.5489 acc_corrupt=0.6459 loss_corrupt=2.2302 wrong_frac=0.4985 init_acc_corrupt=0.4670 acc_corrupt_t_0p0_0p2=0.2747 corrupt_frac_t_0p0_0p2=0.2030 acc_corrupt_t_0p2_0p4=0.4840 corrupt_frac_t_0p2_0p4=0.1931 acc_corrupt_t_0p4_0p6=0.6850 corrupt_frac_t_0p4_0p6=0.2012 acc_corrupt_t_0p6_0p8=0.8334 corrupt_frac_t_0p6_0p8=0.2012 acc_corrupt_t_0p8_1p0=0.9489 corrupt_frac_t_0p8_1p0=0.2025 out_w_norm=137.3240 out_g_norm=0.1680 loss_all=1.5893 init_gold_top10=0.4930 init_gold_top100=0.5009 +step=10500 micro_steps=42000 elapsed=55.4s lr=3.000000e-04 loss=2.2272 loss_recon=2.2272 loss_meanflow=0.0000 mean_model_t=0.4997 mean_corrupt_t=0.4997 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8003 corrupt_frac=0.5507 acc_corrupt=0.6461 loss_corrupt=2.2272 wrong_frac=0.4989 init_acc_corrupt=0.4667 acc_corrupt_t_0p0_0p2=0.2773 corrupt_frac_t_0p0_0p2=0.1974 acc_corrupt_t_0p2_0p4=0.4753 corrupt_frac_t_0p2_0p4=0.1980 acc_corrupt_t_0p4_0p6=0.6881 corrupt_frac_t_0p4_0p6=0.2015 acc_corrupt_t_0p6_0p8=0.8302 corrupt_frac_t_0p6_0p8=0.2033 acc_corrupt_t_0p8_1p0=0.9490 corrupt_frac_t_0p8_1p0=0.2009 out_w_norm=138.0695 out_g_norm=0.1662 loss_all=0.9601 init_gold_top10=0.5734 init_gold_top100=0.5797 +step=10600 micro_steps=42400 elapsed=98.0s lr=3.000000e-04 loss=2.2261 loss_recon=2.2261 loss_meanflow=0.0000 mean_model_t=0.5027 mean_corrupt_t=0.5027 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8018 corrupt_frac=0.5467 acc_corrupt=0.6462 loss_corrupt=2.2261 wrong_frac=0.4985 init_acc_corrupt=0.4670 acc_corrupt_t_0p0_0p2=0.2706 corrupt_frac_t_0p0_0p2=0.1982 acc_corrupt_t_0p2_0p4=0.4828 corrupt_frac_t_0p2_0p4=0.2023 acc_corrupt_t_0p4_0p6=0.6881 corrupt_frac_t_0p4_0p6=0.1988 acc_corrupt_t_0p6_0p8=0.8339 corrupt_frac_t_0p6_0p8=0.1976 acc_corrupt_t_0p8_1p0=0.9495 corrupt_frac_t_0p8_1p0=0.2041 out_w_norm=138.8178 out_g_norm=0.1663 loss_all=1.2571 init_gold_top10=0.5255 init_gold_top100=0.5327 +step=10700 micro_steps=42800 elapsed=126.5s lr=3.000000e-04 loss=2.2359 loss_recon=2.2359 loss_meanflow=0.0000 mean_model_t=0.4989 mean_corrupt_t=0.4989 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7988 corrupt_frac=0.5523 acc_corrupt=0.6445 loss_corrupt=2.2359 wrong_frac=0.5002 init_acc_corrupt=0.4647 acc_corrupt_t_0p0_0p2=0.2761 corrupt_frac_t_0p0_0p2=0.1974 acc_corrupt_t_0p2_0p4=0.4800 corrupt_frac_t_0p2_0p4=0.2073 acc_corrupt_t_0p4_0p6=0.6881 corrupt_frac_t_0p4_0p6=0.1957 acc_corrupt_t_0p6_0p8=0.8315 corrupt_frac_t_0p6_0p8=0.2004 acc_corrupt_t_0p8_1p0=0.9495 corrupt_frac_t_0p8_1p0=0.2001 out_w_norm=139.5640 out_g_norm=0.1648 loss_all=1.3581 init_gold_top10=0.4230 init_gold_top100=0.4305 +step=10800 micro_steps=43200 elapsed=126.5s lr=3.000000e-04 loss=2.2483 loss_recon=2.2483 loss_meanflow=0.0000 mean_model_t=0.4978 mean_corrupt_t=0.4978 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7987 corrupt_frac=0.5493 acc_corrupt=0.6425 loss_corrupt=2.2483 wrong_frac=0.5025 init_acc_corrupt=0.4622 acc_corrupt_t_0p0_0p2=0.2748 corrupt_frac_t_0p0_0p2=0.2038 acc_corrupt_t_0p2_0p4=0.4807 corrupt_frac_t_0p2_0p4=0.2033 acc_corrupt_t_0p4_0p6=0.6858 corrupt_frac_t_0p4_0p6=0.1927 acc_corrupt_t_0p6_0p8=0.8335 corrupt_frac_t_0p6_0p8=0.2009 acc_corrupt_t_0p8_1p0=0.9484 corrupt_frac_t_0p8_1p0=0.2014 out_w_norm=140.2978 out_g_norm=0.1645 loss_all=1.1744 init_gold_top10=0.4611 init_gold_top100=0.4692 +step=10900 micro_steps=43600 elapsed=111.0s lr=3.000000e-04 loss=2.2360 loss_recon=2.2360 loss_meanflow=0.0000 mean_model_t=0.4984 mean_corrupt_t=0.4984 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8006 corrupt_frac=0.5484 acc_corrupt=0.6449 loss_corrupt=2.2360 wrong_frac=0.5017 init_acc_corrupt=0.4632 acc_corrupt_t_0p0_0p2=0.2745 corrupt_frac_t_0p0_0p2=0.1975 acc_corrupt_t_0p2_0p4=0.4827 corrupt_frac_t_0p2_0p4=0.2039 acc_corrupt_t_0p4_0p6=0.6891 corrupt_frac_t_0p4_0p6=0.2060 acc_corrupt_t_0p6_0p8=0.8348 corrupt_frac_t_0p6_0p8=0.1994 acc_corrupt_t_0p8_1p0=0.9488 corrupt_frac_t_0p8_1p0=0.1943 out_w_norm=141.0385 out_g_norm=0.1641 loss_all=1.6743 init_gold_top10=0.4287 init_gold_top100=0.4320 +step=11000 micro_steps=44000 elapsed=126.4s lr=3.000000e-04 loss=2.2256 loss_recon=2.2256 loss_meanflow=0.0000 mean_model_t=0.5014 mean_corrupt_t=0.5014 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8013 corrupt_frac=0.5489 acc_corrupt=0.6463 loss_corrupt=2.2256 wrong_frac=0.4983 init_acc_corrupt=0.4668 acc_corrupt_t_0p0_0p2=0.2735 corrupt_frac_t_0p0_0p2=0.2022 acc_corrupt_t_0p2_0p4=0.4833 corrupt_frac_t_0p2_0p4=0.2007 acc_corrupt_t_0p4_0p6=0.6895 corrupt_frac_t_0p4_0p6=0.1937 acc_corrupt_t_0p6_0p8=0.8350 corrupt_frac_t_0p6_0p8=0.1978 acc_corrupt_t_0p8_1p0=0.9500 corrupt_frac_t_0p8_1p0=0.2056 out_w_norm=141.7777 out_g_norm=0.1621 loss_all=1.0968 init_gold_top10=0.5535 init_gold_top100=0.5596 +step=11100 micro_steps=44400 elapsed=201.1s lr=3.000000e-04 loss=2.2201 loss_recon=2.2201 loss_meanflow=0.0000 mean_model_t=0.5005 mean_corrupt_t=0.5005 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8024 corrupt_frac=0.5462 acc_corrupt=0.6467 loss_corrupt=2.2201 wrong_frac=0.4999 init_acc_corrupt=0.4657 acc_corrupt_t_0p0_0p2=0.2748 corrupt_frac_t_0p0_0p2=0.1988 acc_corrupt_t_0p2_0p4=0.4845 corrupt_frac_t_0p2_0p4=0.2010 acc_corrupt_t_0p4_0p6=0.6862 corrupt_frac_t_0p4_0p6=0.1997 acc_corrupt_t_0p6_0p8=0.8340 corrupt_frac_t_0p6_0p8=0.2014 acc_corrupt_t_0p8_1p0=0.9490 corrupt_frac_t_0p8_1p0=0.2010 out_w_norm=142.5122 out_g_norm=0.1613 loss_all=0.9752 init_gold_top10=0.5595 init_gold_top100=0.5638 +step=11200 micro_steps=44800 elapsed=116.0s lr=3.000000e-04 loss=2.2056 loss_recon=2.2056 loss_meanflow=0.0000 mean_model_t=0.5013 mean_corrupt_t=0.5013 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8022 corrupt_frac=0.5502 acc_corrupt=0.6488 loss_corrupt=2.2056 wrong_frac=0.4970 init_acc_corrupt=0.4685 acc_corrupt_t_0p0_0p2=0.2789 corrupt_frac_t_0p0_0p2=0.1970 acc_corrupt_t_0p2_0p4=0.4808 corrupt_frac_t_0p2_0p4=0.2005 acc_corrupt_t_0p4_0p6=0.6888 corrupt_frac_t_0p4_0p6=0.1985 acc_corrupt_t_0p6_0p8=0.8325 corrupt_frac_t_0p6_0p8=0.2002 acc_corrupt_t_0p8_1p0=0.9509 corrupt_frac_t_0p8_1p0=0.2047 out_w_norm=143.2521 out_g_norm=0.1602 loss_all=1.0516 init_gold_top10=0.5294 init_gold_top100=0.5347 +step=11300 micro_steps=45200 elapsed=121.0s lr=3.000000e-04 loss=2.2170 loss_recon=2.2170 loss_meanflow=0.0000 mean_model_t=0.5003 mean_corrupt_t=0.5003 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8003 corrupt_frac=0.5536 acc_corrupt=0.6478 loss_corrupt=2.2170 wrong_frac=0.4974 init_acc_corrupt=0.4683 acc_corrupt_t_0p0_0p2=0.2706 corrupt_frac_t_0p0_0p2=0.2003 acc_corrupt_t_0p2_0p4=0.4828 corrupt_frac_t_0p2_0p4=0.1971 acc_corrupt_t_0p4_0p6=0.6900 corrupt_frac_t_0p4_0p6=0.1960 acc_corrupt_t_0p6_0p8=0.8347 corrupt_frac_t_0p6_0p8=0.2065 acc_corrupt_t_0p8_1p0=0.9509 corrupt_frac_t_0p8_1p0=0.2016 out_w_norm=143.9891 out_g_norm=0.1582 loss_all=1.5473 init_gold_top10=0.4868 init_gold_top100=0.4903 +step=11400 micro_steps=45600 elapsed=121.1s lr=3.000000e-04 loss=2.2343 loss_recon=2.2343 loss_meanflow=0.0000 mean_model_t=0.5003 mean_corrupt_t=0.5003 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7981 corrupt_frac=0.5550 acc_corrupt=0.6449 loss_corrupt=2.2343 wrong_frac=0.5000 init_acc_corrupt=0.4656 acc_corrupt_t_0p0_0p2=0.2702 corrupt_frac_t_0p0_0p2=0.2046 acc_corrupt_t_0p2_0p4=0.4811 corrupt_frac_t_0p2_0p4=0.1977 acc_corrupt_t_0p4_0p6=0.6880 corrupt_frac_t_0p4_0p6=0.1960 acc_corrupt_t_0p6_0p8=0.8366 corrupt_frac_t_0p6_0p8=0.1993 acc_corrupt_t_0p8_1p0=0.9485 corrupt_frac_t_0p8_1p0=0.2040 out_w_norm=144.7216 out_g_norm=0.1580 loss_all=1.2742 init_gold_top10=0.5059 init_gold_top100=0.5125 +step=11500 micro_steps=46000 elapsed=121.0s lr=3.000000e-04 loss=2.2331 loss_recon=2.2331 loss_meanflow=0.0000 mean_model_t=0.4983 mean_corrupt_t=0.4983 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8006 corrupt_frac=0.5477 acc_corrupt=0.6445 loss_corrupt=2.2331 wrong_frac=0.5026 init_acc_corrupt=0.4623 acc_corrupt_t_0p0_0p2=0.2704 corrupt_frac_t_0p0_0p2=0.2039 acc_corrupt_t_0p2_0p4=0.4880 corrupt_frac_t_0p2_0p4=0.2010 acc_corrupt_t_0p4_0p6=0.6911 corrupt_frac_t_0p4_0p6=0.1956 acc_corrupt_t_0p6_0p8=0.8358 corrupt_frac_t_0p6_0p8=0.2091 acc_corrupt_t_0p8_1p0=0.9511 corrupt_frac_t_0p8_1p0=0.1913 out_w_norm=145.4514 out_g_norm=0.1583 loss_all=1.3271 init_gold_top10=0.4642 init_gold_top100=0.4710 +step=11600 micro_steps=46400 elapsed=103.3s lr=3.000000e-04 loss=2.1934 loss_recon=2.1934 loss_meanflow=0.0000 mean_model_t=0.5037 mean_corrupt_t=0.5037 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8026 corrupt_frac=0.5528 acc_corrupt=0.6510 loss_corrupt=2.1934 wrong_frac=0.4949 init_acc_corrupt=0.4716 acc_corrupt_t_0p0_0p2=0.2736 corrupt_frac_t_0p0_0p2=0.1985 acc_corrupt_t_0p2_0p4=0.4890 corrupt_frac_t_0p2_0p4=0.1962 acc_corrupt_t_0p4_0p6=0.6906 corrupt_frac_t_0p4_0p6=0.1960 acc_corrupt_t_0p6_0p8=0.8350 corrupt_frac_t_0p6_0p8=0.2028 acc_corrupt_t_0p8_1p0=0.9487 corrupt_frac_t_0p8_1p0=0.2069 out_w_norm=146.1795 out_g_norm=0.1560 loss_all=1.2867 init_gold_top10=0.4964 init_gold_top100=0.5009 +step=11700 micro_steps=46800 elapsed=121.2s lr=3.000000e-04 loss=2.1951 loss_recon=2.1951 loss_meanflow=0.0000 mean_model_t=0.5029 mean_corrupt_t=0.5029 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8043 corrupt_frac=0.5467 acc_corrupt=0.6499 loss_corrupt=2.1951 wrong_frac=0.4976 init_acc_corrupt=0.4683 acc_corrupt_t_0p0_0p2=0.2768 corrupt_frac_t_0p0_0p2=0.1945 acc_corrupt_t_0p2_0p4=0.4862 corrupt_frac_t_0p2_0p4=0.1990 acc_corrupt_t_0p4_0p6=0.6892 corrupt_frac_t_0p4_0p6=0.2042 acc_corrupt_t_0p6_0p8=0.8352 corrupt_frac_t_0p6_0p8=0.2034 acc_corrupt_t_0p8_1p0=0.9485 corrupt_frac_t_0p8_1p0=0.1989 out_w_norm=146.9008 out_g_norm=0.1569 loss_all=1.0359 init_gold_top10=0.5389 init_gold_top100=0.5423 +step=11800 micro_steps=47200 elapsed=120.8s lr=3.000000e-04 loss=2.1841 loss_recon=2.1841 loss_meanflow=0.0000 mean_model_t=0.5015 mean_corrupt_t=0.5015 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8039 corrupt_frac=0.5488 acc_corrupt=0.6511 loss_corrupt=2.1841 wrong_frac=0.4963 init_acc_corrupt=0.4701 acc_corrupt_t_0p0_0p2=0.2785 corrupt_frac_t_0p0_0p2=0.1993 acc_corrupt_t_0p2_0p4=0.4893 corrupt_frac_t_0p2_0p4=0.1975 acc_corrupt_t_0p4_0p6=0.6919 corrupt_frac_t_0p4_0p6=0.1948 acc_corrupt_t_0p6_0p8=0.8335 corrupt_frac_t_0p6_0p8=0.2028 acc_corrupt_t_0p8_1p0=0.9492 corrupt_frac_t_0p8_1p0=0.2056 out_w_norm=147.6282 out_g_norm=0.1554 loss_all=1.1672 init_gold_top10=0.5168 init_gold_top100=0.5225 +step=11900 micro_steps=47600 elapsed=122.0s lr=3.000000e-04 loss=2.2245 loss_recon=2.2245 loss_meanflow=0.0000 mean_model_t=0.4974 mean_corrupt_t=0.4974 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8012 corrupt_frac=0.5485 acc_corrupt=0.6459 loss_corrupt=2.2245 wrong_frac=0.5025 init_acc_corrupt=0.4631 acc_corrupt_t_0p0_0p2=0.2759 corrupt_frac_t_0p0_0p2=0.1994 acc_corrupt_t_0p2_0p4=0.4845 corrupt_frac_t_0p2_0p4=0.2017 acc_corrupt_t_0p4_0p6=0.6903 corrupt_frac_t_0p4_0p6=0.2034 acc_corrupt_t_0p6_0p8=0.8349 corrupt_frac_t_0p6_0p8=0.2049 acc_corrupt_t_0p8_1p0=0.9497 corrupt_frac_t_0p8_1p0=0.1929 out_w_norm=148.3479 out_g_norm=0.1558 loss_all=1.1436 init_gold_top10=0.4994 init_gold_top100=0.5058 +step=12000 micro_steps=48000 elapsed=75.7s lr=3.000000e-04 loss=2.2462 loss_recon=2.2462 loss_meanflow=0.0000 mean_model_t=0.4980 mean_corrupt_t=0.4980 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7979 corrupt_frac=0.5520 acc_corrupt=0.6425 loss_corrupt=2.2462 wrong_frac=0.5043 init_acc_corrupt=0.4601 acc_corrupt_t_0p0_0p2=0.2717 corrupt_frac_t_0p0_0p2=0.2069 acc_corrupt_t_0p2_0p4=0.4861 corrupt_frac_t_0p2_0p4=0.2038 acc_corrupt_t_0p4_0p6=0.6928 corrupt_frac_t_0p4_0p6=0.1950 acc_corrupt_t_0p6_0p8=0.8351 corrupt_frac_t_0p6_0p8=0.1956 acc_corrupt_t_0p8_1p0=0.9496 corrupt_frac_t_0p8_1p0=0.1988 out_w_norm=149.0715 out_g_norm=0.1536 loss_all=0.7326 init_gold_top10=0.6241 init_gold_top100=0.6273 +step=12100 micro_steps=48400 elapsed=120.3s lr=3.000000e-04 loss=2.2344 loss_recon=2.2344 loss_meanflow=0.0000 mean_model_t=0.4983 mean_corrupt_t=0.4983 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8000 corrupt_frac=0.5496 acc_corrupt=0.6445 loss_corrupt=2.2344 wrong_frac=0.5029 init_acc_corrupt=0.4627 acc_corrupt_t_0p0_0p2=0.2729 corrupt_frac_t_0p0_0p2=0.2023 acc_corrupt_t_0p2_0p4=0.4838 corrupt_frac_t_0p2_0p4=0.1999 acc_corrupt_t_0p4_0p6=0.6883 corrupt_frac_t_0p4_0p6=0.2033 acc_corrupt_t_0p6_0p8=0.8345 corrupt_frac_t_0p6_0p8=0.1981 acc_corrupt_t_0p8_1p0=0.9510 corrupt_frac_t_0p8_1p0=0.1978 out_w_norm=149.7932 out_g_norm=0.1523 loss_all=1.5042 init_gold_top10=0.3871 init_gold_top100=0.3973 +step=12200 micro_steps=48800 elapsed=83.0s lr=3.000000e-04 loss=2.1988 loss_recon=2.1988 loss_meanflow=0.0000 mean_model_t=0.5025 mean_corrupt_t=0.5025 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8035 corrupt_frac=0.5469 acc_corrupt=0.6492 loss_corrupt=2.1988 wrong_frac=0.4986 init_acc_corrupt=0.4675 acc_corrupt_t_0p0_0p2=0.2771 corrupt_frac_t_0p0_0p2=0.2010 acc_corrupt_t_0p2_0p4=0.4890 corrupt_frac_t_0p2_0p4=0.1969 acc_corrupt_t_0p4_0p6=0.6923 corrupt_frac_t_0p4_0p6=0.2039 acc_corrupt_t_0p6_0p8=0.8351 corrupt_frac_t_0p6_0p8=0.1971 acc_corrupt_t_0p8_1p0=0.9507 corrupt_frac_t_0p8_1p0=0.2022 out_w_norm=150.5110 out_g_norm=0.1530 loss_all=1.5389 init_gold_top10=0.4149 init_gold_top100=0.4265 +step=12300 micro_steps=49200 elapsed=55.6s lr=3.000000e-04 loss=2.2441 loss_recon=2.2441 loss_meanflow=0.0000 mean_model_t=0.4965 mean_corrupt_t=0.4965 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7989 corrupt_frac=0.5493 acc_corrupt=0.6423 loss_corrupt=2.2441 wrong_frac=0.5061 init_acc_corrupt=0.4584 acc_corrupt_t_0p0_0p2=0.2756 corrupt_frac_t_0p0_0p2=0.2014 acc_corrupt_t_0p2_0p4=0.4788 corrupt_frac_t_0p2_0p4=0.2069 acc_corrupt_t_0p4_0p6=0.6912 corrupt_frac_t_0p4_0p6=0.2022 acc_corrupt_t_0p6_0p8=0.8371 corrupt_frac_t_0p6_0p8=0.1958 acc_corrupt_t_0p8_1p0=0.9503 corrupt_frac_t_0p8_1p0=0.1941 out_w_norm=151.2346 out_g_norm=0.1522 loss_all=1.0838 init_gold_top10=0.5257 init_gold_top100=0.5305 +step=12400 micro_steps=49600 elapsed=55.8s lr=3.000000e-04 loss=2.2394 loss_recon=2.2394 loss_meanflow=0.0000 mean_model_t=0.4971 mean_corrupt_t=0.4971 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.7982 corrupt_frac=0.5530 acc_corrupt=0.6435 loss_corrupt=2.2394 wrong_frac=0.5035 init_acc_corrupt=0.4616 acc_corrupt_t_0p0_0p2=0.2718 corrupt_frac_t_0p0_0p2=0.2079 acc_corrupt_t_0p2_0p4=0.4840 corrupt_frac_t_0p2_0p4=0.1986 acc_corrupt_t_0p4_0p6=0.6949 corrupt_frac_t_0p4_0p6=0.2003 acc_corrupt_t_0p6_0p8=0.8352 corrupt_frac_t_0p6_0p8=0.1952 acc_corrupt_t_0p8_1p0=0.9501 corrupt_frac_t_0p8_1p0=0.1994 out_w_norm=151.9504 out_g_norm=0.1509 loss_all=1.2963 init_gold_top10=0.4940 init_gold_top100=0.4995 +step=12500 micro_steps=50000 elapsed=55.6s lr=3.000000e-04 loss=2.2052 loss_recon=2.2052 loss_meanflow=0.0000 mean_model_t=0.4993 mean_corrupt_t=0.4993 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8034 corrupt_frac=0.5464 acc_corrupt=0.6483 loss_corrupt=2.2052 wrong_frac=0.4998 init_acc_corrupt=0.4655 acc_corrupt_t_0p0_0p2=0.2775 corrupt_frac_t_0p0_0p2=0.1996 acc_corrupt_t_0p2_0p4=0.4855 corrupt_frac_t_0p2_0p4=0.2050 acc_corrupt_t_0p4_0p6=0.6949 corrupt_frac_t_0p4_0p6=0.1982 acc_corrupt_t_0p6_0p8=0.8362 corrupt_frac_t_0p6_0p8=0.1962 acc_corrupt_t_0p8_1p0=0.9491 corrupt_frac_t_0p8_1p0=0.2036 out_w_norm=152.6657 out_g_norm=0.1505 loss_all=1.4789 init_gold_top10=0.4659 init_gold_top100=0.4709 +step=12600 micro_steps=50400 elapsed=55.5s lr=3.000000e-04 loss=2.2077 loss_recon=2.2077 loss_meanflow=0.0000 mean_model_t=0.5017 mean_corrupt_t=0.5017 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8030 corrupt_frac=0.5474 acc_corrupt=0.6481 loss_corrupt=2.2077 wrong_frac=0.5001 init_acc_corrupt=0.4646 acc_corrupt_t_0p0_0p2=0.2759 corrupt_frac_t_0p0_0p2=0.2010 acc_corrupt_t_0p2_0p4=0.4869 corrupt_frac_t_0p2_0p4=0.2030 acc_corrupt_t_0p4_0p6=0.6918 corrupt_frac_t_0p4_0p6=0.1965 acc_corrupt_t_0p6_0p8=0.8365 corrupt_frac_t_0p6_0p8=0.1963 acc_corrupt_t_0p8_1p0=0.9514 corrupt_frac_t_0p8_1p0=0.2042 out_w_norm=153.3756 out_g_norm=0.1502 loss_all=0.9648 init_gold_top10=0.5104 init_gold_top100=0.5151 +step=12700 micro_steps=50800 elapsed=55.4s lr=3.000000e-04 loss=2.1883 loss_recon=2.1883 loss_meanflow=0.0000 mean_model_t=0.5004 mean_corrupt_t=0.5004 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8025 corrupt_frac=0.5518 acc_corrupt=0.6502 loss_corrupt=2.1883 wrong_frac=0.4988 init_acc_corrupt=0.4669 acc_corrupt_t_0p0_0p2=0.2790 corrupt_frac_t_0p0_0p2=0.2027 acc_corrupt_t_0p2_0p4=0.4914 corrupt_frac_t_0p2_0p4=0.1913 acc_corrupt_t_0p4_0p6=0.6914 corrupt_frac_t_0p4_0p6=0.2075 acc_corrupt_t_0p6_0p8=0.8350 corrupt_frac_t_0p6_0p8=0.1989 acc_corrupt_t_0p8_1p0=0.9516 corrupt_frac_t_0p8_1p0=0.2002 out_w_norm=154.0860 out_g_norm=0.1501 loss_all=1.4977 init_gold_top10=0.4189 init_gold_top100=0.4250 +step=12800 micro_steps=51200 elapsed=55.4s lr=3.000000e-04 loss=2.1988 loss_recon=2.1988 loss_meanflow=0.0000 mean_model_t=0.5024 mean_corrupt_t=0.5024 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8025 corrupt_frac=0.5512 acc_corrupt=0.6496 loss_corrupt=2.1988 wrong_frac=0.4987 init_acc_corrupt=0.4665 acc_corrupt_t_0p0_0p2=0.2742 corrupt_frac_t_0p0_0p2=0.1992 acc_corrupt_t_0p2_0p4=0.4856 corrupt_frac_t_0p2_0p4=0.2007 acc_corrupt_t_0p4_0p6=0.6935 corrupt_frac_t_0p4_0p6=0.2001 acc_corrupt_t_0p6_0p8=0.8357 corrupt_frac_t_0p6_0p8=0.1983 acc_corrupt_t_0p8_1p0=0.9516 corrupt_frac_t_0p8_1p0=0.2036 out_w_norm=154.7869 out_g_norm=0.1476 loss_all=1.3672 init_gold_top10=0.4773 init_gold_top100=0.4813 +step=12900 micro_steps=51600 elapsed=55.4s lr=3.000000e-04 loss=2.2097 loss_recon=2.2097 loss_meanflow=0.0000 mean_model_t=0.5013 mean_corrupt_t=0.5013 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8011 corrupt_frac=0.5530 acc_corrupt=0.6479 loss_corrupt=2.2097 wrong_frac=0.5005 init_acc_corrupt=0.4650 acc_corrupt_t_0p0_0p2=0.2748 corrupt_frac_t_0p0_0p2=0.1990 acc_corrupt_t_0p2_0p4=0.4853 corrupt_frac_t_0p2_0p4=0.2013 acc_corrupt_t_0p4_0p6=0.6882 corrupt_frac_t_0p4_0p6=0.1995 acc_corrupt_t_0p6_0p8=0.8367 corrupt_frac_t_0p6_0p8=0.1983 acc_corrupt_t_0p8_1p0=0.9502 corrupt_frac_t_0p8_1p0=0.2029 out_w_norm=155.4930 out_g_norm=0.1473 loss_all=1.1169 init_gold_top10=0.5174 init_gold_top100=0.5223 +step=13000 micro_steps=52000 elapsed=55.5s lr=3.000000e-04 loss=2.2149 loss_recon=2.2149 loss_meanflow=0.0000 mean_model_t=0.4973 mean_corrupt_t=0.4973 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8005 corrupt_frac=0.5514 acc_corrupt=0.6462 loss_corrupt=2.2149 wrong_frac=0.5041 init_acc_corrupt=0.4610 acc_corrupt_t_0p0_0p2=0.2780 corrupt_frac_t_0p0_0p2=0.2020 acc_corrupt_t_0p2_0p4=0.4888 corrupt_frac_t_0p2_0p4=0.2033 acc_corrupt_t_0p4_0p6=0.6953 corrupt_frac_t_0p4_0p6=0.2048 acc_corrupt_t_0p6_0p8=0.8359 corrupt_frac_t_0p6_0p8=0.1980 acc_corrupt_t_0p8_1p0=0.9508 corrupt_frac_t_0p8_1p0=0.1925 out_w_norm=156.1953 out_g_norm=0.1480 loss_all=1.7943 init_gold_top10=0.4157 init_gold_top100=0.4269 +step=13100 micro_steps=52400 elapsed=106.1s lr=3.000000e-04 loss=2.2123 loss_recon=2.2123 loss_meanflow=0.0000 mean_model_t=0.4974 mean_corrupt_t=0.4974 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8006 corrupt_frac=0.5526 acc_corrupt=0.6472 loss_corrupt=2.2123 wrong_frac=0.5019 init_acc_corrupt=0.4630 acc_corrupt_t_0p0_0p2=0.2789 corrupt_frac_t_0p0_0p2=0.2035 acc_corrupt_t_0p2_0p4=0.4869 corrupt_frac_t_0p2_0p4=0.2032 acc_corrupt_t_0p4_0p6=0.6923 corrupt_frac_t_0p4_0p6=0.1955 acc_corrupt_t_0p6_0p8=0.8365 corrupt_frac_t_0p6_0p8=0.1964 acc_corrupt_t_0p8_1p0=0.9507 corrupt_frac_t_0p8_1p0=0.2029 out_w_norm=156.8985 out_g_norm=0.1471 loss_all=1.5194 init_gold_top10=0.4165 init_gold_top100=0.4244 +step=13200 micro_steps=52800 elapsed=97.1s lr=3.000000e-04 loss=2.2040 loss_recon=2.2040 loss_meanflow=0.0000 mean_model_t=0.5015 mean_corrupt_t=0.5015 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8020 corrupt_frac=0.5514 acc_corrupt=0.6487 loss_corrupt=2.2040 wrong_frac=0.4996 init_acc_corrupt=0.4656 acc_corrupt_t_0p0_0p2=0.2744 corrupt_frac_t_0p0_0p2=0.1980 acc_corrupt_t_0p2_0p4=0.4860 corrupt_frac_t_0p2_0p4=0.2063 acc_corrupt_t_0p4_0p6=0.6965 corrupt_frac_t_0p4_0p6=0.1988 acc_corrupt_t_0p6_0p8=0.8363 corrupt_frac_t_0p6_0p8=0.1969 acc_corrupt_t_0p8_1p0=0.9510 corrupt_frac_t_0p8_1p0=0.2016 out_w_norm=157.5932 out_g_norm=0.1455 loss_all=1.2480 init_gold_top10=0.4477 init_gold_top100=0.4536 +step=13300 micro_steps=53200 elapsed=55.7s lr=3.000000e-04 loss=2.1882 loss_recon=2.1882 loss_meanflow=0.0000 mean_model_t=0.5006 mean_corrupt_t=0.5006 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8034 corrupt_frac=0.5505 acc_corrupt=0.6505 loss_corrupt=2.1882 wrong_frac=0.4994 init_acc_corrupt=0.4661 acc_corrupt_t_0p0_0p2=0.2788 corrupt_frac_t_0p0_0p2=0.1991 acc_corrupt_t_0p2_0p4=0.4887 corrupt_frac_t_0p2_0p4=0.2021 acc_corrupt_t_0p4_0p6=0.6949 corrupt_frac_t_0p4_0p6=0.1973 acc_corrupt_t_0p6_0p8=0.8363 corrupt_frac_t_0p6_0p8=0.2011 acc_corrupt_t_0p8_1p0=0.9505 corrupt_frac_t_0p8_1p0=0.2014 out_w_norm=158.2836 out_g_norm=0.1455 loss_all=1.2188 init_gold_top10=0.4194 init_gold_top100=0.4249 +step=13400 micro_steps=53600 elapsed=55.6s lr=3.000000e-04 loss=2.1830 loss_recon=2.1830 loss_meanflow=0.0000 mean_model_t=0.4983 mean_corrupt_t=0.4983 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8042 corrupt_frac=0.5475 acc_corrupt=0.6504 loss_corrupt=2.1830 wrong_frac=0.5009 init_acc_corrupt=0.4647 acc_corrupt_t_0p0_0p2=0.2800 corrupt_frac_t_0p0_0p2=0.1986 acc_corrupt_t_0p2_0p4=0.4958 corrupt_frac_t_0p2_0p4=0.2065 acc_corrupt_t_0p4_0p6=0.6924 corrupt_frac_t_0p4_0p6=0.1984 acc_corrupt_t_0p6_0p8=0.8386 corrupt_frac_t_0p6_0p8=0.1978 acc_corrupt_t_0p8_1p0=0.9511 corrupt_frac_t_0p8_1p0=0.1996 out_w_norm=158.9774 out_g_norm=0.1449 loss_all=1.1326 init_gold_top10=0.5622 init_gold_top100=0.5665 +step=13500 micro_steps=54000 elapsed=55.5s lr=3.000000e-04 loss=2.1871 loss_recon=2.1871 loss_meanflow=0.0000 mean_model_t=0.5024 mean_corrupt_t=0.5024 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8027 corrupt_frac=0.5528 acc_corrupt=0.6510 loss_corrupt=2.1871 wrong_frac=0.4977 init_acc_corrupt=0.4683 acc_corrupt_t_0p0_0p2=0.2727 corrupt_frac_t_0p0_0p2=0.2006 acc_corrupt_t_0p2_0p4=0.4882 corrupt_frac_t_0p2_0p4=0.1925 acc_corrupt_t_0p4_0p6=0.6929 corrupt_frac_t_0p4_0p6=0.1983 acc_corrupt_t_0p6_0p8=0.8390 corrupt_frac_t_0p6_0p8=0.2114 acc_corrupt_t_0p8_1p0=0.9509 corrupt_frac_t_0p8_1p0=0.1973 out_w_norm=159.6683 out_g_norm=0.1433 loss_all=1.5648 init_gold_top10=0.4267 init_gold_top100=0.4326 +step=13600 micro_steps=54400 elapsed=55.6s lr=3.000000e-04 loss=2.1716 loss_recon=2.1716 loss_meanflow=0.0000 mean_model_t=0.5047 mean_corrupt_t=0.5047 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8033 corrupt_frac=0.5564 acc_corrupt=0.6535 loss_corrupt=2.1716 wrong_frac=0.4948 init_acc_corrupt=0.4717 acc_corrupt_t_0p0_0p2=0.2716 corrupt_frac_t_0p0_0p2=0.1931 acc_corrupt_t_0p2_0p4=0.4891 corrupt_frac_t_0p2_0p4=0.1989 acc_corrupt_t_0p4_0p6=0.6931 corrupt_frac_t_0p4_0p6=0.1983 acc_corrupt_t_0p6_0p8=0.8371 corrupt_frac_t_0p6_0p8=0.2053 acc_corrupt_t_0p8_1p0=0.9513 corrupt_frac_t_0p8_1p0=0.2050 out_w_norm=160.3536 out_g_norm=0.1423 loss_all=1.1137 init_gold_top10=0.5429 init_gold_top100=0.5500 +step=13700 micro_steps=54800 elapsed=55.6s lr=3.000000e-04 loss=2.1942 loss_recon=2.1942 loss_meanflow=0.0000 mean_model_t=0.5002 mean_corrupt_t=0.5002 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8039 corrupt_frac=0.5482 acc_corrupt=0.6497 loss_corrupt=2.1942 wrong_frac=0.5006 init_acc_corrupt=0.4646 acc_corrupt_t_0p0_0p2=0.2786 corrupt_frac_t_0p0_0p2=0.2010 acc_corrupt_t_0p2_0p4=0.4873 corrupt_frac_t_0p2_0p4=0.1986 acc_corrupt_t_0p4_0p6=0.6939 corrupt_frac_t_0p4_0p6=0.2027 acc_corrupt_t_0p6_0p8=0.8394 corrupt_frac_t_0p6_0p8=0.1976 acc_corrupt_t_0p8_1p0=0.9513 corrupt_frac_t_0p8_1p0=0.2006 out_w_norm=161.0337 out_g_norm=0.1437 loss_all=0.6225 init_gold_top10=0.6512 init_gold_top100=0.6526 +step=13800 micro_steps=55200 elapsed=55.6s lr=3.000000e-04 loss=2.2120 loss_recon=2.2120 loss_meanflow=0.0000 mean_model_t=0.4944 mean_corrupt_t=0.4944 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8011 corrupt_frac=0.5506 acc_corrupt=0.6465 loss_corrupt=2.2120 wrong_frac=0.5047 init_acc_corrupt=0.4601 acc_corrupt_t_0p0_0p2=0.2801 corrupt_frac_t_0p0_0p2=0.2040 acc_corrupt_t_0p2_0p4=0.4876 corrupt_frac_t_0p2_0p4=0.2032 acc_corrupt_t_0p4_0p6=0.6938 corrupt_frac_t_0p4_0p6=0.1967 acc_corrupt_t_0p6_0p8=0.8374 corrupt_frac_t_0p6_0p8=0.2022 acc_corrupt_t_0p8_1p0=0.9513 corrupt_frac_t_0p8_1p0=0.1948 out_w_norm=161.7170 out_g_norm=0.1433 loss_all=1.0916 init_gold_top10=0.5125 init_gold_top100=0.5175 +step=13900 micro_steps=55600 elapsed=55.5s lr=3.000000e-04 loss=2.1948 loss_recon=2.1948 loss_meanflow=0.0000 mean_model_t=0.5008 mean_corrupt_t=0.5008 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8031 corrupt_frac=0.5487 acc_corrupt=0.6489 loss_corrupt=2.1948 wrong_frac=0.5018 init_acc_corrupt=0.4635 acc_corrupt_t_0p0_0p2=0.2757 corrupt_frac_t_0p0_0p2=0.2019 acc_corrupt_t_0p2_0p4=0.4900 corrupt_frac_t_0p2_0p4=0.2014 acc_corrupt_t_0p4_0p6=0.6957 corrupt_frac_t_0p4_0p6=0.2003 acc_corrupt_t_0p6_0p8=0.8379 corrupt_frac_t_0p6_0p8=0.2013 acc_corrupt_t_0p8_1p0=0.9502 corrupt_frac_t_0p8_1p0=0.1982 out_w_norm=162.3952 out_g_norm=0.1422 loss_all=1.1256 init_gold_top10=0.4697 init_gold_top100=0.4763 +step=14000 micro_steps=56000 elapsed=55.6s lr=3.000000e-04 loss=2.1728 loss_recon=2.1728 loss_meanflow=0.0000 mean_model_t=0.5014 mean_corrupt_t=0.5014 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8051 corrupt_frac=0.5495 acc_corrupt=0.6527 loss_corrupt=2.1728 wrong_frac=0.4982 init_acc_corrupt=0.4678 acc_corrupt_t_0p0_0p2=0.2769 corrupt_frac_t_0p0_0p2=0.1970 acc_corrupt_t_0p2_0p4=0.4939 corrupt_frac_t_0p2_0p4=0.2027 acc_corrupt_t_0p4_0p6=0.6955 corrupt_frac_t_0p4_0p6=0.1958 acc_corrupt_t_0p6_0p8=0.8385 corrupt_frac_t_0p6_0p8=0.2079 acc_corrupt_t_0p8_1p0=0.9513 corrupt_frac_t_0p8_1p0=0.1982 out_w_norm=163.0687 out_g_norm=0.1412 loss_all=1.7021 init_gold_top10=0.4614 init_gold_top100=0.4656 +step=14100 micro_steps=56400 elapsed=91.9s lr=3.000000e-04 loss=2.1830 loss_recon=2.1830 loss_meanflow=0.0000 mean_model_t=0.5023 mean_corrupt_t=0.5023 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8023 corrupt_frac=0.5555 acc_corrupt=0.6513 loss_corrupt=2.1830 wrong_frac=0.4975 init_acc_corrupt=0.4683 acc_corrupt_t_0p0_0p2=0.2771 corrupt_frac_t_0p0_0p2=0.1999 acc_corrupt_t_0p2_0p4=0.4824 corrupt_frac_t_0p2_0p4=0.1992 acc_corrupt_t_0p4_0p6=0.6968 corrupt_frac_t_0p4_0p6=0.1966 acc_corrupt_t_0p6_0p8=0.8396 corrupt_frac_t_0p6_0p8=0.2033 acc_corrupt_t_0p8_1p0=0.9509 corrupt_frac_t_0p8_1p0=0.2031 out_w_norm=163.7411 out_g_norm=0.1402 loss_all=1.1132 init_gold_top10=0.5489 init_gold_top100=0.5509 +step=14200 micro_steps=56800 elapsed=110.9s lr=3.000000e-04 loss=2.1952 loss_recon=2.1952 loss_meanflow=0.0000 mean_model_t=0.5004 mean_corrupt_t=0.5004 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8013 corrupt_frac=0.5554 acc_corrupt=0.6495 loss_corrupt=2.1952 wrong_frac=0.5007 init_acc_corrupt=0.4650 acc_corrupt_t_0p0_0p2=0.2786 corrupt_frac_t_0p0_0p2=0.1992 acc_corrupt_t_0p2_0p4=0.4851 corrupt_frac_t_0p2_0p4=0.1989 acc_corrupt_t_0p4_0p6=0.6963 corrupt_frac_t_0p4_0p6=0.2105 acc_corrupt_t_0p6_0p8=0.8384 corrupt_frac_t_0p6_0p8=0.1905 acc_corrupt_t_0p8_1p0=0.9503 corrupt_frac_t_0p8_1p0=0.2018 out_w_norm=164.4192 out_g_norm=0.1399 loss_all=1.6081 init_gold_top10=0.4237 init_gold_top100=0.4327 +step=14300 micro_steps=57200 elapsed=55.8s lr=3.000000e-04 loss=2.1855 loss_recon=2.1855 loss_meanflow=0.0000 mean_model_t=0.5002 mean_corrupt_t=0.5002 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8033 corrupt_frac=0.5512 acc_corrupt=0.6507 loss_corrupt=2.1855 wrong_frac=0.5001 init_acc_corrupt=0.4657 acc_corrupt_t_0p0_0p2=0.2787 corrupt_frac_t_0p0_0p2=0.2001 acc_corrupt_t_0p2_0p4=0.4863 corrupt_frac_t_0p2_0p4=0.1965 acc_corrupt_t_0p4_0p6=0.6959 corrupt_frac_t_0p4_0p6=0.2013 acc_corrupt_t_0p6_0p8=0.8383 corrupt_frac_t_0p6_0p8=0.2076 acc_corrupt_t_0p8_1p0=0.9507 corrupt_frac_t_0p8_1p0=0.1950 out_w_norm=165.0894 out_g_norm=0.1393 loss_all=0.9221 init_gold_top10=0.5172 init_gold_top100=0.5216 +step=14400 micro_steps=57600 elapsed=56.5s lr=3.000000e-04 loss=2.1985 loss_recon=2.1985 loss_meanflow=0.0000 mean_model_t=0.4992 mean_corrupt_t=0.4992 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8018 corrupt_frac=0.5525 acc_corrupt=0.6489 loss_corrupt=2.1985 wrong_frac=0.5018 init_acc_corrupt=0.4629 acc_corrupt_t_0p0_0p2=0.2814 corrupt_frac_t_0p0_0p2=0.2008 acc_corrupt_t_0p2_0p4=0.4886 corrupt_frac_t_0p2_0p4=0.2054 acc_corrupt_t_0p4_0p6=0.6913 corrupt_frac_t_0p4_0p6=0.1958 acc_corrupt_t_0p6_0p8=0.8391 corrupt_frac_t_0p6_0p8=0.2024 acc_corrupt_t_0p8_1p0=0.9525 corrupt_frac_t_0p8_1p0=0.1965 out_w_norm=165.7642 out_g_norm=0.1396 loss_all=1.6168 init_gold_top10=0.4660 init_gold_top100=0.4727 +step=14500 micro_steps=58000 elapsed=55.6s lr=3.000000e-04 loss=2.1655 loss_recon=2.1655 loss_meanflow=0.0000 mean_model_t=0.5024 mean_corrupt_t=0.5024 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8046 corrupt_frac=0.5520 acc_corrupt=0.6536 loss_corrupt=2.1655 wrong_frac=0.4972 init_acc_corrupt=0.4683 acc_corrupt_t_0p0_0p2=0.2787 corrupt_frac_t_0p0_0p2=0.1988 acc_corrupt_t_0p2_0p4=0.4893 corrupt_frac_t_0p2_0p4=0.2019 acc_corrupt_t_0p4_0p6=0.6960 corrupt_frac_t_0p4_0p6=0.1897 acc_corrupt_t_0p6_0p8=0.8416 corrupt_frac_t_0p6_0p8=0.2098 acc_corrupt_t_0p8_1p0=0.9517 corrupt_frac_t_0p8_1p0=0.2008 out_w_norm=166.4376 out_g_norm=0.1384 loss_all=1.2868 init_gold_top10=0.4977 init_gold_top100=0.5042 +step=14600 micro_steps=58400 elapsed=55.6s lr=3.000000e-04 loss=2.2071 loss_recon=2.2071 loss_meanflow=0.0000 mean_model_t=0.4967 mean_corrupt_t=0.4967 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8029 corrupt_frac=0.5476 acc_corrupt=0.6474 loss_corrupt=2.2071 wrong_frac=0.5043 init_acc_corrupt=0.4608 acc_corrupt_t_0p0_0p2=0.2790 corrupt_frac_t_0p0_0p2=0.2040 acc_corrupt_t_0p2_0p4=0.4889 corrupt_frac_t_0p2_0p4=0.2014 acc_corrupt_t_0p4_0p6=0.6947 corrupt_frac_t_0p4_0p6=0.2029 acc_corrupt_t_0p6_0p8=0.8392 corrupt_frac_t_0p6_0p8=0.1990 acc_corrupt_t_0p8_1p0=0.9521 corrupt_frac_t_0p8_1p0=0.1938 out_w_norm=167.0998 out_g_norm=0.1384 loss_all=1.1423 init_gold_top10=0.4943 init_gold_top100=0.4968 +step=14700 micro_steps=58800 elapsed=55.6s lr=3.000000e-04 loss=2.1843 loss_recon=2.1843 loss_meanflow=0.0000 