| {"event": "hub", "repo": "AdaptiveChunking/ali_mem-skel_s2", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 2.7} |
| {"event": "start", "run": "ali_mem-skel_s2", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["skel"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "t": 2.7} |
| {"event": "probe_schedule", "n": 45, "steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 200, 256, 400, 512, 600, 800, 1000, 1024, 1200, 1400, 1600, 1800, 2000, 2048, 2200, 2400, 2600, 2800, 3000, 3200, 3400, 3600, 3800, 4000, 4096, 4200, 4400, 4600, 4800, 5000, 5200, 5400, 5600, 5800, 6000], "t": 2.7} |
| {"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10933.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.628225, "cps": 85.414062, "mean_p": 0.513249, "boundary_entropy": 0.686295, "bpb": 8.12598}, "tr": {"chunks": 8187.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.452245, "cps": 63.960938, "mean_p": 0.498002, "boundary_entropy": 0.686826, "bpb": 8.097536}, "fi": {"chunks": 13540.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.719792, "cps": 105.78125, "mean_p": 0.517169, "boundary_entropy": 0.686661, "bpb": 8.065117}}, "bpb": {"en": 8.12598, "tr": 8.09754, "fi": 8.06512}, "gini_cpc": 0.099077, "gini_cps": 0.109267, "bpic": 1.737048, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7198781817505053, "fi": 1.1457544508645294}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 4.3} |
| {"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.62735, "bpb": 8.11854, "chunk_loss": 0.062086, "F": 0.59833, "G": 0.50836, "chunks": 612.69, "bytes_per_chunk": 1.671, "boundary_entropy": 0.69105, "router_noise": 0.0, "grad_norm": 11.4935, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.027, "probe_sec": 0.0187, "probe_extra_seqs": 0, "n_rows": 0, "extra_fwd_seqs": 0, "wall_s": 0.0, "overhead_frac_est": 0.0, "chunk_mismatch": 0, "n_loo": 0, "mean_delta_loo": NaN, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "read_hit_rate": {"skel": 0.0}, "mean_gate": {"skel": null}, "n_reads": 0, "n_cand": 4583, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 919, "admitted": 919, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 919, "staged_skel": 0, "mean_count": 4.986942291259766, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.019, "probe_overhead_seq_frac": 0.0, "util_loss": -0.0, "alpha": 0.03, "t": 6.1} |
| {"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10909.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.626846, "cps": 85.226562, "mean_p": 0.513143, "boundary_entropy": 0.686304, "bpb": 8.098043}, "tr": {"chunks": 8206.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.453295, "cps": 64.109375, "mean_p": 0.498082, "boundary_entropy": 0.686828, "bpb": 8.079541}, "fi": {"chunks": 13506.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.717984, "cps": 105.515625, "mean_p": 0.517038, "boundary_entropy": 0.686672, "bpb": 8.042084}}, "bpb": {"en": 8.09804, "tr": 8.07954, "fi": 8.04208}, "gini_cpc": 0.098135, "gini_cps": 0.108315, "bpic": 1.739125, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7231362614556248, "fi": 1.1453917229119928}, "inventory": {"size": 919, "admitted": 919, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 919, "staged_skel": 0, "mean_count": 4.986942291259766, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 7.0} |
| {"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10880.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.62518, "cps": 85.0, "mean_p": 0.512981, "boundary_entropy": 0.686318, "bpb": 8.035036}, "tr": {"chunks": 8247.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.45556, "cps": 64.429688, "mean_p": 0.498248, "boundary_entropy": 0.686832, "bpb": 8.035417}, "fi": {"chunks": 13467.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.715911, "cps": 105.210938, "mean_p": 0.5168, "boundary_entropy": 0.686692, "bpb": 7.982606}}, "bpb": {"en": 8.03504, "tr": 8.03542, "fi": 7.98261}, "gini_cpc": 0.096606, "gini_cps": 0.106768, "bpic": 1.740566, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7286864065835692, "fi": 1.1451284419394785}, "inventory": {"size": 1751, "admitted": 1751, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 1751, "staged_skel": 0, "mean_count": 5.095945358276367, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 8.7} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10717.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.615813, "cps": 83.726562, "mean_p": 0.512109, "boundary_entropy": 0.686379, "bpb": 7.823606}, "tr": {"chunks": 8390.