mean_model_t=0.5006 mean_corrupt_t=0.5006 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8029 corrupt_frac=0.5521 acc_corrupt=0.6504 loss_corrupt=2.1843 wrong_frac=0.5008 init_acc_corrupt=0.4650 acc_corrupt_t_0p0_0p2=0.2757 corrupt_frac_t_0p0_0p2=0.2022 acc_corrupt_t_0p2_0p4=0.4885 corrupt_frac_t_0p2_0p4=0.1968 acc_corrupt_t_0p4_0p6=0.6964 corrupt_frac_t_0p4_0p6=0.2029 acc_corrupt_t_0p6_0p8=0.8416 corrupt_frac_t_0p6_0p8=0.1990 acc_corrupt_t_0p8_1p0=0.9513 corrupt_frac_t_0p8_1p0=0.2001 out_w_norm=167.7636 out_g_norm=0.1367 loss_all=1.0071 init_gold_top10=0.4788 init_gold_top100=0.4872 +step=14800 micro_steps=59200 elapsed=55.6s lr=3.000000e-04 loss=2.1520 loss_recon=2.1520 loss_meanflow=0.0000 mean_model_t=0.5035 mean_corrupt_t=0.5035 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8066 corrupt_frac=0.5501 acc_corrupt=0.6557 loss_corrupt=2.1520 wrong_frac=0.4951 init_acc_corrupt=0.4711 acc_corrupt_t_0p0_0p2=0.2803 corrupt_frac_t_0p0_0p2=0.1929 acc_corrupt_t_0p2_0p4=0.4875 corrupt_frac_t_0p2_0p4=0.2000 acc_corrupt_t_0p4_0p6=0.6969 corrupt_frac_t_0p4_0p6=0.2035 acc_corrupt_t_0p6_0p8=0.8385 corrupt_frac_t_0p6_0p8=0.1965 acc_corrupt_t_0p8_1p0=0.9535 corrupt_frac_t_0p8_1p0=0.2075 out_w_norm=168.4231 out_g_norm=0.1356 loss_all=1.1431 init_gold_top10=0.5766 init_gold_top100=0.5813 +step=14900 micro_steps=59600 elapsed=55.9s lr=3.000000e-04 loss=2.1510 loss_recon=2.1510 loss_meanflow=0.0000 mean_model_t=0.5026 mean_corrupt_t=0.5026 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8063 corrupt_frac=0.5511 acc_corrupt=0.6557 loss_corrupt=2.1510 wrong_frac=0.4957 init_acc_corrupt=0.4699 acc_corrupt_t_0p0_0p2=0.2861 corrupt_frac_t_0p0_0p2=0.1990 acc_corrupt_t_0p2_0p4=0.4876 corrupt_frac_t_0p2_0p4=0.1951 acc_corrupt_t_0p4_0p6=0.6982 corrupt_frac_t_0p4_0p6=0.1948 acc_corrupt_t_0p6_0p8=0.8383 corrupt_frac_t_0p6_0p8=0.2089 acc_corrupt_t_0p8_1p0=0.9513 corrupt_frac_t_0p8_1p0=0.2033 out_w_norm=169.0789 out_g_norm=0.1353 loss_all=1.0123 init_gold_top10=0.5301 init_gold_top100=0.5353 +step=15000 micro_steps=60000 elapsed=55.5s lr=3.000000e-04 loss=2.1772 loss_recon=2.1772 loss_meanflow=0.0000 mean_model_t=0.4977 mean_corrupt_t=0.4977 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8038 corrupt_frac=0.5513 acc_corrupt=0.6516 loss_corrupt=2.1772 wrong_frac=0.5008 init_acc_corrupt=0.4642 acc_corrupt_t_0p0_0p2=0.2828 corrupt_frac_t_0p0_0p2=0.1992 acc_corrupt_t_0p2_0p4=0.4893 corrupt_frac_t_0p2_0p4=0.2050 acc_corrupt_t_0p4_0p6=0.6997 corrupt_frac_t_0p4_0p6=0.1958 acc_corrupt_t_0p6_0p8=0.8372 corrupt_frac_t_0p6_0p8=0.2019 acc_corrupt_t_0p8_1p0=0.9524 corrupt_frac_t_0p8_1p0=0.1986 out_w_norm=169.7275 out_g_norm=0.1364 loss_all=1.2788 init_gold_top10=0.4854 init_gold_top100=0.4887 +step=15100 micro_steps=60400 elapsed=84.6s lr=3.000000e-04 loss=2.1545 loss_recon=2.1545 loss_meanflow=0.0000 mean_model_t=0.5029 mean_corrupt_t=0.5029 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8061 corrupt_frac=0.5499 acc_corrupt=0.6544 loss_corrupt=2.1545 wrong_frac=0.4987 init_acc_corrupt=0.4668 acc_corrupt_t_0p0_0p2=0.2855 corrupt_frac_t_0p0_0p2=0.1967 acc_corrupt_t_0p2_0p4=0.4916 corrupt_frac_t_0p2_0p4=0.2010 acc_corrupt_t_0p4_0p6=0.6964 corrupt_frac_t_0p4_0p6=0.2027 acc_corrupt_t_0p6_0p8=0.8403 corrupt_frac_t_0p6_0p8=0.1996 acc_corrupt_t_0p8_1p0=0.9529 corrupt_frac_t_0p8_1p0=0.2000 out_w_norm=170.3808 out_g_norm=0.1352 loss_all=1.2217 init_gold_top10=0.4572 init_gold_top100=0.4630 +step=15200 micro_steps=60800 elapsed=116.3s lr=3.000000e-04 loss=2.1940 loss_recon=2.1940 loss_meanflow=0.0000 mean_model_t=0.4931 mean_corrupt_t=0.4931 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8026 corrupt_frac=0.5494 acc_corrupt=0.6483 loss_corrupt=2.1940 wrong_frac=0.5057 init_acc_corrupt=0.4592 acc_corrupt_t_0p0_0p2=0.2835 corrupt_frac_t_0p0_0p2=0.2062 acc_corrupt_t_0p2_0p4=0.4937 corrupt_frac_t_0p2_0p4=0.2065 acc_corrupt_t_0p4_0p6=0.6980 corrupt_frac_t_0p4_0p6=0.1960 acc_corrupt_t_0p6_0p8=0.8396 corrupt_frac_t_0p6_0p8=0.1986 acc_corrupt_t_0p8_1p0=0.9523 corrupt_frac_t_0p8_1p0=0.1951 out_w_norm=171.0327 out_g_norm=0.1359 loss_all=1.9575 init_gold_top10=0.4621 init_gold_top100=0.4669 +step=15300 micro_steps=61200 elapsed=56.9s lr=3.000000e-04 loss=2.1823 loss_recon=2.1823 loss_meanflow=0.0000 mean_model_t=0.5002 mean_corrupt_t=0.5002 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8041 corrupt_frac=0.5501 acc_corrupt=0.6510 loss_corrupt=2.1823 wrong_frac=0.5002 init_acc_corrupt=0.4645 acc_corrupt_t_0p0_0p2=0.2792 corrupt_frac_t_0p0_0p2=0.2065 acc_corrupt_t_0p2_0p4=0.4926 corrupt_frac_t_0p2_0p4=0.2008 acc_corrupt_t_0p4_0p6=0.7004 corrupt_frac_t_0p4_0p6=0.1948 acc_corrupt_t_0p6_0p8=0.8416 corrupt_frac_t_0p6_0p8=0.1917 acc_corrupt_t_0p8_1p0=0.9520 corrupt_frac_t_0p8_1p0=0.2068 out_w_norm=171.6814 out_g_norm=0.1343 loss_all=1.4501 init_gold_top10=0.4540 init_gold_top100=0.4598 +step=15400 micro_steps=61600 elapsed=55.4s lr=3.000000e-04 loss=2.1537 loss_recon=2.1537 loss_meanflow=0.0000 mean_model_t=0.5024 mean_corrupt_t=0.5024 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8055 corrupt_frac=0.5521 acc_corrupt=0.6548 loss_corrupt=2.1537 wrong_frac=0.4972 init_acc_corrupt=0.4683 acc_corrupt_t_0p0_0p2=0.2828 corrupt_frac_t_0p0_0p2=0.1968 acc_corrupt_t_0p2_0p4=0.4907 corrupt_frac_t_0p2_0p4=0.1986 acc_corrupt_t_0p4_0p6=0.6981 corrupt_frac_t_0p4_0p6=0.2057 acc_corrupt_t_0p6_0p8=0.8408 corrupt_frac_t_0p6_0p8=0.1999 acc_corrupt_t_0p8_1p0=0.9520 corrupt_frac_t_0p8_1p0=0.2001 out_w_norm=172.3251 out_g_norm=0.1343 loss_all=1.0365 init_gold_top10=0.5561 init_gold_top100=0.5583 +step=15500 micro_steps=62000 elapsed=55.4s lr=3.000000e-04 loss=2.1403 loss_recon=2.1403 loss_meanflow=0.0000 mean_model_t=0.5019 mean_corrupt_t=0.5019 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8083 corrupt_frac=0.5458 acc_corrupt=0.6560 loss_corrupt=2.1403 wrong_frac=0.4977 init_acc_corrupt=0.4677 acc_corrupt_t_0p0_0p2=0.2854 corrupt_frac_t_0p0_0p2=0.1978 acc_corrupt_t_0p2_0p4=0.4960 corrupt_frac_t_0p2_0p4=0.2016 acc_corrupt_t_0p4_0p6=0.6965 corrupt_frac_t_0p4_0p6=0.1936 acc_corrupt_t_0p6_0p8=0.8416 corrupt_frac_t_0p6_0p8=0.2062 acc_corrupt_t_0p8_1p0=0.9508 corrupt_frac_t_0p8_1p0=0.2014 out_w_norm=172.9647 out_g_norm=0.1336 loss_all=1.2289 init_gold_top10=0.4852 init_gold_top100=0.4913 +step=15600 micro_steps=62400 elapsed=55.4s lr=3.000000e-04 loss=2.1255 loss_recon=2.1255 loss_meanflow=0.0000 mean_model_t=0.5047 mean_corrupt_t=0.5047 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8093 corrupt_frac=0.5478 acc_corrupt=0.6586 loss_corrupt=2.1255 wrong_frac=0.4950 init_acc_corrupt=0.4711 acc_corrupt_t_0p0_0p2=0.2841 corrupt_frac_t_0p0_0p2=0.1921 acc_corrupt_t_0p2_0p4=0.4980 corrupt_frac_t_0p2_0p4=0.2008 acc_corrupt_t_0p4_0p6=0.6953 corrupt_frac_t_0p4_0p6=0.2003 acc_corrupt_t_0p6_0p8=0.8419 corrupt_frac_t_0p6_0p8=0.2037 acc_corrupt_t_0p8_1p0=0.9517 corrupt_frac_t_0p8_1p0=0.2031 out_w_norm=173.5968 out_g_norm=0.1324 loss_all=1.0189 init_gold_top10=0.5307 init_gold_top100=0.5357 +step=15700 micro_steps=62800 elapsed=55.4s lr=3.000000e-04 loss=2.1736 loss_recon=2.1736 loss_meanflow=0.0000 mean_model_t=0.5025 mean_corrupt_t=0.5025 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8032 corrupt_frac=0.5544 acc_corrupt=0.6524 loss_corrupt=2.1736 wrong_frac=0.4982 init_acc_corrupt=0.4676 acc_corrupt_t_0p0_0p2=0.2757 corrupt_frac_t_0p0_0p2=0.2045 acc_corrupt_t_0p2_0p4=0.4912 corrupt_frac_t_0p2_0p4=0.1957 acc_corrupt_t_0p4_0p6=0.7004 corrupt_frac_t_0p4_0p6=0.1969 acc_corrupt_t_0p6_0p8=0.8414 corrupt_frac_t_0p6_0p8=0.1962 acc_corrupt_t_0p8_1p0=0.9528 corrupt_frac_t_0p8_1p0=0.2066 out_w_norm=174.2329 out_g_norm=0.1316 loss_all=0.7402 init_gold_top10=0.5752 init_gold_top100=0.5801 +step=15800 micro_steps=63200 elapsed=55.4s lr=3.000000e-04 loss=2.1750 loss_recon=2.1750 loss_meanflow=0.0000 mean_model_t=0.4988 mean_corrupt_t=0.4988 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8032 corrupt_frac=0.5533 acc_corrupt=0.6515 loss_corrupt=2.1750 wrong_frac=0.5008 init_acc_corrupt=0.4637 acc_corrupt_t_0p0_0p2=0.2814 corrupt_frac_t_0p0_0p2=0.2001 acc_corrupt_t_0p2_0p4=0.4904 corrupt_frac_t_0p2_0p4=0.2034 acc_corrupt_t_0p4_0p6=0.6984 corrupt_frac_t_0p4_0p6=0.1998 acc_corrupt_t_0p6_0p8=0.8401 corrupt_frac_t_0p6_0p8=0.1962 acc_corrupt_t_0p8_1p0=0.9517 corrupt_frac_t_0p8_1p0=0.2021 out_w_norm=174.8751 out_g_norm=0.1321 loss_all=0.7646 init_gold_top10=0.5875 init_gold_top100=0.5905 +step=15900 micro_steps=63600 elapsed=55.3s lr=3.000000e-04 loss=2.1411 loss_recon=2.1411 loss_meanflow=0.0000 mean_model_t=0.5021 mean_corrupt_t=0.5021 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8083 corrupt_frac=0.5482 acc_corrupt=0.6568 loss_corrupt=2.1411 wrong_frac=0.4973 init_acc_corrupt=0.4682 acc_corrupt_t_0p0_0p2=0.2855 corrupt_frac_t_0p0_0p2=0.1936 acc_corrupt_t_0p2_0p4=0.4917 corrupt_frac_t_0p2_0p4=0.2018 acc_corrupt_t_0p4_0p6=0.7001 corrupt_frac_t_0p4_0p6=0.2039 acc_corrupt_t_0p6_0p8=0.8414 corrupt_frac_t_0p6_0p8=0.2029 acc_corrupt_t_0p8_1p0=0.9519 corrupt_frac_t_0p8_1p0=0.1988 out_w_norm=175.5112 out_g_norm=0.1314 loss_all=1.3576 init_gold_top10=0.4244 init_gold_top100=0.4313 +step=16000 micro_steps=64000 elapsed=55.3s lr=3.000000e-04 loss=2.1910 loss_recon=2.1910 loss_meanflow=0.0000 mean_model_t=0.5002 mean_corrupt_t=0.5002 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8034 corrupt_frac=0.5501 acc_corrupt=0.6496 loss_corrupt=2.1910 wrong_frac=0.5024 init_acc_corrupt=0.4628 acc_corrupt_t_0p0_0p2=0.2796 corrupt_frac_t_0p0_0p2=0.2005 acc_corrupt_t_0p2_0p4=0.4892 corrupt_frac_t_0p2_0p4=0.2074 acc_corrupt_t_0p4_0p6=0.6966 corrupt_frac_t_0p4_0p6=0.1934 acc_corrupt_t_0p6_0p8=0.8407 corrupt_frac_t_0p6_0p8=0.2022 acc_corrupt_t_0p8_1p0=0.9523 corrupt_frac_t_0p8_1p0=0.1970 out_w_norm=176.1433 out_g_norm=0.1314 loss_all=1.7813 init_gold_top10=0.4612 init_gold_top100=0.4672 +step=16100 micro_steps=64400 elapsed=110.0s lr=3.000000e-04 loss=2.1537 loss_recon=2.1537 loss_meanflow=0.0000 mean_model_t=0.4986 mean_corrupt_t=0.4986 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8063 corrupt_frac=0.5478 acc_corrupt=0.6537 loss_corrupt=2.1537 wrong_frac=0.5009 init_acc_corrupt=0.4645 acc_corrupt_t_0p0_0p2=0.2839 corrupt_frac_t_0p0_0p2=0.1951 acc_corrupt_t_0p2_0p4=0.4923 corrupt_frac_t_0p2_0p4=0.2044 acc_corrupt_t_0p4_0p6=0.6986 corrupt_frac_t_0p4_0p6=0.2027 acc_corrupt_t_0p6_0p8=0.8394 corrupt_frac_t_0p6_0p8=0.2014 acc_corrupt_t_0p8_1p0=0.9521 corrupt_frac_t_0p8_1p0=0.1965 out_w_norm=176.7754 out_g_norm=0.1314 loss_all=1.5428 init_gold_top10=0.4131 init_gold_top100=0.4229 +step=16200 micro_steps=64800 elapsed=92.9s lr=3.000000e-04 loss=2.1462 loss_recon=2.1462 loss_meanflow=0.0000 mean_model_t=0.5021 mean_corrupt_t=0.5021 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8066 corrupt_frac=0.5506 acc_corrupt=0.6557 loss_corrupt=2.1462 wrong_frac=0.4969 init_acc_corrupt=0.4688 acc_corrupt_t_0p0_0p2=0.2805 corrupt_frac_t_0p0_0p2=0.2007 acc_corrupt_t_0p2_0p4=0.4955 corrupt_frac_t_0p2_0p4=0.1959 acc_corrupt_t_0p4_0p6=0.6995 corrupt_frac_t_0p4_0p6=0.2017 acc_corrupt_t_0p6_0p8=0.8426 corrupt_frac_t_0p6_0p8=0.1983 acc_corrupt_t_0p8_1p0=0.9525 corrupt_frac_t_0p8_1p0=0.2049 out_w_norm=177.4045 out_g_norm=0.1298 loss_all=1.2087 init_gold_top10=0.5233 init_gold_top100=0.5317 +step=16300 micro_steps=65200 elapsed=55.6s lr=3.000000e-04 loss=2.1450 loss_recon=2.1450 loss_meanflow=0.0000 mean_model_t=0.5055 mean_corrupt_t=0.5055 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8069 corrupt_frac=0.5503 acc_corrupt=0.6559 loss_corrupt=2.1450 wrong_frac=0.4970 init_acc_corrupt=0.4687 acc_corrupt_t_0p0_0p2=0.2791 corrupt_frac_t_0p0_0p2=0.1951 acc_corrupt_t_0p2_0p4=0.4933 corrupt_frac_t_0p2_0p4=0.2016 acc_corrupt_t_0p4_0p6=0.7012 corrupt_frac_t_0p4_0p6=0.1989 acc_corrupt_t_0p6_0p8=0.8428 corrupt_frac_t_0p6_0p8=0.2061 acc_corrupt_t_0p8_1p0=0.9519 corrupt_frac_t_0p8_1p0=0.1989 out_w_norm=178.0323 out_g_norm=0.1300 loss_all=1.2170 init_gold_top10=0.4526 init_gold_top100=0.4593 +step=16400 micro_steps=65600 elapsed=55.3s lr=3.000000e-04 loss=2.1364 loss_recon=2.1364 loss_meanflow=0.0000 mean_model_t=0.5017 mean_corrupt_t=0.5017 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8075 corrupt_frac=0.5493 acc_corrupt=0.6568 loss_corrupt=2.1364 wrong_frac=0.4969 init_acc_corrupt=0.4690 acc_corrupt_t_0p0_0p2=0.2859 corrupt_frac_t_0p0_0p2=0.1976 acc_corrupt_t_0p2_0p4=0.4912 corrupt_frac_t_0p2_0p4=0.1956 acc_corrupt_t_0p4_0p6=0.6988 corrupt_frac_t_0p4_0p6=0.2027 acc_corrupt_t_0p6_0p8=0.8399 corrupt_frac_t_0p6_0p8=0.1963 acc_corrupt_t_0p8_1p0=0.9513 corrupt_frac_t_0p8_1p0=0.2078 out_w_norm=178.6554 out_g_norm=0.1303 loss_all=0.7634 init_gold_top10=0.6115 init_gold_top100=0.6135 +step=16500 micro_steps=66000 elapsed=55.2s lr=3.000000e-04 loss=2.1400 loss_recon=2.1400 loss_meanflow=0.0000 mean_model_t=0.5027 mean_corrupt_t=0.5027 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8077 corrupt_frac=0.5479 acc_corrupt=0.6562 loss_corrupt=2.1400 wrong_frac=0.4975 init_acc_corrupt=0.4680 acc_corrupt_t_0p0_0p2=0.2826 corrupt_frac_t_0p0_0p2=0.1990 acc_corrupt_t_0p2_0p4=0.4929 corrupt_frac_t_0p2_0p4=0.2003 acc_corrupt_t_0p4_0p6=0.7032 corrupt_frac_t_0p4_0p6=0.1963 acc_corrupt_t_0p6_0p8=0.8434 corrupt_frac_t_0p6_0p8=0.1991 acc_corrupt_t_0p8_1p0=0.9513 corrupt_frac_t_0p8_1p0=0.2052 out_w_norm=179.2760 out_g_norm=0.1292 loss_all=1.2640 init_gold_top10=0.4774 init_gold_top100=0.4814 +step=16600 micro_steps=66400 elapsed=55.2s lr=3.000000e-04 loss=2.1511 loss_recon=2.1511 loss_meanflow=0.0000 mean_model_t=0.4987 mean_corrupt_t=0.4987 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8066 corrupt_frac=0.5488 acc_corrupt=0.6546 loss_corrupt=2.1511 wrong_frac=0.5008 init_acc_corrupt=0.4644 acc_corrupt_t_0p0_0p2=0.2872 corrupt_frac_t_0p0_0p2=0.1985 acc_corrupt_t_0p2_0p4=0.4923 corrupt_frac_t_0p2_0p4=0.2030 acc_corrupt_t_0p4_0p6=0.7010 corrupt_frac_t_0p4_0p6=0.2011 acc_corrupt_t_0p6_0p8=0.8427 corrupt_frac_t_0p6_0p8=0.1990 acc_corrupt_t_0p8_1p0=0.9525 corrupt_frac_t_0p8_1p0=0.1984 out_w_norm=179.8948 out_g_norm=0.1298 loss_all=1.6949 init_gold_top10=0.3811 init_gold_top100=0.3869 +step=16700 micro_steps=66800 elapsed=55.3s lr=3.000000e-04 loss=2.1447 loss_recon=2.1447 loss_meanflow=0.0000 mean_model_t=0.5014 mean_corrupt_t=0.5014 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8062 corrupt_frac=0.5511 acc_corrupt=0.6554 loss_corrupt=2.1447 wrong_frac=0.4989 init_acc_corrupt=0.4667 acc_corrupt_t_0p0_0p2=0.2849 corrupt_frac_t_0p0_0p2=0.1995 acc_corrupt_t_0p2_0p4=0.4944 corrupt_frac_t_0p2_0p4=0.1998 acc_corrupt_t_0p4_0p6=0.7018 corrupt_frac_t_0p4_0p6=0.2004 acc_corrupt_t_0p6_0p8=0.8408 corrupt_frac_t_0p6_0p8=0.1994 acc_corrupt_t_0p8_1p0=0.9530 corrupt_frac_t_0p8_1p0=0.2010 out_w_norm=180.5139 out_g_norm=0.1285 loss_all=1.1103 init_gold_top10=0.4919 init_gold_top100=0.4971 +step=16800 micro_steps=67200 elapsed=55.4s lr=3.000000e-04 loss=2.1592 loss_recon=2.1592 loss_meanflow=0.0000 mean_model_t=0.5012 mean_corrupt_t=0.5012 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8060 corrupt_frac=0.5491 acc_corrupt=0.6536 loss_corrupt=2.1592 wrong_frac=0.5008 init_acc_corrupt=0.4645 acc_corrupt_t_0p0_0p2=0.2829 corrupt_frac_t_0p0_0p2=0.1993 acc_corrupt_t_0p2_0p4=0.4940 corrupt_frac_t_0p2_0p4=0.2052 acc_corrupt_t_0p4_0p6=0.7013 corrupt_frac_t_0p4_0p6=0.1978 acc_corrupt_t_0p6_0p8=0.8426 corrupt_frac_t_0p6_0p8=0.2025 acc_corrupt_t_0p8_1p0=0.9526 corrupt_frac_t_0p8_1p0=0.1968 out_w_norm=181.1301 out_g_norm=0.1280 loss_all=1.3353 init_gold_top10=0.4830 init_gold_top100=0.4898 +step=16900 micro_steps=67600 elapsed=55.4s lr=3.000000e-04 loss=2.1491 loss_recon=2.1491 loss_meanflow=0.0000 mean_model_t=0.4996 mean_corrupt_t=0.4996 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8056 corrupt_frac=0.5518 acc_corrupt=0.6546 loss_corrupt=2.1491 wrong_frac=0.5000 init_acc_corrupt=0.4658 acc_corrupt_t_0p0_0p2=0.2818 corrupt_frac_t_0p0_0p2=0.1939 acc_corrupt_t_0p2_0p4=0.4964 corrupt_frac_t_0p2_0p4=0.2101 acc_corrupt_t_0p4_0p6=0.7019 corrupt_frac_t_0p4_0p6=0.2045 acc_corrupt_t_0p6_0p8=0.8418 corrupt_frac_t_0p6_0p8=0.1942 acc_corrupt_t_0p8_1p0=0.9519 corrupt_frac_t_0p8_1p0=0.1988 out_w_norm=181.7404 out_g_norm=0.1281 loss_all=1.2911 init_gold_top10=0.4646 init_gold_top100=0.4730 +step=17000 micro_steps=68000 elapsed=55.3s lr=3.000000e-04 loss=2.1308 loss_recon=2.1308 loss_meanflow=0.0000 mean_model_t=0.5017 mean_corrupt_t=0.5017 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8101 corrupt_frac=0.5437 acc_corrupt=0.6577 loss_corrupt=2.1308 wrong_frac=0.4984 init_acc_corrupt=0.4678 acc_corrupt_t_0p0_0p2=0.2838 corrupt_frac_t_0p0_0p2=0.1972 acc_corrupt_t_0p2_0p4=0.4978 corrupt_frac_t_0p2_0p4=0.1980 acc_corrupt_t_0p4_0p6=0.7026 corrupt_frac_t_0p4_0p6=0.2049 acc_corrupt_t_0p6_0p8=0.8415 corrupt_frac_t_0p6_0p8=0.2036 acc_corrupt_t_0p8_1p0=0.9531 corrupt_frac_t_0p8_1p0=0.1984 out_w_norm=182.3490 out_g_norm=0.1278 loss_all=1.2377 init_gold_top10=0.5916 init_gold_top100=0.5951 +step=17100 micro_steps=68400 elapsed=98.0s lr=3.000000e-04 loss=2.1111 loss_recon=2.1111 loss_meanflow=0.0000 mean_model_t=0.5017 mean_corrupt_t=0.5017 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8110 corrupt_frac=0.5446 acc_corrupt=0.6596 loss_corrupt=2.1111 wrong_frac=0.4969 init_acc_corrupt=0.4693 acc_corrupt_t_0p0_0p2=0.2879 corrupt_frac_t_0p0_0p2=0.1944 acc_corrupt_t_0p2_0p4=0.5003 corrupt_frac_t_0p2_0p4=0.2004 acc_corrupt_t_0p4_0p6=0.6998 corrupt_frac_t_0p4_0p6=0.2015 acc_corrupt_t_0p6_0p8=0.8437 corrupt_frac_t_0p6_0p8=0.2014 acc_corrupt_t_0p8_1p0=0.9515 corrupt_frac_t_0p8_1p0=0.2028 out_w_norm=182.9590 out_g_norm=0.1269 loss_all=0.9167 init_gold_top10=0.5828 init_gold_top100=0.5894 +step=17200 micro_steps=68800 elapsed=104.6s lr=3.000000e-04 loss=2.1132 loss_recon=2.1132 loss_meanflow=0.0000 mean_model_t=0.5042 mean_corrupt_t=0.5042 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8086 corrupt_frac=0.5526 acc_corrupt=0.6602 loss_corrupt=2.1132 wrong_frac=0.4942 init_acc_corrupt=0.4717 acc_corrupt_t_0p0_0p2=0.2875 corrupt_frac_t_0p0_0p2=0.1867 acc_corrupt_t_0p2_0p4=0.4928 corrupt_frac_t_0p2_0p4=0.2091 acc_corrupt_t_0p4_0p6=0.6986 corrupt_frac_t_0p4_0p6=0.1971 acc_corrupt_t_0p6_0p8=0.8431 corrupt_frac_t_0p6_0p8=0.2026 acc_corrupt_t_0p8_1p0=0.9534 corrupt_frac_t_0p8_1p0=0.2050 out_w_norm=183.5602 out_g_norm=0.1252 loss_all=1.6001 init_gold_top10=0.4351 init_gold_top100=0.4424 +step=17300 micro_steps=69200 elapsed=55.3s lr=3.000000e-04 loss=2.0949 loss_recon=2.0949 loss_meanflow=0.0000 mean_model_t=0.5065 mean_corrupt_t=0.5065 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8104 corrupt_frac=0.5514 acc_corrupt=0.6629 loss_corrupt=2.0949 wrong_frac=0.4916 init_acc_corrupt=0.4745 acc_corrupt_t_0p0_0p2=0.2858 corrupt_frac_t_0p0_0p2=0.1914 acc_corrupt_t_0p2_0p4=0.4984 corrupt_frac_t_0p2_0p4=0.1953 acc_corrupt_t_0p4_0p6=0.7032 corrupt_frac_t_0p4_0p6=0.2048 acc_corrupt_t_0p6_0p8=0.8417 corrupt_frac_t_0p6_0p8=0.2013 acc_corrupt_t_0p8_1p0=0.9529 corrupt_frac_t_0p8_1p0=0.2071 out_w_norm=184.1579 out_g_norm=0.1251 loss_all=1.2162 init_gold_top10=0.4742 init_gold_top100=0.4800 +step=17400 micro_steps=69600 elapsed=55.3s lr=3.000000e-04 loss=2.1355 loss_recon=2.1355 loss_meanflow=0.0000 mean_model_t=0.5016 mean_corrupt_t=0.5016 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8087 corrupt_frac=0.5459 acc_corrupt=0.6565 loss_corrupt=2.1355 wrong_frac=0.4993 init_acc_corrupt=0.4665 acc_corrupt_t_0p0_0p2=0.2852 corrupt_frac_t_0p0_0p2=0.2009 acc_corrupt_t_0p2_0p4=0.4947 corrupt_frac_t_0p2_0p4=0.1933 acc_corrupt_t_0p4_0p6=0.7004 corrupt_frac_t_0p4_0p6=0.2061 acc_corrupt_t_0p6_0p8=0.8443 corrupt_frac_t_0p6_0p8=0.2023 acc_corrupt_t_0p8_1p0=0.9522 corrupt_frac_t_0p8_1p0=0.1990 out_w_norm=184.7580 out_g_norm=0.1256 loss_all=0.9814 init_gold_top10=0.5159 init_gold_top100=0.5212 +step=17500 micro_steps=70000 elapsed=55.3s lr=3.000000e-04 loss=2.1552 loss_recon=2.1552 loss_meanflow=0.0000 mean_model_t=0.5005 mean_corrupt_t=0.5005 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8054 corrupt_frac=0.5521 acc_corrupt=0.6540 loss_corrupt=2.1552 wrong_frac=0.5011 init_acc_corrupt=0.4638 acc_corrupt_t_0p0_0p2=0.2812 corrupt_frac_t_0p0_0p2=0.2012 acc_corrupt_t_0p2_0p4=0.4956 corrupt_frac_t_0p2_0p4=0.2016 acc_corrupt_t_0p4_0p6=0.7018 corrupt_frac_t_0p4_0p6=0.1959 acc_corrupt_t_0p6_0p8=0.8432 corrupt_frac_t_0p6_0p8=0.2057 acc_corrupt_t_0p8_1p0=0.9523 corrupt_frac_t_0p8_1p0=0.1961 out_w_norm=185.3536 out_g_norm=0.1249 loss_all=0.9326 init_gold_top10=0.5210 init_gold_top100=0.5266 +step=17600 micro_steps=70400 elapsed=55.3s lr=3.000000e-04 loss=2.1364 loss_recon=2.1364 loss_meanflow=0.0000 mean_model_t=0.5011 mean_corrupt_t=0.5011 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8072 corrupt_frac=0.5511 acc_corrupt=0.6571 loss_corrupt=2.1364 wrong_frac=0.4972 init_acc_corrupt=0.4685 acc_corrupt_t_0p0_0p2=0.2836 corrupt_frac_t_0p0_0p2=0.2022 acc_corrupt_t_0p2_0p4=0.4962 corrupt_frac_t_0p2_0p4=0.1984 acc_corrupt_t_0p4_0p6=0.7068 corrupt_frac_t_0p4_0p6=0.1973 acc_corrupt_t_0p6_0p8=0.8423 corrupt_frac_t_0p6_0p8=0.1979 acc_corrupt_t_0p8_1p0=0.9537 corrupt_frac_t_0p8_1p0=0.2051 out_w_norm=185.9519 out_g_norm=0.1244 loss_all=1.2206 init_gold_top10=0.5121 init_gold_top100=0.5165 +step=17700 micro_steps=70800 elapsed=55.3s lr=3.000000e-04 loss=2.1307 loss_recon=2.1307 loss_meanflow=0.0000 mean_model_t=0.4987 mean_corrupt_t=0.4987 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8080 corrupt_frac=0.5477 acc_corrupt=0.6567 loss_corrupt=2.1307 wrong_frac=0.5005 init_acc_corrupt=0.4651 acc_corrupt_t_0p0_0p2=0.2883 corrupt_frac_t_0p0_0p2=0.2000 acc_corrupt_t_0p2_0p4=0.4983 corrupt_frac_t_0p2_0p4=0.1980 acc_corrupt_t_0p4_0p6=0.6992 corrupt_frac_t_0p4_0p6=0.2016 acc_corrupt_t_0p6_0p8=0.8437 corrupt_frac_t_0p6_0p8=0.2012 acc_corrupt_t_0p8_1p0=0.9517 corrupt_frac_t_0p8_1p0=0.1997 out_w_norm=186.5332 out_g_norm=0.1250 loss_all=0.9894 init_gold_top10=0.4327 init_gold_top100=0.4378 +step=17800 micro_steps=71200 elapsed=55.5s lr=3.000000e-04 loss=2.1271 loss_recon=2.1271 loss_meanflow=0.0000 mean_model_t=0.5011 mean_corrupt_t=0.5011 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8069 corrupt_frac=0.5523 acc_corrupt=0.6575 loss_corrupt=2.1271 wrong_frac=0.4983 init_acc_corrupt=0.4679 acc_corrupt_t_0p0_0p2=0.2844 corrupt_frac_t_0p0_0p2=0.2001 acc_corrupt_t_0p2_0p4=0.4985 corrupt_frac_t_0p2_0p4=0.1945 acc_corrupt_t_0p4_0p6=0.7006 corrupt_frac_t_0p4_0p6=0.2021 acc_corrupt_t_0p6_0p8=0.8427 corrupt_frac_t_0p6_0p8=0.2035 acc_corrupt_t_0p8_1p0=0.9520 corrupt_frac_t_0p8_1p0=0.2008 out_w_norm=187.1258 out_g_norm=0.1242 loss_all=1.2792 init_gold_top10=0.3856 init_gold_top100=0.3927 +step=17900 micro_steps=71600 elapsed=55.5s lr=3.000000e-04 loss=2.1208 loss_recon=2.1208 loss_meanflow=0.0000 mean_model_t=0.5024 mean_corrupt_t=0.5024 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8082 corrupt_frac=0.5511 acc_corrupt=0.6587 loss_corrupt=2.1208 wrong_frac=0.4964 init_acc_corrupt=0.4690 acc_corrupt_t_0p0_0p2=0.2851 corrupt_frac_t_0p0_0p2=0.1964 acc_corrupt_t_0p2_0p4=0.4956 corrupt_frac_t_0p2_0p4=0.2004 acc_corrupt_t_0p4_0p6=0.7042 corrupt_frac_t_0p4_0p6=0.1985 acc_corrupt_t_0p6_0p8=0.8429 corrupt_frac_t_0p6_0p8=0.2021 acc_corrupt_t_0p8_1p0=0.9534 corrupt_frac_t_0p8_1p0=0.2037 out_w_norm=187.7200 out_g_norm=0.1229 loss_all=0.9684 init_gold_top10=0.5295 init_gold_top100=0.5383 +step=18000 micro_steps=72000 elapsed=55.5s lr=3.000000e-04 loss=2.1417 loss_recon=2.1417 loss_meanflow=0.0000 mean_model_t=0.4991 mean_corrupt_t=0.4991 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8061 corrupt_frac=0.5519 acc_corrupt=0.6554 loss_corrupt=2.1417 wrong_frac=0.5004 init_acc_corrupt=0.4651 acc_corrupt_t_0p0_0p2=0.2832 corrupt_frac_t_0p0_0p2=0.2034 acc_corrupt_t_0p2_0p4=0.5004 corrupt_frac_t_0p2_0p4=0.1953 acc_corrupt_t_0p4_0p6=0.6998 corrupt_frac_t_0p4_0p6=0.2034 acc_corrupt_t_0p6_0p8=0.8447 corrupt_frac_t_0p6_0p8=0.1952 acc_corrupt_t_0p8_1p0=0.9513 corrupt_frac_t_0p8_1p0=0.2027 out_w_norm=188.3058 out_g_norm=0.1231 loss_all=1.1764 init_gold_top10=0.5219 init_gold_top100=0.5256 +step=18100 micro_steps=72400 elapsed=87.2s lr=3.000000e-04 loss=2.1557 loss_recon=2.1557 loss_meanflow=0.0000 mean_model_t=0.4990 mean_corrupt_t=0.4990 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8052 corrupt_frac=0.5522 acc_corrupt=0.6541 loss_corrupt=2.1557 wrong_frac=0.5012 init_acc_corrupt=0.4640 acc_corrupt_t_0p0_0p2=0.2853 corrupt_frac_t_0p0_0p2=0.2008 acc_corrupt_t_0p2_0p4=0.4958 corrupt_frac_t_0p2_0p4=0.2021 acc_corrupt_t_0p4_0p6=0.7002 corrupt_frac_t_0p4_0p6=0.2015 acc_corrupt_t_0p6_0p8=0.8422 corrupt_frac_t_0p6_0p8=0.1978 acc_corrupt_t_0p8_1p0=0.9541 corrupt_frac_t_0p8_1p0=0.1991 out_w_norm=188.8931 out_g_norm=0.1228 loss_all=1.4938 init_gold_top10=0.4904 init_gold_top100=0.4963 +step=18200 micro_steps=72800 elapsed=115.6s lr=3.000000e-04 loss=2.1394 loss_recon=2.1394 loss_meanflow=0.0000 mean_model_t=0.5013 mean_corrupt_t=0.5013 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8062 corrupt_frac=0.5535 acc_corrupt=0.6567 loss_corrupt=2.1394 wrong_frac=0.4985 init_acc_corrupt=0.4674 acc_corrupt_t_0p0_0p2=0.2820 corrupt_frac_t_0p0_0p2=0.1991 acc_corrupt_t_0p2_0p4=0.5003 corrupt_frac_t_0p2_0p4=0.1991 acc_corrupt_t_0p4_0p6=0.7012 corrupt_frac_t_0p4_0p6=0.2043 acc_corrupt_t_0p6_0p8=0.8443 corrupt_frac_t_0p6_0p8=0.1982 acc_corrupt_t_0p8_1p0=0.9529 corrupt_frac_t_0p8_1p0=0.2007 out_w_norm=189.4813 out_g_norm=0.1215 loss_all=1.3989 init_gold_top10=0.4034 init_gold_top100=0.4116 +step=18300 micro_steps=73200 elapsed=55.7s lr=3.000000e-04 loss=2.1362 loss_recon=2.1362 loss_meanflow=0.0000 mean_model_t=0.4990 mean_corrupt_t=0.4990 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8084 corrupt_frac=0.5468 acc_corrupt=0.6561 loss_corrupt=2.1362 wrong_frac=0.5018 init_acc_corrupt=0.4632 acc_corrupt_t_0p0_0p2=0.2912 corrupt_frac_t_0p0_0p2=0.1986 acc_corrupt_t_0p2_0p4=0.4959 corrupt_frac_t_0p2_0p4=0.2010 acc_corrupt_t_0p4_0p6=0.7016 corrupt_frac_t_0p4_0p6=0.2070 acc_corrupt_t_0p6_0p8=0.8421 corrupt_frac_t_0p6_0p8=0.1987 acc_corrupt_t_0p8_1p0=0.9526 corrupt_frac_t_0p8_1p0=0.1967 out_w_norm=190.0654 out_g_norm=0.1224 loss_all=0.9277 init_gold_top10=0.5509 init_gold_top100=0.5550 +step=18400 micro_steps=73600 elapsed=55.5s lr=3.000000e-04 loss=2.1671 loss_recon=2.1671 loss_meanflow=0.0000 mean_model_t=0.4955 mean_corrupt_t=0.4955 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8035 corrupt_frac=0.5532 acc_corrupt=0.6516 loss_corrupt=2.1671 wrong_frac=0.5049 init_acc_corrupt=0.4597 acc_corrupt_t_0p0_0p2=0.2876 corrupt_frac_t_0p0_0p2=0.1998 acc_corrupt_t_0p2_0p4=0.4919 corrupt_frac_t_0p2_0p4=0.2072 acc_corrupt_t_0p4_0p6=0.7006 corrupt_frac_t_0p4_0p6=0.2062 acc_corrupt_t_0p6_0p8=0.8419 corrupt_frac_t_0p6_0p8=0.1864 acc_corrupt_t_0p8_1p0=0.9522 corrupt_frac_t_0p8_1p0=0.2004 out_w_norm=190.6413 out_g_norm=0.1220 loss_all=1.1219 init_gold_top10=0.4733 init_gold_top100=0.4800 +step=18500 micro_steps=74000 elapsed=55.5s lr=3.000000e-04 loss=2.1222 loss_recon=2.1222 loss_meanflow=0.0000 mean_model_t=0.5010 mean_corrupt_t=0.5010 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8092 corrupt_frac=0.5472 acc_corrupt=0.6581 loss_corrupt=2.1222 wrong_frac=0.4989 init_acc_corrupt=0.4672 acc_corrupt_t_0p0_0p2=0.2881 corrupt_frac_t_0p0_0p2=0.2030 acc_corrupt_t_0p2_0p4=0.5015 corrupt_frac_t_0p2_0p4=0.1938 acc_corrupt_t_0p4_0p6=0.7016 corrupt_frac_t_0p4_0p6=0.2051 acc_corrupt_t_0p6_0p8=0.8452 corrupt_frac_t_0p6_0p8=0.1975 acc_corrupt_t_0p8_1p0=0.9529 corrupt_frac_t_0p8_1p0=0.2016 out_w_norm=191.2196 out_g_norm=0.1218 loss_all=1.3138 init_gold_top10=0.4293 init_gold_top100=0.4378 +step=18600 micro_steps=74400 elapsed=55.5s lr=3.000000e-04 loss=2.1170 loss_recon=2.1170 loss_meanflow=0.0000 mean_model_t=0.5025 mean_corrupt_t=0.5025 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8095 corrupt_frac=0.5482 acc_corrupt=0.6593 loss_corrupt=2.1170 wrong_frac=0.4969 init_acc_corrupt=0.4689 acc_corrupt_t_0p0_0p2=0.2905 corrupt_frac_t_0p0_0p2=0.2012 acc_corrupt_t_0p2_0p4=0.4963 corrupt_frac_t_0p2_0p4=0.1943 acc_corrupt_t_0p4_0p6=0.7051 corrupt_frac_t_0p4_0p6=0.2005 acc_corrupt_t_0p6_0p8=0.8425 corrupt_frac_t_0p6_0p8=0.1989 acc_corrupt_t_0p8_1p0=0.9523 corrupt_frac_t_0p8_1p0=0.2055 out_w_norm=191.7959 out_g_norm=0.1213 loss_all=1.0330 init_gold_top10=0.4394 init_gold_top100=0.4490 +step=18700 micro_steps=74800 elapsed=55.6s lr=3.000000e-04 loss=2.1449 loss_recon=2.1449 loss_meanflow=0.0000 mean_model_t=0.4982 mean_corrupt_t=0.4982 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8075 corrupt_frac=0.5481 acc_corrupt=0.6554 loss_corrupt=2.1449 wrong_frac=0.5018 init_acc_corrupt=0.4632 acc_corrupt_t_0p0_0p2=0.2858 corrupt_frac_t_0p0_0p2=0.1995 acc_corrupt_t_0p2_0p4=0.4963 corrupt_frac_t_0p2_0p4=0.2060 acc_corrupt_t_0p4_0p6=0.7060 corrupt_frac_t_0p4_0p6=0.1977 acc_corrupt_t_0p6_0p8=0.8433 corrupt_frac_t_0p6_0p8=0.2010 acc_corrupt_t_0p8_1p0=0.9528 corrupt_frac_t_0p8_1p0=0.1967 out_w_norm=192.3699 out_g_norm=0.1207 loss_all=0.8939 init_gold_top10=0.5505 init_gold_top100=0.5580 +step=18800 micro_steps=75200 elapsed=57.1s lr=3.000000e-04 loss=2.1472 loss_recon=2.1472 loss_meanflow=0.0000 mean_model_t=0.4989 mean_corrupt_t=0.4989 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8057 corrupt_frac=0.5517 acc_corrupt=0.6546 loss_corrupt=2.1472 wrong_frac=0.5021 init_acc_corrupt=0.4638 acc_corrupt_t_0p0_0p2=0.2863 corrupt_frac_t_0p0_0p2=0.2024 acc_corrupt_t_0p2_0p4=0.4972 corrupt_frac_t_0p2_0p4=0.1970 acc_corrupt_t_0p4_0p6=0.7008 corrupt_frac_t_0p4_0p6=0.2066 acc_corrupt_t_0p6_0p8=0.8436 corrupt_frac_t_0p6_0p8=0.1966 acc_corrupt_t_0p8_1p0=0.9527 corrupt_frac_t_0p8_1p0=0.1974 out_w_norm=192.9412 out_g_norm=0.1212 loss_all=0.8950 init_gold_top10=0.5424 init_gold_top100=0.5482 +step=18900 micro_steps=75600 elapsed=55.6s lr=3.000000e-04 loss=2.0923 loss_recon=2.0923 loss_meanflow=0.0000 mean_model_t=0.5058 mean_corrupt_t=0.5058 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8101 corrupt_frac=0.5524 acc_corrupt=0.6627 loss_corrupt=2.0923 wrong_frac=0.4936 init_acc_corrupt=0.4728 acc_corrupt_t_0p0_0p2=0.2865 corrupt_frac_t_0p0_0p2=0.1924 acc_corrupt_t_0p2_0p4=0.5016 corrupt_frac_t_0p2_0p4=0.1977 acc_corrupt_t_0p4_0p6=0.7029 corrupt_frac_t_0p4_0p6=0.2034 acc_corrupt_t_0p6_0p8=0.8436 corrupt_frac_t_0p6_0p8=0.2026 acc_corrupt_t_0p8_1p0=0.9524 corrupt_frac_t_0p8_1p0=0.2044 out_w_norm=193.5230 out_g_norm=0.1198 loss_all=1.0672 init_gold_top10=0.5289 init_gold_top100=0.5319 +step=19000 micro_steps=76000 elapsed=55.6s lr=3.000000e-04 loss=2.1360 loss_recon=2.1360 loss_meanflow=0.0000 mean_model_t=0.5018 mean_corrupt_t=0.5018 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8068 corrupt_frac=0.5530 acc_corrupt=0.6571 loss_corrupt=2.1360 wrong_frac=0.4988 init_acc_corrupt=0.4669 acc_corrupt_t_0p0_0p2=0.2830 corrupt_frac_t_0p0_0p2=0.1968 acc_corrupt_t_0p2_0p4=0.4980 corrupt_frac_t_0p2_0p4=0.2018 acc_corrupt_t_0p4_0p6=0.7012 corrupt_frac_t_0p4_0p6=0.2033 acc_corrupt_t_0p6_0p8=0.8433 corrupt_frac_t_0p6_0p8=0.1948 acc_corrupt_t_0p8_1p0=0.9529 corrupt_frac_t_0p8_1p0=0.2048 out_w_norm=194.0959 out_g_norm=0.1196 loss_all=1.1546 init_gold_top10=0.5585 init_gold_top100=0.5633 +step=19100 micro_steps=76400 elapsed=116.3s lr=3.000000e-04 loss=2.1502 loss_recon=2.1502 loss_meanflow=0.0000 mean_model_t=0.4985 mean_corrupt_t=0.4985 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8051 corrupt_frac=0.5536 acc_corrupt=0.6544 loss_corrupt=2.1502 wrong_frac=0.5024 init_acc_corrupt=0.4624 acc_corrupt_t_0p0_0p2=0.2820 corrupt_frac_t_0p0_0p2=0.1980 acc_corrupt_t_0p2_0p4=0.4980 corrupt_frac_t_0p2_0p4=0.2065 acc_corrupt_t_0p4_0p6=0.7024 corrupt_frac_t_0p4_0p6=0.1983 acc_corrupt_t_0p6_0p8=0.8434 corrupt_frac_t_0p6_0p8=0.2032 acc_corrupt_t_0p8_1p0=0.9536 corrupt_frac_t_0p8_1p0=0.1940 out_w_norm=194.6702 out_g_norm=0.1192 loss_all=1.2710 init_gold_top10=0.4713 init_gold_top100=0.4788 +step=19200 micro_steps=76800 elapsed=86.2s lr=3.000000e-04 loss=2.0919 loss_recon=2.0919 loss_meanflow=0.0000 mean_model_t=0.5040 mean_corrupt_t=0.5040 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8121 corrupt_frac=0.5457 acc_corrupt=0.6623 loss_corrupt=2.0919 wrong_frac=0.4952 init_acc_corrupt=0.4717 acc_corrupt_t_0p0_0p2=0.2863 corrupt_frac_t_0p0_0p2=0.1949 acc_corrupt_t_0p2_0p4=0.5036 corrupt_frac_t_0p2_0p4=0.1967 acc_corrupt_t_0p4_0p6=0.7021 corrupt_frac_t_0p4_0p6=0.2017 acc_corrupt_t_0p6_0p8=0.8446 corrupt_frac_t_0p6_0p8=0.2021 acc_corrupt_t_0p8_1p0=0.9536 corrupt_frac_t_0p8_1p0=0.2050 out_w_norm=195.2411 out_g_norm=0.1192 loss_all=0.6054 init_gold_top10=0.6149 init_gold_top100=0.6177 +step=19300 micro_steps=77200 elapsed=55.4s lr=3.000000e-04 loss=2.1508 loss_recon=2.1508 loss_meanflow=0.0000 mean_model_t=0.4980 mean_corrupt_t=0.4980 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8067 corrupt_frac=0.5492 acc_corrupt=0.6545 loss_corrupt=2.1508 wrong_frac=0.5018 init_acc_corrupt=0.4629 acc_corrupt_t_0p0_0p2=0.2850 corrupt_frac_t_0p0_0p2=0.2045 acc_corrupt_t_0p2_0p4=0.4989 corrupt_frac_t_0p2_0p4=0.1998 acc_corrupt_t_0p4_0p6=0.7018 corrupt_frac_t_0p4_0p6=0.2004 acc_corrupt_t_0p6_0p8=0.8441 corrupt_frac_t_0p6_0p8=0.1960 acc_corrupt_t_0p8_1p0=0.9535 corrupt_frac_t_0p8_1p0=0.1999 out_w_norm=195.8075 out_g_norm=0.1188 loss_all=1.1057 init_gold_top10=0.5422 init_gold_top100=0.5471 +step=19400 micro_steps=77600 elapsed=55.4s lr=3.000000e-04 loss=2.1307 loss_recon=2.1307 loss_meanflow=0.0000 mean_model_t=0.5001 mean_corrupt_t=0.5001 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8087 corrupt_frac=0.5473 acc_corrupt=0.6572 loss_corrupt=2.1307 wrong_frac=0.5001 init_acc_corrupt=0.4655 acc_corrupt_t_0p0_0p2=0.2846 corrupt_frac_t_0p0_0p2=0.2002 acc_corrupt_t_0p2_0p4=0.4995 corrupt_frac_t_0p2_0p4=0.2020 acc_corrupt_t_0p4_0p6=0.7059 corrupt_frac_t_0p4_0p6=0.2022 acc_corrupt_t_0p6_0p8=0.8457 corrupt_frac_t_0p6_0p8=0.1959 acc_corrupt_t_0p8_1p0=0.9539 corrupt_frac_t_0p8_1p0=0.2012 out_w_norm=196.3726 out_g_norm=0.1180 loss_all=1.3110 init_gold_top10=0.4275 init_gold_top100=0.4342 +step=19500 micro_steps=78000 elapsed=55.4s lr=3.000000e-04 loss=2.1395 loss_recon=2.1395 loss_meanflow=0.0000 mean_model_t=0.4986 mean_corrupt_t=0.4986 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8068 corrupt_frac=0.5517 acc_corrupt=0.6561 loss_corrupt=2.1395 wrong_frac=0.5009 init_acc_corrupt=0.4640 acc_corrupt_t_0p0_0p2=0.2869 corrupt_frac_t_0p0_0p2=0.2009 acc_corrupt_t_0p2_0p4=0.5029 corrupt_frac_t_0p2_0p4=0.2020 acc_corrupt_t_0p4_0p6=0.7002 corrupt_frac_t_0p4_0p6=0.1993 acc_corrupt_t_0p6_0p8=0.8436 corrupt_frac_t_0p6_0p8=0.1987 acc_corrupt_t_0p8_1p0=0.9531 corrupt_frac_t_0p8_1p0=0.1990 out_w_norm=196.9418 out_g_norm=0.1183 loss_all=0.9406 init_gold_top10=0.5897 init_gold_top100=0.5928 +step=19600 micro_steps=78400 elapsed=55.4s lr=3.000000e-04 loss=2.1421 loss_recon=2.1421 loss_meanflow=0.0000 mean_model_t=0.4959 mean_corrupt_t=0.4959 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8070 corrupt_frac=0.5491 acc_corrupt=0.6549 loss_corrupt=2.1421 wrong_frac=0.5041 init_acc_corrupt=0.4610 acc_corrupt_t_0p0_0p2=0.2884 corrupt_frac_t_0p0_0p2=0.2021 acc_corrupt_t_0p2_0p4=0.4960 corrupt_frac_t_0p2_0p4=0.1991 acc_corrupt_t_0p4_0p6=0.7040 corrupt_frac_t_0p4_0p6=0.2063 acc_corrupt_t_0p6_0p8=0.8440 corrupt_frac_t_0p6_0p8=0.1972 acc_corrupt_t_0p8_1p0=0.9524 corrupt_frac_t_0p8_1p0=0.1957 out_w_norm=197.5004 out_g_norm=0.1182 loss_all=1.1042 init_gold_top10=0.4874 init_gold_top100=0.4924 +step=19700 micro_steps=78800 elapsed=55.4s lr=3.000000e-04 loss=2.1415 loss_recon=2.1415 loss_meanflow=0.0000 mean_model_t=0.5003 mean_corrupt_t=0.5003 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8062 corrupt_frac=0.5525 acc_corrupt=0.6557 loss_corrupt=2.1415 wrong_frac=0.5010 init_acc_corrupt=0.4646 acc_corrupt_t_0p0_0p2=0.2861 corrupt_frac_t_0p0_0p2=0.2014 acc_corrupt_t_0p2_0p4=0.4937 corrupt_frac_t_0p2_0p4=0.1980 acc_corrupt_t_0p4_0p6=0.7050 corrupt_frac_t_0p4_0p6=0.2047 acc_corrupt_t_0p6_0p8=0.8449 corrupt_frac_t_0p6_0p8=0.1992 acc_corrupt_t_0p8_1p0=0.9533 corrupt_frac_t_0p8_1p0=0.1977 out_w_norm=198.0594 out_g_norm=0.1171 loss_all=1.0772 init_gold_top10=0.5133 init_gold_top100=0.5187 +step=19800 micro_steps=79200 elapsed=55.4s lr=3.000000e-04 loss=2.1286 loss_recon=2.1286 loss_meanflow=0.0000 mean_model_t=0.5008 mean_corrupt_t=0.5008 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8077 corrupt_frac=0.5511 acc_corrupt=0.6574 loss_corrupt=2.1286 wrong_frac=0.4994 init_acc_corrupt=0.4657 acc_corrupt_t_0p0_0p2=0.2838 corrupt_frac_t_0p0_0p2=0.2004 acc_corrupt_t_0p2_0p4=0.4992 corrupt_frac_t_0p2_0p4=0.2004 acc_corrupt_t_0p4_0p6=0.7034 corrupt_frac_t_0p4_0p6=0.1953 acc_corrupt_t_0p6_0p8=0.8453 corrupt_frac_t_0p6_0p8=0.2054 acc_corrupt_t_0p8_1p0=0.9548 corrupt_frac_t_0p8_1p0=0.1989 out_w_norm=198.6215 out_g_norm=0.1167 loss_all=1.0427 init_gold_top10=0.5674 init_gold_top100=0.5723 +step=19900 micro_steps=79600 elapsed=55.4s lr=3.000000e-04 loss=2.1213 loss_recon=2.1213 loss_meanflow=0.0000 mean_model_t=0.5040 mean_corrupt_t=0.5040 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8077 corrupt_frac=0.5538 acc_corrupt=0.6591 loss_corrupt=2.1213 wrong_frac=0.4972 init_acc_corrupt=0.4682 acc_corrupt_t_0p0_0p2=0.2873 corrupt_frac_t_0p0_0p2=0.1979 acc_corrupt_t_0p2_0p4=0.4963 corrupt_frac_t_0p2_0p4=0.2007 acc_corrupt_t_0p4_0p6=0.7048 corrupt_frac_t_0p4_0p6=0.1961 acc_corrupt_t_0p6_0p8=0.8445 corrupt_frac_t_0p6_0p8=0.2037 acc_corrupt_t_0p8_1p0=0.9542 corrupt_frac_t_0p8_1p0=0.2016 out_w_norm=199.1770 out_g_norm=0.1162 loss_all=0.9346 init_gold_top10=0.5982 init_gold_top100=0.6023 +step=20000 micro_steps=80000 elapsed=55.4s lr=3.000000e-04 loss=2.1063 loss_recon=2.1063 loss_meanflow=0.0000 mean_model_t=0.5026 mean_corrupt_t=0.5026 mean_loss_t_weight=1.0000 linear_soft_target_mean_conf=0.0000 prior_center_loss_beta=0.0000 rollout_train_applied=0.0000 grad_enabled_before_rollout=1.0000 grad_enabled_after_rollout=1.0000 logits_requires_grad=1.0000 raw_loss_requires_grad=1.0000 acc_all=0.8103 corrupt_frac=0.5484 acc_corrupt=0.6605 loss_corrupt=2.1063 wrong_frac=0.4979 init_acc_corrupt=0.4679 acc_corrupt_t_0p0_0p2=0.2916 corrupt_frac_t_0p0_0p2=0.1994 acc_corrupt_t_0p2_0p4=0.5002 corrupt_frac_t_0p2_0p4=0.1963 acc_corrupt_t_0p4_0p6=0.7026 corrupt_frac_t_0p4_0p6=0.1989 acc_corrupt_t_0p6_0p8=0.8459 corrupt_frac_t_0p6_0p8=0.2066 acc_corrupt_t_0p8_1p0=0.9538 corrupt_frac_t_0p8_1p0=0.1988 out_w_norm=199.7329 out_g_norm=0.1163 loss_all=1.3640 init_gold_top10=0.4185 init_gold_top100=0.4252 diff --git a/LTA_openwebtext_dualt/logs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526_watcher.sh b/LTA_openwebtext_dualt/logs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526_watcher.sh new file mode 100644 index 0000000000000000000000000000000000000000..2bfe47075c46d380660b33950910dc45f1d20ada --- /dev/null +++ b/LTA_openwebtext_dualt/logs/lta_owt_bert_absrope_time4_dirichlet_len1024_C1_to_1024_mask1_sameT_gbs512_b32_8gpu_1m_save10k_20260526_watcher.sh @@ -0,0 +1,117 @@ +#!/usr/bin/env bash +set -euo pipefail + +cd /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt +export PYTHONPATH="$(pwd)${PYTHONPATH:+:$PYTHONPATH}" +export TOKENIZERS_PARALLELISM=false +export PYTHONUNBUFFERED=1 + +: "${RUN_DIR:?RUN_DIR is required}" +: "${OUT_BASE:?OUT_BASE is required}" +: "${LOG_DIR:?LOG_DIR is required}" +: "${TOKENIZER_PATH:?TOKENIZER_PATH is required}" +: "${SCORER:?SCORER is required}" + +RUN_STEM="$(basename "${RUN_DIR}")" +TEMP_TAG="${ENDPOINT_TEMP//./p}" +PROCESSED_FILE="${LOG_DIR}/processed_${RUN_STEM}_steps${STEPS}_c${CMIN}_${CMAX}_gumbel_t${TEMP_TAG}_n${N_SAMPLES}.txt" + +mkdir -p "${OUT_BASE}" "${LOG_DIR}" +touch "${PROCESSED_FILE}" + +echo "[watch-gumbel] run_dir=${RUN_DIR}" +echo "[watch-gumbel] out_base=${OUT_BASE}" +echo "[watch-gumbel] interval=${STEP_INTERVAL} max_len=${MAX_LEN} steps=${STEPS} c=${CMIN}->${CMAX} decode_mode=${DECODE_MODE:-sde_gumbel} temp=${ENDPOINT_TEMP} top_p=${ENDPOINT_TOP_P} tau=${GUMBEL_TAU_START}->${GUMBEL_TAU_END} n=${N_SAMPLES}" + +while true; do + shopt -s nullglob + ckpts=("${RUN_DIR}"/step_*.pt) + shopt -u nullglob + + if (( ${#ckpts[@]} == 0 )); then + echo "[watch-gumbel] $(date +%F_%T) no ckpt yet" + sleep "${SLEEP_SECONDS}" + continue + fi + + printf "%s\n" "${ckpts[@]}" | sort | while read -r ckpt; do + base="$(basename "${ckpt}")" + step="${base#step_}" + step="${step%.pt}" + step_num=$((10#${step})) + if (( step_num % STEP_INTERVAL != 0 )); then + continue + fi + if grep -Fxq "${ckpt}" "${PROCESSED_FILE}"; then + continue + fi + + out_dir="${OUT_BASE}/step_${step}" + log_file="${LOG_DIR}/infer_${RUN_STEM}_step_${step}.log" + mkdir -p "${out_dir}" + + echo "[watch-gumbel] $(date +%F_%T) infer ${ckpt} -> ${out_dir}" | tee -a "${log_file}" + if [[ "${DECODE_MODE:-sde_gumbel}" == "dual_line_probe" ]]; then + CUDA_VISIBLE_DEVICES="${WATCH_CUDA_VISIBLE_DEVICES}" python scripts/infer_softkl_decode_probe.py \ + --checkpoint "${ckpt}" \ + --tokenizer_path "${TOKENIZER_PATH}" \ + --scorer "${SCORER}" \ + --score \ + --out_dir "${out_dir}" \ + --max_lens "${MAX_LEN}" \ + --n_samples "${N_SAMPLES}" \ + --batch_size "${DECODE_BATCH}" \ + --steps "${STEPS}" \ + --decode_rule dual_line_resample \ + --c_min "${CMIN}" \ + --c_max "${CMAX}" \ + --input_noise_dirichlet_concentration "${CMIN}" \ + --anchor_mode state \ + --model_t_mode flow \ + --time_schedule uniform \ + --support_power 1.0 \ + --semantic_power "${DUAL_SEMANTIC_POWER}" \ + --early_temp "${DUAL_EARLY_TEMP}" \ + --late_temp "${DUAL_LATE_TEMP}" \ + --temp_end "${DUAL_TEMP_END}" \ + --temp_power "${DUAL_TEMP_POWER}" \ + --final_from blend \ + --final_decode argmax \ + --seed 20260524 \ + 2>&1 | tee -a "${log_file}" + else + CUDA_VISIBLE_DEVICES="${WATCH_CUDA_VISIBLE_DEVICES}" python scripts/eval_lm1b_c1024_fullycoupled_sde_genppl.py \ + --checkpoint "${ckpt}" \ + --tokenizer_path "${TOKENIZER_PATH}" \ + --scorer "${SCORER}" \ + --out_dir "${out_dir}" \ + --n_samples "${N_SAMPLES}" \ + --max_len "${MAX_LEN}" \ + --steps "${STEPS}" \ + --batch_size "${DECODE_BATCH}" \ + --score_batch "${SCORE_BATCH}" \ + --score_max_length "${SCORE_MAX_LENGTH}" \ + --concentration_min "${CMIN}" \ + --concentration_max "${CMAX}" \ + --endpoint_temp "${ENDPOINT_TEMP}" \ + --endpoint_projection gumbel_softmax \ + --endpoint_top_p "${ENDPOINT_TOP_P}" \ + --gumbel_tau_start "${GUMBEL_TAU_START}" \ + --gumbel_tau_end "${GUMBEL_TAU_END}" \ + --model_t_mode support_t \ + --mean_mode endpoint_only \ + --semantic_power 1.0 \ + --noise_init dirichlet \ + --noise_dirichlet_concentration "${CMIN}" \ + --sde_resample dirichlet \ + --final_from blend_0.5 \ + --seed 20260524 \ + 2>&1 | tee -a "${log_file}" + fi + + echo "${ckpt}" >> "${PROCESSED_FILE}" + echo "[watch-gumbel] $(date +%F_%T) done step_${step}" | tee -a "${log_file}" + done + + sleep "${SLEEP_SECONDS}" +done diff --git a/LTA_openwebtext_dualt/logs/lta_owt_gpt2cached_len1024_rollout1_p05_b64_mlpckpt2_bench4gpu_20260513_161437.log b/LTA_openwebtext_dualt/logs/lta_owt_gpt2cached_len1024_rollout1_p05_b64_mlpckpt2_bench4gpu_20260513_161437.log new file mode 100644 index 0000000000000000000000000000000000000000..94b071500361272da705f6b4f9f8773da205b2a5 --- /dev/null +++ b/LTA_openwebtext_dualt/logs/lta_owt_gpt2cached_len1024_rollout1_p05_b64_mlpckpt2_bench4gpu_20260513_161437.log @@ -0,0 +1,150 @@ +NCCL version 2.25.1+cuda12.8 +{ + "device": "cuda:0", + "rank": 0, + "world_size": 4, + "samples": "owt_cached_chunks:8734897", + "vocab_size": 50257, + "tokenizer_vocab_size": 50257, + "save_dir": "runs/lta_owt_gpt2cached_len1024_rollout1_p05_b64_mlpckpt2_bench4gpu_20260513_161437", + "batch_size": 64, + "grad_accum": 1, + "effective_batch_size": 256, + "global_batch_size": 256, + "lr_schedule": "cosine", + "optimizer": "adamw", + "warmup_steps": 2, + "min_lr": 6e-05, + "weight_decay": 0.1, + "adamw_param_groups": "nanogpt", + "adam_beta1": 0.9, + "adam_beta2": 0.95, + "adam_eps": 1e-08, + "muon_momentum": 0.95, + "muon_ns_steps": 5, + "muon_update_scale": 1.0, + "ema_decay": 0.0, + "ema_start_step": 0, + "model_type": "ddit", + "dual_t": true, + "corrupt_t_mode": "same", + "corrupt_min_t": 0.0, + "corrupt_max_t": 1.0, + "prefix_block_prob": 0.0, + "prefix_block_len": 128, + "dirichlet_endpoint_mode": "categorical_dual_t", + "dirichlet_semantic_t_mode": "same", + "dirichlet_semantic_t_value": 0.0, + "categorical_wrong_from_full_vocab": true, + "categorical_wrong_from_batch_valid_tokens": false, + "mask_mixture_original_prob": 0.0, + "mask_mixture_lowk_prob": 0.0, + "mask_mixture_lowcorrupt_prob": 0.0, + "mask_mixture_block_prob": 0.0, + "mask_mixture_all_prob": 0.0, + "mask_mixture_lowk_clean_tokens": "1,2,4,8,16,32,64", + "mask_mixture_lowcorrupt_tokens": "1,2,4,8,16,32,64", + "mask_mixture_block_tokens": "64,128", + "simplex_bridge_sampler": "dirichlet", + "logistic_normal_sigma_min": 0.18, + "logistic_normal_sigma_max": 2.2, + "logistic_normal_tau_min": 0.65, + "logistic_normal_tau_max": 1.15, + "torch_compile": false, + "compile_mode": "max-autotune", + "state_format": "prob", + "target_loss": "hard_ce", + "meanflow_weight": 0.0, + "rollout_train_prob": 0.5, + "rollout_train_steps": 1, + "rollout_train_infer_steps": 64, + "rollout_train_temp": 1.45, + "rollout_train_max_gamma": 1.0, + "rollout_train_corrupt_only": true, + "rollout_train_samplewise": false, + "rollout_train_compute_always": false, + "bridge_noise_init": "logistic_normal", + "noise_sigma": -1.0, + "allow_tf32": true, + "activation_checkpointing": true, + "activation_checkpoint_interval": 1, + "activation_checkpoint_scope": "mlp", + "ddp_static_graph": false, + "ddp_gradient_as_bucket_view": true, + "blocking_data_transfer": false, + "dataloader_prefetch_factor": 4, + "full_train_stats": false, + "record_pad_truncate": false, + "record_add_eos": false, + "record_add_special_tokens": false, + "record_pad_token": "pad", + "record_shuffle_buffer": 10000, + "wrap": true, + "wrap_mode": "stream", + "wrap_record_buffer_size": 200, + "owt_cached_chunks": true, + "owt_chunk_cache_dir": "/e2e-data/evad-tech-vla/wanghan58/data/small_benchmarks/langflow_2604_11748/openwebtext_lta_cached_chunks/gpt2_len1024_train_minus_100k", + "owt_chunk_cache_rebuild": false, + "owt_chunk_cache_write_batch": 4096, + "owt_exact_repeat_per_chunk": 0, + "online_chunk_shuffle": false, + "online_chunk_shuffle_buffer": 10000, + "openwebtext_split": "train_minus_100k", + "detokenizer": "auto", + "resolved_detokenizer": null, + "num_workers": 8, + "latest_every": 0, + "resume_path": "" +} +[rank0]: Traceback (most recent call last): +[rank0]: File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1546, in +[rank0]: main() +[rank0]: File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1465, in main +[rank0]: loss.backward() +[rank0]: File "/usr/local/lib/python3.12/dist-packages/torch/_tensor.py", line 648, in backward +[rank0]: torch.autograd.backward( +[rank0]: File "/usr/local/lib/python3.12/dist-packages/torch/autograd/__init__.py", line 347, in backward +[rank0]: _engine_run_backward( +[rank0]: File "/usr/local/lib/python3.12/dist-packages/torch/autograd/graph.py", line 823, in _engine_run_backward +[rank0]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass +[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank0]: RuntimeError: CUDA error: CUBLAS_STATUS_ALLOC_FAILED when calling `cublasCreate(handle)` +[rank2]: Traceback (most recent call last): +[rank2]: File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1546, in +[rank2]: main() +[rank2]: File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1465, in main +[rank2]: loss.backward() +[rank2]: File "/usr/local/lib/python3.12/dist-packages/torch/_tensor.py", line 648, in backward +[rank2]: torch.autograd.backward( +[rank2]: File "/usr/local/lib/python3.12/dist-packages/torch/autograd/__init__.py", line 347, in backward +[rank2]: _engine_run_backward( +[rank2]: File "/usr/local/lib/python3.12/dist-packages/torch/autograd/graph.py", line 823, in _engine_run_backward +[rank2]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass +[rank2]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank2]: RuntimeError: CUDA error: CUBLAS_STATUS_ALLOC_FAILED when calling `cublasCreate(handle)` +[rank3]: Traceback (most recent call last): +[rank3]: File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1546, in +[rank3]: main() +[rank3]: File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1465, in main +[rank3]: loss.backward() +[rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/_tensor.py", line 648, in backward +[rank3]: torch.autograd.backward( +[rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/autograd/__init__.py", line 347, in backward +[rank3]: _engine_run_backward( +[rank3]: File "/usr/local/lib/python3.12/dist-packages/torch/autograd/graph.py", line 823, in _engine_run_backward +[rank3]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass +[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank3]: RuntimeError: CUDA error: CUBLAS_STATUS_ALLOC_FAILED when calling `cublasCreate(handle)` +[rank1]: Traceback (most recent call last): +[rank1]: File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1546, in +[rank1]: main() +[rank1]: File "/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/train.py", line 1465, in main +[rank1]: loss.backward() +[rank1]: File "/usr/local/lib/python3.12/dist-packages/torch/_tensor.py", line 648, in backward +[rank1]: torch.autograd.backward( +[rank1]: File "/usr/local/lib/python3.12/dist-packages/torch/autograd/__init__.py", line 347, in backward +[rank1]: _engine_run_backward( +[rank1]: File "/usr/local/lib/python3.12/dist-packages/torch/autograd/graph.py", line 823, in _engine_run_backward +[rank1]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass +[rank1]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ +[rank1]: RuntimeError: t() expects a tensor with <= 2 dimensions, but self is 3D diff --git a/LTA_openwebtext_dualt/logs/owt_lowk64plus_cleanbridge_2k_watcher.nohup b/LTA_openwebtext_dualt/logs/owt_lowk64plus_cleanbridge_2k_watcher.nohup new file mode 100644 index 0000000000000000000000000000000000000000..4a993094336ec7ce3bb0d67112d047191ca578f1 --- /dev/null +++ b/LTA_openwebtext_dualt/logs/owt_lowk64plus_cleanbridge_2k_watcher.nohup @@ -0,0 +1,29 @@ +[watcher] start 2026-05-12T00:37:29+00:00 run_dir=runs/lta_owt_c1024_len1024_t0to1_lowk64plus_cleanbridge_buf1000_gbs128_4gpu_2k +[watcher] poll 2026-05-12T00:46:32+00:00 step=500 last=-1 +[watcher] infer step=500 out=docs/lta_samples/metrics_20260512/owt_lowk64plus_cleanbridge_2k_rolling_sync_probe/step500_sync_probe_n16 +[watcher] done step=500 +[watcher] poll 2026-05-12T00:48:43+00:00 step=500 last=500 +[watcher] poll 2026-05-12T00:50:16+00:00 step=500 last=500 +[watcher] poll 2026-05-12T00:51:49+00:00 step=500 last=500 +[watcher] poll 2026-05-12T00:53:23+00:00 step=500 last=500 +[watcher] poll 2026-05-12T00:54:56+00:00 step=500 last=500 +[watcher] poll 2026-05-12T00:56:29+00:00 step=1000 last=500 +[watcher] infer step=1000 out=docs/lta_samples/metrics_20260512/owt_lowk64plus_cleanbridge_2k_rolling_sync_probe/step1000_sync_probe_n16 +[watcher] done step=1000 +[watcher] poll 2026-05-12T00:58:40+00:00 step=1000 last=1000 +[watcher] poll 2026-05-12T01:00:13+00:00 step=1000 last=1000 +[watcher] poll 2026-05-12T01:01:46+00:00 step=1000 last=1000 +[watcher] poll 2026-05-12T01:03:20+00:00 step=1000 last=1000 +[watcher] poll 2026-05-12T01:04:53+00:00 step=1500 last=1000 +[watcher] infer step=1500 out=docs/lta_samples/metrics_20260512/owt_lowk64plus_cleanbridge_2k_rolling_sync_probe/step1500_sync_probe_n16 +[watcher] done step=1500 +[watcher] poll 2026-05-12T01:07:04+00:00 step=1500 last=1500 +[watcher] poll 2026-05-12T01:08:37+00:00 step=1500 last=1500 +[watcher] poll 2026-05-12T01:10:10+00:00 step=1500 last=1500 +[watcher] poll 2026-05-12T01:11:44+00:00 step=1500 last=1500 +[watcher] poll 2026-05-12T01:13:17+00:00 step=1500 last=1500 +[watcher] poll 2026-05-12T01:14:50+00:00 step=1500 last=1500 +[watcher] poll 2026-05-12T01:16:23+00:00 step=2000 last=1500 +[watcher] infer step=2000 out=docs/lta_samples/metrics_20260512/owt_lowk64plus_cleanbridge_2k_rolling_sync_probe/step2000_sync_probe_n16 +[watcher] done step=2000 +[watcher] reached max step 2000; exiting diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/pass/arrayprint.py b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/pass/arrayprint.py new file mode 100644 index 0000000000000000000000000000000000000000..6c704c755570d1508424af92a0eb5aa1353666a0 --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/pass/arrayprint.py @@ -0,0 +1,37 @@ +import numpy as np + +AR = np.arange(10) +AR.setflags(write=False) + +with np.printoptions(): + np.set_printoptions( + precision=1, + threshold=2, + edgeitems=3, + linewidth=4, + suppress=False, + nanstr="Bob", + infstr="Bill", + formatter={}, + sign="+", + floatmode="unique", + ) + np.get_printoptions() + str(AR) + + np.array2string( + AR, + max_line_width=5, + precision=2, + suppress_small=True, + separator=";", + prefix="test", + threshold=5, + floatmode="fixed", + suffix="?", + legacy="1.13", + ) + np.format_float_scientific(1, precision=5) + np.format_float_positional(1, trim="k") + np.array_repr(AR) + np.array_str(AR) diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/pass/numeric.py b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/pass/numeric.py new file mode 100644 index 0000000000000000000000000000000000000000..c4a73c1e9b7c2792da739047b1d7c88c22c6acfb --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/pass/numeric.py @@ -0,0 +1,90 @@ +""" +Tests for :mod:`numpy.core.numeric`. + +Does not include tests which fall under ``array_constructors``. + +""" + +from __future__ import annotations + +import numpy as np + +class SubClass(np.ndarray): + ... + +i8 = np.int64(1) + +A = np.arange(27).reshape(3, 3, 3) +B: list[list[list[int]]] = A.tolist() +C = np.empty((27, 27)).view(SubClass) + +np.count_nonzero(i8) +np.count_nonzero(A) +np.count_nonzero(B) +np.count_nonzero(A, keepdims=True) +np.count_nonzero(A, axis=0) + +np.isfortran(i8) +np.isfortran(A) + +np.argwhere(i8) +np.argwhere(A) + +np.flatnonzero(i8) +np.flatnonzero(A) + +np.correlate(B[0][0], A.ravel(), mode="valid") +np.correlate(A.ravel(), A.ravel(), mode="same") + +np.convolve(B[0][0], A.ravel(), mode="valid") +np.convolve(A.ravel(), A.ravel(), mode="same") + +np.outer(i8, A) +np.outer(B, A) +np.outer(A, A) +np.outer(A, A, out=C) + +np.tensordot(B, A) +np.tensordot(A, A) +np.tensordot(A, A, axes=0) +np.tensordot(A, A, axes=(0, 1)) + +np.isscalar(i8) +np.isscalar(A) +np.isscalar(B) + +np.roll(A, 1) +np.roll(A, (1, 2)) +np.roll(B, 1) + +np.rollaxis(A, 0, 1) + +np.moveaxis(A, 0, 1) +np.moveaxis(A, (0, 1), (1, 2)) + +np.cross(B, A) +np.cross(A, A) + +np.indices([0, 1, 2]) +np.indices([0, 1, 2], sparse=False) +np.indices([0, 1, 2], sparse=True) + +np.binary_repr(1) + +np.base_repr(1) + +np.allclose(i8, A) +np.allclose(B, A) +np.allclose(A, A) + +np.isclose(i8, A) +np.isclose(B, A) +np.isclose(A, A) + +np.array_equal(i8, A) +np.array_equal(B, A) +np.array_equal(A, A) + +np.array_equiv(i8, A) +np.array_equiv(B, A) +np.array_equiv(A, A) diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/pass/random.py b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/pass/random.py new file mode 100644 index 0000000000000000000000000000000000000000..6a4d99f12b1304773533c1cbdbd45a2d52a44d8b --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/pass/random.py @@ -0,0 +1,1499 @@ +from __future__ import annotations + +from typing import Any +import numpy as np + +SEED_NONE = None +SEED_INT = 4579435749574957634658964293569 +SEED_ARR: np.ndarray[Any, np.dtype[np.int64]] = np.array([1, 2, 3, 4], dtype=np.int64) +SEED_ARRLIKE: list[int] = [1, 2, 3, 4] +SEED_SEED_SEQ: np.random.SeedSequence = np.random.SeedSequence(0) +SEED_MT19937: np.random.MT19937 = np.random.MT19937(0) +SEED_PCG64: np.random.PCG64 = np.random.PCG64(0) +SEED_PHILOX: np.random.Philox = np.random.Philox(0) +SEED_SFC64: np.random.SFC64 = np.random.SFC64(0) + +# default rng +np.random.default_rng() +np.random.default_rng(SEED_NONE) +np.random.default_rng(SEED_INT) +np.random.default_rng(SEED_ARR) +np.random.default_rng(SEED_ARRLIKE) +np.random.default_rng(SEED_SEED_SEQ) +np.random.default_rng(SEED_MT19937) +np.random.default_rng(SEED_PCG64) +np.random.default_rng(SEED_PHILOX) +np.random.default_rng(SEED_SFC64) + +# Seed Sequence +np.random.SeedSequence(SEED_NONE) +np.random.SeedSequence(SEED_INT) +np.random.SeedSequence(SEED_ARR) +np.random.SeedSequence(SEED_ARRLIKE) + +# Bit Generators +np.random.MT19937(SEED_NONE) +np.random.MT19937(SEED_INT) +np.random.MT19937(SEED_ARR) +np.random.MT19937(SEED_ARRLIKE) +np.random.MT19937(SEED_SEED_SEQ) + +np.random.PCG64(SEED_NONE) +np.random.PCG64(SEED_INT) +np.random.PCG64(SEED_ARR) +np.random.PCG64(SEED_ARRLIKE) +np.random.PCG64(SEED_SEED_SEQ) + +np.random.Philox(SEED_NONE) +np.random.Philox(SEED_INT) +np.random.Philox(SEED_ARR) +np.random.Philox(SEED_ARRLIKE) +np.random.Philox(SEED_SEED_SEQ) + +np.random.SFC64(SEED_NONE) +np.random.SFC64(SEED_INT) +np.random.SFC64(SEED_ARR) +np.random.SFC64(SEED_ARRLIKE) +np.random.SFC64(SEED_SEED_SEQ) + +seed_seq: np.random.bit_generator.SeedSequence = np.random.SeedSequence(SEED_NONE) +seed_seq.spawn(10) +seed_seq.generate_state(3) +seed_seq.generate_state(3, "u4") +seed_seq.generate_state(3, "uint32") +seed_seq.generate_state(3, "u8") +seed_seq.generate_state(3, "uint64") +seed_seq.generate_state(3, np.uint32) +seed_seq.generate_state(3, np.uint64) + + +def_gen: np.random.Generator = np.random.default_rng() + +D_arr_0p1: np.ndarray[Any, np.dtype[np.float64]] = np.array([0.1]) +D_arr_0p5: np.ndarray[Any, np.dtype[np.float64]] = np.array([0.5]) +D_arr_0p9: np.ndarray[Any, np.dtype[np.float64]] = np.array([0.9]) +D_arr_1p5: np.ndarray[Any, np.dtype[np.float64]] = np.array([1.5]) +I_arr_10: np.ndarray[Any, np.dtype[np.int_]] = np.array([10], dtype=np.int_) +I_arr_20: np.ndarray[Any, np.dtype[np.int_]] = np.array([20], dtype=np.int_) +D_arr_like_0p1: list[float] = [0.1] +D_arr_like_0p5: list[float] = [0.5] +D_arr_like_0p9: list[float] = [0.9] +D_arr_like_1p5: list[float] = [1.5] +I_arr_like_10: list[int] = [10] +I_arr_like_20: list[int] = [20] +D_2D_like: list[list[float]] = [[1, 2], [2, 3], [3, 4], [4, 5.1]] +D_2D: np.ndarray[Any, np.dtype[np.float64]] = np.array(D_2D_like) + +S_out: np.ndarray[Any, np.dtype[np.float32]] = np.empty(1, dtype=np.float32) +D_out: np.ndarray[Any, np.dtype[np.float64]] = np.empty(1) + +def_gen.standard_normal() +def_gen.standard_normal(dtype=np.float32) +def_gen.standard_normal(dtype="float32") +def_gen.standard_normal(dtype="double") +def_gen.standard_normal(dtype=np.float64) +def_gen.standard_normal(size=None) +def_gen.standard_normal(size=1) +def_gen.standard_normal(size=1, dtype=np.float32) +def_gen.standard_normal(size=1, dtype="f4") +def_gen.standard_normal(size=1, dtype="float32", out=S_out) +def_gen.standard_normal(dtype=np.float32, out=S_out) +def_gen.standard_normal(size=1, dtype=np.float64) +def_gen.standard_normal(size=1, dtype="float64") +def_gen.standard_normal(size=1, dtype="f8") +def_gen.standard_normal(out=D_out) +def_gen.standard_normal(size=1, dtype="float64") +def_gen.standard_normal(size=1, dtype="float64", out=D_out) + +def_gen.random() +def_gen.random(dtype=np.float32) +def_gen.random(dtype="float32") +def_gen.random(dtype="double") +def_gen.random(dtype=np.float64) +def_gen.random(size=None) +def_gen.random(size=1) +def_gen.random(size=1, dtype=np.float32) +def_gen.random(size=1, dtype="f4") +def_gen.random(size=1, dtype="float32", out=S_out) +def_gen.random(dtype=np.float32, out=S_out) +def_gen.random(size=1, dtype=np.float64) +def_gen.random(size=1, dtype="float64") +def_gen.random(size=1, dtype="f8") +def_gen.random(out=D_out) +def_gen.random(size=1, dtype="float64") +def_gen.random(size=1, dtype="float64", out=D_out) + +def_gen.standard_cauchy() +def_gen.standard_cauchy(size=None) +def_gen.standard_cauchy(size=1) + +def_gen.standard_exponential() +def_gen.standard_exponential(method="inv") +def_gen.standard_exponential(dtype=np.float32) +def_gen.standard_exponential(dtype="float32") +def_gen.standard_exponential(dtype="double") +def_gen.standard_exponential(dtype=np.float64) +def_gen.standard_exponential(size=None) +def_gen.standard_exponential(size=None, method="inv") +def_gen.standard_exponential(size=1, method="inv") +def_gen.standard_exponential(size=1, dtype=np.float32) +def_gen.standard_exponential(size=1, dtype="f4", method="inv") +def_gen.standard_exponential(size=1, dtype="float32", out=S_out) +def_gen.standard_exponential(dtype=np.float32, out=S_out) +def_gen.standard_exponential(size=1, dtype=np.float64, method="inv") +def_gen.standard_exponential(size=1, dtype="float64") +def_gen.standard_exponential(size=1, dtype="f8") +def_gen.standard_exponential(out=D_out) +def_gen.standard_exponential(size=1, dtype="float64") +def_gen.standard_exponential(size=1, dtype="float64", out=D_out) + +def_gen.zipf(1.5) +def_gen.zipf(1.5, size=None) +def_gen.zipf(1.5, size=1) +def_gen.zipf(D_arr_1p5) +def_gen.zipf(D_arr_1p5, size=1) +def_gen.zipf(D_arr_like_1p5) +def_gen.zipf(D_arr_like_1p5, size=1) + +def_gen.weibull(0.5) +def_gen.weibull(0.5, size=None) +def_gen.weibull(0.5, size=1) +def_gen.weibull(D_arr_0p5) +def_gen.weibull(D_arr_0p5, size=1) +def_gen.weibull(D_arr_like_0p5) +def_gen.weibull(D_arr_like_0p5, size=1) + +def_gen.standard_t(0.5) +def_gen.standard_t(0.5, size=None) +def_gen.standard_t(0.5, size=1) +def_gen.standard_t(D_arr_0p5) +def_gen.standard_t(D_arr_0p5, size=1) +def_gen.standard_t(D_arr_like_0p5) +def_gen.standard_t(D_arr_like_0p5, size=1) + +def_gen.poisson(0.5) +def_gen.poisson(0.5, size=None) +def_gen.poisson(0.5, size=1) +def_gen.poisson(D_arr_0p5) +def_gen.poisson(D_arr_0p5, size=1) +def_gen.poisson(D_arr_like_0p5) +def_gen.poisson(D_arr_like_0p5, size=1) + +def_gen.power(0.5) +def_gen.power(0.5, size=None) +def_gen.power(0.5, size=1) +def_gen.power(D_arr_0p5) +def_gen.power(D_arr_0p5, size=1) +def_gen.power(D_arr_like_0p5) +def_gen.power(D_arr_like_0p5, size=1) + +def_gen.pareto(0.5) +def_gen.pareto(0.5, size=None) +def_gen.pareto(0.5, size=1) +def_gen.pareto(D_arr_0p5) +def_gen.pareto(D_arr_0p5, size=1) +def_gen.pareto(D_arr_like_0p5) +def_gen.pareto(D_arr_like_0p5, size=1) + +def_gen.chisquare(0.5) +def_gen.chisquare(0.5, size=None) +def_gen.chisquare(0.5, size=1) +def_gen.chisquare(D_arr_0p5) +def_gen.chisquare(D_arr_0p5, size=1) +def_gen.chisquare(D_arr_like_0p5) +def_gen.chisquare(D_arr_like_0p5, size=1) + +def_gen.exponential(0.5) +def_gen.exponential(0.5, size=None) +def_gen.exponential(0.5, size=1) +def_gen.exponential(D_arr_0p5) +def_gen.exponential(D_arr_0p5, size=1) +def_gen.exponential(D_arr_like_0p5) +def_gen.exponential(D_arr_like_0p5, size=1) + +def_gen.geometric(0.5) +def_gen.geometric(0.5, size=None) +def_gen.geometric(0.5, size=1) +def_gen.geometric(D_arr_0p5) +def_gen.geometric(D_arr_0p5, size=1) +def_gen.geometric(D_arr_like_0p5) +def_gen.geometric(D_arr_like_0p5, size=1) + +def_gen.logseries(0.5) +def_gen.logseries(0.5, size=None) +def_gen.logseries(0.5, size=1) +def_gen.logseries(D_arr_0p5) +def_gen.logseries(D_arr_0p5, size=1) +def_gen.logseries(D_arr_like_0p5) +def_gen.logseries(D_arr_like_0p5, size=1) + +def_gen.rayleigh(0.5) +def_gen.rayleigh(0.5, size=None) +def_gen.rayleigh(0.5, size=1) +def_gen.rayleigh(D_arr_0p5) +def_gen.rayleigh(D_arr_0p5, size=1) +def_gen.rayleigh(D_arr_like_0p5) +def_gen.rayleigh(D_arr_like_0p5, size=1) + +def_gen.standard_gamma(0.5) +def_gen.standard_gamma(0.5, size=None) +def_gen.standard_gamma(0.5, dtype="float32") +def_gen.standard_gamma(0.5, size=None, dtype="float32") +def_gen.standard_gamma(0.5, size=1) +def_gen.standard_gamma(D_arr_0p5) +def_gen.standard_gamma(D_arr_0p5, dtype="f4") +def_gen.standard_gamma(0.5, size=1, dtype="float32", out=S_out) +def_gen.standard_gamma(D_arr_0p5, dtype=np.float32, out=S_out) +def_gen.standard_gamma(D_arr_0p5, size=1) +def_gen.standard_gamma(D_arr_like_0p5) +def_gen.standard_gamma(D_arr_like_0p5, size=1) +def_gen.standard_gamma(0.5, out=D_out) +def_gen.standard_gamma(D_arr_like_0p5, out=D_out) +def_gen.standard_gamma(D_arr_like_0p5, size=1) +def_gen.standard_gamma(D_arr_like_0p5, size=1, out=D_out, dtype=np.float64) + +def_gen.vonmises(0.5, 0.5) +def_gen.vonmises(0.5, 0.5, size=None) +def_gen.vonmises(0.5, 0.5, size=1) +def_gen.vonmises(D_arr_0p5, 0.5) +def_gen.vonmises(0.5, D_arr_0p5) +def_gen.vonmises(D_arr_0p5, 0.5, size=1) +def_gen.vonmises(0.5, D_arr_0p5, size=1) +def_gen.vonmises(D_arr_like_0p5, 0.5) +def_gen.vonmises(0.5, D_arr_like_0p5) +def_gen.vonmises(D_arr_0p5, D_arr_0p5) +def_gen.vonmises(D_arr_like_0p5, D_arr_like_0p5) +def_gen.vonmises(D_arr_0p5, D_arr_0p5, size=1) +def_gen.vonmises(D_arr_like_0p5, D_arr_like_0p5, size=1) + +def_gen.wald(0.5, 0.5) +def_gen.wald(0.5, 0.5, size=None) +def_gen.wald(0.5, 0.5, size=1) +def_gen.wald(D_arr_0p5, 0.5) +def_gen.wald(0.5, D_arr_0p5) +def_gen.wald(D_arr_0p5, 0.5, size=1) +def_gen.wald(0.5, D_arr_0p5, size=1) +def_gen.wald(D_arr_like_0p5, 0.5) +def_gen.wald(0.5, D_arr_like_0p5) +def_gen.wald(D_arr_0p5, D_arr_0p5) +def_gen.wald(D_arr_like_0p5, D_arr_like_0p5) +def_gen.wald(D_arr_0p5, D_arr_0p5, size=1) +def_gen.wald(D_arr_like_0p5, D_arr_like_0p5, size=1) + +def_gen.uniform(0.5, 0.5) +def_gen.uniform(0.5, 0.5, size=None) +def_gen.uniform(0.5, 0.5, size=1) +def_gen.uniform(D_arr_0p5, 0.5) +def_gen.uniform(0.5, D_arr_0p5) +def_gen.uniform(D_arr_0p5, 0.5, size=1) +def_gen.uniform(0.5, D_arr_0p5, size=1) +def_gen.uniform(D_arr_like_0p5, 0.5) +def_gen.uniform(0.5, D_arr_like_0p5) +def_gen.uniform(D_arr_0p5, D_arr_0p5) +def_gen.uniform(D_arr_like_0p5, D_arr_like_0p5) +def_gen.uniform(D_arr_0p5, D_arr_0p5, size=1) +def_gen.uniform(D_arr_like_0p5, D_arr_like_0p5, size=1) + +def_gen.beta(0.5, 0.5) +def_gen.beta(0.5, 0.5, size=None) +def_gen.beta(0.5, 0.5, size=1) +def_gen.beta(D_arr_0p5, 0.5) +def_gen.beta(0.5, D_arr_0p5) +def_gen.beta(D_arr_0p5, 0.5, size=1) +def_gen.beta(0.5, D_arr_0p5, size=1) +def_gen.beta(D_arr_like_0p5, 0.5) +def_gen.beta(0.5, D_arr_like_0p5) +def_gen.beta(D_arr_0p5, D_arr_0p5) +def_gen.beta(D_arr_like_0p5, D_arr_like_0p5) +def_gen.beta(D_arr_0p5, D_arr_0p5, size=1) +def_gen.beta(D_arr_like_0p5, D_arr_like_0p5, size=1) + +def_gen.f(0.5, 0.5) +def_gen.f(0.5, 0.5, size=None) +def_gen.f(0.5, 0.5, size=1) +def_gen.f(D_arr_0p5, 0.5) +def_gen.f(0.5, D_arr_0p5) +def_gen.f(D_arr_0p5, 0.5, size=1) +def_gen.f(0.5, D_arr_0p5, size=1) +def_gen.f(D_arr_like_0p5, 0.5) +def_gen.f(0.5, D_arr_like_0p5) +def_gen.f(D_arr_0p5, D_arr_0p5) +def_gen.f(D_arr_like_0p5, D_arr_like_0p5) +def_gen.f(D_arr_0p5, D_arr_0p5, size=1) +def_gen.f(D_arr_like_0p5, D_arr_like_0p5, size=1) + +def_gen.gamma(0.5, 0.5) +def_gen.gamma(0.5, 0.5, size=None) +def_gen.gamma(0.5, 0.5, size=1) +def_gen.gamma(D_arr_0p5, 0.5) +def_gen.gamma(0.5, D_arr_0p5) +def_gen.gamma(D_arr_0p5, 0.5, size=1) +def_gen.gamma(0.5, D_arr_0p5, size=1) +def_gen.gamma(D_arr_like_0p5, 0.5) +def_gen.gamma(0.5, D_arr_like_0p5) +def_gen.gamma(D_arr_0p5, D_arr_0p5) +def_gen.gamma(D_arr_like_0p5, D_arr_like_0p5) +def_gen.gamma(D_arr_0p5, D_arr_0p5, size=1) +def_gen.gamma(D_arr_like_0p5, D_arr_like_0p5, size=1) + +def_gen.gumbel(0.5, 0.5) +def_gen.gumbel(0.5, 0.5, size=None) +def_gen.gumbel(0.5, 0.5, size=1) +def_gen.gumbel(D_arr_0p5, 0.5) +def_gen.gumbel(0.5, D_arr_0p5) +def_gen.gumbel(D_arr_0p5, 0.5, size=1) +def_gen.gumbel(0.5, D_arr_0p5, size=1) +def_gen.gumbel(D_arr_like_0p5, 0.5) +def_gen.gumbel(0.5, D_arr_like_0p5) +def_gen.gumbel(D_arr_0p5, D_arr_0p5) +def_gen.gumbel(D_arr_like_0p5, D_arr_like_0p5) +def_gen.gumbel(D_arr_0p5, D_arr_0p5, size=1) +def_gen.gumbel(D_arr_like_0p5, D_arr_like_0p5, size=1) + +def_gen.laplace(0.5, 0.5) +def_gen.laplace(0.5, 0.5, size=None) +def_gen.laplace(0.5, 0.5, size=1) +def_gen.laplace(D_arr_0p5, 0.5) +def_gen.laplace(0.5, D_arr_0p5) +def_gen.laplace(D_arr_0p5, 0.5, size=1) +def_gen.laplace(0.5, D_arr_0p5, size=1) +def_gen.laplace(D_arr_like_0p5, 0.5) +def_gen.laplace(0.5, D_arr_like_0p5) +def_gen.laplace(D_arr_0p5, D_arr_0p5) +def_gen.laplace(D_arr_like_0p5, D_arr_like_0p5) +def_gen.laplace(D_arr_0p5, D_arr_0p5, size=1) +def_gen.laplace(D_arr_like_0p5, D_arr_like_0p5, size=1) + +def_gen.logistic(0.5, 0.5) +def_gen.logistic(0.5, 0.5, size=None) +def_gen.logistic(0.5, 0.5, size=1) +def_gen.logistic(D_arr_0p5, 0.5) +def_gen.logistic(0.5, D_arr_0p5) +def_gen.logistic(D_arr_0p5, 0.5, size=1) +def_gen.logistic(0.5, D_arr_0p5, size=1) +def_gen.logistic(D_arr_like_0p5, 0.5) +def_gen.logistic(0.5, D_arr_like_0p5) +def_gen.logistic(D_arr_0p5, D_arr_0p5) +def_gen.logistic(D_arr_like_0p5, D_arr_like_0p5) +def_gen.logistic(D_arr_0p5, D_arr_0p5, size=1) +def_gen.logistic(D_arr_like_0p5, D_arr_like_0p5, size=1) + +def_gen.lognormal(0.5, 0.5) +def_gen.lognormal(0.5, 0.5, size=None) +def_gen.lognormal(0.5, 0.5, size=1) +def_gen.lognormal(D_arr_0p5, 0.5) +def_gen.lognormal(0.5, D_arr_0p5) +def_gen.lognormal(D_arr_0p5, 0.5, size=1) +def_gen.lognormal(0.5, D_arr_0p5, size=1) +def_gen.lognormal(D_arr_like_0p5, 0.5) +def_gen.lognormal(0.5, D_arr_like_0p5) +def_gen.lognormal(D_arr_0p5, D_arr_0p5) +def_gen.lognormal(D_arr_like_0p5, D_arr_like_0p5) +def_gen.lognormal(D_arr_0p5, D_arr_0p5, size=1) +def_gen.lognormal(D_arr_like_0p5, D_arr_like_0p5, size=1) + +def_gen.noncentral_chisquare(0.5, 0.5) +def_gen.noncentral_chisquare(0.5, 0.5, size=None) +def_gen.noncentral_chisquare(0.5, 0.5, size=1) +def_gen.noncentral_chisquare(D_arr_0p5, 0.5) +def_gen.noncentral_chisquare(0.5, D_arr_0p5) +def_gen.noncentral_chisquare(D_arr_0p5, 0.5, size=1) +def_gen.noncentral_chisquare(0.5, D_arr_0p5, size=1) +def_gen.noncentral_chisquare(D_arr_like_0p5, 0.5) +def_gen.noncentral_chisquare(0.5, D_arr_like_0p5) +def_gen.noncentral_chisquare(D_arr_0p5, D_arr_0p5) +def_gen.noncentral_chisquare(D_arr_like_0p5, D_arr_like_0p5) +def_gen.noncentral_chisquare(D_arr_0p5, D_arr_0p5, size=1) +def_gen.noncentral_chisquare(D_arr_like_0p5, D_arr_like_0p5, size=1) + +def_gen.normal(0.5, 0.5) +def_gen.normal(0.5, 0.5, size=None) +def_gen.normal(0.5, 0.5, size=1) +def_gen.normal(D_arr_0p5, 0.5) +def_gen.normal(0.5, D_arr_0p5) +def_gen.normal(D_arr_0p5, 0.5, size=1) +def_gen.normal(0.5, D_arr_0p5, size=1) +def_gen.normal(D_arr_like_0p5, 0.5) +def_gen.normal(0.5, D_arr_like_0p5) +def_gen.normal(D_arr_0p5, D_arr_0p5) +def_gen.normal(D_arr_like_0p5, D_arr_like_0p5) +def_gen.normal(D_arr_0p5, D_arr_0p5, size=1) +def_gen.normal(D_arr_like_0p5, D_arr_like_0p5, size=1) + +def_gen.triangular(0.1, 0.5, 0.9) +def_gen.triangular(0.1, 0.5, 0.9, size=None) +def_gen.triangular(0.1, 0.5, 0.9, size=1) +def_gen.triangular(D_arr_0p1, 0.5, 0.9) +def_gen.triangular(0.1, D_arr_0p5, 0.9) +def_gen.triangular(D_arr_0p1, 0.5, D_arr_like_0p9, size=1) +def_gen.triangular(0.1, D_arr_0p5, 0.9, size=1) +def_gen.triangular(D_arr_like_0p1, 0.5, D_arr_0p9) +def_gen.triangular(0.5, D_arr_like_0p5, 0.9) +def_gen.triangular(D_arr_0p1, D_arr_0p5, 0.9) +def_gen.triangular(D_arr_like_0p1, D_arr_like_0p5, 0.9) +def_gen.triangular(D_arr_0p1, D_arr_0p5, D_arr_0p9, size=1) +def_gen.triangular(D_arr_like_0p1, D_arr_like_0p5, D_arr_like_0p9, size=1) + +def_gen.noncentral_f(0.1, 0.5, 0.9) +def_gen.noncentral_f(0.1, 0.5, 0.9, size=None) +def_gen.noncentral_f(0.1, 0.5, 0.9, size=1) +def_gen.noncentral_f(D_arr_0p1, 0.5, 0.9) +def_gen.noncentral_f(0.1, D_arr_0p5, 0.9) +def_gen.noncentral_f(D_arr_0p1, 0.5, D_arr_like_0p9, size=1) +def_gen.noncentral_f(0.1, D_arr_0p5, 0.9, size=1) +def_gen.noncentral_f(D_arr_like_0p1, 0.5, D_arr_0p9) +def_gen.noncentral_f(0.5, D_arr_like_0p5, 0.9) +def_gen.noncentral_f(D_arr_0p1, D_arr_0p5, 0.9) +def_gen.noncentral_f(D_arr_like_0p1, D_arr_like_0p5, 0.9) +def_gen.noncentral_f(D_arr_0p1, D_arr_0p5, D_arr_0p9, size=1) +def_gen.noncentral_f(D_arr_like_0p1, D_arr_like_0p5, D_arr_like_0p9, size=1) + +def_gen.binomial(10, 0.5) +def_gen.binomial(10, 0.5, size=None) +def_gen.binomial(10, 0.5, size=1) +def_gen.binomial(I_arr_10, 0.5) +def_gen.binomial(10, D_arr_0p5) +def_gen.binomial(I_arr_10, 0.5, size=1) +def_gen.binomial(10, D_arr_0p5, size=1) +def_gen.binomial(I_arr_like_10, 0.5) +def_gen.binomial(10, D_arr_like_0p5) +def_gen.binomial(I_arr_10, D_arr_0p5) +def_gen.binomial(I_arr_like_10, D_arr_like_0p5) +def_gen.binomial(I_arr_10, D_arr_0p5, size=1) +def_gen.binomial(I_arr_like_10, D_arr_like_0p5, size=1) + +def_gen.negative_binomial(10, 0.5) +def_gen.negative_binomial(10, 0.5, size=None) +def_gen.negative_binomial(10, 0.5, size=1) +def_gen.negative_binomial(I_arr_10, 0.5) +def_gen.negative_binomial(10, D_arr_0p5) +def_gen.negative_binomial(I_arr_10, 0.5, size=1) +def_gen.negative_binomial(10, D_arr_0p5, size=1) +def_gen.negative_binomial(I_arr_like_10, 0.5) +def_gen.negative_binomial(10, D_arr_like_0p5) +def_gen.negative_binomial(I_arr_10, D_arr_0p5) +def_gen.negative_binomial(I_arr_like_10, D_arr_like_0p5) +def_gen.negative_binomial(I_arr_10, D_arr_0p5, size=1) +def_gen.negative_binomial(I_arr_like_10, D_arr_like_0p5, size=1) + +def_gen.hypergeometric(20, 20, 10) +def_gen.hypergeometric(20, 20, 10, size=None) +def_gen.hypergeometric(20, 20, 10, size=1) +def_gen.hypergeometric(I_arr_20, 20, 10) +def_gen.hypergeometric(20, I_arr_20, 10) +def_gen.hypergeometric(I_arr_20, 20, I_arr_like_10, size=1) +def_gen.hypergeometric(20, I_arr_20, 10, size=1) +def_gen.hypergeometric(I_arr_like_20, 20, I_arr_10) +def_gen.hypergeometric(20, I_arr_like_20, 10) +def_gen.hypergeometric(I_arr_20, I_arr_20, 10) +def_gen.hypergeometric(I_arr_like_20, I_arr_like_20, 10) +def_gen.hypergeometric(I_arr_20, I_arr_20, I_arr_10, size=1) +def_gen.hypergeometric(I_arr_like_20, I_arr_like_20, I_arr_like_10, size=1) + +I_int64_100: np.ndarray[Any, np.dtype[np.int64]] = np.array([100], dtype=np.int64) + +def_gen.integers(0, 100) +def_gen.integers(100) +def_gen.integers([100]) +def_gen.integers(0, [100]) + +I_bool_low: np.ndarray[Any, np.dtype[np.bool_]] = np.array([0], dtype=np.bool_) +I_bool_low_like: list[int] = [0] +I_bool_high_open: np.ndarray[Any, np.dtype[np.bool_]] = np.array([1], dtype=np.bool_) +I_bool_high_closed: np.ndarray[Any, np.dtype[np.bool_]] = np.array([1], dtype=np.bool_) + +def_gen.integers(2, dtype=bool) +def_gen.integers(0, 2, dtype=bool) +def_gen.integers(1, dtype=bool, endpoint=True) +def_gen.integers(0, 1, dtype=bool, endpoint=True) +def_gen.integers(I_bool_low_like, 1, dtype=bool, endpoint=True) +def_gen.integers(I_bool_high_open, dtype=bool) +def_gen.integers(I_bool_low, I_bool_high_open, dtype=bool) +def_gen.integers(0, I_bool_high_open, dtype=bool) +def_gen.integers(I_bool_high_closed, dtype=bool, endpoint=True) +def_gen.integers(I_bool_low, I_bool_high_closed, dtype=bool, endpoint=True) +def_gen.integers(0, I_bool_high_closed, dtype=bool, endpoint=True) + +def_gen.integers(2, dtype=np.bool_) +def_gen.integers(0, 2, dtype=np.bool_) +def_gen.integers(1, dtype=np.bool_, endpoint=True) +def_gen.integers(0, 1, dtype=np.bool_, endpoint=True) +def_gen.integers(I_bool_low_like, 1, dtype=np.bool_, endpoint=True) +def_gen.integers(I_bool_high_open, dtype=np.bool_) +def_gen.integers(I_bool_low, I_bool_high_open, dtype=np.bool_) +def_gen.integers(0, I_bool_high_open, dtype=np.bool_) +def_gen.integers(I_bool_high_closed, dtype=np.bool_, endpoint=True) +def_gen.integers(I_bool_low, I_bool_high_closed, dtype=np.bool_, endpoint=True) +def_gen.integers(0, I_bool_high_closed, dtype=np.bool_, endpoint=True) + +I_u1_low: np.ndarray[Any, np.dtype[np.uint8]] = np.array([0], dtype=np.uint8) +I_u1_low_like: list[int] = [0] +I_u1_high_open: np.ndarray[Any, np.dtype[np.uint8]] = np.array([255], dtype=np.uint8) +I_u1_high_closed: np.ndarray[Any, np.dtype[np.uint8]] = np.array([255], dtype=np.uint8) + +def_gen.integers(256, dtype="u1") +def_gen.integers(0, 256, dtype="u1") +def_gen.integers(255, dtype="u1", endpoint=True) +def_gen.integers(0, 255, dtype="u1", endpoint=True) +def_gen.integers(I_u1_low_like, 255, dtype="u1", endpoint=True) +def_gen.integers(I_u1_high_open, dtype="u1") +def_gen.integers(I_u1_low, I_u1_high_open, dtype="u1") +def_gen.integers(0, I_u1_high_open, dtype="u1") +def_gen.integers(I_u1_high_closed, dtype="u1", endpoint=True) +def_gen.integers(I_u1_low, I_u1_high_closed, dtype="u1", endpoint=True) +def_gen.integers(0, I_u1_high_closed, dtype="u1", endpoint=True) + +def_gen.integers(256, dtype="uint8") +def_gen.integers(0, 256, dtype="uint8") +def_gen.integers(255, dtype="uint8", endpoint=True) +def_gen.integers(0, 255, dtype="uint8", endpoint=True) +def_gen.integers(I_u1_low_like, 255, dtype="uint8", endpoint=True) +def_gen.integers(I_u1_high_open, dtype="uint8") +def_gen.integers(I_u1_low, I_u1_high_open, dtype="uint8") +def_gen.integers(0, I_u1_high_open, dtype="uint8") +def_gen.integers(I_u1_high_closed, dtype="uint8", endpoint=True) +def_gen.integers(I_u1_low, I_u1_high_closed, dtype="uint8", endpoint=True) +def_gen.integers(0, I_u1_high_closed, dtype="uint8", endpoint=True) + +def_gen.integers(256, dtype=np.uint8) +def_gen.integers(0, 256, dtype=np.uint8) +def_gen.integers(255, dtype=np.uint8, endpoint=True) +def_gen.integers(0, 255, dtype=np.uint8, endpoint=True) +def_gen.integers(I_u1_low_like, 255, dtype=np.uint8, endpoint=True) +def_gen.integers(I_u1_high_open, dtype=np.uint8) +def_gen.integers(I_u1_low, I_u1_high_open, dtype=np.uint8) +def_gen.integers(0, I_u1_high_open, dtype=np.uint8) +def_gen.integers(I_u1_high_closed, dtype=np.uint8, endpoint=True) +def_gen.integers(I_u1_low, I_u1_high_closed, dtype=np.uint8, endpoint=True) +def_gen.integers(0, I_u1_high_closed, dtype=np.uint8, endpoint=True) + +I_u2_low: np.ndarray[Any, np.dtype[np.uint16]] = np.array([0], dtype=np.uint16) +I_u2_low_like: list[int] = [0] +I_u2_high_open: np.ndarray[Any, np.dtype[np.uint16]] = np.array([65535], dtype=np.uint16) +I_u2_high_closed: np.ndarray[Any, np.dtype[np.uint16]] = np.array([65535], dtype=np.uint16) + +def_gen.integers(65536, dtype="u2") +def_gen.integers(0, 65536, dtype="u2") +def_gen.integers(65535, dtype="u2", endpoint=True) +def_gen.integers(0, 65535, dtype="u2", endpoint=True) +def_gen.integers(I_u2_low_like, 65535, dtype="u2", endpoint=True) +def_gen.integers(I_u2_high_open, dtype="u2") +def_gen.integers(I_u2_low, I_u2_high_open, dtype="u2") +def_gen.integers(0, I_u2_high_open, dtype="u2") +def_gen.integers(I_u2_high_closed, dtype="u2", endpoint=True) +def_gen.integers(I_u2_low, I_u2_high_closed, dtype="u2", endpoint=True) +def_gen.integers(0, I_u2_high_closed, dtype="u2", endpoint=True) + +def_gen.integers(65536, dtype="uint16") +def_gen.integers(0, 65536, dtype="uint16") +def_gen.integers(65535, dtype="uint16", endpoint=True) +def_gen.integers(0, 65535, dtype="uint16", endpoint=True) +def_gen.integers(I_u2_low_like, 65535, dtype="uint16", endpoint=True) +def_gen.integers(I_u2_high_open, dtype="uint16") +def_gen.integers(I_u2_low, I_u2_high_open, dtype="uint16") +def_gen.integers(0, I_u2_high_open, dtype="uint16") +def_gen.integers(I_u2_high_closed, dtype="uint16", endpoint=True) +def_gen.integers(I_u2_low, I_u2_high_closed, dtype="uint16", endpoint=True) +def_gen.integers(0, I_u2_high_closed, dtype="uint16", endpoint=True) + +def_gen.integers(65536, dtype=np.uint16) +def_gen.integers(0, 65536, dtype=np.uint16) +def_gen.integers(65535, dtype=np.uint16, endpoint=True) +def_gen.integers(0, 65535, dtype=np.uint16, endpoint=True) +def_gen.integers(I_u2_low_like, 65535, dtype=np.uint16, endpoint=True) +def_gen.integers(I_u2_high_open, dtype=np.uint16) +def_gen.integers(I_u2_low, I_u2_high_open, dtype=np.uint16) +def_gen.integers(0, I_u2_high_open, dtype=np.uint16) +def_gen.integers(I_u2_high_closed, dtype=np.uint16, endpoint=True) +def_gen.integers(I_u2_low, I_u2_high_closed, dtype=np.uint16, endpoint=True) +def_gen.integers(0, I_u2_high_closed, dtype=np.uint16, endpoint=True) + +I_u4_low: np.ndarray[Any, np.dtype[np.uint32]] = np.array([0], dtype=np.uint32) +I_u4_low_like: list[int] = [0] +I_u4_high_open: np.ndarray[Any, np.dtype[np.uint32]] = np.array([4294967295], dtype=np.uint32) +I_u4_high_closed: np.ndarray[Any, np.dtype[np.uint32]] = np.array([4294967295], dtype=np.uint32) + +def_gen.integers(4294967296, dtype="u4") +def_gen.integers(0, 4294967296, dtype="u4") +def_gen.integers(4294967295, dtype="u4", endpoint=True) +def_gen.integers(0, 4294967295, dtype="u4", endpoint=True) +def_gen.integers(I_u4_low_like, 4294967295, dtype="u4", endpoint=True) +def_gen.integers(I_u4_high_open, dtype="u4") +def_gen.integers(I_u4_low, I_u4_high_open, dtype="u4") +def_gen.integers(0, I_u4_high_open, dtype="u4") +def_gen.integers(I_u4_high_closed, dtype="u4", endpoint=True) +def_gen.integers(I_u4_low, I_u4_high_closed, dtype="u4", endpoint=True) +def_gen.integers(0, I_u4_high_closed, dtype="u4", endpoint=True) + +def_gen.integers(4294967296, dtype="uint32") +def_gen.integers(0, 4294967296, dtype="uint32") +def_gen.integers(4294967295, dtype="uint32", endpoint=True) +def_gen.integers(0, 4294967295, dtype="uint32", endpoint=True) +def_gen.integers(I_u4_low_like, 4294967295, dtype="uint32", endpoint=True) +def_gen.integers(I_u4_high_open, dtype="uint32") +def_gen.integers(I_u4_low, I_u4_high_open, dtype="uint32") +def_gen.integers(0, I_u4_high_open, dtype="uint32") +def_gen.integers(I_u4_high_closed, dtype="uint32", endpoint=True) +def_gen.integers(I_u4_low, I_u4_high_closed, dtype="uint32", endpoint=True) +def_gen.integers(0, I_u4_high_closed, dtype="uint32", endpoint=True) + +def_gen.integers(4294967296, dtype=np.uint32) +def_gen.integers(0, 4294967296, dtype=np.uint32) +def_gen.integers(4294967295, dtype=np.uint32, endpoint=True) +def_gen.integers(0, 4294967295, dtype=np.uint32, endpoint=True) +def_gen.integers(I_u4_low_like, 4294967295, dtype=np.uint32, endpoint=True) +def_gen.integers(I_u4_high_open, dtype=np.uint32) +def_gen.integers(I_u4_low, I_u4_high_open, dtype=np.uint32) +def_gen.integers(0, I_u4_high_open, dtype=np.uint32) +def_gen.integers(I_u4_high_closed, dtype=np.uint32, endpoint=True) +def_gen.integers(I_u4_low, I_u4_high_closed, dtype=np.uint32, endpoint=True) +def_gen.integers(0, I_u4_high_closed, dtype=np.uint32, endpoint=True) + +I_u8_low: np.ndarray[Any, np.dtype[np.uint64]] = np.array([0], dtype=np.uint64) +I_u8_low_like: list[int] = [0] +I_u8_high_open: np.ndarray[Any, np.dtype[np.uint64]] = np.array([18446744073709551615], dtype=np.uint64) +I_u8_high_closed: np.ndarray[Any, np.dtype[np.uint64]] = np.array([18446744073709551615], dtype=np.uint64) + +def_gen.integers(18446744073709551616, dtype="u8") +def_gen.integers(0, 18446744073709551616, dtype="u8") +def_gen.integers(18446744073709551615, dtype="u8", endpoint=True) +def_gen.integers(0, 18446744073709551615, dtype="u8", endpoint=True) +def_gen.integers(I_u8_low_like, 18446744073709551615, dtype="u8", endpoint=True) +def_gen.integers(I_u8_high_open, dtype="u8") +def_gen.integers(I_u8_low, I_u8_high_open, dtype="u8") +def_gen.integers(0, I_u8_high_open, dtype="u8") +def_gen.integers(I_u8_high_closed, dtype="u8", endpoint=True) +def_gen.integers(I_u8_low, I_u8_high_closed, dtype="u8", endpoint=True) +def_gen.integers(0, I_u8_high_closed, dtype="u8", endpoint=True) + +def_gen.integers(18446744073709551616, dtype="uint64") +def_gen.integers(0, 18446744073709551616, dtype="uint64") +def_gen.integers(18446744073709551615, dtype="uint64", endpoint=True) +def_gen.integers(0, 18446744073709551615, dtype="uint64", endpoint=True) +def_gen.integers(I_u8_low_like, 18446744073709551615, dtype="uint64", endpoint=True) +def_gen.integers(I_u8_high_open, dtype="uint64") +def_gen.integers(I_u8_low, I_u8_high_open, dtype="uint64") +def_gen.integers(0, I_u8_high_open, dtype="uint64") +def_gen.integers(I_u8_high_closed, dtype="uint64", endpoint=True) +def_gen.integers(I_u8_low, I_u8_high_closed, dtype="uint64", endpoint=True) +def_gen.integers(0, I_u8_high_closed, dtype="uint64", endpoint=True) + +def_gen.integers(18446744073709551616, dtype=np.uint64) +def_gen.integers(0, 18446744073709551616, dtype=np.uint64) +def_gen.integers(18446744073709551615, dtype=np.uint64, endpoint=True) +def_gen.integers(0, 18446744073709551615, dtype=np.uint64, endpoint=True) +def_gen.integers(I_u8_low_like, 18446744073709551615, dtype=np.uint64, endpoint=True) +def_gen.integers(I_u8_high_open, dtype=np.uint64) +def_gen.integers(I_u8_low, I_u8_high_open, dtype=np.uint64) +def_gen.integers(0, I_u8_high_open, dtype=np.uint64) +def_gen.integers(I_u8_high_closed, dtype=np.uint64, endpoint=True) +def_gen.integers(I_u8_low, I_u8_high_closed, dtype=np.uint64, endpoint=True) +def_gen.integers(0, I_u8_high_closed, dtype=np.uint64, endpoint=True) + +I_i1_low: np.ndarray[Any, np.dtype[np.int8]] = np.array([-128], dtype=np.int8) +I_i1_low_like: list[int] = [-128] +I_i1_high_open: np.ndarray[Any, np.dtype[np.int8]] = np.array([127], dtype=np.int8) +I_i1_high_closed: np.ndarray[Any, np.dtype[np.int8]] = np.array([127], dtype=np.int8) + +def_gen.integers(128, dtype="i1") +def_gen.integers(-128, 128, dtype="i1") +def_gen.integers(127, dtype="i1", endpoint=True) +def_gen.integers(-128, 127, dtype="i1", endpoint=True) +def_gen.integers(I_i1_low_like, 127, dtype="i1", endpoint=True) +def_gen.integers(I_i1_high_open, dtype="i1") +def_gen.integers(I_i1_low, I_i1_high_open, dtype="i1") +def_gen.integers(-128, I_i1_high_open, dtype="i1") +def_gen.integers(I_i1_high_closed, dtype="i1", endpoint=True) +def_gen.integers(I_i1_low, I_i1_high_closed, dtype="i1", endpoint=True) +def_gen.integers(-128, I_i1_high_closed, dtype="i1", endpoint=True) + +def_gen.integers(128, dtype="int8") +def_gen.integers(-128, 128, dtype="int8") +def_gen.integers(127, dtype="int8", endpoint=True) +def_gen.integers(-128, 127, dtype="int8", endpoint=True) +def_gen.integers(I_i1_low_like, 127, dtype="int8", endpoint=True) +def_gen.integers(I_i1_high_open, dtype="int8") +def_gen.integers(I_i1_low, I_i1_high_open, dtype="int8") +def_gen.integers(-128, I_i1_high_open, dtype="int8") +def_gen.integers(I_i1_high_closed, dtype="int8", endpoint=True) +def_gen.integers(I_i1_low, I_i1_high_closed, dtype="int8", endpoint=True) +def_gen.integers(-128, I_i1_high_closed, dtype="int8", endpoint=True) + +def_gen.integers(128, dtype=np.int8) +def_gen.integers(-128, 128, dtype=np.int8) +def_gen.integers(127, dtype=np.int8, endpoint=True) +def_gen.integers(-128, 127, dtype=np.int8, endpoint=True) +def_gen.integers(I_i1_low_like, 127, dtype=np.int8, endpoint=True) +def_gen.integers(I_i1_high_open, dtype=np.int8) +def_gen.integers(I_i1_low, I_i1_high_open, dtype=np.int8) +def_gen.integers(-128, I_i1_high_open, dtype=np.int8) +def_gen.integers(I_i1_high_closed, dtype=np.int8, endpoint=True) +def_gen.integers(I_i1_low, I_i1_high_closed, dtype=np.int8, endpoint=True) +def_gen.integers(-128, I_i1_high_closed, dtype=np.int8, endpoint=True) + +I_i2_low: np.ndarray[Any, np.dtype[np.int16]] = np.array([-32768], dtype=np.int16) +I_i2_low_like: list[int] = [-32768] +I_i2_high_open: np.ndarray[Any, np.dtype[np.int16]] = np.array([32767], dtype=np.int16) +I_i2_high_closed: np.ndarray[Any, np.dtype[np.int16]] = np.array([32767], dtype=np.int16) + +def_gen.integers(32768, dtype="i2") +def_gen.integers(-32768, 32768, dtype="i2") +def_gen.integers(32767, dtype="i2", endpoint=True) +def_gen.integers(-32768, 32767, dtype="i2", endpoint=True) +def_gen.integers(I_i2_low_like, 32767, dtype="i2", endpoint=True) +def_gen.integers(I_i2_high_open, dtype="i2") +def_gen.integers(I_i2_low, I_i2_high_open, dtype="i2") +def_gen.integers(-32768, I_i2_high_open, dtype="i2") +def_gen.integers(I_i2_high_closed, dtype="i2", endpoint=True) +def_gen.integers(I_i2_low, I_i2_high_closed, dtype="i2", endpoint=True) +def_gen.integers(-32768, I_i2_high_closed, dtype="i2", endpoint=True) + +def_gen.integers(32768, dtype="int16") +def_gen.integers(-32768, 32768, dtype="int16") +def_gen.integers(32767, dtype="int16", endpoint=True) +def_gen.integers(-32768, 32767, dtype="int16", endpoint=True) +def_gen.integers(I_i2_low_like, 32767, dtype="int16", endpoint=True) +def_gen.integers(I_i2_high_open, dtype="int16") +def_gen.integers(I_i2_low, I_i2_high_open, dtype="int16") +def_gen.integers(-32768, I_i2_high_open, dtype="int16") +def_gen.integers(I_i2_high_closed, dtype="int16", endpoint=True) +def_gen.integers(I_i2_low, I_i2_high_closed, dtype="int16", endpoint=True) +def_gen.integers(-32768, I_i2_high_closed, dtype="int16", endpoint=True) + +def_gen.integers(32768, dtype=np.int16) +def_gen.integers(-32768, 32768, dtype=np.int16) +def_gen.integers(32767, dtype=np.int16, endpoint=True) +def_gen.integers(-32768, 32767, dtype=np.int16, endpoint=True) +def_gen.integers(I_i2_low_like, 32767, dtype=np.int16, endpoint=True) +def_gen.integers(I_i2_high_open, dtype=np.int16) +def_gen.integers(I_i2_low, I_i2_high_open, dtype=np.int16) +def_gen.integers(-32768, I_i2_high_open, dtype=np.int16) +def_gen.integers(I_i2_high_closed, dtype=np.int16, endpoint=True) +def_gen.integers(I_i2_low, I_i2_high_closed, dtype=np.int16, endpoint=True) +def_gen.integers(-32768, I_i2_high_closed, dtype=np.int16, endpoint=True) + +I_i4_low: np.ndarray[Any, np.dtype[np.int32]] = np.array([-2147483648], dtype=np.int32) +I_i4_low_like: list[int] = [-2147483648] +I_i4_high_open: np.ndarray[Any, np.dtype[np.int32]] = np.array([2147483647], dtype=np.int32) +I_i4_high_closed: np.ndarray[Any, np.dtype[np.int32]] = np.array([2147483647], dtype=np.int32) + +def_gen.integers(2147483648, dtype="i4") +def_gen.integers(-2147483648, 2147483648, dtype="i4") +def_gen.integers(2147483647, dtype="i4", endpoint=True) +def_gen.integers(-2147483648, 2147483647, dtype="i4", endpoint=True) +def_gen.integers(I_i4_low_like, 2147483647, dtype="i4", endpoint=True) +def_gen.integers(I_i4_high_open, dtype="i4") +def_gen.integers(I_i4_low, I_i4_high_open, dtype="i4") +def_gen.integers(-2147483648, I_i4_high_open, dtype="i4") +def_gen.integers(I_i4_high_closed, dtype="i4", endpoint=True) +def_gen.integers(I_i4_low, I_i4_high_closed, dtype="i4", endpoint=True) +def_gen.integers(-2147483648, I_i4_high_closed, dtype="i4", endpoint=True) + +def_gen.integers(2147483648, dtype="int32") +def_gen.integers(-2147483648, 2147483648, dtype="int32") +def_gen.integers(2147483647, dtype="int32", endpoint=True) +def_gen.integers(-2147483648, 2147483647, dtype="int32", endpoint=True) +def_gen.integers(I_i4_low_like, 2147483647, dtype="int32", endpoint=True) +def_gen.integers(I_i4_high_open, dtype="int32") +def_gen.integers(I_i4_low, I_i4_high_open, dtype="int32") +def_gen.integers(-2147483648, I_i4_high_open, dtype="int32") +def_gen.integers(I_i4_high_closed, dtype="int32", endpoint=True) +def_gen.integers(I_i4_low, I_i4_high_closed, dtype="int32", endpoint=True) +def_gen.integers(-2147483648, I_i4_high_closed, dtype="int32", endpoint=True) + +def_gen.integers(2147483648, dtype=np.int32) +def_gen.integers(-2147483648, 2147483648, dtype=np.int32) +def_gen.integers(2147483647, dtype=np.int32, endpoint=True) +def_gen.integers(-2147483648, 2147483647, dtype=np.int32, endpoint=True) +def_gen.integers(I_i4_low_like, 