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.463459, "cps": 65.546875, "mean_p": 0.498676, "boundary_entropy": 0.686846, "bpb": 7.881871}, "fi": {"chunks": 13215.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.702514, "cps": 103.242188, "mean_p": 0.515686, "boundary_entropy": 0.686774, "bpb": 7.780405}}, "bpb": {"en": 7.82361, "tr": 7.88187, "fi": 7.7804}, "gini_cpc": 0.089444, "gini_cps": 0.099519, "bpic": 1.755213, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7525966812891371, "fi": 1.14079122920813}, "inventory": {"size": 2917, "admitted": 2917, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 2917, "staged_skel": 0, "mean_count": 6.1021599769592285, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 10.8} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10329.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.593518, "cps": 80.695312, "mean_p": 0.510366, "boundary_entropy": 0.686482, "bpb": 7.545944}, "tr": {"chunks": 8506.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.469867, "cps": 66.453125, "mean_p": 0.499199, "boundary_entropy": 0.686869, "bpb": 7.676336}, "fi": {"chunks": 12731.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.676785, "cps": 99.460938, "mean_p": 0.513635, "boundary_entropy": 0.686908, "bpb": 7.524371}}, "bpb": {"en": 7.54594, "tr": 7.67634, "fi": 7.52437}, "gini_cpc": 0.079271, "gini_cps": 0.089231, "bpic": 1.79725, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7916635868887795, "fi": 1.1402930503658901}, "inventory": {"size": 3795, "admitted": 3795, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 3795, "staged_skel": 0, "mean_count": 7.169960975646973, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 16.8} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 9638.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.553813, "cps": 75.296875, "mean_p": 0.507315, "boundary_entropy": 0.686613, "bpb": 7.294618}, "tr": {"chunks": 8636.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.477048, "cps": 67.46875, "mean_p": 0.499602, "boundary_entropy": 0.686901, "bpb": 7.478655}, "fi": {"chunks": 11862.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.630588, "cps": 92.671875, "mean_p": 0.510265, "boundary_entropy": 0.687081, "bpb": 7.299749}}, "bpb": {"en": 7.29462, "tr": 7.47866, "fi": 7.29975}, "gini_cpc": 0.061609, "gini_cps": 0.071365, "bpic": 1.882533, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8613889186381628, "fi": 1.1386316053612187}, "inventory": {"size": 4903, "admitted": 4903, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 4903, "staged_skel": 0, "mean_count": 7.478686332702637, "n_util_obs": 1, "util_count_spearman": NaN}, "t": 18.8} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3756.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.215825, "cps": 29.34375, "mean_p": 0.484458, "boundary_entropy": 0.686166, "bpb": 6.724935}, "tr": {"chunks": 7306.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.40358, "cps": 57.078125, "mean_p": 0.494463, "boundary_entropy": 0.686997, "bpb": 7.013801}, "fi": {"chunks": 4572.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.243049, "cps": 35.71875, "mean_p": 0.486442, "boundary_entropy": 0.686895, "bpb": 6.786972}}, "bpb": {"en": 6.72494, "tr": 7.0138, "fi": 6.78697}, "gini_cpc": 0.145132, "gini_cps": 0.151379, "bpic": 3.628758, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.8699399195833426, "fi": 1.1261412710902787}, "inventory": {"size": 14835, "admitted": 14835, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 14835, "staged_skel": 0, "mean_count": 4.882035255432129, "n_util_obs": 4, "util_count_spearman": 0.4}, "t": 21.9} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2611.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.150032, "cps": 20.398438, "mean_p": 0.486768, "boundary_entropy": 0.68695, "bpb": 6.304832}, "tr": {"chunks": 4648.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.256753, "cps": 36.3125, "mean_p": 0.490177, "boundary_entropy": 0.687559, "bpb": 6.651757}, "fi": {"chunks": 1384.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.073574, "cps": 10.8125, "mean_p": 0.479302, "boundary_entropy": 0.686963, "bpb": 6.421109}}, "bpb": {"en": 6.30483, "tr": 6.65176, "fi": 6.42111}, "gini_cpc": 0.254225, "gini_cps": 0.251764, "bpic": 6.563925, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.711326266927452, "fi": 0.49038983013129395}, "inventory": {"size": 36347, "admitted": 36347, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 36347, "staged_skel": 0, "mean_count": 3.3309764862060547, "n_util_obs": 6, "util_count_spearman": -0.7142857142857143}, "t": 32.4} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 36.2} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.2126, "bpb": 6.0775, "chunk_loss": 0.034047, "F": 0.22351, "G": 0.48835, "chunks": 228.88, "bytes_per_chunk": 4.474, "boundary_entropy": 0.6916, "router_noise": 0.0, "grad_norm": 2.4794, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 0.611, "probe_sec": 0.0307, "probe_extra_seqs": 15, "n_rows": 2, "extra_fwd_seqs": 15, "wall_s": 0.0238, "overhead_frac_est": 0.15625, "chunk_mismatch": 0, "n_loo": 13, "mean_delta_loo": -0.00458, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.03084, "trunc_abs_frac": 0.20883, "mean_delta_full": 0.09204, "read_hit_rate": {"skel": 0.54444}, "mean_gate": {"skel": 0.11917}, "n_reads": 1948, "n_cand": 3578, "read_suppressed_frac": 0.1052, "dropout_reads_logged": 1948, "inventory": {"size": 64578, "admitted": 64578, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 64578, "staged_skel": 0, "mean_count": 2.7373409271240234, "n_util_obs": 13, "util_count_spearman": -0.4010989010989011}, "probe_overhead_sec_total": 2.535, "probe_overhead_seq_frac": 0.4786, "util_loss": 0.02402, "alpha": 0.03, "t": 36.7} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4809.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.276332, "cps": 37.570312, "mean_p": 0.485229, "boundary_entropy": 0.685511, "bpb": 5.552486}, "tr": {"chunks": 5395.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.298017, "cps": 42.148438, "mean_p": 0.486556, "boundary_entropy": 0.686483, "bpb": 5.956141}, "fi": {"chunks": 4703.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.250013, "cps": 36.742188, "mean_p": 0.483942, "boundary_entropy": 0.686326, "bpb": 5.717873}}, "bpb": {"en": 5.55249, "tr": 5.95614, "fi": 5.71787}, "gini_cpc": 0.038821, "gini_cps": 0.030947, "bpic": 3.805729, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0784753935760056, "fi": 0.9047580136295534}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 3.065704345703125, "n_util_obs": 15, "util_count_spearman": -0.3464285714285714}, "t": 40.2} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 48.7} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.25058, "bpb": 4.68959, "chunk_loss": 0.029486, "F": 0.15817, "G": 0.45236, "chunks": 161.97, "bytes_per_chunk": 6.322, "boundary_entropy": 0.68192, "router_noise": 0.0, "grad_norm": 1.6661, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.245, "probe_sec": 0.0513, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0365, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01195, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.26112, "trunc_abs_frac": 0.5735, "mean_delta_full": -0.09882, "read_hit_rate": {"skel": 0.53369}, "mean_gate": {"skel": 0.1195}, "n_reads": 1798, "n_cand": 3369, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 1798, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 4.3301849365234375, "n_util_obs": 24, "util_count_spearman": -0.5947826086956521}, "probe_overhead_sec_total": 4.742, "probe_overhead_seq_frac": 0.4882, "util_loss": 0.000564, "alpha": 0.03, "t": 48.9} |
| {"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3735.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.214618, "cps": 29.179688, "mean_p": 0.414449, "boundary_entropy": 0.643366, "bpb": 4.150663}, "tr": {"chunks": 4971.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.274595, "cps": 38.835938, "mean_p": 0.437624, "boundary_entropy": 0.659174, "bpb": 4.392612}, "fi": {"chunks": 3924.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.208601, "cps": 30.65625, "mean_p": 0.428402, "boundary_entropy": 0.659633, "bpb": 4.344803}}, "bpb": {"en": 4.15066, "tr": 4.39261, "fi": 4.3448}, "gini_cpc": 0.063048, "gini_cps": 0.065241, "bpic": 4.491845, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2794600375761456, "fi": 0.9719650063760438}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 5.3640594482421875, "n_util_obs": 30, "util_count_spearman": -0.32458286985539486}, "t": 54.9} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 63.1} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.79044, "bpb": 4.02575, "chunk_loss": 0.032099, "F": 0.21024, "G": 0.38685, "chunks": 215.28, "bytes_per_chunk": 4.757, "boundary_entropy": 0.62015, "router_noise": 0.0, "grad_norm": 1.8841, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.3, "probe_sec": 0.0586, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0204, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0044, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.44498, "trunc_abs_frac": 0.44498, "mean_delta_full": -0.10205, "read_hit_rate": {"skel": 0.6389}, "mean_gate": {"skel": 0.12261}, "n_reads": 2562, "n_cand": 4010, "read_suppressed_frac": 0.1038, "dropout_reads_logged": 2562, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 6.190887451171875, "n_util_obs": 36, "util_count_spearman": -0.16962676962676962}, "probe_overhead_sec_total": 7.22, "probe_overhead_seq_frac": 0.4921, "util_loss": 0.000311, "alpha": 0.03, "t": 63.9} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 80.7} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3528.