2147483647, dtype=np.int32, endpoint=True) +def_gen.integers(I_i4_high_open, dtype=np.int32) +def_gen.integers(I_i4_low, I_i4_high_open, dtype=np.int32) +def_gen.integers(-2147483648, I_i4_high_open, dtype=np.int32) +def_gen.integers(I_i4_high_closed, dtype=np.int32, endpoint=True) +def_gen.integers(I_i4_low, I_i4_high_closed, dtype=np.int32, endpoint=True) +def_gen.integers(-2147483648, I_i4_high_closed, dtype=np.int32, endpoint=True) + +I_i8_low: np.ndarray[Any, np.dtype[np.int64]] = np.array([-9223372036854775808], dtype=np.int64) +I_i8_low_like: list[int] = [-9223372036854775808] +I_i8_high_open: np.ndarray[Any, np.dtype[np.int64]] = np.array([9223372036854775807], dtype=np.int64) +I_i8_high_closed: np.ndarray[Any, np.dtype[np.int64]] = np.array([9223372036854775807], dtype=np.int64) + +def_gen.integers(9223372036854775808, dtype="i8") +def_gen.integers(-9223372036854775808, 9223372036854775808, dtype="i8") +def_gen.integers(9223372036854775807, dtype="i8", endpoint=True) +def_gen.integers(-9223372036854775808, 9223372036854775807, dtype="i8", endpoint=True) +def_gen.integers(I_i8_low_like, 9223372036854775807, dtype="i8", endpoint=True) +def_gen.integers(I_i8_high_open, dtype="i8") +def_gen.integers(I_i8_low, I_i8_high_open, dtype="i8") +def_gen.integers(-9223372036854775808, I_i8_high_open, dtype="i8") +def_gen.integers(I_i8_high_closed, dtype="i8", endpoint=True) +def_gen.integers(I_i8_low, I_i8_high_closed, dtype="i8", endpoint=True) +def_gen.integers(-9223372036854775808, I_i8_high_closed, dtype="i8", endpoint=True) + +def_gen.integers(9223372036854775808, dtype="int64") +def_gen.integers(-9223372036854775808, 9223372036854775808, dtype="int64") +def_gen.integers(9223372036854775807, dtype="int64", endpoint=True) +def_gen.integers(-9223372036854775808, 9223372036854775807, dtype="int64", endpoint=True) +def_gen.integers(I_i8_low_like, 9223372036854775807, dtype="int64", endpoint=True) +def_gen.integers(I_i8_high_open, dtype="int64") +def_gen.integers(I_i8_low, I_i8_high_open, dtype="int64") +def_gen.integers(-9223372036854775808, I_i8_high_open, dtype="int64") +def_gen.integers(I_i8_high_closed, dtype="int64", endpoint=True) +def_gen.integers(I_i8_low, I_i8_high_closed, dtype="int64", endpoint=True) +def_gen.integers(-9223372036854775808, I_i8_high_closed, dtype="int64", endpoint=True) + +def_gen.integers(9223372036854775808, dtype=np.int64) +def_gen.integers(-9223372036854775808, 9223372036854775808, dtype=np.int64) +def_gen.integers(9223372036854775807, dtype=np.int64, endpoint=True) +def_gen.integers(-9223372036854775808, 9223372036854775807, dtype=np.int64, endpoint=True) +def_gen.integers(I_i8_low_like, 9223372036854775807, dtype=np.int64, endpoint=True) +def_gen.integers(I_i8_high_open, dtype=np.int64) +def_gen.integers(I_i8_low, I_i8_high_open, dtype=np.int64) +def_gen.integers(-9223372036854775808, I_i8_high_open, dtype=np.int64) +def_gen.integers(I_i8_high_closed, dtype=np.int64, endpoint=True) +def_gen.integers(I_i8_low, I_i8_high_closed, dtype=np.int64, endpoint=True) +def_gen.integers(-9223372036854775808, I_i8_high_closed, dtype=np.int64, endpoint=True) + + +def_gen.bit_generator + +def_gen.bytes(2) + +def_gen.choice(5) +def_gen.choice(5, 3) +def_gen.choice(5, 3, replace=True) +def_gen.choice(5, 3, p=[1 / 5] * 5) +def_gen.choice(5, 3, p=[1 / 5] * 5, replace=False) + +def_gen.choice(["pooh", "rabbit", "piglet", "Christopher"]) +def_gen.choice(["pooh", "rabbit", "piglet", "Christopher"], 3) +def_gen.choice(["pooh", "rabbit", "piglet", "Christopher"], 3, p=[1 / 4] * 4) +def_gen.choice(["pooh", "rabbit", "piglet", "Christopher"], 3, replace=True) +def_gen.choice(["pooh", "rabbit", "piglet", "Christopher"], 3, replace=False, p=np.array([1 / 8, 1 / 8, 1 / 2, 1 / 4])) + +def_gen.dirichlet([0.5, 0.5]) +def_gen.dirichlet(np.array([0.5, 0.5])) +def_gen.dirichlet(np.array([0.5, 0.5]), size=3) + +def_gen.multinomial(20, [1 / 6.0] * 6) +def_gen.multinomial(20, np.array([0.5, 0.5])) +def_gen.multinomial(20, [1 / 6.0] * 6, size=2) +def_gen.multinomial([[10], [20]], [1 / 6.0] * 6, size=(2, 2)) +def_gen.multinomial(np.array([[10], [20]]), np.array([0.5, 0.5]), size=(2, 2)) + +def_gen.multivariate_hypergeometric([3, 5, 7], 2) +def_gen.multivariate_hypergeometric(np.array([3, 5, 7]), 2) +def_gen.multivariate_hypergeometric(np.array([3, 5, 7]), 2, size=4) +def_gen.multivariate_hypergeometric(np.array([3, 5, 7]), 2, size=(4, 7)) +def_gen.multivariate_hypergeometric([3, 5, 7], 2, method="count") +def_gen.multivariate_hypergeometric(np.array([3, 5, 7]), 2, method="marginals") + +def_gen.multivariate_normal([0.0], [[1.0]]) +def_gen.multivariate_normal([0.0], np.array([[1.0]])) +def_gen.multivariate_normal(np.array([0.0]), [[1.0]]) +def_gen.multivariate_normal([0.0], np.array([[1.0]])) + +def_gen.permutation(10) +def_gen.permutation([1, 2, 3, 4]) +def_gen.permutation(np.array([1, 2, 3, 4])) +def_gen.permutation(D_2D, axis=1) +def_gen.permuted(D_2D) +def_gen.permuted(D_2D_like) +def_gen.permuted(D_2D, axis=1) +def_gen.permuted(D_2D, out=D_2D) +def_gen.permuted(D_2D_like, out=D_2D) +def_gen.permuted(D_2D_like, out=D_2D) +def_gen.permuted(D_2D, axis=1, out=D_2D) + +def_gen.shuffle(np.arange(10)) +def_gen.shuffle([1, 2, 3, 4, 5]) +def_gen.shuffle(D_2D, axis=1) + +def_gen.__str__() +def_gen.__repr__() +def_gen_state: dict[str, Any] +def_gen_state = def_gen.__getstate__() +def_gen.__setstate__(def_gen_state) + +# RandomState +random_st: np.random.RandomState = np.random.RandomState() + +random_st.standard_normal() +random_st.standard_normal(size=None) +random_st.standard_normal(size=1) + +random_st.random() +random_st.random(size=None) +random_st.random(size=1) + +random_st.standard_cauchy() +random_st.standard_cauchy(size=None) +random_st.standard_cauchy(size=1) + +random_st.standard_exponential() +random_st.standard_exponential(size=None) +random_st.standard_exponential(size=1) + +random_st.zipf(1.5) +random_st.zipf(1.5, size=None) +random_st.zipf(1.5, size=1) +random_st.zipf(D_arr_1p5) +random_st.zipf(D_arr_1p5, size=1) +random_st.zipf(D_arr_like_1p5) +random_st.zipf(D_arr_like_1p5, size=1) + +random_st.weibull(0.5) +random_st.weibull(0.5, size=None) +random_st.weibull(0.5, size=1) +random_st.weibull(D_arr_0p5) +random_st.weibull(D_arr_0p5, size=1) +random_st.weibull(D_arr_like_0p5) +random_st.weibull(D_arr_like_0p5, size=1) + +random_st.standard_t(0.5) +random_st.standard_t(0.5, size=None) +random_st.standard_t(0.5, size=1) +random_st.standard_t(D_arr_0p5) +random_st.standard_t(D_arr_0p5, size=1) +random_st.standard_t(D_arr_like_0p5) +random_st.standard_t(D_arr_like_0p5, size=1) + +random_st.poisson(0.5) +random_st.poisson(0.5, size=None) +random_st.poisson(0.5, size=1) +random_st.poisson(D_arr_0p5) +random_st.poisson(D_arr_0p5, size=1) +random_st.poisson(D_arr_like_0p5) +random_st.poisson(D_arr_like_0p5, size=1) + +random_st.power(0.5) +random_st.power(0.5, size=None) +random_st.power(0.5, size=1) +random_st.power(D_arr_0p5) +random_st.power(D_arr_0p5, size=1) +random_st.power(D_arr_like_0p5) +random_st.power(D_arr_like_0p5, size=1) + +random_st.pareto(0.5) +random_st.pareto(0.5, size=None) +random_st.pareto(0.5, size=1) +random_st.pareto(D_arr_0p5) +random_st.pareto(D_arr_0p5, size=1) +random_st.pareto(D_arr_like_0p5) +random_st.pareto(D_arr_like_0p5, size=1) + +random_st.chisquare(0.5) +random_st.chisquare(0.5, size=None) +random_st.chisquare(0.5, size=1) +random_st.chisquare(D_arr_0p5) +random_st.chisquare(D_arr_0p5, size=1) +random_st.chisquare(D_arr_like_0p5) +random_st.chisquare(D_arr_like_0p5, size=1) + +random_st.exponential(0.5) +random_st.exponential(0.5, size=None) +random_st.exponential(0.5, size=1) +random_st.exponential(D_arr_0p5) +random_st.exponential(D_arr_0p5, size=1) +random_st.exponential(D_arr_like_0p5) +random_st.exponential(D_arr_like_0p5, size=1) + +random_st.geometric(0.5) +random_st.geometric(0.5, size=None) +random_st.geometric(0.5, size=1) +random_st.geometric(D_arr_0p5) +random_st.geometric(D_arr_0p5, size=1) +random_st.geometric(D_arr_like_0p5) +random_st.geometric(D_arr_like_0p5, size=1) + +random_st.logseries(0.5) +random_st.logseries(0.5, size=None) +random_st.logseries(0.5, size=1) +random_st.logseries(D_arr_0p5) +random_st.logseries(D_arr_0p5, size=1) +random_st.logseries(D_arr_like_0p5) +random_st.logseries(D_arr_like_0p5, size=1) + +random_st.rayleigh(0.5) +random_st.rayleigh(0.5, size=None) +random_st.rayleigh(0.5, size=1) +random_st.rayleigh(D_arr_0p5) +random_st.rayleigh(D_arr_0p5, size=1) +random_st.rayleigh(D_arr_like_0p5) +random_st.rayleigh(D_arr_like_0p5, size=1) + +random_st.standard_gamma(0.5) +random_st.standard_gamma(0.5, size=None) +random_st.standard_gamma(0.5, size=1) +random_st.standard_gamma(D_arr_0p5) +random_st.standard_gamma(D_arr_0p5, size=1) +random_st.standard_gamma(D_arr_like_0p5) +random_st.standard_gamma(D_arr_like_0p5, size=1) +random_st.standard_gamma(D_arr_like_0p5, size=1) + +random_st.vonmises(0.5, 0.5) +random_st.vonmises(0.5, 0.5, size=None) +random_st.vonmises(0.5, 0.5, size=1) +random_st.vonmises(D_arr_0p5, 0.5) +random_st.vonmises(0.5, D_arr_0p5) +random_st.vonmises(D_arr_0p5, 0.5, size=1) +random_st.vonmises(0.5, D_arr_0p5, size=1) +random_st.vonmises(D_arr_like_0p5, 0.5) +random_st.vonmises(0.5, D_arr_like_0p5) +random_st.vonmises(D_arr_0p5, D_arr_0p5) +random_st.vonmises(D_arr_like_0p5, D_arr_like_0p5) +random_st.vonmises(D_arr_0p5, D_arr_0p5, size=1) +random_st.vonmises(D_arr_like_0p5, D_arr_like_0p5, size=1) + +random_st.wald(0.5, 0.5) +random_st.wald(0.5, 0.5, size=None) +random_st.wald(0.5, 0.5, size=1) +random_st.wald(D_arr_0p5, 0.5) +random_st.wald(0.5, D_arr_0p5) +random_st.wald(D_arr_0p5, 0.5, size=1) +random_st.wald(0.5, D_arr_0p5, size=1) +random_st.wald(D_arr_like_0p5, 0.5) +random_st.wald(0.5, D_arr_like_0p5) +random_st.wald(D_arr_0p5, D_arr_0p5) +random_st.wald(D_arr_like_0p5, D_arr_like_0p5) +random_st.wald(D_arr_0p5, D_arr_0p5, size=1) +random_st.wald(D_arr_like_0p5, D_arr_like_0p5, size=1) + +random_st.uniform(0.5, 0.5) +random_st.uniform(0.5, 0.5, size=None) +random_st.uniform(0.5, 0.5, size=1) +random_st.uniform(D_arr_0p5, 0.5) +random_st.uniform(0.5, D_arr_0p5) +random_st.uniform(D_arr_0p5, 0.5, size=1) +random_st.uniform(0.5, D_arr_0p5, size=1) +random_st.uniform(D_arr_like_0p5, 0.5) +random_st.uniform(0.5, D_arr_like_0p5) +random_st.uniform(D_arr_0p5, D_arr_0p5) +random_st.uniform(D_arr_like_0p5, D_arr_like_0p5) +random_st.uniform(D_arr_0p5, D_arr_0p5, size=1) +random_st.uniform(D_arr_like_0p5, D_arr_like_0p5, size=1) + +random_st.beta(0.5, 0.5) +random_st.beta(0.5, 0.5, size=None) +random_st.beta(0.5, 0.5, size=1) +random_st.beta(D_arr_0p5, 0.5) +random_st.beta(0.5, D_arr_0p5) +random_st.beta(D_arr_0p5, 0.5, size=1) +random_st.beta(0.5, D_arr_0p5, size=1) +random_st.beta(D_arr_like_0p5, 0.5) +random_st.beta(0.5, D_arr_like_0p5) +random_st.beta(D_arr_0p5, D_arr_0p5) +random_st.beta(D_arr_like_0p5, D_arr_like_0p5) +random_st.beta(D_arr_0p5, D_arr_0p5, size=1) +random_st.beta(D_arr_like_0p5, D_arr_like_0p5, size=1) + +random_st.f(0.5, 0.5) +random_st.f(0.5, 0.5, size=None) +random_st.f(0.5, 0.5, size=1) +random_st.f(D_arr_0p5, 0.5) +random_st.f(0.5, D_arr_0p5) +random_st.f(D_arr_0p5, 0.5, size=1) +random_st.f(0.5, D_arr_0p5, size=1) +random_st.f(D_arr_like_0p5, 0.5) +random_st.f(0.5, D_arr_like_0p5) +random_st.f(D_arr_0p5, D_arr_0p5) +random_st.f(D_arr_like_0p5, D_arr_like_0p5) +random_st.f(D_arr_0p5, D_arr_0p5, size=1) +random_st.f(D_arr_like_0p5, D_arr_like_0p5, size=1) + +random_st.gamma(0.5, 0.5) +random_st.gamma(0.5, 0.5, size=None) +random_st.gamma(0.5, 0.5, size=1) +random_st.gamma(D_arr_0p5, 0.5) +random_st.gamma(0.5, D_arr_0p5) +random_st.gamma(D_arr_0p5, 0.5, size=1) +random_st.gamma(0.5, D_arr_0p5, size=1) +random_st.gamma(D_arr_like_0p5, 0.5) +random_st.gamma(0.5, D_arr_like_0p5) +random_st.gamma(D_arr_0p5, D_arr_0p5) +random_st.gamma(D_arr_like_0p5, D_arr_like_0p5) +random_st.gamma(D_arr_0p5, D_arr_0p5, size=1) +random_st.gamma(D_arr_like_0p5, D_arr_like_0p5, size=1) + +random_st.gumbel(0.5, 0.5) +random_st.gumbel(0.5, 0.5, size=None) +random_st.gumbel(0.5, 0.5, size=1) +random_st.gumbel(D_arr_0p5, 0.5) +random_st.gumbel(0.5, D_arr_0p5) +random_st.gumbel(D_arr_0p5, 0.5, size=1) +random_st.gumbel(0.5, D_arr_0p5, size=1) +random_st.gumbel(D_arr_like_0p5, 0.5) +random_st.gumbel(0.5, D_arr_like_0p5) +random_st.gumbel(D_arr_0p5, D_arr_0p5) +random_st.gumbel(D_arr_like_0p5, D_arr_like_0p5) +random_st.gumbel(D_arr_0p5, D_arr_0p5, size=1) +random_st.gumbel(D_arr_like_0p5, D_arr_like_0p5, size=1) + +random_st.laplace(0.5, 0.5) +random_st.laplace(0.5, 0.5, size=None) +random_st.laplace(0.5, 0.5, size=1) +random_st.laplace(D_arr_0p5, 0.5) +random_st.laplace(0.5, D_arr_0p5) +random_st.laplace(D_arr_0p5, 0.5, size=1) +random_st.laplace(0.5, D_arr_0p5, size=1) +random_st.laplace(D_arr_like_0p5, 0.5) +random_st.laplace(0.5, D_arr_like_0p5) +random_st.laplace(D_arr_0p5, D_arr_0p5) +random_st.laplace(D_arr_like_0p5, D_arr_like_0p5) +random_st.laplace(D_arr_0p5, D_arr_0p5, size=1) +random_st.laplace(D_arr_like_0p5, D_arr_like_0p5, size=1) + +random_st.logistic(0.5, 0.5) +random_st.logistic(0.5, 0.5, size=None) +random_st.logistic(0.5, 0.5, size=1) +random_st.logistic(D_arr_0p5, 0.5) +random_st.logistic(0.5, D_arr_0p5) +random_st.logistic(D_arr_0p5, 0.5, size=1) +random_st.logistic(0.5, D_arr_0p5, size=1) +random_st.logistic(D_arr_like_0p5, 0.5) +random_st.logistic(0.5, D_arr_like_0p5) +random_st.logistic(D_arr_0p5, D_arr_0p5) +random_st.logistic(D_arr_like_0p5, D_arr_like_0p5) +random_st.logistic(D_arr_0p5, D_arr_0p5, size=1) +random_st.logistic(D_arr_like_0p5, D_arr_like_0p5, size=1) + +random_st.lognormal(0.5, 0.5) +random_st.lognormal(0.5, 0.5, size=None) +random_st.lognormal(0.5, 0.5, size=1) +random_st.lognormal(D_arr_0p5, 0.5) +random_st.lognormal(0.5, D_arr_0p5) +random_st.lognormal(D_arr_0p5, 0.5, size=1) +random_st.lognormal(0.5, D_arr_0p5, size=1) +random_st.lognormal(D_arr_like_0p5, 0.5) +random_st.lognormal(0.5, D_arr_like_0p5) +random_st.lognormal(D_arr_0p5, D_arr_0p5) +random_st.lognormal(D_arr_like_0p5, D_arr_like_0p5) +random_st.lognormal(D_arr_0p5, D_arr_0p5, size=1) +random_st.lognormal(D_arr_like_0p5, D_arr_like_0p5, size=1) + +random_st.noncentral_chisquare(0.5, 0.5) +random_st.noncentral_chisquare(0.5, 0.5, size=None) +random_st.noncentral_chisquare(0.5, 0.5, size=1) +random_st.noncentral_chisquare(D_arr_0p5, 0.5) +random_st.noncentral_chisquare(0.5, D_arr_0p5) +random_st.noncentral_chisquare(D_arr_0p5, 0.5, size=1) +random_st.noncentral_chisquare(0.5, D_arr_0p5, size=1) +random_st.noncentral_chisquare(D_arr_like_0p5, 0.5) +random_st.noncentral_chisquare(0.5, D_arr_like_0p5) +random_st.noncentral_chisquare(D_arr_0p5, D_arr_0p5) +random_st.noncentral_chisquare(D_arr_like_0p5, D_arr_like_0p5) +random_st.noncentral_chisquare(D_arr_0p5, D_arr_0p5, size=1) +random_st.noncentral_chisquare(D_arr_like_0p5, D_arr_like_0p5, size=1) + +random_st.normal(0.5, 0.5) +random_st.normal(0.5, 0.5, size=None) +random_st.normal(0.5, 0.5, size=1) +random_st.normal(D_arr_0p5, 0.5) +random_st.normal(0.5, D_arr_0p5) +random_st.normal(D_arr_0p5, 0.5, size=1) +random_st.normal(0.5, D_arr_0p5, size=1) +random_st.normal(D_arr_like_0p5, 0.5) +random_st.normal(0.5, D_arr_like_0p5) +random_st.normal(D_arr_0p5, D_arr_0p5) +random_st.normal(D_arr_like_0p5, D_arr_like_0p5) +random_st.normal(D_arr_0p5, D_arr_0p5, size=1) +random_st.normal(D_arr_like_0p5, D_arr_like_0p5, size=1) + +random_st.triangular(0.1, 0.5, 0.9) +random_st.triangular(0.1, 0.5, 0.9, size=None) +random_st.triangular(0.1, 0.5, 0.9, size=1) +random_st.triangular(D_arr_0p1, 0.5, 0.9) +random_st.triangular(0.1, D_arr_0p5, 0.9) +random_st.triangular(D_arr_0p1, 0.5, D_arr_like_0p9, size=1) +random_st.triangular(0.1, D_arr_0p5, 0.9, size=1) +random_st.triangular(D_arr_like_0p1, 0.5, D_arr_0p9) +random_st.triangular(0.5, D_arr_like_0p5, 0.9) +random_st.triangular(D_arr_0p1, D_arr_0p5, 0.9) +random_st.triangular(D_arr_like_0p1, D_arr_like_0p5, 0.9) +random_st.triangular(D_arr_0p1, D_arr_0p5, D_arr_0p9, size=1) +random_st.triangular(D_arr_like_0p1, D_arr_like_0p5, D_arr_like_0p9, size=1) + +random_st.noncentral_f(0.1, 0.5, 0.9) +random_st.noncentral_f(0.1, 0.5, 0.9, size=None) +random_st.noncentral_f(0.1, 0.5, 0.9, size=1) +random_st.noncentral_f(D_arr_0p1, 0.5, 0.9) +random_st.noncentral_f(0.1, D_arr_0p5, 0.9) +random_st.noncentral_f(D_arr_0p1, 0.5, D_arr_like_0p9, size=1) +random_st.noncentral_f(0.1, D_arr_0p5, 0.9, size=1) +random_st.noncentral_f(D_arr_like_0p1, 0.5, D_arr_0p9) +random_st.noncentral_f(0.5, D_arr_like_0p5, 0.9) +random_st.noncentral_f(D_arr_0p1, D_arr_0p5, 0.9) +random_st.noncentral_f(D_arr_like_0p1, D_arr_like_0p5, 0.9) +random_st.noncentral_f(D_arr_0p1, D_arr_0p5, D_arr_0p9, size=1) +random_st.noncentral_f(D_arr_like_0p1, D_arr_like_0p5, D_arr_like_0p9, size=1) + +random_st.binomial(10, 0.5) +random_st.binomial(10, 0.5, size=None) +random_st.binomial(10, 0.5, size=1) +random_st.binomial(I_arr_10, 0.5) +random_st.binomial(10, D_arr_0p5) +random_st.binomial(I_arr_10, 0.5, size=1) +random_st.binomial(10, D_arr_0p5, size=1) +random_st.binomial(I_arr_like_10, 0.5) +random_st.binomial(10, D_arr_like_0p5) +random_st.binomial(I_arr_10, D_arr_0p5) +random_st.binomial(I_arr_like_10, D_arr_like_0p5) +random_st.binomial(I_arr_10, D_arr_0p5, size=1) +random_st.binomial(I_arr_like_10, D_arr_like_0p5, size=1) + +random_st.negative_binomial(10, 0.5) +random_st.negative_binomial(10, 0.5, size=None) +random_st.negative_binomial(10, 0.5, size=1) +random_st.negative_binomial(I_arr_10, 0.5) +random_st.negative_binomial(10, D_arr_0p5) +random_st.negative_binomial(I_arr_10, 0.5, size=1) +random_st.negative_binomial(10, D_arr_0p5, size=1) +random_st.negative_binomial(I_arr_like_10, 0.5) +random_st.negative_binomial(10, D_arr_like_0p5) +random_st.negative_binomial(I_arr_10, D_arr_0p5) +random_st.negative_binomial(I_arr_like_10, D_arr_like_0p5) +random_st.negative_binomial(I_arr_10, D_arr_0p5, size=1) +random_st.negative_binomial(I_arr_like_10, D_arr_like_0p5, size=1) + +random_st.hypergeometric(20, 20, 10) +random_st.hypergeometric(20, 20, 10, size=None) +random_st.hypergeometric(20, 20, 10, size=1) +random_st.hypergeometric(I_arr_20, 20, 10) +random_st.hypergeometric(20, I_arr_20, 10) +random_st.hypergeometric(I_arr_20, 20, I_arr_like_10, size=1) +random_st.hypergeometric(20, I_arr_20, 10, size=1) +random_st.hypergeometric(I_arr_like_20, 20, I_arr_10) +random_st.hypergeometric(20, I_arr_like_20, 10) +random_st.hypergeometric(I_arr_20, I_arr_20, 10) +random_st.hypergeometric(I_arr_like_20, I_arr_like_20, 10) +random_st.hypergeometric(I_arr_20, I_arr_20, I_arr_10, size=1) +random_st.hypergeometric(I_arr_like_20, I_arr_like_20, I_arr_like_10, size=1) + +random_st.randint(0, 100) +random_st.randint(100) +random_st.randint([100]) +random_st.randint(0, [100]) + +random_st.randint(2, dtype=bool) +random_st.randint(0, 2, dtype=bool) +random_st.randint(I_bool_high_open, dtype=bool) +random_st.randint(I_bool_low, I_bool_high_open, dtype=bool) +random_st.randint(0, I_bool_high_open, dtype=bool) + +random_st.randint(2, dtype=np.bool_) +random_st.randint(0, 2, dtype=np.bool_) +random_st.randint(I_bool_high_open, dtype=np.bool_) +random_st.randint(I_bool_low, I_bool_high_open, dtype=np.bool_) +random_st.randint(0, I_bool_high_open, dtype=np.bool_) + +random_st.randint(256, dtype="u1") +random_st.randint(0, 256, dtype="u1") +random_st.randint(I_u1_high_open, dtype="u1") +random_st.randint(I_u1_low, I_u1_high_open, dtype="u1") +random_st.randint(0, I_u1_high_open, dtype="u1") + +random_st.randint(256, dtype="uint8") +random_st.randint(0, 256, dtype="uint8") +random_st.randint(I_u1_high_open, dtype="uint8") +random_st.randint(I_u1_low, I_u1_high_open, dtype="uint8") +random_st.randint(0, I_u1_high_open, dtype="uint8") + +random_st.randint(256, dtype=np.uint8) +random_st.randint(0, 256, dtype=np.uint8) +random_st.randint(I_u1_high_open, dtype=np.uint8) +random_st.randint(I_u1_low, I_u1_high_open, dtype=np.uint8) +random_st.randint(0, I_u1_high_open, dtype=np.uint8) + +random_st.randint(65536, dtype="u2") +random_st.randint(0, 65536, dtype="u2") +random_st.randint(I_u2_high_open, dtype="u2") +random_st.randint(I_u2_low, I_u2_high_open, dtype="u2") +random_st.randint(0, I_u2_high_open, dtype="u2") + +random_st.randint(65536, dtype="uint16") +random_st.randint(0, 65536, dtype="uint16") +random_st.randint(I_u2_high_open, dtype="uint16") +random_st.randint(I_u2_low, I_u2_high_open, dtype="uint16") +random_st.randint(0, I_u2_high_open, dtype="uint16") + +random_st.randint(65536, dtype=np.uint16) +random_st.randint(0, 65536, dtype=np.uint16) +random_st.randint(I_u2_high_open, dtype=np.uint16) +random_st.randint(I_u2_low, I_u2_high_open, dtype=np.uint16) +random_st.randint(0, I_u2_high_open, dtype=np.uint16) + +random_st.randint(4294967296, dtype="u4") +random_st.randint(0, 4294967296, dtype="u4") +random_st.randint(I_u4_high_open, dtype="u4") +random_st.randint(I_u4_low, I_u4_high_open, dtype="u4") +random_st.randint(0, I_u4_high_open, dtype="u4") + +random_st.randint(4294967296, dtype="uint32") +random_st.randint(0, 4294967296, dtype="uint32") +random_st.randint(I_u4_high_open, dtype="uint32") +random_st.randint(I_u4_low, I_u4_high_open, dtype="uint32") +random_st.randint(0, I_u4_high_open, dtype="uint32") + +random_st.randint(4294967296, dtype=np.uint32) +random_st.randint(0, 4294967296, dtype=np.uint32) +random_st.randint(I_u4_high_open, dtype=np.uint32) +random_st.randint(I_u4_low, I_u4_high_open, dtype=np.uint32) +random_st.randint(0, I_u4_high_open, dtype=np.uint32) + + +random_st.randint(18446744073709551616, dtype="u8") +random_st.randint(0, 18446744073709551616, dtype="u8") +random_st.randint(I_u8_high_open, dtype="u8") +random_st.randint(I_u8_low, I_u8_high_open, dtype="u8") +random_st.randint(0, I_u8_high_open, dtype="u8") + +random_st.randint(18446744073709551616, dtype="uint64") +random_st.randint(0, 18446744073709551616, dtype="uint64") +random_st.randint(I_u8_high_open, dtype="uint64") +random_st.randint(I_u8_low, I_u8_high_open, dtype="uint64") +random_st.randint(0, I_u8_high_open, dtype="uint64") + +random_st.randint(18446744073709551616, dtype=np.uint64) +random_st.randint(0, 18446744073709551616, dtype=np.uint64) +random_st.randint(I_u8_high_open, dtype=np.uint64) +random_st.randint(I_u8_low, I_u8_high_open, dtype=np.uint64) +random_st.randint(0, I_u8_high_open, dtype=np.uint64) + +random_st.randint(128, dtype="i1") +random_st.randint(-128, 128, dtype="i1") +random_st.randint(I_i1_high_open, dtype="i1") +random_st.randint(I_i1_low, I_i1_high_open, dtype="i1") +random_st.randint(-128, I_i1_high_open, dtype="i1") + +random_st.randint(128, dtype="int8") +random_st.randint(-128, 128, dtype="int8") +random_st.randint(I_i1_high_open, dtype="int8") +random_st.randint(I_i1_low, I_i1_high_open, dtype="int8") +random_st.randint(-128, I_i1_high_open, dtype="int8") + +random_st.randint(128, dtype=np.int8) +random_st.randint(-128, 128, dtype=np.int8) +random_st.randint(I_i1_high_open, dtype=np.int8) +random_st.randint(I_i1_low, I_i1_high_open, dtype=np.int8) +random_st.randint(-128, I_i1_high_open, dtype=np.int8) + +random_st.randint(32768, dtype="i2") +random_st.randint(-32768, 32768, dtype="i2") +random_st.randint(I_i2_high_open, dtype="i2") +random_st.randint(I_i2_low, I_i2_high_open, dtype="i2") +random_st.randint(-32768, I_i2_high_open, dtype="i2") +random_st.randint(32768, dtype="int16") +random_st.randint(-32768, 32768, dtype="int16") +random_st.randint(I_i2_high_open, dtype="int16") +random_st.randint(I_i2_low, I_i2_high_open, dtype="int16") +random_st.randint(-32768, I_i2_high_open, dtype="int16") +random_st.randint(32768, dtype=np.int16) +random_st.randint(-32768, 32768, dtype=np.int16) +random_st.randint(I_i2_high_open, dtype=np.int16) +random_st.randint(I_i2_low, I_i2_high_open, dtype=np.int16) +random_st.randint(-32768, I_i2_high_open, dtype=np.int16) + +random_st.randint(2147483648, dtype="i4") +random_st.randint(-2147483648, 2147483648, dtype="i4") +random_st.randint(I_i4_high_open, dtype="i4") +random_st.randint(I_i4_low, I_i4_high_open, dtype="i4") +random_st.randint(-2147483648, I_i4_high_open, dtype="i4") + +random_st.randint(2147483648, dtype="int32") +random_st.randint(-2147483648, 2147483648, dtype="int32") +random_st.randint(I_i4_high_open, dtype="int32") +random_st.randint(I_i4_low, I_i4_high_open, dtype="int32") +random_st.randint(-2147483648, I_i4_high_open, dtype="int32") + +random_st.randint(2147483648, dtype=np.int32) +random_st.randint(-2147483648, 2147483648, dtype=np.int32) +random_st.randint(I_i4_high_open, dtype=np.int32) +random_st.randint(I_i4_low, I_i4_high_open, dtype=np.int32) +random_st.randint(-2147483648, I_i4_high_open, dtype=np.int32) + +random_st.randint(9223372036854775808, dtype="i8") +random_st.randint(-9223372036854775808, 9223372036854775808, dtype="i8") +random_st.randint(I_i8_high_open, dtype="i8") +random_st.randint(I_i8_low, I_i8_high_open, dtype="i8") +random_st.randint(-9223372036854775808, I_i8_high_open, dtype="i8") + +random_st.randint(9223372036854775808, dtype="int64") +random_st.randint(-9223372036854775808, 9223372036854775808, dtype="int64") +random_st.randint(I_i8_high_open, dtype="int64") +random_st.randint(I_i8_low, I_i8_high_open, dtype="int64") +random_st.randint(-9223372036854775808, I_i8_high_open, dtype="int64") + +random_st.randint(9223372036854775808, dtype=np.int64) +random_st.randint(-9223372036854775808, 9223372036854775808, dtype=np.int64) +random_st.randint(I_i8_high_open, dtype=np.int64) +random_st.randint(I_i8_low, I_i8_high_open, dtype=np.int64) +random_st.randint(-9223372036854775808, I_i8_high_open, dtype=np.int64) + +bg: np.random.BitGenerator = random_st._bit_generator + +random_st.bytes(2) + +random_st.choice(5) +random_st.choice(5, 3) +random_st.choice(5, 3, replace=True) +random_st.choice(5, 3, p=[1 / 5] * 5) +random_st.choice(5, 3, p=[1 / 5] * 5, replace=False) + +random_st.choice(["pooh", "rabbit", "piglet", "Christopher"]) +random_st.choice(["pooh", "rabbit", "piglet", "Christopher"], 3) +random_st.choice(["pooh", "rabbit", "piglet", "Christopher"], 3, p=[1 / 4] * 4) +random_st.choice(["pooh", "rabbit", "piglet", "Christopher"], 3, replace=True) +random_st.choice(["pooh", "rabbit", "piglet", "Christopher"], 3, replace=False, p=np.array([1 / 8, 1 / 8, 1 / 2, 1 / 4])) + +random_st.dirichlet([0.5, 0.5]) +random_st.dirichlet(np.array([0.5, 0.5])) +random_st.dirichlet(np.array([0.5, 0.5]), size=3) + +random_st.multinomial(20, [1 / 6.0] * 6) +random_st.multinomial(20, np.array([0.5, 0.5])) +random_st.multinomial(20, [1 / 6.0] * 6, size=2) + +random_st.multivariate_normal([0.0], [[1.0]]) +random_st.multivariate_normal([0.0], np.array([[1.0]])) +random_st.multivariate_normal(np.array([0.0]), [[1.0]]) +random_st.multivariate_normal([0.0], np.array([[1.0]])) + +random_st.permutation(10) +random_st.permutation([1, 2, 3, 4]) +random_st.permutation(np.array([1, 2, 3, 4])) +random_st.permutation(D_2D) + +random_st.shuffle(np.arange(10)) +random_st.shuffle([1, 2, 3, 4, 5]) +random_st.shuffle(D_2D) + +np.random.RandomState(SEED_PCG64) +np.random.RandomState(0) +np.random.RandomState([0, 1, 2]) +random_st.__str__() +random_st.__repr__() +random_st_state = random_st.__getstate__() +random_st.__setstate__(random_st_state) +random_st.seed() +random_st.seed(1) +random_st.seed([0, 1]) +random_st_get_state = random_st.get_state() +random_st_get_state_legacy = random_st.get_state(legacy=True) +random_st.set_state(random_st_get_state) + +random_st.rand() +random_st.rand(1) +random_st.rand(1, 2) +random_st.randn() +random_st.randn(1) +random_st.randn(1, 2) +random_st.random_sample() +random_st.random_sample(1) +random_st.random_sample(size=(1, 2)) + +random_st.tomaxint() +random_st.tomaxint(1) +random_st.tomaxint((1,)) + +np.random.set_bit_generator(SEED_PCG64) +np.random.get_bit_generator() diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/pass/simple.py b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/pass/simple.py new file mode 100644 index 0000000000000000000000000000000000000000..80116870287e4faa58f1640974536ae0ee6250d0 --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/pass/simple.py @@ -0,0 +1,165 @@ +"""Simple expression that should pass with mypy.""" +import operator + +import numpy as np +from collections.abc import Iterable + +# Basic checks +array = np.array([1, 2]) + + +def ndarray_func(x): + # type: (np.ndarray) -> np.ndarray + return x + + +ndarray_func(np.array([1, 2])) +array == 1 +array.dtype == float + +# Dtype construction +np.dtype(float) +np.dtype(np.float64) +np.dtype(None) +np.dtype("float64") +np.dtype(np.dtype(float)) +np.dtype(("U", 10)) +np.dtype((np.int32, (2, 2))) +# Define the arguments on the previous line to prevent bidirectional +# type inference in mypy from broadening the types. +two_tuples_dtype = [("R", "u1"), ("G", "u1"), ("B", "u1")] +np.dtype(two_tuples_dtype) + +three_tuples_dtype = [("R", "u1", 2)] +np.dtype(three_tuples_dtype) + +mixed_tuples_dtype = [("R", "u1"), ("G", np.str_, 1)] +np.dtype(mixed_tuples_dtype) + +shape_tuple_dtype = [("R", "u1", (2, 2))] +np.dtype(shape_tuple_dtype) + +shape_like_dtype = [("R", "u1", (2, 2)), ("G", np.str_, 1)] +np.dtype(shape_like_dtype) + +object_dtype = [("field1", object)] +np.dtype(object_dtype) + +np.dtype((np.int32, (np.int8, 4))) + +# Dtype comparison +np.dtype(float) == float +np.dtype(float) != np.float64 +np.dtype(float) < None +np.dtype(float) <= "float64" +np.dtype(float) > np.dtype(float) +np.dtype(float) >= np.dtype(("U", 10)) + +# Iteration and indexing +def iterable_func(x): + # type: (Iterable) -> Iterable + return x + + +iterable_func(array) +[element for element in array] +iter(array) +zip(array, array) +array[1] +array[:] +array[...] +array[:] = 0 + +array_2d = np.ones((3, 3)) +array_2d[:2, :2] +array_2d[..., 0] +array_2d[:2, :2] = 0 + +# Other special methods +len(array) +str(array) +array_scalar = np.array(1) +int(array_scalar) +float(array_scalar) +# currently does not work due to https://github.com/python/typeshed/issues/1904 +# complex(array_scalar) +bytes(array_scalar) +operator.index(array_scalar) +bool(array_scalar) + +# comparisons +array < 1 +array <= 1 +array == 1 +array != 1 +array > 1 +array >= 1 +1 < array +1 <= array +1 == array +1 != array +1 > array +1 >= array + +# binary arithmetic +array + 1 +1 + array +array += 1 + +array - 1 +1 - array +array -= 1 + +array * 1 +1 * array +array *= 1 + +nonzero_array = np.array([1, 2]) +array / 1 +1 / nonzero_array +float_array = np.array([1.0, 2.0]) +float_array /= 1 + +array // 1 +1 // nonzero_array +array //= 1 + +array % 1 +1 % nonzero_array +array %= 1 + +divmod(array, 1) +divmod(1, nonzero_array) + +array ** 1 +1 ** array +array **= 1 + +array << 1 +1 << array +array <<= 1 + +array >> 1 +1 >> array +array >>= 1 + +array & 1 +1 & array +array &= 1 + +array ^ 1 +1 ^ array +array ^= 1 + +array | 1 +1 | array +array |= 1 + +# unary arithmetic +-array ++array +abs(array) +~array + +# Other methods +np.array([1, 2]).transpose() diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/arithmetic.pyi b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/arithmetic.pyi new file mode 100644 index 0000000000000000000000000000000000000000..6291fda6cefceeea0129e4006d1cf77c2e92d609 --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/arithmetic.pyi @@ -0,0 +1,516 @@ +import sys +from typing import Any + +import numpy as np +import numpy.typing as npt +from numpy._typing import _32Bit,_64Bit, _128Bit + +if sys.version_info >= (3, 11): + from typing import assert_type +else: + from typing_extensions import assert_type + +# Can't directly import `np.float128` as it is not available on all platforms +f16: np.floating[_128Bit] + +c16 = np.complex128() +f8 = np.float64() +i8 = np.int64() +u8 = np.uint64() + +c8 = np.complex64() +f4 = np.float32() +i4 = np.int32() +u4 = np.uint32() + +dt = np.datetime64(0, "D") +td = np.timedelta64(0, "D") + +b_ = np.bool_() + +b = bool() +c = complex() +f = float() +i = int() + +AR_b: npt.NDArray[np.bool_] +AR_u: npt.NDArray[np.uint32] +AR_i: npt.NDArray[np.int64] +AR_f: npt.NDArray[np.float64] +AR_c: npt.NDArray[np.complex128] +AR_m: npt.NDArray[np.timedelta64] +AR_M: npt.NDArray[np.datetime64] +AR_O: npt.NDArray[np.object_] +AR_number: npt.NDArray[np.number[Any]] + +AR_LIKE_b: list[bool] +AR_LIKE_u: list[np.uint32] +AR_LIKE_i: list[int] +AR_LIKE_f: list[float] +AR_LIKE_c: list[complex] +AR_LIKE_m: list[np.timedelta64] +AR_LIKE_M: list[np.datetime64] +AR_LIKE_O: list[np.object_] + +# Array subtraction + +assert_type(AR_number - AR_number, npt.NDArray[np.number[Any]]) + +assert_type(AR_b - AR_LIKE_u, npt.NDArray[np.unsignedinteger[Any]]) +assert_type(AR_b - AR_LIKE_i, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_b - AR_LIKE_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_b - AR_LIKE_c, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_b - AR_LIKE_m, npt.NDArray[np.timedelta64]) +assert_type(AR_b - AR_LIKE_O, Any) + +assert_type(AR_LIKE_u - AR_b, npt.NDArray[np.unsignedinteger[Any]]) +assert_type(AR_LIKE_i - AR_b, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_LIKE_f - AR_b, npt.NDArray[np.floating[Any]]) +assert_type(AR_LIKE_c - AR_b, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_LIKE_m - AR_b, npt.NDArray[np.timedelta64]) +assert_type(AR_LIKE_M - AR_b, npt.NDArray[np.datetime64]) +assert_type(AR_LIKE_O - AR_b, Any) + +assert_type(AR_u - AR_LIKE_b, npt.NDArray[np.unsignedinteger[Any]]) +assert_type(AR_u - AR_LIKE_u, npt.NDArray[np.unsignedinteger[Any]]) +assert_type(AR_u - AR_LIKE_i, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_u - AR_LIKE_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_u - AR_LIKE_c, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_u - AR_LIKE_m, npt.NDArray[np.timedelta64]) +assert_type(AR_u - AR_LIKE_O, Any) + +assert_type(AR_LIKE_b - AR_u, npt.NDArray[np.unsignedinteger[Any]]) +assert_type(AR_LIKE_u - AR_u, npt.NDArray[np.unsignedinteger[Any]]) +assert_type(AR_LIKE_i - AR_u, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_LIKE_f - AR_u, npt.NDArray[np.floating[Any]]) +assert_type(AR_LIKE_c - AR_u, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_LIKE_m - AR_u, npt.NDArray[np.timedelta64]) +assert_type(AR_LIKE_M - AR_u, npt.NDArray[np.datetime64]) +assert_type(AR_LIKE_O - AR_u, Any) + +assert_type(AR_i - AR_LIKE_b, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_i - AR_LIKE_u, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_i - AR_LIKE_i, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_i - AR_LIKE_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_i - AR_LIKE_c, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_i - AR_LIKE_m, npt.NDArray[np.timedelta64]) +assert_type(AR_i - AR_LIKE_O, Any) + +assert_type(AR_LIKE_b - AR_i, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_LIKE_u - AR_i, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_LIKE_i - AR_i, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_LIKE_f - AR_i, npt.NDArray[np.floating[Any]]) +assert_type(AR_LIKE_c - AR_i, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_LIKE_m - AR_i, npt.NDArray[np.timedelta64]) +assert_type(AR_LIKE_M - AR_i, npt.NDArray[np.datetime64]) +assert_type(AR_LIKE_O - AR_i, Any) + +assert_type(AR_f - AR_LIKE_b, npt.NDArray[np.floating[Any]]) +assert_type(AR_f - AR_LIKE_u, npt.NDArray[np.floating[Any]]) +assert_type(AR_f - AR_LIKE_i, npt.NDArray[np.floating[Any]]) +assert_type(AR_f - AR_LIKE_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_f - AR_LIKE_c, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_f - AR_LIKE_O, Any) + +assert_type(AR_LIKE_b - AR_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_LIKE_u - AR_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_LIKE_i - AR_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_LIKE_f - AR_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_LIKE_c - AR_f, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_LIKE_O - AR_f, Any) + +assert_type(AR_c - AR_LIKE_b, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_c - AR_LIKE_u, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_c - AR_LIKE_i, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_c - AR_LIKE_f, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_c - AR_LIKE_c, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_c - AR_LIKE_O, Any) + +assert_type(AR_LIKE_b - AR_c, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_LIKE_u - AR_c, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_LIKE_i - AR_c, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_LIKE_f - AR_c, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_LIKE_c - AR_c, npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(AR_LIKE_O - AR_c, Any) + +assert_type(AR_m - AR_LIKE_b, npt.NDArray[np.timedelta64]) +assert_type(AR_m - AR_LIKE_u, npt.NDArray[np.timedelta64]) +assert_type(AR_m - AR_LIKE_i, npt.NDArray[np.timedelta64]) +assert_type(AR_m - AR_LIKE_m, npt.NDArray[np.timedelta64]) +assert_type(AR_m - AR_LIKE_O, Any) + +assert_type(AR_LIKE_b - AR_m, npt.NDArray[np.timedelta64]) +assert_type(AR_LIKE_u - AR_m, npt.NDArray[np.timedelta64]) +assert_type(AR_LIKE_i - AR_m, npt.NDArray[np.timedelta64]) +assert_type(AR_LIKE_m - AR_m, npt.NDArray[np.timedelta64]) +assert_type(AR_LIKE_M - AR_m, npt.NDArray[np.datetime64]) +assert_type(AR_LIKE_O - AR_m, Any) + +assert_type(AR_M - AR_LIKE_b, npt.NDArray[np.datetime64]) +assert_type(AR_M - AR_LIKE_u, npt.NDArray[np.datetime64]) +assert_type(AR_M - AR_LIKE_i, npt.NDArray[np.datetime64]) +assert_type(AR_M - AR_LIKE_m, npt.NDArray[np.datetime64]) +assert_type(AR_M - AR_LIKE_M, npt.NDArray[np.timedelta64]) +assert_type(AR_M - AR_LIKE_O, Any) + +assert_type(AR_LIKE_M - AR_M, npt.NDArray[np.timedelta64]) +assert_type(AR_LIKE_O - AR_M, Any) + +assert_type(AR_O - AR_LIKE_b, Any) +assert_type(AR_O - AR_LIKE_u, Any) +assert_type(AR_O - AR_LIKE_i, Any) +assert_type(AR_O - AR_LIKE_f, Any) +assert_type(AR_O - AR_LIKE_c, Any) +assert_type(AR_O - AR_LIKE_m, Any) +assert_type(AR_O - AR_LIKE_M, Any) +assert_type(AR_O - AR_LIKE_O, Any) + +assert_type(AR_LIKE_b - AR_O, Any) +assert_type(AR_LIKE_u - AR_O, Any) +assert_type(AR_LIKE_i - AR_O, Any) +assert_type(AR_LIKE_f - AR_O, Any) +assert_type(AR_LIKE_c - AR_O, Any) +assert_type(AR_LIKE_m - AR_O, Any) +assert_type(AR_LIKE_M - AR_O, Any) +assert_type(AR_LIKE_O - AR_O, Any) + +# Array floor division + +assert_type(AR_b // AR_LIKE_b, npt.NDArray[np.int8]) +assert_type(AR_b // AR_LIKE_u, npt.NDArray[np.unsignedinteger[Any]]) +assert_type(AR_b // AR_LIKE_i, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_b // AR_LIKE_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_b // AR_LIKE_O, Any) + +assert_type(AR_LIKE_b // AR_b, npt.NDArray[np.int8]) +assert_type(AR_LIKE_u // AR_b, npt.NDArray[np.unsignedinteger[Any]]) +assert_type(AR_LIKE_i // AR_b, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_LIKE_f // AR_b, npt.NDArray[np.floating[Any]]) +assert_type(AR_LIKE_O // AR_b, Any) + +assert_type(AR_u // AR_LIKE_b, npt.NDArray[np.unsignedinteger[Any]]) +assert_type(AR_u // AR_LIKE_u, npt.NDArray[np.unsignedinteger[Any]]) +assert_type(AR_u // AR_LIKE_i, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_u // AR_LIKE_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_u // AR_LIKE_O, Any) + +assert_type(AR_LIKE_b // AR_u, npt.NDArray[np.unsignedinteger[Any]]) +assert_type(AR_LIKE_u // AR_u, npt.NDArray[np.unsignedinteger[Any]]) +assert_type(AR_LIKE_i // AR_u, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_LIKE_f // AR_u, npt.NDArray[np.floating[Any]]) +assert_type(AR_LIKE_m // AR_u, npt.NDArray[np.timedelta64]) +assert_type(AR_LIKE_O // AR_u, Any) + +assert_type(AR_i // AR_LIKE_b, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_i // AR_LIKE_u, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_i // AR_LIKE_i, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_i // AR_LIKE_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_i // AR_LIKE_O, Any) + +assert_type(AR_LIKE_b // AR_i, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_LIKE_u // AR_i, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_LIKE_i // AR_i, npt.NDArray[np.signedinteger[Any]]) +assert_type(AR_LIKE_f // AR_i, npt.NDArray[np.floating[Any]]) +assert_type(AR_LIKE_m // AR_i, npt.NDArray[np.timedelta64]) +assert_type(AR_LIKE_O // AR_i, Any) + +assert_type(AR_f // AR_LIKE_b, npt.NDArray[np.floating[Any]]) +assert_type(AR_f // AR_LIKE_u, npt.NDArray[np.floating[Any]]) +assert_type(AR_f // AR_LIKE_i, npt.NDArray[np.floating[Any]]) +assert_type(AR_f // AR_LIKE_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_f // AR_LIKE_O, Any) + +assert_type(AR_LIKE_b // AR_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_LIKE_u // AR_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_LIKE_i // AR_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_LIKE_f // AR_f, npt.NDArray[np.floating[Any]]) +assert_type(AR_LIKE_m // AR_f, npt.NDArray[np.timedelta64]) +assert_type(AR_LIKE_O // AR_f, Any) + +assert_type(AR_m // AR_LIKE_u, npt.NDArray[np.timedelta64]) +assert_type(AR_m // AR_LIKE_i, npt.NDArray[np.timedelta64]) +assert_type(AR_m // AR_LIKE_f, npt.NDArray[np.timedelta64]) +assert_type(AR_m // AR_LIKE_m, npt.NDArray[np.int64]) +assert_type(AR_m // AR_LIKE_O, Any) + +assert_type(AR_LIKE_m // AR_m, npt.NDArray[np.int64]) +assert_type(AR_LIKE_O // AR_m, Any) + +assert_type(AR_O // AR_LIKE_b, Any) +assert_type(AR_O // AR_LIKE_u, Any) +assert_type(AR_O // AR_LIKE_i, Any) +assert_type(AR_O // AR_LIKE_f, Any) +assert_type(AR_O // AR_LIKE_m, Any) +assert_type(AR_O // AR_LIKE_M, Any) +assert_type(AR_O // AR_LIKE_O, Any) + +assert_type(AR_LIKE_b // AR_O, Any) +assert_type(AR_LIKE_u // AR_O, Any) +assert_type(AR_LIKE_i // AR_O, Any) +assert_type(AR_LIKE_f // AR_O, Any) +assert_type(AR_LIKE_m // AR_O, Any) +assert_type(AR_LIKE_M // AR_O, Any) +assert_type(AR_LIKE_O // AR_O, Any) + +# unary ops + +assert_type(-f16, np.floating[_128Bit]) +assert_type(-c16, np.complex128) +assert_type(-c8, np.complex64) +assert_type(-f8, np.float64) +assert_type(-f4, np.float32) +assert_type(-i8, np.int64) +assert_type(-i4, np.int32) +assert_type(-u8, np.uint64) +assert_type(-u4, np.uint32) +assert_type(-td, np.timedelta64) +assert_type(-AR_f, npt.NDArray[np.float64]) + +assert_type(+f16, np.floating[_128Bit]) +assert_type(+c16, np.complex128) +assert_type(+c8, np.complex64) +assert_type(+f8, np.float64) +assert_type(+f4, np.float32) +assert_type(+i8, np.int64) +assert_type(+i4, np.int32) +assert_type(+u8, np.uint64) +assert_type(+u4, np.uint32) +assert_type(+td, np.timedelta64) +assert_type(+AR_f, npt.NDArray[np.float64]) + +assert_type(abs(f16), np.floating[_128Bit]) +assert_type(abs(c16), np.float64) +assert_type(abs(c8), np.float32) +assert_type(abs(f8), np.float64) +assert_type(abs(f4), np.float32) +assert_type(abs(i8), np.int64) +assert_type(abs(i4), np.int32) +assert_type(abs(u8), np.uint64) +assert_type(abs(u4), np.uint32) +assert_type(abs(td), np.timedelta64) +assert_type(abs(b_), np.bool_) + +# Time structures + +assert_type(dt + td, np.datetime64) +assert_type(dt + i, np.datetime64) +assert_type(dt + i4, np.datetime64) +assert_type(dt + i8, np.datetime64) +assert_type(dt - dt, np.timedelta64) +assert_type(dt - i, np.datetime64) +assert_type(dt - i4, np.datetime64) +assert_type(dt - i8, np.datetime64) + +assert_type(td + td, np.timedelta64) +assert_type(td + i, np.timedelta64) +assert_type(td + i4, np.timedelta64) +assert_type(td + i8, np.timedelta64) +assert_type(td - td, np.timedelta64) +assert_type(td - i, np.timedelta64) +assert_type(td - i4, np.timedelta64) +assert_type(td - i8, np.timedelta64) +assert_type(td / f, np.timedelta64) +assert_type(td / f4, np.timedelta64) +assert_type(td / f8, np.timedelta64) +assert_type(td / td, np.float64) +assert_type(td // td, np.int64) + +# boolean + +assert_type(b_ / b, np.float64) +assert_type(b_ / b_, np.float64) +assert_type(b_ / i, np.float64) +assert_type(b_ / i8, np.float64) +assert_type(b_ / i4, np.float64) +assert_type(b_ / u8, np.float64) +assert_type(b_ / u4, np.float64) +assert_type(b_ / f, np.float64) +assert_type(b_ / f16, np.floating[_128Bit]) +assert_type(b_ / f8, np.float64) +assert_type(b_ / f4, np.float32) +assert_type(b_ / c, np.complex128) +assert_type(b_ / c16, np.complex128) +assert_type(b_ / c8, np.complex64) + +assert_type(b / b_, np.float64) +assert_type(b_ / b_, np.float64) +assert_type(i / b_, np.float64) +assert_type(i8 / b_, np.float64) +assert_type(i4 / b_, np.float64) +assert_type(u8 / b_, np.float64) +assert_type(u4 / b_, np.float64) +assert_type(f / b_, np.float64) +assert_type(f16 / b_, np.floating[_128Bit]) +assert_type(f8 / b_, np.float64) +assert_type(f4 / b_, np.float32) +assert_type(c / b_, np.complex128) +assert_type(c16 / b_, np.complex128) +assert_type(c8 / b_, np.complex64) + +# Complex + +assert_type(c16 + f16, np.complexfloating[_64Bit | _128Bit, _64Bit | _128Bit]) +assert_type(c16 + c16, np.complex128) +assert_type(c16 + f8, np.complex128) +assert_type(c16 + i8, np.complex128) +assert_type(c16 + c8, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(c16 + f4, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(c16 + i4, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(c16 + b_, np.complex128) +assert_type(c16 + b, np.complex128) +assert_type(c16 + c, np.complex128) +assert_type(c16 + f, np.complex128) +assert_type(c16 + AR_f, npt.NDArray[np.complexfloating[Any, Any]]) + +assert_type(f16 + c16, np.complexfloating[_64Bit | _128Bit, _64Bit | _128Bit]) +assert_type(c16 + c16, np.complex128) +assert_type(f8 + c16, np.complex128) +assert_type(i8 + c16, np.complex128) +assert_type(c8 + c16, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(f4 + c16, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(i4 + c16, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(b_ + c16, np.complex128) +assert_type(b + c16, np.complex128) +assert_type(c + c16, np.complex128) +assert_type(f + c16, np.complex128) +assert_type(AR_f + c16, npt.NDArray[np.complexfloating[Any, Any]]) + +assert_type(c8 + f16, np.complexfloating[_32Bit | _128Bit, _32Bit | _128Bit]) +assert_type(c8 + c16, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(c8 + f8, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(c8 + i8, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(c8 + c8, np.complex64) +assert_type(c8 + f4, np.complex64) +assert_type(c8 + i4, np.complex64) +assert_type(c8 + b_, np.complex64) +assert_type(c8 + b, np.complex64) +assert_type(c8 + c, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(c8 + f, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(c8 + AR_f, npt.NDArray[np.complexfloating[Any, Any]]) + +assert_type(f16 + c8, np.complexfloating[_32Bit | _128Bit, _32Bit | _128Bit]) +assert_type(c16 + c8, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(f8 + c8, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(i8 + c8, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(c8 + c8, np.complex64) +assert_type(f4 + c8, np.complex64) +assert_type(i4 + c8, np.complex64) +assert_type(b_ + c8, np.complex64) +assert_type(b + c8, np.complex64) +assert_type(c + c8, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(f + c8, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(AR_f + c8, npt.NDArray[np.complexfloating[Any, Any]]) + +# Float + +assert_type(f8 + f16, np.floating[_64Bit | _128Bit]) +assert_type(f8 + f8, np.float64) +assert_type(f8 + i8, np.float64) +assert_type(f8 + f4, np.floating[_32Bit | _64Bit]) +assert_type(f8 + i4, np.floating[_32Bit | _64Bit]) +assert_type(f8 + b_, np.float64) +assert_type(f8 + b, np.float64) +assert_type(f8 + c, np.complex128) +assert_type(f8 + f, np.float64) +assert_type(f8 + AR_f, npt.NDArray[np.floating[Any]]) + +assert_type(f16 + f8, np.floating[_64Bit | _128Bit]) +assert_type(f8 + f8, np.float64) +assert_type(i8 + f8, np.float64) +assert_type(f4 + f8, np.floating[_32Bit | _64Bit]) +assert_type(i4 + f8, np.floating[_32Bit | _64Bit]) +assert_type(b_ + f8, np.float64) +assert_type(b + f8, np.float64) +assert_type(c + f8, np.complex128) +assert_type(f + f8, np.float64) +assert_type(AR_f + f8, npt.NDArray[np.floating[Any]]) + +assert_type(f4 + f16, np.floating[_32Bit | _128Bit]) +assert_type(f4 + f8, np.floating[_32Bit | _64Bit]) +assert_type(f4 + i8, np.floating[_32Bit | _64Bit]) +assert_type(f4 + f4, np.float32) +assert_type(f4 + i4, np.float32) +assert_type(f4 + b_, np.float32) +assert_type(f4 + b, np.float32) +assert_type(f4 + c, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(f4 + f, np.floating[_32Bit | _64Bit]) +assert_type(f4 + AR_f, npt.NDArray[np.floating[Any]]) + +assert_type(f16 + f4, np.floating[_32Bit | _128Bit]) +assert_type(f8 + f4, np.floating[_32Bit | _64Bit]) +assert_type(i8 + f4, np.floating[_32Bit | _64Bit]) +assert_type(f4 + f4, np.float32) +assert_type(i4 + f4, np.float32) +assert_type(b_ + f4, np.float32) +assert_type(b + f4, np.float32) +assert_type(c + f4, np.complexfloating[_32Bit | _64Bit, _32Bit | _64Bit]) +assert_type(f + f4, np.floating[_32Bit | _64Bit]) +assert_type(AR_f + f4, npt.NDArray[np.floating[Any]]) + +# Int + +assert_type(i8 + i8, np.int64) +assert_type(i8 + u8, Any) +assert_type(i8 + i4, np.signedinteger[_32Bit | _64Bit]) +assert_type(i8 + u4, Any) +assert_type(i8 + b_, np.int64) +assert_type(i8 + b, np.int64) +assert_type(i8 + c, np.complex128) +assert_type(i8 + f, np.float64) +assert_type(i8 + AR_f, npt.NDArray[np.floating[Any]]) + +assert_type(u8 + u8, np.uint64) +assert_type(u8 + i4, Any) +assert_type(u8 + u4, np.unsignedinteger[_32Bit | _64Bit]) +assert_type(u8 + b_, np.uint64) +assert_type(u8 + b, np.uint64) +assert_type(u8 + c, np.complex128) +assert_type(u8 + f, np.float64) +assert_type(u8 + AR_f, npt.NDArray[np.floating[Any]]) + +assert_type(i8 + i8, np.int64) +assert_type(u8 + i8, Any) +assert_type(i4 + i8, np.signedinteger[_32Bit | _64Bit]) +assert_type(u4 + i8, Any) +assert_type(b_ + i8, np.int64) +assert_type(b + i8, np.int64) +assert_type(c + i8, np.complex128) +assert_type(f + i8, np.float64) +assert_type(AR_f + i8, npt.NDArray[np.floating[Any]]) + +assert_type(u8 + u8, np.uint64) +assert_type(i4 + u8, Any) +assert_type(u4 + u8, np.unsignedinteger[_32Bit | _64Bit]) +assert_type(b_ + u8, np.uint64) +assert_type(b + u8, np.uint64) +assert_type(c + u8, np.complex128) +assert_type(f + u8, np.float64) +assert_type(AR_f + u8, npt.NDArray[np.floating[Any]]) + +assert_type(i4 + i8, np.signedinteger[_32Bit | _64Bit]) +assert_type(i4 + i4, np.int32) +assert_type(i4 + b_, np.int32) +assert_type(i4 + b, np.int32) +assert_type(i4 + AR_f, npt.NDArray[np.floating[Any]]) + +assert_type(u4 + i8, Any) +assert_type(u4 + i4, Any) +assert_type(u4 + u8, np.unsignedinteger[_32Bit | _64Bit]) +assert_type(u4 + u4, np.uint32) +assert_type(u4 + b_, np.uint32) +assert_type(u4 + b, np.uint32) +assert_type(u4 + AR_f, npt.NDArray[np.floating[Any]]) + +assert_type(i8 + i4, np.signedinteger[_32Bit | _64Bit]) +assert_type(i4 + i4, np.int32) +assert_type(b_ + i4, np.int32) +assert_type(b + i4, np.int32) +assert_type(AR_f + i4, npt.NDArray[np.floating[Any]]) + +assert_type(i8 + u4, Any) +assert_type(i4 + u4, Any) +assert_type(u8 + u4, np.unsignedinteger[_32Bit | _64Bit]) +assert_type(u4 + u4, np.uint32) +assert_type(b_ + u4, np.uint32) +assert_type(b + u4, np.uint32) +assert_type(AR_f + u4, npt.NDArray[np.floating[Any]]) diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/array_constructors.pyi b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/array_constructors.pyi new file mode 100644 index 0000000000000000000000000000000000000000..0bfbc63093a331accc4339347b48004aec683c9f --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/array_constructors.pyi @@ -0,0 +1,221 @@ +import sys +from typing import Any, TypeVar +from pathlib import Path +from collections import deque + +import numpy as np +import numpy.typing as npt + +if sys.version_info >= (3, 11): + from typing import assert_type +else: + from typing_extensions import assert_type + +_SCT = TypeVar("_SCT", bound=np.generic, covariant=True) + +class SubClass(np.ndarray[Any, np.dtype[_SCT]]): ... + +i8: np.int64 + +A: npt.NDArray[np.float64] +B: SubClass[np.float64] +C: list[int] + +def func(i: int, j: int, **kwargs: Any) -> SubClass[np.float64]: ... + +assert_type(np.empty_like(A), npt.NDArray[np.float64]) +assert_type(np.empty_like(B), SubClass[np.float64]) +assert_type(np.empty_like([1, 1.0]), npt.NDArray[Any]) +assert_type(np.empty_like(A, dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.empty_like(A, dtype='c16'), npt.NDArray[Any]) + +assert_type(np.array(A), npt.NDArray[np.float64]) +assert_type(np.array(B), npt.NDArray[np.float64]) +assert_type(np.array(B, subok=True), SubClass[np.float64]) +assert_type(np.array([1, 1.0]), npt.NDArray[Any]) +assert_type(np.array(deque([1, 2, 3])), npt.NDArray[Any]) +assert_type(np.array(A, dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.array(A, dtype='c16'), npt.NDArray[Any]) +assert_type(np.array(A, like=A), npt.NDArray[np.float64]) + +assert_type(np.zeros([1, 5, 6]), npt.NDArray[np.float64]) +assert_type(np.zeros([1, 5, 6], dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.zeros([1, 5, 6], dtype='c16'), npt.NDArray[Any]) + +assert_type(np.empty([1, 5, 6]), npt.NDArray[np.float64]) +assert_type(np.empty([1, 5, 6], dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.empty([1, 5, 6], dtype='c16'), npt.NDArray[Any]) + +assert_type(np.concatenate(A), npt.NDArray[np.float64]) +assert_type(np.concatenate([A, A]), Any) +assert_type(np.concatenate([[1], A]), npt.NDArray[Any]) +assert_type(np.concatenate([[1], [1]]), npt.NDArray[Any]) +assert_type(np.concatenate((A, A)), npt.NDArray[np.float64]) +assert_type(np.concatenate(([1], [1])), npt.NDArray[Any]) +assert_type(np.concatenate([1, 1.0]), npt.NDArray[Any]) +assert_type(np.concatenate(A, dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.concatenate(A, dtype='c16'), npt.NDArray[Any]) +assert_type(np.concatenate([1, 1.0], out=A), npt.NDArray[np.float64]) + +assert_type(np.asarray(A), npt.NDArray[np.float64]) +assert_type(np.asarray(B), npt.NDArray[np.float64]) +assert_type(np.asarray([1, 1.0]), npt.NDArray[Any]) +assert_type(np.asarray(A, dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.asarray(A, dtype='c16'), npt.NDArray[Any]) + +assert_type(np.asanyarray(A), npt.NDArray[np.float64]) +assert_type(np.asanyarray(B), SubClass[np.float64]) +assert_type(np.asanyarray([1, 1.0]), npt.NDArray[Any]) +assert_type(np.asanyarray(A, dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.asanyarray(A, dtype='c16'), npt.NDArray[Any]) + +assert_type(np.ascontiguousarray(A), npt.NDArray[np.float64]) +assert_type(np.ascontiguousarray(B), npt.NDArray[np.float64]) +assert_type(np.ascontiguousarray([1, 1.0]), npt.NDArray[Any]) +assert_type(np.ascontiguousarray(A, dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.ascontiguousarray(A, dtype='c16'), npt.NDArray[Any]) + +assert_type(np.asfortranarray(A), npt.NDArray[np.float64]) +assert_type(np.asfortranarray(B), npt.NDArray[np.float64]) +assert_type(np.asfortranarray([1, 1.0]), npt.NDArray[Any]) +assert_type(np.asfortranarray(A, dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.asfortranarray(A, dtype='c16'), npt.NDArray[Any]) + +assert_type(np.fromstring("1 1 1", sep=" "), npt.NDArray[np.float64]) +assert_type(np.fromstring(b"1 1 1", sep=" "), npt.NDArray[np.float64]) +assert_type(np.fromstring("1 1 1", dtype=np.int64, sep=" "), npt.NDArray[np.int64]) +assert_type(np.fromstring(b"1 1 1", dtype=np.int64, sep=" "), npt.NDArray[np.int64]) +assert_type(np.fromstring("1 1 1", dtype="c16", sep=" "), npt.NDArray[Any]) +assert_type(np.fromstring(b"1 1 1", dtype="c16", sep=" "), npt.NDArray[Any]) + +assert_type(np.fromfile("test.txt", sep=" "), npt.NDArray[np.float64]) +assert_type(np.fromfile("test.txt", dtype=np.int64, sep=" "), npt.NDArray[np.int64]) +assert_type(np.fromfile("test.txt", dtype="c16", sep=" "), npt.NDArray[Any]) +with open("test.txt") as f: + assert_type(np.fromfile(f, sep=" "), npt.NDArray[np.float64]) + assert_type(np.fromfile(b"test.txt", sep=" "), npt.NDArray[np.float64]) + assert_type(np.fromfile(Path("test.txt"), sep=" "), npt.NDArray[np.float64]) + +assert_type(np.fromiter("12345", np.float64), npt.NDArray[np.float64]) +assert_type(np.fromiter("12345", float), npt.NDArray[Any]) + +assert_type(np.frombuffer(A), npt.NDArray[np.float64]) +assert_type(np.frombuffer(A, dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.frombuffer(A, dtype="c16"), npt.NDArray[Any]) + +assert_type(np.arange(False, True), npt.NDArray[np.signedinteger[Any]]) +assert_type(np.arange(10), npt.NDArray[np.signedinteger[Any]]) +assert_type(np.arange(0, 10, step=2), npt.NDArray[np.signedinteger[Any]]) +assert_type(np.arange(10.0), npt.NDArray[np.floating[Any]]) +assert_type(np.arange(start=0, stop=10.0), npt.NDArray[np.floating[Any]]) +assert_type(np.arange(np.timedelta64(0)), npt.NDArray[np.timedelta64]) +assert_type(np.arange(0, np.timedelta64(10)), npt.NDArray[np.timedelta64]) +assert_type(np.arange(np.datetime64("0"), np.datetime64("10")), npt.NDArray[np.datetime64]) +assert_type(np.arange(10, dtype=np.float64), npt.NDArray[np.float64]) +assert_type(np.arange(0, 10, step=2, dtype=np.int16), npt.NDArray[np.int16]) +assert_type(np.arange(10, dtype=int), npt.NDArray[Any]) +assert_type(np.arange(0, 10, dtype="f8"), npt.NDArray[Any]) + +assert_type(np.require(A), npt.NDArray[np.float64]) +assert_type(np.require(B), SubClass[np.float64]) +assert_type(np.require(B, requirements=None), SubClass[np.float64]) +assert_type(np.require(B, dtype=int), np.ndarray[Any, Any]) +assert_type(np.require(B, requirements="E"), np.ndarray[Any, Any]) +assert_type(np.require(B, requirements=["ENSUREARRAY"]), np.ndarray[Any, Any]) +assert_type(np.require(B, requirements={"F", "E"}), np.ndarray[Any, Any]) +assert_type(np.require(B, requirements=["C", "OWNDATA"]), SubClass[np.float64]) +assert_type(np.require(B, requirements="W"), SubClass[np.float64]) +assert_type(np.require(B, requirements="A"), SubClass[np.float64]) +assert_type(np.require(C), np.ndarray[Any, Any]) + +assert_type(np.linspace(0, 10), npt.NDArray[np.floating[Any]]) +assert_type(np.linspace(0, 10j), npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(np.linspace(0, 10, dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.linspace(0, 10, dtype=int), npt.NDArray[Any]) +assert_type(np.linspace(0, 10, retstep=True), tuple[npt.NDArray[np.floating[Any]], np.floating[Any]]) +assert_type(np.linspace(0j, 10, retstep=True), tuple[npt.NDArray[np.complexfloating[Any, Any]], np.complexfloating[Any, Any]]) +assert_type(np.linspace(0, 10, retstep=True, dtype=np.int64), tuple[npt.NDArray[np.int64], np.int64]) +assert_type(np.linspace(0j, 10, retstep=True, dtype=int), tuple[npt.NDArray[Any], Any]) + +assert_type(np.logspace(0, 10), npt.NDArray[np.floating[Any]]) +assert_type(np.logspace(0, 10j), npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(np.logspace(0, 10, dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.logspace(0, 10, dtype=int), npt.NDArray[Any]) + +assert_type(np.geomspace(0, 10), npt.NDArray[np.floating[Any]]) +assert_type(np.geomspace(0, 10j), npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(np.geomspace(0, 10, dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.geomspace(0, 10, dtype=int), npt.NDArray[Any]) + +assert_type(np.zeros_like(A), npt.NDArray[np.float64]) +assert_type(np.zeros_like(C), npt.NDArray[Any]) +assert_type(np.zeros_like(A, dtype=float), npt.NDArray[Any]) +assert_type(np.zeros_like(B), SubClass[np.float64]) +assert_type(np.zeros_like(B, dtype=np.int64), npt.NDArray[np.int64]) + +assert_type(np.ones_like(A), npt.NDArray[np.float64]) +assert_type(np.ones_like(C), npt.NDArray[Any]) +assert_type(np.ones_like(A, dtype=float), npt.NDArray[Any]) +assert_type(np.ones_like(B), SubClass[np.float64]) +assert_type(np.ones_like(B, dtype=np.int64), npt.NDArray[np.int64]) + +assert_type(np.full_like(A, i8), npt.NDArray[np.float64]) +assert_type(np.full_like(C, i8), npt.NDArray[Any]) +assert_type(np.full_like(A, i8, dtype=int), npt.NDArray[Any]) +assert_type(np.full_like(B, i8), SubClass[np.float64]) +assert_type(np.full_like(B, i8, dtype=np.int64), npt.NDArray[np.int64]) + +assert_type(np.ones(1), npt.NDArray[np.float64]) +assert_type(np.ones([1, 1, 1]), npt.NDArray[np.float64]) +assert_type(np.ones(5, dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.ones(5, dtype=int), npt.NDArray[Any]) + +assert_type(np.full(1, i8), npt.NDArray[Any]) +assert_type(np.full([1, 1, 1], i8), npt.NDArray[Any]) +assert_type(np.full(1, i8, dtype=np.float64), npt.NDArray[np.float64]) +assert_type(np.full(1, i8, dtype=float), npt.NDArray[Any]) + +assert_type(np.indices([1, 2, 3]), npt.NDArray[np.int_]) +assert_type(np.indices([1, 2, 