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202724, "cps": 27.5625, "mean_p": 0.376068, "boundary_entropy": 0.589523, "bpb": 3.645356}, "tr": {"chunks": 3745.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206872, "cps": 29.257812, "mean_p": 0.36234, "boundary_entropy": 0.581378, "bpb": 3.816028}, "fi": {"chunks": 4309.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229068, "cps": 33.664062, "mean_p": 0.409002, "boundary_entropy": 0.635892, "bpb": 3.864379}}, "bpb": {"en": 3.64536, "tr": 3.81603, "fi": 3.86438}, "gini_cpc": 0.0275, "gini_cps": 0.044955, "bpic": 4.89829, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0204619465860696, "fi": 1.1299524142228259}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 7.7770538330078125, "n_util_obs": 44, "util_count_spearman": -0.0773784355179704}, "t": 81.4} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.65171, "tr": 3.82742, "fi": 3.86118}, "bytes_per_chunk": {"en": 5.362, "tr": 6.5, "fi": 4.898}, "bpb_mean": 3.7801, "frozen_inventory": true, "t": 82.6} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.53958, "bpb": 3.66384, "chunk_loss": 0.030442, "F": 0.17499, "G": 0.36368, "chunks": 179.19, "bytes_per_chunk": 5.715, "boundary_entropy": 0.58842, "router_noise": 0.0, "grad_norm": 1.685, "lr": 0.000201, "router_trainable": true, "sec_per_step": 0.406, "probe_sec": 0.0326, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02097, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02896, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -3.28682, "trunc_abs_frac": 0.80547, "mean_delta_full": -0.0063, "read_hit_rate": {"skel": 0.6116}, "mean_gate": {"skel": 0.1265}, "n_reads": 2288, "n_cand": 3741, "read_suppressed_frac": 0.097, "dropout_reads_logged": 2288, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 7.812469482421875, "n_util_obs": 44, "util_count_spearman": -0.06892177589852008}, "probe_overhead_sec_total": 9.705, "probe_overhead_seq_frac": 0.493, "util_loss": 0.000616, "alpha": 0.03, "t": 84.2} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 94.4} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.38528, "bpb": 3.44123, "chunk_loss": 0.031089, "F": 0.19547, "G": 0.33946, "chunks": 200.16, "bytes_per_chunk": 5.116, "boundary_entropy": 0.54086, "router_noise": 0.0, "grad_norm": 1.5896, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.206, "probe_sec": 0.0393, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03159, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04058, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.30269, "trunc_abs_frac": 0.91314, "mean_delta_full": -0.07935, "read_hit_rate": {"skel": 0.65792}, "mean_gate": {"skel": 0.12753}, "n_reads": 2858, "n_cand": 4344, "read_suppressed_frac": 0.0892, "dropout_reads_logged": 2858, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 9.742568969726562, "n_util_obs": 59, "util_count_spearman": -0.15073056691992986}, "probe_overhead_sec_total": 11.635, "probe_overhead_seq_frac": 0.4944, "util_loss": 0.000846, "alpha": 0.03, "t": 94.6} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3421.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196575, "cps": 26.726562, "mean_p": 0.340035, "boundary_entropy": 0.528744, "bpb": 3.35731}, "tr": {"chunks": 4338.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.239629, "cps": 33.890625, "mean_p": 0.331541, "boundary_entropy": 0.531472, "bpb": 3.519999}, "fi": {"chunks": 4259.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.22641, "cps": 33.273438, "mean_p": 0.366489, "boundary_entropy": 0.583007, "bpb": 3.66983}}, "bpb": {"en": 3.35731, "tr": 3.52, "fi": 3.66983}, "gini_cpc": 0.043317, "gini_cps": 0.050868, "bpic": 4.720586, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2190177861544265, "fi": 1.151772758980931}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 9.947921752929688, "n_util_obs": 60, "util_count_spearman": -0.20022228396776884}, "t": 96.1} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 107.5} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.29196, "bpb": 3.3066, "chunk_loss": 0.030462, "F": 0.18182, "G": 0.301, "chunks": 186.19, "bytes_per_chunk": 