3], sparse=True), tuple[npt.NDArray[np.int_], ...]) + +assert_type(np.fromfunction(func, (3, 5)), SubClass[np.float64]) + +assert_type(np.identity(10), npt.NDArray[np.float64]) +assert_type(np.identity(10, dtype=np.int64), npt.NDArray[np.int64]) +assert_type(np.identity(10, dtype=int), npt.NDArray[Any]) + +assert_type(np.atleast_1d(A), npt.NDArray[np.float64]) +assert_type(np.atleast_1d(C), npt.NDArray[Any]) +assert_type(np.atleast_1d(A, A), list[npt.NDArray[Any]]) +assert_type(np.atleast_1d(A, C), list[npt.NDArray[Any]]) +assert_type(np.atleast_1d(C, C), list[npt.NDArray[Any]]) + +assert_type(np.atleast_2d(A), npt.NDArray[np.float64]) + +assert_type(np.atleast_3d(A), npt.NDArray[np.float64]) + +assert_type(np.vstack([A, A]), np.ndarray[Any, Any]) +assert_type(np.vstack([A, A], dtype=np.float64), npt.NDArray[np.float64]) +assert_type(np.vstack([A, C]), npt.NDArray[Any]) +assert_type(np.vstack([C, C]), npt.NDArray[Any]) + +assert_type(np.hstack([A, A]), np.ndarray[Any, Any]) +assert_type(np.hstack([A, A], dtype=np.float64), npt.NDArray[np.float64]) + +assert_type(np.stack([A, A]), Any) +assert_type(np.stack([A, A], dtype=np.float64), npt.NDArray[np.float64]) +assert_type(np.stack([A, C]), npt.NDArray[Any]) +assert_type(np.stack([C, C]), npt.NDArray[Any]) +assert_type(np.stack([A, A], axis=0), Any) +assert_type(np.stack([A, A], out=B), SubClass[np.float64]) + +assert_type(np.block([[A, A], [A, A]]), npt.NDArray[Any]) +assert_type(np.block(C), npt.NDArray[Any]) + +if sys.version_info >= (3, 12): + from collections.abc import Buffer + + def create_array(obj: npt.ArrayLike) -> npt.NDArray[Any]: ... + + buffer: Buffer + assert_type(create_array(buffer), npt.NDArray[Any]) diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/datasource.pyi b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/datasource.pyi new file mode 100644 index 0000000000000000000000000000000000000000..865722d8c9448ffa601ce20838315c5df1c80852 --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/datasource.pyi @@ -0,0 +1,29 @@ +import sys +from pathlib import Path +from typing import IO, Any + +import numpy as np + +if sys.version_info >= (3, 11): + from typing import assert_type +else: + from typing_extensions import assert_type + +path1: Path +path2: str + +d1 = np.DataSource(path1) +d2 = np.DataSource(path2) +d3 = np.DataSource(None) + +assert_type(d1.abspath("..."), str) +assert_type(d2.abspath("..."), str) +assert_type(d3.abspath("..."), str) + +assert_type(d1.exists("..."), bool) +assert_type(d2.exists("..."), bool) +assert_type(d3.exists("..."), bool) + +assert_type(d1.open("...", "r"), IO[Any]) +assert_type(d2.open("...", encoding="utf8"), IO[Any]) +assert_type(d3.open("...", newline="/n"), IO[Any]) diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/fromnumeric.pyi b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/fromnumeric.pyi new file mode 100644 index 0000000000000000000000000000000000000000..aec21ec22c93335245a77810081e8eb700a52e0d --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/fromnumeric.pyi @@ -0,0 +1,305 @@ +"""Tests for :mod:`core.fromnumeric`.""" + +import sys +from typing import Any + +import numpy as np +import numpy.typing as npt + +if sys.version_info >= (3, 11): + from typing import assert_type +else: + from typing_extensions import assert_type + +class NDArraySubclass(npt.NDArray[np.complex128]): + ... + +AR_b: npt.NDArray[np.bool_] +AR_f4: npt.NDArray[np.float32] +AR_c16: npt.NDArray[np.complex128] +AR_u8: npt.NDArray[np.uint64] +AR_i8: npt.NDArray[np.int64] +AR_O: npt.NDArray[np.object_] +AR_subclass: NDArraySubclass + +b: np.bool_ +f4: np.float32 +i8: np.int64 +f: float + +assert_type(np.take(b, 0), np.bool_) +assert_type(np.take(f4, 0), np.float32) +assert_type(np.take(f, 0), Any) +assert_type(np.take(AR_b, 0), np.bool_) +assert_type(np.take(AR_f4, 0), np.float32) +assert_type(np.take(AR_b, [0]), npt.NDArray[np.bool_]) +assert_type(np.take(AR_f4, [0]), npt.NDArray[np.float32]) +assert_type(np.take([1], [0]), npt.NDArray[Any]) +assert_type(np.take(AR_f4, [0], out=AR_subclass), NDArraySubclass) + +assert_type(np.reshape(b, 1), npt.NDArray[np.bool_]) +assert_type(np.reshape(f4, 1), npt.NDArray[np.float32]) +assert_type(np.reshape(f, 1), npt.NDArray[Any]) +assert_type(np.reshape(AR_b, 1), npt.NDArray[np.bool_]) +assert_type(np.reshape(AR_f4, 1), npt.NDArray[np.float32]) + +assert_type(np.choose(1, [True, True]), Any) +assert_type(np.choose([1], [True, True]), npt.NDArray[Any]) +assert_type(np.choose([1], AR_b), npt.NDArray[np.bool_]) +assert_type(np.choose([1], AR_b, out=AR_f4), npt.NDArray[np.float32]) + +assert_type(np.repeat(b, 1), npt.NDArray[np.bool_]) +assert_type(np.repeat(f4, 1), npt.NDArray[np.float32]) +assert_type(np.repeat(f, 1), npt.NDArray[Any]) +assert_type(np.repeat(AR_b, 1), npt.NDArray[np.bool_]) +assert_type(np.repeat(AR_f4, 1), npt.NDArray[np.float32]) + +# TODO: array_bdd tests for np.put() + +assert_type(np.swapaxes([[0, 1]], 0, 0), npt.NDArray[Any]) +assert_type(np.swapaxes(AR_b, 0, 0), npt.NDArray[np.bool_]) +assert_type(np.swapaxes(AR_f4, 0, 0), npt.NDArray[np.float32]) + +assert_type(np.transpose(b), npt.NDArray[np.bool_]) +assert_type(np.transpose(f4), npt.NDArray[np.float32]) +assert_type(np.transpose(f), npt.NDArray[Any]) +assert_type(np.transpose(AR_b), npt.NDArray[np.bool_]) +assert_type(np.transpose(AR_f4), npt.NDArray[np.float32]) + +assert_type(np.partition(b, 0, axis=None), npt.NDArray[np.bool_]) +assert_type(np.partition(f4, 0, axis=None), npt.NDArray[np.float32]) +assert_type(np.partition(f, 0, axis=None), npt.NDArray[Any]) +assert_type(np.partition(AR_b, 0), npt.NDArray[np.bool_]) +assert_type(np.partition(AR_f4, 0), npt.NDArray[np.float32]) + +assert_type(np.argpartition(b, 0), npt.NDArray[np.intp]) +assert_type(np.argpartition(f4, 0), npt.NDArray[np.intp]) +assert_type(np.argpartition(f, 0), npt.NDArray[np.intp]) +assert_type(np.argpartition(AR_b, 0), npt.NDArray[np.intp]) +assert_type(np.argpartition(AR_f4, 0), npt.NDArray[np.intp]) + +assert_type(np.sort([2, 1], 0), npt.NDArray[Any]) +assert_type(np.sort(AR_b, 0), npt.NDArray[np.bool_]) +assert_type(np.sort(AR_f4, 0), npt.NDArray[np.float32]) + +assert_type(np.argsort(AR_b, 0), npt.NDArray[np.intp]) +assert_type(np.argsort(AR_f4, 0), npt.NDArray[np.intp]) + +assert_type(np.argmax(AR_b), np.intp) +assert_type(np.argmax(AR_f4), np.intp) +assert_type(np.argmax(AR_b, axis=0), Any) +assert_type(np.argmax(AR_f4, axis=0), Any) +assert_type(np.argmax(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.argmin(AR_b), np.intp) +assert_type(np.argmin(AR_f4), np.intp) +assert_type(np.argmin(AR_b, axis=0), Any) +assert_type(np.argmin(AR_f4, axis=0), Any) +assert_type(np.argmin(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.searchsorted(AR_b[0], 0), np.intp) +assert_type(np.searchsorted(AR_f4[0], 0), np.intp) +assert_type(np.searchsorted(AR_b[0], [0]), npt.NDArray[np.intp]) +assert_type(np.searchsorted(AR_f4[0], [0]), npt.NDArray[np.intp]) + +assert_type(np.resize(b, (5, 5)), npt.NDArray[np.bool_]) +assert_type(np.resize(f4, (5, 5)), npt.NDArray[np.float32]) +assert_type(np.resize(f, (5, 5)), npt.NDArray[Any]) +assert_type(np.resize(AR_b, (5, 5)), npt.NDArray[np.bool_]) +assert_type(np.resize(AR_f4, (5, 5)), npt.NDArray[np.float32]) + +assert_type(np.squeeze(b), np.bool_) +assert_type(np.squeeze(f4), np.float32) +assert_type(np.squeeze(f), npt.NDArray[Any]) +assert_type(np.squeeze(AR_b), npt.NDArray[np.bool_]) +assert_type(np.squeeze(AR_f4), npt.NDArray[np.float32]) + +assert_type(np.diagonal(AR_b), npt.NDArray[np.bool_]) +assert_type(np.diagonal(AR_f4), npt.NDArray[np.float32]) + +assert_type(np.trace(AR_b), Any) +assert_type(np.trace(AR_f4), Any) +assert_type(np.trace(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.ravel(b), npt.NDArray[np.bool_]) +assert_type(np.ravel(f4), npt.NDArray[np.float32]) +assert_type(np.ravel(f), npt.NDArray[Any]) +assert_type(np.ravel(AR_b), npt.NDArray[np.bool_]) +assert_type(np.ravel(AR_f4), npt.NDArray[np.float32]) + +assert_type(np.nonzero(b), tuple[npt.NDArray[np.intp], ...]) +assert_type(np.nonzero(f4), tuple[npt.NDArray[np.intp], ...]) +assert_type(np.nonzero(f), tuple[npt.NDArray[np.intp], ...]) +assert_type(np.nonzero(AR_b), tuple[npt.NDArray[np.intp], ...]) +assert_type(np.nonzero(AR_f4), tuple[npt.NDArray[np.intp], ...]) + +assert_type(np.shape(b), tuple[int, ...]) +assert_type(np.shape(f4), tuple[int, ...]) +assert_type(np.shape(f), tuple[int, ...]) +assert_type(np.shape(AR_b), tuple[int, ...]) +assert_type(np.shape(AR_f4), tuple[int, ...]) + +assert_type(np.compress([True], b), npt.NDArray[np.bool_]) +assert_type(np.compress([True], f4), npt.NDArray[np.float32]) +assert_type(np.compress([True], f), npt.NDArray[Any]) +assert_type(np.compress([True], AR_b), npt.NDArray[np.bool_]) +assert_type(np.compress([True], AR_f4), npt.NDArray[np.float32]) + +assert_type(np.clip(b, 0, 1.0), np.bool_) +assert_type(np.clip(f4, -1, 1), np.float32) +assert_type(np.clip(f, 0, 1), Any) +assert_type(np.clip(AR_b, 0, 1), npt.NDArray[np.bool_]) +assert_type(np.clip(AR_f4, 0, 1), npt.NDArray[np.float32]) +assert_type(np.clip([0], 0, 1), npt.NDArray[Any]) +assert_type(np.clip(AR_b, 0, 1, out=AR_subclass), NDArraySubclass) + +assert_type(np.sum(b), np.bool_) +assert_type(np.sum(f4), np.float32) +assert_type(np.sum(f), Any) +assert_type(np.sum(AR_b), np.bool_) +assert_type(np.sum(AR_f4), np.float32) +assert_type(np.sum(AR_b, axis=0), Any) +assert_type(np.sum(AR_f4, axis=0), Any) +assert_type(np.sum(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.all(b), np.bool_) +assert_type(np.all(f4), np.bool_) +assert_type(np.all(f), np.bool_) +assert_type(np.all(AR_b), np.bool_) +assert_type(np.all(AR_f4), np.bool_) +assert_type(np.all(AR_b, axis=0), Any) +assert_type(np.all(AR_f4, axis=0), Any) +assert_type(np.all(AR_b, keepdims=True), Any) +assert_type(np.all(AR_f4, keepdims=True), Any) +assert_type(np.all(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.any(b), np.bool_) +assert_type(np.any(f4), np.bool_) +assert_type(np.any(f), np.bool_) +assert_type(np.any(AR_b), np.bool_) +assert_type(np.any(AR_f4), np.bool_) +assert_type(np.any(AR_b, axis=0), Any) +assert_type(np.any(AR_f4, axis=0), Any) +assert_type(np.any(AR_b, keepdims=True), Any) +assert_type(np.any(AR_f4, keepdims=True), Any) +assert_type(np.any(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.cumsum(b), npt.NDArray[np.bool_]) +assert_type(np.cumsum(f4), npt.NDArray[np.float32]) +assert_type(np.cumsum(f), npt.NDArray[Any]) +assert_type(np.cumsum(AR_b), npt.NDArray[np.bool_]) +assert_type(np.cumsum(AR_f4), npt.NDArray[np.float32]) +assert_type(np.cumsum(f, dtype=float), npt.NDArray[Any]) +assert_type(np.cumsum(f, dtype=np.float64), npt.NDArray[np.float64]) +assert_type(np.cumsum(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.ptp(b), np.bool_) +assert_type(np.ptp(f4), np.float32) +assert_type(np.ptp(f), Any) +assert_type(np.ptp(AR_b), np.bool_) +assert_type(np.ptp(AR_f4), np.float32) +assert_type(np.ptp(AR_b, axis=0), Any) +assert_type(np.ptp(AR_f4, axis=0), Any) +assert_type(np.ptp(AR_b, keepdims=True), Any) +assert_type(np.ptp(AR_f4, keepdims=True), Any) +assert_type(np.ptp(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.amax(b), np.bool_) +assert_type(np.amax(f4), np.float32) +assert_type(np.amax(f), Any) +assert_type(np.amax(AR_b), np.bool_) +assert_type(np.amax(AR_f4), np.float32) +assert_type(np.amax(AR_b, axis=0), Any) +assert_type(np.amax(AR_f4, axis=0), Any) +assert_type(np.amax(AR_b, keepdims=True), Any) +assert_type(np.amax(AR_f4, keepdims=True), Any) +assert_type(np.amax(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.amin(b), np.bool_) +assert_type(np.amin(f4), np.float32) +assert_type(np.amin(f), Any) +assert_type(np.amin(AR_b), np.bool_) +assert_type(np.amin(AR_f4), np.float32) +assert_type(np.amin(AR_b, axis=0), Any) +assert_type(np.amin(AR_f4, axis=0), Any) +assert_type(np.amin(AR_b, keepdims=True), Any) +assert_type(np.amin(AR_f4, keepdims=True), Any) +assert_type(np.amin(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.prod(AR_b), np.int_) +assert_type(np.prod(AR_u8), np.uint64) +assert_type(np.prod(AR_i8), np.int64) +assert_type(np.prod(AR_f4), np.floating[Any]) +assert_type(np.prod(AR_c16), np.complexfloating[Any, Any]) +assert_type(np.prod(AR_O), Any) +assert_type(np.prod(AR_f4, axis=0), Any) +assert_type(np.prod(AR_f4, keepdims=True), Any) +assert_type(np.prod(AR_f4, dtype=np.float64), np.float64) +assert_type(np.prod(AR_f4, dtype=float), Any) +assert_type(np.prod(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.cumprod(AR_b), npt.NDArray[np.int_]) +assert_type(np.cumprod(AR_u8), npt.NDArray[np.uint64]) +assert_type(np.cumprod(AR_i8), npt.NDArray[np.int64]) +assert_type(np.cumprod(AR_f4), npt.NDArray[np.floating[Any]]) +assert_type(np.cumprod(AR_c16), npt.NDArray[np.complexfloating[Any, Any]]) +assert_type(np.cumprod(AR_O), npt.NDArray[np.object_]) +assert_type(np.cumprod(AR_f4, axis=0), npt.NDArray[np.floating[Any]]) +assert_type(np.cumprod(AR_f4, dtype=np.float64), npt.NDArray[np.float64]) +assert_type(np.cumprod(AR_f4, dtype=float), npt.NDArray[Any]) +assert_type(np.cumprod(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.ndim(b), int) +assert_type(np.ndim(f4), int) +assert_type(np.ndim(f), int) +assert_type(np.ndim(AR_b), int) +assert_type(np.ndim(AR_f4), int) + +assert_type(np.size(b), int) +assert_type(np.size(f4), int) +assert_type(np.size(f), int) +assert_type(np.size(AR_b), int) +assert_type(np.size(AR_f4), int) + +assert_type(np.around(b), np.float16) +assert_type(np.around(f), Any) +assert_type(np.around(i8), np.int64) +assert_type(np.around(f4), np.float32) +assert_type(np.around(AR_b), npt.NDArray[np.float16]) +assert_type(np.around(AR_i8), npt.NDArray[np.int64]) +assert_type(np.around(AR_f4), npt.NDArray[np.float32]) +assert_type(np.around([1.5]), npt.NDArray[Any]) +assert_type(np.around(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.mean(AR_b), np.floating[Any]) +assert_type(np.mean(AR_i8), np.floating[Any]) +assert_type(np.mean(AR_f4), np.floating[Any]) +assert_type(np.mean(AR_c16), np.complexfloating[Any, Any]) +assert_type(np.mean(AR_O), Any) +assert_type(np.mean(AR_f4, axis=0), Any) +assert_type(np.mean(AR_f4, keepdims=True), Any) +assert_type(np.mean(AR_f4, dtype=float), Any) +assert_type(np.mean(AR_f4, dtype=np.float64), np.float64) +assert_type(np.mean(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.std(AR_b), np.floating[Any]) +assert_type(np.std(AR_i8), np.floating[Any]) +assert_type(np.std(AR_f4), np.floating[Any]) +assert_type(np.std(AR_c16), np.floating[Any]) +assert_type(np.std(AR_O), Any) +assert_type(np.std(AR_f4, axis=0), Any) +assert_type(np.std(AR_f4, keepdims=True), Any) +assert_type(np.std(AR_f4, dtype=float), Any) +assert_type(np.std(AR_f4, dtype=np.float64), np.float64) +assert_type(np.std(AR_f4, out=AR_subclass), NDArraySubclass) + +assert_type(np.var(AR_b), np.floating[Any]) +assert_type(np.var(AR_i8), np.floating[Any]) +assert_type(np.var(AR_f4), np.floating[Any]) +assert_type(np.var(AR_c16), np.floating[Any]) +assert_type(np.var(AR_O), Any) +assert_type(np.var(AR_f4, axis=0), Any) +assert_type(np.var(AR_f4, keepdims=True), Any) +assert_type(np.var(AR_f4, dtype=float), Any) +assert_type(np.var(AR_f4, dtype=np.float64), np.float64) +assert_type(np.var(AR_f4, out=AR_subclass), NDArraySubclass) diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/memmap.pyi b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/memmap.pyi new file mode 100644 index 0000000000000000000000000000000000000000..53278ff1122b126c90ac45ecbd1b0b26b09a037b --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/memmap.pyi @@ -0,0 +1,25 @@ +import sys +from typing import Any + +import numpy as np + +if sys.version_info >= (3, 11): + from typing import assert_type +else: + from typing_extensions import assert_type + +memmap_obj: np.memmap[Any, np.dtype[np.str_]] + +assert_type(np.memmap.__array_priority__, float) +assert_type(memmap_obj.__array_priority__, float) +assert_type(memmap_obj.filename, str | None) +assert_type(memmap_obj.offset, int) +assert_type(memmap_obj.mode, str) +assert_type(memmap_obj.flush(), None) + +assert_type(np.memmap("file.txt", offset=5), np.memmap[Any, np.dtype[np.uint8]]) +assert_type(np.memmap(b"file.txt", dtype=np.float64, shape=(10, 3)), np.memmap[Any, np.dtype[np.float64]]) +with open("file.txt", "rb") as f: + assert_type(np.memmap(f, dtype=float, order="K"), np.memmap[Any, np.dtype[Any]]) + +assert_type(memmap_obj.__array_finalize__(object()), None) diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/mod.pyi b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/mod.pyi new file mode 100644 index 0000000000000000000000000000000000000000..48fee893cd895fe4ab7cda95421f90a0c587167d --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/mod.pyi @@ -0,0 +1,148 @@ +import sys +from typing import Any + +import numpy as np +import numpy.typing as npt +from numpy._typing import _32Bit, _64Bit + +if sys.version_info >= (3, 11): + from typing import assert_type +else: + from typing_extensions import assert_type + +f8 = np.float64() +i8 = np.int64() +u8 = np.uint64() + +f4 = np.float32() +i4 = np.int32() +u4 = np.uint32() + +td = np.timedelta64(0, "D") +b_ = np.bool_() + +b = bool() +f = float() +i = int() + +AR_b: npt.NDArray[np.bool_] +AR_m: npt.NDArray[np.timedelta64] + +# Time structures + +assert_type(td % td, np.timedelta64) +assert_type(AR_m % td, npt.NDArray[np.timedelta64]) +assert_type(td % AR_m, npt.NDArray[np.timedelta64]) + +assert_type(divmod(td, td), tuple[np.int64, np.timedelta64]) +assert_type(divmod(AR_m, td), tuple[npt.NDArray[np.int64], npt.NDArray[np.timedelta64]]) +assert_type(divmod(td, AR_m), tuple[npt.NDArray[np.int64], npt.NDArray[np.timedelta64]]) + +# Bool + +assert_type(b_ % b, np.int8) +assert_type(b_ % i, np.int_) +assert_type(b_ % f, np.float64) +assert_type(b_ % b_, np.int8) +assert_type(b_ % i8, np.int64) +assert_type(b_ % u8, np.uint64) +assert_type(b_ % f8, np.float64) +assert_type(b_ % AR_b, npt.NDArray[np.int8]) + +assert_type(divmod(b_, b), tuple[np.int8, np.int8]) +assert_type(divmod(b_, i), tuple[np.int_, np.int_]) +assert_type(divmod(b_, f), tuple[np.float64, np.float64]) +assert_type(divmod(b_, b_), tuple[np.int8, np.int8]) +assert_type(divmod(b_, i8), tuple[np.int64, np.int64]) +assert_type(divmod(b_, u8), tuple[np.uint64, np.uint64]) +assert_type(divmod(b_, f8), tuple[np.float64, np.float64]) +assert_type(divmod(b_, AR_b), tuple[npt.NDArray[np.int8], npt.NDArray[np.int8]]) + +assert_type(b % b_, np.int8) +assert_type(i % b_, np.int_) +assert_type(f % b_, np.float64) +assert_type(b_ % b_, np.int8) +assert_type(i8 % b_, np.int64) +assert_type(u8 % b_, np.uint64) +assert_type(f8 % b_, np.float64) +assert_type(AR_b % b_, npt.NDArray[np.int8]) + +assert_type(divmod(b, b_), tuple[np.int8, np.int8]) +assert_type(divmod(i, b_), tuple[np.int_, np.int_]) +assert_type(divmod(f, b_), tuple[np.float64, np.float64]) +assert_type(divmod(b_, b_), tuple[np.int8, np.int8]) +assert_type(divmod(i8, b_), tuple[np.int64, np.int64]) +assert_type(divmod(u8, b_), tuple[np.uint64, np.uint64]) +assert_type(divmod(f8, b_), tuple[np.float64, np.float64]) +assert_type(divmod(AR_b, b_), tuple[npt.NDArray[np.int8], npt.NDArray[np.int8]]) + +# int + +assert_type(i8 % b, np.int64) +assert_type(i8 % f, np.float64) +assert_type(i8 % i8, np.int64) +assert_type(i8 % f8, np.float64) +assert_type(i4 % i8, np.signedinteger[_32Bit | _64Bit]) +assert_type(i4 % f8, np.floating[_32Bit | _64Bit]) +assert_type(i4 % i4, np.int32) +assert_type(i4 % f4, np.float32) +assert_type(i8 % AR_b, npt.NDArray[np.signedinteger[Any]]) + +assert_type(divmod(i8, b), tuple[np.int64, np.int64]) +assert_type(divmod(i8, f), tuple[np.float64, np.float64]) +assert_type(divmod(i8, i8), tuple[np.int64, np.int64]) +assert_type(divmod(i8, f8), tuple[np.float64, np.float64]) +assert_type(divmod(i8, i4), tuple[np.signedinteger[_32Bit | _64Bit], np.signedinteger[_32Bit | _64Bit]]) +assert_type(divmod(i8, f4), tuple[np.floating[_32Bit | _64Bit], np.floating[_32Bit | _64Bit]]) +assert_type(divmod(i4, i4), tuple[np.int32, np.int32]) +assert_type(divmod(i4, f4), tuple[np.float32, np.float32]) +assert_type(divmod(i8, AR_b), tuple[npt.NDArray[np.signedinteger[Any]], npt.NDArray[np.signedinteger[Any]]]) + +assert_type(b % i8, np.int64) +assert_type(f % i8, np.float64) +assert_type(i8 % i8, np.int64) +assert_type(f8 % i8, np.float64) +assert_type(i8 % i4, np.signedinteger[_32Bit | _64Bit]) +assert_type(f8 % i4, np.floating[_32Bit | _64Bit]) +assert_type(i4 % i4, np.int32) +assert_type(f4 % i4, np.float32) +assert_type(AR_b % i8, npt.NDArray[np.signedinteger[Any]]) + +assert_type(divmod(b, i8), tuple[np.int64, np.int64]) +assert_type(divmod(f, i8), tuple[np.float64, np.float64]) +assert_type(divmod(i8, i8), tuple[np.int64, np.int64]) +assert_type(divmod(f8, i8), tuple[np.float64, np.float64]) +assert_type(divmod(i4, i8), tuple[np.signedinteger[_32Bit | _64Bit], np.signedinteger[_32Bit | _64Bit]]) +assert_type(divmod(f4, i8), tuple[np.floating[_32Bit | _64Bit], np.floating[_32Bit | _64Bit]]) +assert_type(divmod(i4, i4), tuple[np.int32, np.int32]) +assert_type(divmod(f4, i4), tuple[np.float32, np.float32]) +assert_type(divmod(AR_b, i8), tuple[npt.NDArray[np.signedinteger[Any]], npt.NDArray[np.signedinteger[Any]]]) + +# float + +assert_type(f8 % b, np.float64) +assert_type(f8 % f, np.float64) +assert_type(i8 % f4, np.floating[_32Bit | _64Bit]) +assert_type(f4 % f4, np.float32) +assert_type(f8 % AR_b, npt.NDArray[np.floating[Any]]) + +assert_type(divmod(f8, b), tuple[np.float64, np.float64]) +assert_type(divmod(f8, f), tuple[np.float64, np.float64]) +assert_type(divmod(f8, f8), tuple[np.float64, np.float64]) +assert_type(divmod(f8, f4), tuple[np.floating[_32Bit | _64Bit], np.floating[_32Bit | _64Bit]]) +assert_type(divmod(f4, f4), tuple[np.float32, np.float32]) +assert_type(divmod(f8, AR_b), tuple[npt.NDArray[np.floating[Any]], npt.NDArray[np.floating[Any]]]) + +assert_type(b % f8, np.float64) +assert_type(f % f8, np.float64) +assert_type(f8 % f8, np.float64) +assert_type(f8 % f8, np.float64) +assert_type(f4 % f4, np.float32) +assert_type(AR_b % f8, npt.NDArray[np.floating[Any]]) + +assert_type(divmod(b, f8), tuple[np.float64, np.float64]) +assert_type(divmod(f, f8), tuple[np.float64, np.float64]) +assert_type(divmod(f8, f8), tuple[np.float64, np.float64]) +assert_type(divmod(f4, f8), tuple[np.floating[_32Bit | _64Bit], np.floating[_32Bit | _64Bit]]) +assert_type(divmod(f4, f4), tuple[np.float32, np.float32]) +assert_type(divmod(AR_b, f8), tuple[npt.NDArray[np.floating[Any]], npt.NDArray[np.floating[Any]]]) diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/ndarray_shape_manipulation.pyi b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/ndarray_shape_manipulation.pyi new file mode 100644 index 0000000000000000000000000000000000000000..9a41a90f1ee92baa07605d5d202530623369bd8a --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/ndarray_shape_manipulation.pyi @@ -0,0 +1,44 @@ +import sys +from typing import Any + +import numpy as np +import numpy.typing as npt + +if sys.version_info >= (3, 11): + from typing import assert_type +else: + from typing_extensions import assert_type + +nd: npt.NDArray[np.int64] + +# reshape +assert_type(nd.reshape(), npt.NDArray[np.int64]) +assert_type(nd.reshape(4), npt.NDArray[np.int64]) +assert_type(nd.reshape(2, 2), npt.NDArray[np.int64]) +assert_type(nd.reshape((2, 2)), npt.NDArray[np.int64]) + +assert_type(nd.reshape((2, 2), order="C"), npt.NDArray[np.int64]) +assert_type(nd.reshape(4, order="C"), npt.NDArray[np.int64]) + +# resize does not return a value + +# transpose +assert_type(nd.transpose(), npt.NDArray[np.int64]) +assert_type(nd.transpose(1, 0), npt.NDArray[np.int64]) +assert_type(nd.transpose((1, 0)), npt.NDArray[np.int64]) + +# swapaxes +assert_type(nd.swapaxes(0, 1), npt.NDArray[np.int64]) + +# flatten +assert_type(nd.flatten(), npt.NDArray[np.int64]) +assert_type(nd.flatten("C"), npt.NDArray[np.int64]) + +# ravel +assert_type(nd.ravel(), npt.NDArray[np.int64]) +assert_type(nd.ravel("C"), npt.NDArray[np.int64]) + +# squeeze +assert_type(nd.squeeze(), npt.NDArray[np.int64]) +assert_type(nd.squeeze(0), npt.NDArray[np.int64]) +assert_type(nd.squeeze((0, 2)), npt.NDArray[np.int64]) diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/nditer.pyi b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/nditer.pyi new file mode 100644 index 0000000000000000000000000000000000000000..589453e777f222fa409ae10226921bf848164cd3 --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/nditer.pyi @@ -0,0 +1,55 @@ +import sys +from typing import Any + +import numpy as np +import numpy.typing as npt + +if sys.version_info >= (3, 11): + from typing import assert_type +else: + from typing_extensions import assert_type + +nditer_obj: np.nditer + +assert_type(np.nditer([0, 1], flags=["c_index"]), np.nditer) +assert_type(np.nditer([0, 1], op_flags=[["readonly", "readonly"]]), np.nditer) +assert_type(np.nditer([0, 1], op_dtypes=np.int_), np.nditer) +assert_type(np.nditer([0, 1], order="C", casting="no"), np.nditer) + +assert_type(nditer_obj.dtypes, tuple[np.dtype[Any], ...]) +assert_type(nditer_obj.finished, bool) +assert_type(nditer_obj.has_delayed_bufalloc, bool) +assert_type(nditer_obj.has_index, bool) +assert_type(nditer_obj.has_multi_index, bool) +assert_type(nditer_obj.index, int) +assert_type(nditer_obj.iterationneedsapi, bool) +assert_type(nditer_obj.iterindex, int) +assert_type(nditer_obj.iterrange, tuple[int, ...]) +assert_type(nditer_obj.itersize, int) +assert_type(nditer_obj.itviews, tuple[npt.NDArray[Any], ...]) +assert_type(nditer_obj.multi_index, tuple[int, ...]) +assert_type(nditer_obj.ndim, int) +assert_type(nditer_obj.nop, int) +assert_type(nditer_obj.operands, tuple[npt.NDArray[Any], ...]) +assert_type(nditer_obj.shape, tuple[int, ...]) +assert_type(nditer_obj.value, tuple[npt.NDArray[Any], ...]) + +assert_type(nditer_obj.close(), None) +assert_type(nditer_obj.copy(), np.nditer) +assert_type(nditer_obj.debug_print(), None) +assert_type(nditer_obj.enable_external_loop(), None) +assert_type(nditer_obj.iternext(), bool) +assert_type(nditer_obj.remove_axis(0), None) +assert_type(nditer_obj.remove_multi_index(), None) +assert_type(nditer_obj.reset(), None) + +assert_type(len(nditer_obj), int) +assert_type(iter(nditer_obj), np.nditer) +assert_type(next(nditer_obj), tuple[npt.NDArray[Any], ...]) +assert_type(nditer_obj.