5.5, "boundary_entropy": 0.47657, "router_noise": 0.0, "grad_norm": 1.6076, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.264, "probe_sec": 0.0649, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05866, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 4e-05, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.10882, "trunc_abs_frac": 0.11744, "mean_delta_full": -0.19238, "read_hit_rate": {"skel": 0.63103}, "mean_gate": {"skel": 0.1336}, "n_reads": 2550, "n_cand": 4041, "read_suppressed_frac": 0.0945, "dropout_reads_logged": 2550, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 11.68255615234375, "n_util_obs": 68, "util_count_spearman": -0.04660839027369546}, "probe_overhead_sec_total": 13.902, "probe_overhead_seq_frac": 0.4953, "util_loss": 0.000941, "alpha": 0.03, "t": 107.8} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 118.2} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.18037, "bpb": 3.14561, "chunk_loss": 0.030969, "F": 0.20273, "G": 0.28696, "chunks": 207.59, "bytes_per_chunk": 4.933, "boundary_entropy": 0.4293, "router_noise": 0.0, "grad_norm": 1.2471, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.213, "probe_sec": 0.0584, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05413, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03212, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.81439, "trunc_abs_frac": 0.81439, "mean_delta_full": 0.12134, "read_hit_rate": {"skel": 0.70177}, "mean_gate": {"skel": 0.14923}, "n_reads": 3019, "n_cand": 4302, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 3019, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 14.045440673828125, "n_util_obs": 76, "util_count_spearman": 0.07956254272043746}, "probe_overhead_sec_total": 16.308, "probe_overhead_seq_frac": 0.496, "util_loss": 0.00121, "alpha": 0.03, "t": 118.4} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 128.6} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3387.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194622, "cps": 26.460938, "mean_p": 0.280027, "boundary_entropy": 0.392397, "bpb": 2.915264}, "tr": {"chunks": 3556.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196432, "cps": 27.78125, "mean_p": 0.278399, "boundary_entropy": 0.438151, "bpb": 3.202803}, "fi": {"chunks": 4311.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229174, "cps": 33.679688, "mean_p": 0.283955, "boundary_entropy": 0.448798, "bpb": 3.370768}}, "bpb": {"en": 2.91526, "tr": 3.2028, "fi": 3.37077}, "gini_cpc": 0.03714, "gini_cps": 0.054736, "bpic": 5.041052, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0092996541246337, "fi": 1.1775383577603495}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 16.467803955078125, "n_util_obs": 90, "util_count_spearman": 0.12771124007079063}, "t": 129.2} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.92578, "tr": 3.20973, "fi": 3.36431}, "bytes_per_chunk": {"en": 5.176, "tr": 5.593, "fi": 4.861}, "bpb_mean": 3.1666, "frozen_inventory": true, "t": 130.4} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.09842, "bpb": 3.02738, "chunk_loss": 0.030494, "F": 0.18683, "G": 0.27193, "chunks": 191.31, "bytes_per_chunk": 5.352, "boundary_entropy": 0.39759, "router_noise": 0.0, "grad_norm": 1.1303, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 0.26, "probe_sec": 0.0398, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03096, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09188, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 22.39453, "trunc_abs_frac": 1.07481, "mean_delta_full": -0.0026, "read_hit_rate": {"skel": 0.66824}, "mean_gate": {"skel": 0.17684}, "n_reads": 2981, "n_cand": 4461, "read_suppressed_frac": 0.109, "dropout_reads_logged": 2981, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 16.514083862304688, "n_util_obs": 90, "util_count_spearman": 0.11615521385925774}, "probe_overhead_sec_total": 18.829, "probe_overhead_seq_frac": 0.4965, "util_loss": 0.001597, "alpha": 0.03, "t": 131.4} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 141.1} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.00456, "bpb": 2.89197, "chunk_loss": 0.030905, "F": 0.20663, "G": 0.26595, "chunks": 211.59, "bytes_per_chunk": 4.839, "boundary_entropy": 0.34457, "router_noise": 0.0, "grad_norm": 1.7478, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 0.197, "probe_sec": 0.0251, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02004, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08382, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 7.942, "trunc_abs_frac": 1.80033, "mean_delta_full": -0.01017, "read_hit_rate": {"skel": 0.71177}, "mean_gate": {"skel": 0.19268}, "n_reads": 3635, "n_cand": 5107, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 3635, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 19.183441162109375, "n_util_obs": 100, "util_count_spearman": 0.22509450945094509}, "probe_overhead_sec_total": 21.065, "probe_overhead_seq_frac": 0.4969, "util_loss": 0.00219, "alpha": 0.03, "t": 141.3} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 150.9} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.89936, "bpb": 2.7402, "chunk_loss": 0.031191, "F": 0.21753, "G": 0.26989, "chunks": 222.75, "bytes_per_chunk": 4.597, "boundary_entropy": 0.31238, "router_noise": 0.0, "grad_norm": 1.3416, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.196, "probe_sec": 0.0286, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02346, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04272, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -2.43914, "trunc_abs_frac": 0.50691, "mean_delta_full": 0.02791, "read_hit_rate": {"skel": 0.74805}, "mean_gate": {"skel": 0.20342}, "n_reads": 4133, "n_cand": 5525, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 4133, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 22.030471801757812, "n_util_obs": 105, "util_count_spearman": 0.18461682006171823}, "probe_overhead_sec_total": 22.934, "probe_overhead_seq_frac": 0.4972, "util_loss": 0.003297, "alpha": 0.03, "t": 151.1} |
| {"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3606.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207206, "cps": 28.171875, "mean_p": 0.264557, "boundary_entropy": 0.314349, "bpb": 2.691783}, "tr": {"chunks": 3588.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198199, "cps": 28.03125, "mean_p": 0.257707, "boundary_entropy": 0.356695, "bpb": 3.059382}, "fi": {"chunks": 4141.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220137, "cps": 32.351562, "mean_p": 0.267538, "boundary_entropy": 0.350972, "bpb": 3.21921}}, "bpb": {"en": 2.69178, "tr": 3.05938, "fi": 3.21921}, "gini_cpc": 0.02338, "gini_cps": 0.032525, "bpic": 5.005029, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9565337117435697, "fi": 1.062409009278831}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 22.655075073242188, "n_util_obs": 107, "util_count_spearman": 0.21754549239809357}, "t": 153.7} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 161.8} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.9053, "bpb": 2.74877, "chunk_loss": 0.030823, "F": 0.20483, "G": 0.2584, "chunks": 209.75, "bytes_per_chunk": 4.882, "boundary_entropy": 0.3026, "router_noise": 0.0, "grad_norm": 1.6162, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 0.218, "probe_sec": 0.0281, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0227, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.25362, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.33975, "trunc_abs_frac": 0.6042, "mean_delta_full": 0.32617, "read_hit_rate": {"skel": 0.71929}, "mean_gate": {"skel": 0.19494}, "n_reads": 3800, "n_cand": 5283, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 3800, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 24.876632690429688, "n_util_obs": 114, "util_count_spearman": 0.2824897885625536}, "probe_overhead_sec_total": 24.872, "probe_overhead_seq_frac": 0.4974, "util_loss": 0.003638, "alpha": 0.03, "t": 162.0} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 172.4} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3609.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207378, "cps": 28.195312, "mean_p": 0.256366, "boundary_entropy": 0.287216, "bpb": 2.583698}, "tr": {"chunks": 3897.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215268, "cps": 30.445312, "mean_p": 0.247716, "boundary_entropy": 0.297387, "bpb": 2.927345}, "fi": {"chunks": 4454.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.236776, "cps": 34.796875, "mean_p": 0.267699, "boundary_entropy": 0.308373, "bpb": 3.136999}}, "bpb": {"en": 2.5837, "tr": 2.92734, "fi": 3.137}, "gini_cpc": 0.029721, "gini_cps": 0.047101, "bpic": 4.743478, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0380471789096004, "fi": 1.141761954336599}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 27.734634399414062, "n_util_obs": 123, "util_count_spearman": 0.3022976478671945}, "t": 173.1} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.59682, "tr": 2.93478, "fi": 3.13636}, "bytes_per_chunk": {"en": 4.851, "tr": 5.164, "fi": 4.568}, "bpb_mean": 2.88932, "frozen_inventory": true, "t": 174.7} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.88637, "bpb": 2.72146, "chunk_loss": 0.030739, "F": 0.20395, "G": 