__copy__(), np.nditer) +with nditer_obj as f: + assert_type(f, np.nditer) +assert_type(nditer_obj[0], npt.NDArray[Any]) +assert_type(nditer_obj[:], tuple[npt.NDArray[Any], ...]) +nditer_obj[0] = 0 +nditer_obj[:] = [0, 1] diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/warnings_and_errors.pyi b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/warnings_and_errors.pyi new file mode 100644 index 0000000000000000000000000000000000000000..b026e4f6e3b03dd0a4750a74ce49f7ec0336ee4d --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/.venv_qwen35_uv/lib/python3.12/site-packages/numpy/typing/tests/data/reveal/warnings_and_errors.pyi @@ -0,0 +1,16 @@ +import sys + +import numpy as np + +if sys.version_info >= (3, 11): + from typing import assert_type +else: + from typing_extensions import assert_type + +assert_type(np.ModuleDeprecationWarning(), np.ModuleDeprecationWarning) +assert_type(np.VisibleDeprecationWarning(), np.VisibleDeprecationWarning) +assert_type(np.ComplexWarning(), np.ComplexWarning) +assert_type(np.RankWarning(), np.RankWarning) +assert_type(np.TooHardError(), np.TooHardError) +assert_type(np.AxisError("test"), np.AxisError) +assert_type(np.AxisError(5, 1), np.AxisError) diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/logs/decode1_4_uniform_online_20260610/lr2e3_online_latest_decode1_4_uniform.log b/LTA_openwebtext_dualt/mini_owt_logdirichlet/logs/decode1_4_uniform_online_20260610/lr2e3_online_latest_decode1_4_uniform.log new file mode 100644 index 0000000000000000000000000000000000000000..f3c7f5a0c792260647660115a5e9d985935171c8 --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/logs/decode1_4_uniform_online_20260610/lr2e3_online_latest_decode1_4_uniform.log @@ -0,0 +1,51 @@ +checkpoint=runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_8gpu_50ep_lr2e3_ema0p9999_elfopt_t5embed_unfixed_norm_stateprobadd_selfcond_ce_fast_20260606_144231/step_292000.pt +use_ema=0 +step=292000 +decode_steps=1 4 +n=64 chunk_n=8 gpu=0 +out_base=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610 +[2026-06-09T22:56:14+00:00] infer step=292000 decode=1 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr2e3_ema0p9999_online_uniform_step292000_sc1p0_decode1_n64 +[2026-06-09T22:56:14+00:00] run decode=1 chunk=0 n=8 seed=123 +[2026-06-09T22:56:18+00:00] done decode=1 chunk=0 +[2026-06-09T22:56:18+00:00] run decode=1 chunk=1 n=8 seed=124 +[2026-06-09T22:56:22+00:00] done decode=1 chunk=1 +[2026-06-09T22:56:22+00:00] run decode=1 chunk=2 n=8 seed=125 +[2026-06-09T22:56:26+00:00] done decode=1 chunk=2 +[2026-06-09T22:56:26+00:00] run decode=1 chunk=3 n=8 seed=126 +[2026-06-09T22:56:30+00:00] done decode=1 chunk=3 +[2026-06-09T22:56:30+00:00] run decode=1 chunk=4 n=8 seed=127 +[2026-06-09T22:56:35+00:00] done decode=1 chunk=4 +[2026-06-09T22:56:35+00:00] run decode=1 chunk=5 n=8 seed=128 +[2026-06-09T22:56:39+00:00] done decode=1 chunk=5 +[2026-06-09T22:56:39+00:00] run decode=1 chunk=6 n=8 seed=129 +[2026-06-09T22:56:43+00:00] done decode=1 chunk=6 +[2026-06-09T22:56:43+00:00] run decode=1 chunk=7 n=8 seed=130 +[2026-06-09T22:56:47+00:00] done decode=1 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr2e3_ema0p9999_online_uniform_step292000_sc1p0_decode1_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 1.096572649728622 0.11783789575524715 0.0004955050612302683 0.001132663174288546 0.9815247398598429 0 0 79304 14127 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr2e3_ema0p9999_online_uniform_step292000_sc1p0_decode1_n64/sc1p0 +sc1p0 pre_eos 1.096572649728622 0.11783789575524715 0.0004955050612302683 0.001132663174288546 0.9815247398598429 0 0 79304 14127 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr2e3_ema0p9999_online_uniform_step292000_sc1p0_decode1_n64/sc1p0 +[2026-06-09T22:57:02+00:00] infer step=292000 decode=4 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr2e3_ema0p9999_online_uniform_step292000_sc1p0_decode4_n64 +[2026-06-09T22:57:02+00:00] run decode=4 chunk=0 n=8 seed=123 +[2026-06-09T22:57:07+00:00] done decode=4 chunk=0 +[2026-06-09T22:57:07+00:00] run decode=4 chunk=1 n=8 seed=124 +[2026-06-09T22:57:11+00:00] done decode=4 chunk=1 +[2026-06-09T22:57:11+00:00] run decode=4 chunk=2 n=8 seed=125 +[2026-06-09T22:57:16+00:00] done decode=4 chunk=2 +[2026-06-09T22:57:16+00:00] run decode=4 chunk=3 n=8 seed=126 +[2026-06-09T22:57:20+00:00] done decode=4 chunk=3 +[2026-06-09T22:57:20+00:00] run decode=4 chunk=4 n=8 seed=127 +[2026-06-09T22:57:25+00:00] done decode=4 chunk=4 +[2026-06-09T22:57:25+00:00] run decode=4 chunk=5 n=8 seed=128 +[2026-06-09T22:57:29+00:00] done decode=4 chunk=5 +[2026-06-09T22:57:29+00:00] run decode=4 chunk=6 n=8 seed=129 +[2026-06-09T22:57:34+00:00] done decode=4 chunk=6 +[2026-06-09T22:57:34+00:00] run decode=4 chunk=7 n=8 seed=130 +[2026-06-09T22:57:39+00:00] done decode=4 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr2e3_ema0p9999_online_uniform_step292000_sc1p0_decode4_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 3.9144810000584296 1.655964156095804 0.002519893899204244 0.010636886920077455 0.42931034482758623 8 8 53648 37700 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr2e3_ema0p9999_online_uniform_step292000_sc1p0_decode4_n64/sc1p0 +sc1p0 pre_eos 4.001451443769343 1.6315027573148633 0.002560163850486431 0.010779927774483911 0.4361441237502358 0 0 52447 37107 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr2e3_ema0p9999_online_uniform_step292000_sc1p0_decode4_n64/sc1p0 +[2026-06-09T22:57:50+00:00] done diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/logs/decode1_4_uniform_online_20260610/lr6e4_online_latest_decode1_4_uniform.log b/LTA_openwebtext_dualt/mini_owt_logdirichlet/logs/decode1_4_uniform_online_20260610/lr6e4_online_latest_decode1_4_uniform.log new file mode 100644 index 0000000000000000000000000000000000000000..90e6f94e107999fe43d0c249ca06fd7233aa212a --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/logs/decode1_4_uniform_online_20260610/lr6e4_online_latest_decode1_4_uniform.log @@ -0,0 +1,51 @@ +checkpoint=runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_8gpu_50ep_lr6e4_ema0p9999_elfopt_t5embed_unfixed_norm_stateprobadd_selfcond_ce_fast_20260606_144245/step_297000.pt +use_ema=0 +step=297000 +decode_steps=1 4 +n=64 chunk_n=8 gpu=1 +out_base=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610 +[2026-06-09T22:56:14+00:00] infer step=297000 decode=1 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr6e4_ema0p9999_online_uniform_step297000_sc1p0_decode1_n64 +[2026-06-09T22:56:14+00:00] run decode=1 chunk=0 n=8 seed=123 +[2026-06-09T22:56:18+00:00] done decode=1 chunk=0 +[2026-06-09T22:56:18+00:00] run decode=1 chunk=1 n=8 seed=124 +[2026-06-09T22:56:23+00:00] done decode=1 chunk=1 +[2026-06-09T22:56:23+00:00] run decode=1 chunk=2 n=8 seed=125 +[2026-06-09T22:56:27+00:00] done decode=1 chunk=2 +[2026-06-09T22:56:27+00:00] run decode=1 chunk=3 n=8 seed=126 +[2026-06-09T22:56:31+00:00] done decode=1 chunk=3 +[2026-06-09T22:56:31+00:00] run decode=1 chunk=4 n=8 seed=127 +[2026-06-09T22:56:35+00:00] done decode=1 chunk=4 +[2026-06-09T22:56:35+00:00] run decode=1 chunk=5 n=8 seed=128 +[2026-06-09T22:56:40+00:00] done decode=1 chunk=5 +[2026-06-09T22:56:40+00:00] run decode=1 chunk=6 n=8 seed=129 +[2026-06-09T22:56:44+00:00] done decode=1 chunk=6 +[2026-06-09T22:56:44+00:00] run decode=1 chunk=7 n=8 seed=130 +[2026-06-09T22:56:48+00:00] done decode=1 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr6e4_ema0p9999_online_uniform_step297000_sc1p0_decode1_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 1.1010033551856535 0.10100969198731712 0.0004690706537672237 0.0008209217778820218 0.9846379360891234 0 0 82237 17055 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr6e4_ema0p9999_online_uniform_step297000_sc1p0_decode1_n64/sc1p0 +sc1p0 pre_eos 1.1010033551856535 0.10100969198731712 0.0004690706537672237 0.0008209217778820218 0.9846379360891234 0 0 82237 17055 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr6e4_ema0p9999_online_uniform_step297000_sc1p0_decode1_n64/sc1p0 +[2026-06-09T22:57:03+00:00] infer step=297000 decode=4 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr6e4_ema0p9999_online_uniform_step297000_sc1p0_decode4_n64 +[2026-06-09T22:57:03+00:00] run decode=4 chunk=0 n=8 seed=123 +[2026-06-09T22:57:08+00:00] done decode=4 chunk=0 +[2026-06-09T22:57:08+00:00] run decode=4 chunk=1 n=8 seed=124 +[2026-06-09T22:57:12+00:00] done decode=4 chunk=1 +[2026-06-09T22:57:12+00:00] run decode=4 chunk=2 n=8 seed=125 +[2026-06-09T22:57:16+00:00] done decode=4 chunk=2 +[2026-06-09T22:57:16+00:00] run decode=4 chunk=3 n=8 seed=126 +[2026-06-09T22:57:21+00:00] done decode=4 chunk=3 +[2026-06-09T22:57:21+00:00] run decode=4 chunk=4 n=8 seed=127 +[2026-06-09T22:57:25+00:00] done decode=4 chunk=4 +[2026-06-09T22:57:25+00:00] run decode=4 chunk=5 n=8 seed=128 +[2026-06-09T22:57:30+00:00] done decode=4 chunk=5 +[2026-06-09T22:57:30+00:00] run decode=4 chunk=6 n=8 seed=129 +[2026-06-09T22:57:34+00:00] done decode=4 chunk=6 +[2026-06-09T22:57:34+00:00] run decode=4 chunk=7 n=8 seed=130 +[2026-06-09T22:57:39+00:00] done decode=4 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr6e4_ema0p9999_online_uniform_step297000_sc1p0_decode4_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 1.862719202099405 0.4595035966522467 0.0006571908910285801 0.0022008589463387795 0.8955524988537368 4 4 38889 65430 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr6e4_ema0p9999_online_uniform_step297000_sc1p0_decode4_n64/sc1p0 +sc1p0 pre_eos 1.8667632184252236 0.44590311903537977 0.0006596609649459231 0.002193789887088856 0.8989184628365422 0 0 38395 65185 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260610/owt_t5_elftokenized_full_pow1_unfixed_norm_stateprobadd_selfcond_ce_fast_lr6e4_ema0p9999_online_uniform_step297000_sc1p0_decode4_n64/sc1p0 +[2026-06-09T22:57:49+00:00] done diff --git a/LTA_openwebtext_dualt/mini_owt_logdirichlet/logs/infer_lr3e3_170k_logitnorm64_focused_20260612/gpu3.log b/LTA_openwebtext_dualt/mini_owt_logdirichlet/logs/infer_lr3e3_170k_logitnorm64_focused_20260612/gpu3.log new file mode 100644 index 0000000000000000000000000000000000000000..54991fe6d11857dc5ccabd88eeb6b243ef055f3b --- /dev/null +++ b/LTA_openwebtext_dualt/mini_owt_logdirichlet/logs/infer_lr3e3_170k_logitnorm64_focused_20260612/gpu3.log @@ -0,0 +1,352 @@ +===== START \2026-06-11T22:34:52+00:00 gpu=3 m=-0.5 s=1.4 gamma=0.25 ===== +checkpoint=runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck128_unfixed_norm_stateprobadd_selfcond_ce_fast_20260609_230847/step_170000.pt +use_ema=1 +step=170000 +decode_steps=64 +n=64 chunk_n=8 gpu=3 +out_base=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612 +[2026-06-11T22:34:52+00:00] infer step=170000 decode=64 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s1p4_sc1p0_decode64_n64 +[2026-06-11T22:34:52+00:00] run decode=64 chunk=0 n=8 seed=123 +[2026-06-11T22:35:02+00:00] done decode=64 chunk=0 +[2026-06-11T22:35:02+00:00] run decode=64 chunk=1 n=8 seed=124 +[2026-06-11T22:35:12+00:00] done decode=64 chunk=1 +[2026-06-11T22:35:12+00:00] run decode=64 chunk=2 n=8 seed=125 +[2026-06-11T22:35:21+00:00] done decode=64 chunk=2 +[2026-06-11T22:35:21+00:00] run decode=64 chunk=3 n=8 seed=126 +[2026-06-11T22:35:31+00:00] done decode=64 chunk=3 +[2026-06-11T22:35:31+00:00] run decode=64 chunk=4 n=8 seed=127 +[2026-06-11T22:35:41+00:00] done decode=64 chunk=4 +[2026-06-11T22:35:41+00:00] run decode=64 chunk=5 n=8 seed=128 +[2026-06-11T22:35:50+00:00] done decode=64 chunk=5 +[2026-06-11T22:35:50+00:00] run decode=64 chunk=6 n=8 seed=129 +[2026-06-11T22:36:00+00:00] done decode=64 chunk=6 +[2026-06-11T22:36:00+00:00] run decode=64 chunk=7 n=8 seed=130 +[2026-06-11T22:36:10+00:00] done decode=64 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s1p4_sc1p0_decode64_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 14.323413610616049 5.009203366356262 0.07500610053684724 0.4101758506565803 0.03593216203025866 62 62 64109 65568 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s1p4_sc1p0_decode64_n64/sc1p0 +sc1p0 pre_eos 16.11984853009216 5.038465452975301 0.07745990736948234 0.4236337571088741 0.03711522102145625 0 0 59907 63478 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s1p4_sc1p0_decode64_n64/sc1p0 +[2026-06-11T22:36:42+00:00] done +===== DONE \2026-06-11T22:34:52+00:00 gpu=3 m=-0.5 s=1.4 gamma=0.25 ===== +===== START \2026-06-11T22:34:52+00:00 gpu=3 m=-0.5 s=1.6 gamma=0.25 ===== +checkpoint=runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck128_unfixed_norm_stateprobadd_selfcond_ce_fast_20260609_230847/step_170000.pt +use_ema=1 +step=170000 +decode_steps=64 +n=64 chunk_n=8 gpu=3 +out_base=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612 +[2026-06-11T22:36:42+00:00] infer step=170000 decode=64 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s1p6_sc1p0_decode64_n64 +[2026-06-11T22:36:42+00:00] run decode=64 chunk=0 n=8 seed=123 +[2026-06-11T22:36:52+00:00] done decode=64 chunk=0 +[2026-06-11T22:36:52+00:00] run decode=64 chunk=1 n=8 seed=124 +[2026-06-11T22:37:02+00:00] done decode=64 chunk=1 +[2026-06-11T22:37:02+00:00] run decode=64 chunk=2 n=8 seed=125 +[2026-06-11T22:37:11+00:00] done decode=64 chunk=2 +[2026-06-11T22:37:11+00:00] run decode=64 chunk=3 n=8 seed=126 +[2026-06-11T22:37:21+00:00] done decode=64 chunk=3 +[2026-06-11T22:37:21+00:00] run decode=64 chunk=4 n=8 seed=127 +[2026-06-11T22:37:31+00:00] done decode=64 chunk=4 +[2026-06-11T22:37:31+00:00] run decode=64 chunk=5 n=8 seed=128 +[2026-06-11T22:37:41+00:00] done decode=64 chunk=5 +[2026-06-11T22:37:41+00:00] run decode=64 chunk=6 n=8 seed=129 +[2026-06-11T22:37:50+00:00] done decode=64 chunk=6 +[2026-06-11T22:37:50+00:00] run decode=64 chunk=7 n=8 seed=130 +[2026-06-11T22:38:00+00:00] done decode=64 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s1p6_sc1p0_decode64_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 15.92467574106307 5.044005213841317 0.07881645608252456 0.42744655201355086 0.03688274430812428 63 63 63695 65532 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s1p6_sc1p0_decode64_n64/sc1p0 +sc1p0 pre_eos 17.980638543037145 5.073226190698163 0.08132027337721646 0.4410481724697249 0.03806179332934396 0 0 59610 63502 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s1p6_sc1p0_decode64_n64/sc1p0 +[2026-06-11T22:38:31+00:00] done +===== DONE \2026-06-11T22:34:52+00:00 gpu=3 m=-0.5 s=1.6 gamma=0.25 ===== +===== START \2026-06-11T22:34:52+00:00 gpu=3 m=-0.5 s=1.8 gamma=0.25 ===== +checkpoint=runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck128_unfixed_norm_stateprobadd_selfcond_ce_fast_20260609_230847/step_170000.pt +use_ema=1 +step=170000 +decode_steps=64 +n=64 chunk_n=8 gpu=3 +out_base=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612 +[2026-06-11T22:38:31+00:00] infer step=170000 decode=64 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s1p8_sc1p0_decode64_n64 +[2026-06-11T22:38:31+00:00] run decode=64 chunk=0 n=8 seed=123 +[2026-06-11T22:38:41+00:00] done decode=64 chunk=0 +[2026-06-11T22:38:41+00:00] run decode=64 chunk=1 n=8 seed=124 +[2026-06-11T22:38:51+00:00] done decode=64 chunk=1 +[2026-06-11T22:38:51+00:00] run decode=64 chunk=2 n=8 seed=125 +[2026-06-11T22:39:00+00:00] done decode=64 chunk=2 +[2026-06-11T22:39:00+00:00] run decode=64 chunk=3 n=8 seed=126 +[2026-06-11T22:39:10+00:00] done decode=64 chunk=3 +[2026-06-11T22:39:10+00:00] run decode=64 chunk=4 n=8 seed=127 +[2026-06-11T22:39:20+00:00] done decode=64 chunk=4 +[2026-06-11T22:39:20+00:00] run decode=64 chunk=5 n=8 seed=128 +[2026-06-11T22:39:29+00:00] done decode=64 chunk=5 +[2026-06-11T22:39:29+00:00] run decode=64 chunk=6 n=8 seed=129 +[2026-06-11T22:39:39+00:00] done decode=64 chunk=6 +[2026-06-11T22:39:39+00:00] run decode=64 chunk=7 n=8 seed=130 +[2026-06-11T22:39:49+00:00] done decode=64 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s1p8_sc1p0_decode64_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 16.511156990368196 5.021153347378399 0.07948354013094638 0.43199230792710847 0.03784930482426018 64 64 63897 65523 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s1p8_sc1p0_decode64_n64/sc1p0 +sc1p0 pre_eos 19.003033006817592 5.056660543916634 0.08224998815296887 0.4470508324645373 0.03917418294975279 0 0 59441 63307 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s1p8_sc1p0_decode64_n64/sc1p0 +[2026-06-11T22:40:18+00:00] done +===== DONE \2026-06-11T22:34:52+00:00 gpu=3 m=-0.5 s=1.8 gamma=0.25 ===== +===== START \2026-06-11T22:34:52+00:00 gpu=3 m=-0.5 s=2.0 gamma=0.25 ===== +checkpoint=runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck128_unfixed_norm_stateprobadd_selfcond_ce_fast_20260609_230847/step_170000.pt +use_ema=1 +step=170000 +decode_steps=64 +n=64 chunk_n=8 gpu=3 +out_base=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612 +[2026-06-11T22:40:18+00:00] infer step=170000 decode=64 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s2p0_sc1p0_decode64_n64 +[2026-06-11T22:40:18+00:00] run decode=64 chunk=0 n=8 seed=123 +[2026-06-11T22:40:27+00:00] done decode=64 chunk=0 +[2026-06-11T22:40:27+00:00] run decode=64 chunk=1 n=8 seed=124 +[2026-06-11T22:40:37+00:00] done decode=64 chunk=1 +[2026-06-11T22:40:37+00:00] run decode=64 chunk=2 n=8 seed=125 +[2026-06-11T22:40:47+00:00] done decode=64 chunk=2 +[2026-06-11T22:40:47+00:00] run decode=64 chunk=3 n=8 seed=126 +[2026-06-11T22:40:56+00:00] done decode=64 chunk=3 +[2026-06-11T22:40:56+00:00] run decode=64 chunk=4 n=8 seed=127 +[2026-06-11T22:41:06+00:00] done decode=64 chunk=4 +[2026-06-11T22:41:06+00:00] run decode=64 chunk=5 n=8 seed=128 +[2026-06-11T22:41:16+00:00] done decode=64 chunk=5 +[2026-06-11T22:41:16+00:00] run decode=64 chunk=6 n=8 seed=129 +[2026-06-11T22:41:26+00:00] done decode=64 chunk=6 +[2026-06-11T22:41:26+00:00] run decode=64 chunk=7 n=8 seed=130 +[2026-06-11T22:41:35+00:00] done decode=64 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s2p0_sc1p0_decode64_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 19.298351448530592 5.110950318985606 0.08453284081022087 0.4514898033947979 0.03542808297589791 62 62 63303 65513 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s2p0_sc1p0_decode64_n64/sc1p0 +sc1p0 pre_eos 22.248318536053922 5.146961055265112 0.08728895054624565 0.46623155505107833 0.036589787649961375 0 0 59118 63433 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p5_s2p0_sc1p0_decode64_n64/sc1p0 +[2026-06-11T22:41:49+00:00] done +===== DONE \2026-06-11T22:34:52+00:00 gpu=3 m=-0.5 s=2.0 gamma=0.25 ===== +===== START \2026-06-11T22:34:52+00:00 gpu=3 m=-0.4 s=1.4 gamma=0.25 ===== +checkpoint=runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck128_unfixed_norm_stateprobadd_selfcond_ce_fast_20260609_230847/step_170000.pt +use_ema=1 +step=170000 +decode_steps=64 +n=64 chunk_n=8 gpu=3 +out_base=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612 +[2026-06-11T22:41:49+00:00] infer step=170000 decode=64 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s1p4_sc1p0_decode64_n64 +[2026-06-11T22:41:49+00:00] run decode=64 chunk=0 n=8 seed=123 +[2026-06-11T22:41:59+00:00] done decode=64 chunk=0 +[2026-06-11T22:41:59+00:00] run decode=64 chunk=1 n=8 seed=124 +[2026-06-11T22:42:09+00:00] done decode=64 chunk=1 +[2026-06-11T22:42:09+00:00] run decode=64 chunk=2 n=8 seed=125 +[2026-06-11T22:42:19+00:00] done decode=64 chunk=2 +[2026-06-11T22:42:19+00:00] run decode=64 chunk=3 n=8 seed=126 +[2026-06-11T22:42:28+00:00] done decode=64 chunk=3 +[2026-06-11T22:42:28+00:00] run decode=64 chunk=4 n=8 seed=127 +[2026-06-11T22:42:38+00:00] done decode=64 chunk=4 +[2026-06-11T22:42:38+00:00] run decode=64 chunk=5 n=8 seed=128 +[2026-06-11T22:42:48+00:00] done decode=64 chunk=5 +[2026-06-11T22:42:48+00:00] run decode=64 chunk=6 n=8 seed=129 +[2026-06-11T22:42:57+00:00] done decode=64 chunk=6 +[2026-06-11T22:42:57+00:00] run decode=64 chunk=7 n=8 seed=130 +[2026-06-11T22:43:07+00:00] done decode=64 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s1p4_sc1p0_decode64_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 15.483493569064475 5.033153009064646 0.08120904800452095 0.43044354838709675 0.036167580529378525 64 64 63185 65473 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s1p4_sc1p0_decode64_n64/sc1p0 +sc1p0 pre_eos 17.4396064742211 5.063226024759697 0.08378514689194301 0.44411871325673397 0.037321901399571304 0 0 59110 63448 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s1p4_sc1p0_decode64_n64/sc1p0 +[2026-06-11T22:43:38+00:00] done +===== DONE \2026-06-11T22:34:52+00:00 gpu=3 m=-0.4 s=1.4 gamma=0.25 ===== +===== START \2026-06-11T22:34:52+00:00 gpu=3 m=-0.4 s=1.6 gamma=0.25 ===== +checkpoint=runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck128_unfixed_norm_stateprobadd_selfcond_ce_fast_20260609_230847/step_170000.pt +use_ema=1 +step=170000 +decode_steps=64 +n=64 chunk_n=8 gpu=3 +out_base=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612 +[2026-06-11T22:43:38+00:00] infer step=170000 decode=64 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s1p6_sc1p0_decode64_n64 +[2026-06-11T22:43:38+00:00] run decode=64 chunk=0 n=8 seed=123 +[2026-06-11T22:43:48+00:00] done decode=64 chunk=0 +[2026-06-11T22:43:48+00:00] run decode=64 chunk=1 n=8 seed=124 +[2026-06-11T22:43:58+00:00] done decode=64 chunk=1 +[2026-06-11T22:43:58+00:00] run decode=64 chunk=2 n=8 seed=125 +[2026-06-11T22:44:08+00:00] done decode=64 chunk=2 +[2026-06-11T22:44:08+00:00] run decode=64 chunk=3 n=8 seed=126 +[2026-06-11T22:44:18+00:00] done decode=64 chunk=3 +[2026-06-11T22:44:18+00:00] run decode=64 chunk=4 n=8 seed=127 +[2026-06-11T22:44:27+00:00] done decode=64 chunk=4 +[2026-06-11T22:44:27+00:00] run decode=64 chunk=5 n=8 seed=128 +[2026-06-11T22:44:37+00:00] done decode=64 chunk=5 +[2026-06-11T22:44:37+00:00] run decode=64 chunk=6 n=8 seed=129 +[2026-06-11T22:44:47+00:00] done decode=64 chunk=6 +[2026-06-11T22:44:47+00:00] run decode=64 chunk=7 n=8 seed=130 +[2026-06-11T22:44:57+00:00] done decode=64 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s1p6_sc1p0_decode64_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 16.66348509590687 5.050779955667837 0.07936846922431547 0.43019495377856426 0.036015559453893675 63 64 63791 65555 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s1p6_sc1p0_decode64_n64/sc1p0 +sc1p0 pre_eos 19.053423138624822 5.082727122518139 0.08202589129440704 0.444621400864108 0.037228590800864096 0 0 59497 63419 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s1p6_sc1p0_decode64_n64/sc1p0 +[2026-06-11T22:45:27+00:00] done +===== DONE \2026-06-11T22:34:52+00:00 gpu=3 m=-0.4 s=1.6 gamma=0.25 ===== +===== START \2026-06-11T22:34:52+00:00 gpu=3 m=-0.4 s=1.8 gamma=0.25 ===== +checkpoint=runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck128_unfixed_norm_stateprobadd_selfcond_ce_fast_20260609_230847/step_170000.pt +use_ema=1 +step=170000 +decode_steps=64 +n=64 chunk_n=8 gpu=3 +out_base=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612 +[2026-06-11T22:45:27+00:00] infer step=170000 decode=64 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s1p8_sc1p0_decode64_n64 +[2026-06-11T22:45:27+00:00] run decode=64 chunk=0 n=8 seed=123 +[2026-06-11T22:45:37+00:00] done decode=64 chunk=0 +[2026-06-11T22:45:37+00:00] run decode=64 chunk=1 n=8 seed=124 +[2026-06-11T22:45:47+00:00] done decode=64 chunk=1 +[2026-06-11T22:45:47+00:00] run decode=64 chunk=2 n=8 seed=125 +[2026-06-11T22:45:56+00:00] done decode=64 chunk=2 +[2026-06-11T22:45:56+00:00] run decode=64 chunk=3 n=8 seed=126 +[2026-06-11T22:46:06+00:00] done decode=64 chunk=3 +[2026-06-11T22:46:06+00:00] run decode=64 chunk=4 n=8 seed=127 +[2026-06-11T22:46:16+00:00] done decode=64 chunk=4 +[2026-06-11T22:46:16+00:00] run decode=64 chunk=5 n=8 seed=128 +[2026-06-11T22:46:25+00:00] done decode=64 chunk=5 +[2026-06-11T22:46:25+00:00] run decode=64 chunk=6 n=8 seed=129 +[2026-06-11T22:46:35+00:00] done decode=64 chunk=6 +[2026-06-11T22:46:35+00:00] run decode=64 chunk=7 n=8 seed=130 +[2026-06-11T22:46:45+00:00] done decode=64 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s1p8_sc1p0_decode64_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 17.12205216748313 5.048381543880808 0.08074211955692533 0.43983155582002104 0.0353818925269293 63 63 63840 65542 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s1p8_sc1p0_decode64_n64/sc1p0 +sc1p0 pre_eos 19.704070943129636 5.0846337107330655 0.08349244922756466 0.454836673504813 0.0365940256584242 0 0 59468 63371 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s1p8_sc1p0_decode64_n64/sc1p0 +[2026-06-11T22:47:14+00:00] done +===== DONE \2026-06-11T22:34:52+00:00 gpu=3 m=-0.4 s=1.8 gamma=0.25 ===== +===== START \2026-06-11T22:34:52+00:00 gpu=3 m=-0.4 s=2.0 gamma=0.25 ===== +checkpoint=runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck128_unfixed_norm_stateprobadd_selfcond_ce_fast_20260609_230847/step_170000.pt +use_ema=1 +step=170000 +decode_steps=64 +n=64 chunk_n=8 gpu=3 +out_base=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612 +[2026-06-11T22:47:14+00:00] infer step=170000 decode=64 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s2p0_sc1p0_decode64_n64 +[2026-06-11T22:47:14+00:00] run decode=64 chunk=0 n=8 seed=123 +[2026-06-11T22:47:24+00:00] done decode=64 chunk=0 +[2026-06-11T22:47:24+00:00] run decode=64 chunk=1 n=8 seed=124 +[2026-06-11T22:47:34+00:00] done decode=64 chunk=1 +[2026-06-11T22:47:34+00:00] run decode=64 chunk=2 n=8 seed=125 +[2026-06-11T22:47:44+00:00] done decode=64 chunk=2 +[2026-06-11T22:47:44+00:00] run decode=64 chunk=3 n=8 seed=126 +[2026-06-11T22:47:53+00:00] done decode=64 chunk=3 +[2026-06-11T22:47:53+00:00] run decode=64 chunk=4 n=8 seed=127 +[2026-06-11T22:48:03+00:00] done decode=64 chunk=4 +[2026-06-11T22:48:03+00:00] run decode=64 chunk=5 n=8 seed=128 +[2026-06-11T22:48:13+00:00] done decode=64 chunk=5 +[2026-06-11T22:48:13+00:00] run decode=64 chunk=6 n=8 seed=129 +[2026-06-11T22:48:23+00:00] done decode=64 chunk=6 +[2026-06-11T22:48:23+00:00] run decode=64 chunk=7 n=8 seed=130 +[2026-06-11T22:48:32+00:00] done decode=64 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s2p0_sc1p0_decode64_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 19.332869239573228 5.078073205362629 0.08395001449518608 0.45138698159846197 0.03645157844947283 63 63 63394 65539 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s2p0_sc1p0_decode64_n64/sc1p0 +sc1p0 pre_eos 22.574710721942342 5.115753742175942 0.08684738163274972 0.46698768550678876 0.037716486951579545 0 0 58973 63341 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p4_s2p0_sc1p0_decode64_n64/sc1p0 +[2026-06-11T22:49:02+00:00] done +===== DONE \2026-06-11T22:34:52+00:00 gpu=3 m=-0.4 s=2.0 gamma=0.25 ===== +===== START \2026-06-11T22:34:52+00:00 gpu=3 m=-0.3 s=1.4 gamma=0.25 ===== +checkpoint=runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck128_unfixed_norm_stateprobadd_selfcond_ce_fast_20260609_230847/step_170000.pt +use_ema=1 +step=170000 +decode_steps=64 +n=64 chunk_n=8 gpu=3 +out_base=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612 +[2026-06-11T22:49:02+00:00] infer step=170000 decode=64 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p3_s1p4_sc1p0_decode64_n64 +[2026-06-11T22:49:02+00:00] run decode=64 chunk=0 n=8 seed=123 +[2026-06-11T22:49:11+00:00] done decode=64 chunk=0 +[2026-06-11T22:49:11+00:00] run decode=64 chunk=1 n=8 seed=124 +[2026-06-11T22:49:21+00:00] done decode=64 chunk=1 +[2026-06-11T22:49:21+00:00] run decode=64 chunk=2 n=8 seed=125 +[2026-06-11T22:49:31+00:00] done decode=64 chunk=2 +[2026-06-11T22:49:31+00:00] run decode=64 chunk=3 n=8 seed=126 +[2026-06-11T22:49:41+00:00] done decode=64 chunk=3 +[2026-06-11T22:49:41+00:00] run decode=64 chunk=4 n=8 seed=127 +[2026-06-11T22:49:50+00:00] done decode=64 chunk=4 +[2026-06-11T22:49:50+00:00] run decode=64 chunk=5 n=8 seed=128 +[2026-06-11T22:50:00+00:00] done decode=64 chunk=5 +[2026-06-11T22:50:00+00:00] run decode=64 chunk=6 n=8 seed=129 +[2026-06-11T22:50:10+00:00] done decode=64 chunk=6 +[2026-06-11T22:50:10+00:00] run decode=64 chunk=7 n=8 seed=130 +[2026-06-11T22:50:20+00:00] done decode=64 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p3_s1p4_sc1p0_decode64_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 16.928129598265656 5.071954158306068 0.0861945144778058 0.4528299003728378 0.03430361372144549 62 62 62556 65445 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p3_s1p4_sc1p0_decode64_n64/sc1p0 +sc1p0 pre_eos 19.28985262070039 5.103880944521719 0.08894075347326258 0.46729325679682077 0.03540283538075789 0 0 58467 63413 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p3_s1p4_sc1p0_decode64_n64/sc1p0 +[2026-06-11T22:50:49+00:00] done +===== DONE \2026-06-11T22:34:52+00:00 gpu=3 m=-0.3 s=1.4 gamma=0.25 ===== +===== START \2026-06-11T22:34:52+00:00 gpu=3 m=-0.3 s=1.6 gamma=0.25 ===== +checkpoint=runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck128_unfixed_norm_stateprobadd_selfcond_ce_fast_20260609_230847/step_170000.pt +use_ema=1 +step=170000 +decode_steps=64 +n=64 chunk_n=8 gpu=3 +out_base=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612 +[2026-06-11T22:50:49+00:00] infer step=170000 decode=64 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p3_s1p6_sc1p0_decode64_n64 +[2026-06-11T22:50:49+00:00] run decode=64 chunk=0 n=8 seed=123 +[2026-06-11T22:50:59+00:00] done decode=64 chunk=0 +[2026-06-11T22:50:59+00:00] run decode=64 chunk=1 n=8 seed=124 +[2026-06-11T22:51:08+00:00] done decode=64 chunk=1 +[2026-06-11T22:51:08+00:00] run decode=64 chunk=2 n=8 seed=125 +[2026-06-11T22:51:18+00:00] done decode=64 chunk=2 +[2026-06-11T22:51:18+00:00] run decode=64 chunk=3 n=8 seed=126 +[2026-06-11T22:51:28+00:00] done decode=64 chunk=3 +[2026-06-11T22:51:28+00:00] run decode=64 chunk=4 n=8 seed=127 +[2026-06-11T22:51:37+00:00] done decode=64 chunk=4 +[2026-06-11T22:51:37+00:00] run decode=64 chunk=5 n=8 seed=128 +[2026-06-11T22:51:47+00:00] done decode=64 chunk=5 +[2026-06-11T22:51:47+00:00] run decode=64 chunk=6 n=8 seed=129 +[2026-06-11T22:51:57+00:00] done decode=64 chunk=6 +[2026-06-11T22:51:57+00:00] run decode=64 chunk=7 n=8 seed=130 +[2026-06-11T22:52:07+00:00] done decode=64 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p3_s1p6_sc1p0_decode64_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 17.3935578124969 5.063727260360727 0.08130180585873696 0.43846553092752033 0.03590346364621655 64 64 63426 65509 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p3_s1p6_sc1p0_decode64_n64/sc1p0 +sc1p0 pre_eos 19.776817910921572 5.090824696528666 0.08390980287105466 0.45255278915852504 0.03706213264839823 0 0 59306 63461 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p3_s1p6_sc1p0_decode64_n64/sc1p0 +[2026-06-11T22:52:21+00:00] done +===== DONE \2026-06-11T22:34:52+00:00 gpu=3 m=-0.3 s=1.6 gamma=0.25 ===== +===== START \2026-06-11T22:34:52+00:00 gpu=3 m=-0.3 s=1.8 gamma=0.25 ===== +checkpoint=runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck128_unfixed_norm_stateprobadd_selfcond_ce_fast_20260609_230847/step_170000.pt +use_ema=1 +step=170000 +decode_steps=64 +n=64 chunk_n=8 gpu=3 +out_base=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612 +[2026-06-11T22:52:21+00:00] infer step=170000 decode=64 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p3_s1p8_sc1p0_decode64_n64 +[2026-06-11T22:52:21+00:00] run decode=64 chunk=0 n=8 seed=123 +[2026-06-11T22:52:31+00:00] done decode=64 chunk=0 +[2026-06-11T22:52:31+00:00] run decode=64 chunk=1 n=8 seed=124 +[2026-06-11T22:52:40+00:00] done decode=64 chunk=1 +[2026-06-11T22:52:40+00:00] run decode=64 chunk=2 n=8 seed=125 +[2026-06-11T22:52:50+00:00] done decode=64 chunk=2 +[2026-06-11T22:52:50+00:00] run decode=64 chunk=3 n=8 seed=126 +[2026-06-11T22:53:00+00:00] done decode=64 chunk=3 +[2026-06-11T22:53:00+00:00] run decode=64 chunk=4 n=8 seed=127 +[2026-06-11T22:53:09+00:00] done decode=64 chunk=4 +[2026-06-11T22:53:09+00:00] run decode=64 chunk=5 n=8 seed=128 +[2026-06-11T22:53:19+00:00] done decode=64 chunk=5 +[2026-06-11T22:53:19+00:00] run decode=64 chunk=6 n=8 seed=129 +[2026-06-11T22:53:29+00:00] done decode=64 chunk=6 +[2026-06-11T22:53:29+00:00] run decode=64 chunk=7 n=8 seed=130 +[2026-06-11T22:53:39+00:00] done decode=64 chunk=7 +merged 64 samples -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p3_s1p8_sc1p0_decode64_n64/sc1p0/samples64.txt +loading scorer /e2e-data/evad-tech-vla/wanghan58/models/flowtext_scorers/gpt2-large-standard dtype=fp32 device=cuda +run kind ppl mean_entropy distinct_1 distinct_2 top_token_mass eos_rows eos_total ppl_tokens t5_tokens path +sc1p0 raw_full 17.7391049377987 5.071205544786745 0.08228292385167099 0.4488547055502144 0.035754616206317716 64 64 63672 65530 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p3_s1p8_sc1p0_decode64_n64/sc1p0 +sc1p0 pre_eos 20.245708500544218 5.104042020201639 0.08494044242768009 0.4633746671498574 0.03691624125543581 0 0 59522 63468 /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p3_s1p8_sc1p0_decode64_n64/sc1p0 +[2026-06-11T22:53:53+00:00] done +===== DONE \2026-06-11T22:34:52+00:00 gpu=3 m=-0.3 s=1.8 gamma=0.25 ===== +===== START \2026-06-11T22:34:52+00:00 gpu=3 m=-0.3 s=2.0 gamma=0.25 ===== +checkpoint=runs/owt_t5_elftokenized_full_len1024_C1_to_1024_pow1_d768_l12_h12_gbs512_2x8gpu_50ep_lr3e3_ema0p9999_elfopt_not5_bottleneck128_unfixed_norm_stateprobadd_selfcond_ce_fast_20260609_230847/step_170000.pt +use_ema=1 +step=170000 +decode_steps=64 +n=64 chunk_n=8 gpu=3 +out_base=/e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612 +[2026-06-11T22:53:53+00:00] infer step=170000 decode=64 -> /e2e-data/evad-tech-vla/wanghan58/workspace/LTA_openwebtext_dualt/docs/lta_samples/metrics_20260612/owt_t5_not5_bottleneck128_norm_stateprobadd_selfcond_ce_fast_lr3e3_ema0p9999_logitnorm64_focused_ppl24_ent515_step170000_ema_dgamma0p25_tschedlogit_normal_mn0p3_s2p0_sc1p0_decode64_n64 +[2026-06-11T22:53:53+00:00] run decode=64 chunk=0 n=8 seed=123 +[2026-06-11T22:54:03+00:00] done decode=64 chunk=0 +[2026-06-11T22:54:03+00:00] run decode=64 chunk=1 n=8 seed=124