0.25084, "chunks": 208.84, "bytes_per_chunk": 4.903, "boundary_entropy": 0.28321, "router_noise": 0.0, "grad_norm": 1.1756, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 0.278, "probe_sec": 0.0709, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06347, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01104, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.03675, "trunc_abs_frac": 0.21499, "mean_delta_full": 0.27588, "read_hit_rate": {"skel": 0.72279}, "mean_gate": {"skel": 0.20309}, "n_reads": 3851, "n_cand": 5328, "read_suppressed_frac": 0.1059, "dropout_reads_logged": 3851, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 27.792510986328125, "n_util_obs": 123, "util_count_spearman": 0.3006387788086814}, "probe_overhead_sec_total": 27.114, "probe_overhead_seq_frac": 0.4977, "util_loss": 0.00438, "alpha": 0.03, "t": 175.8} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 187.7} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.8379, "bpb": 2.65153, "chunk_loss": 0.030831, "F": 0.20709, "G": 0.25373, "chunks": 212.06, "bytes_per_chunk": 4.829, "boundary_entropy": 0.27028, "router_noise": 0.0, "grad_norm": 1.2217, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.244, "probe_sec": 0.0257, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02112, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16633, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.4376, "trunc_abs_frac": 0.80168, "mean_delta_full": 0.15145, "read_hit_rate": {"skel": 0.7319}, "mean_gate": {"skel": 0.19902}, "n_reads": 3912, "n_cand": 5345, "read_suppressed_frac": 0.0969, "dropout_reads_logged": 3912, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 30.744583129882812, "n_util_obs": 136, "util_count_spearman": 0.24411584398677338}, "probe_overhead_sec_total": 29.923, "probe_overhead_seq_frac": 0.4978, "util_loss": 0.00486, "alpha": 0.03, "t": 188.0} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 201.0} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.9865, "bpb": 2.86591, "chunk_loss": 0.031188, "F": 0.21942, "G": 0.25699, "chunks": 224.69, "bytes_per_chunk": 4.557, "boundary_entropy": 0.26162, "router_noise": 0.0, "grad_norm": 1.2695, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.0296, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02193, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.034, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.1964, "trunc_abs_frac": 10.88822, "mean_delta_full": 0.08028, "read_hit_rate": {"skel": 0.72007}, "mean_gate": {"skel": 0.20253}, "n_reads": 3861, "n_cand": 5362, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 3861, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 33.717803955078125, "n_util_obs": 148, "util_count_spearman": 0.22135907594844512}, "probe_overhead_sec_total": 32.767, "probe_overhead_seq_frac": 0.498, "util_loss": 0.005472, "alpha": 0.03, "t": 201.2} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 214.8} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.78073, "bpb": 2.56905, "chunk_loss": 0.031018, "F": 0.2175, "G": 0.25568, "chunks": 222.72, "bytes_per_chunk": 4.598, "boundary_entropy": 0.25272, "router_noise": 0.0, "grad_norm": 1.2122, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.278, "probe_sec": 0.141, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13198, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04147, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.06887, "trunc_abs_frac": 6.39134, "mean_delta_full": 0.20589, "read_hit_rate": {"skel": 0.74872}, "mean_gate": {"skel": 0.1975}, "n_reads": 4103, "n_cand": 5480, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 4103, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 36.720672607421875, "n_util_obs": 168, "util_count_spearman": 0.20411683234348943}, "probe_overhead_sec_total": 35.95, "probe_overhead_seq_frac": 0.4981, "util_loss": 0.006093, "alpha": 0.03, "t": 215.1} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 227.2} |
| {"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3731.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.214388, "cps": 29.148438, "mean_p": 0.256465, "boundary_entropy": 0.255849, "bpb": 2.433336}, "tr": {"chunks": 3931.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.217146, "cps": 30.710938, "mean_p": 0.233218, "boundary_entropy": 0.256321, "bpb": 2.75861}, "fi": {"chunks": 4463.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.237255, "cps": 34.867188, "mean_p": 0.258698, "boundary_entropy": 0.265701, "bpb": 2.974328}}, "bpb": {"en": 2.43334, "tr": 2.75861, "fi": 2.97433}, "gini_cpc": 0.022794, "gini_cps": 0.040247, "bpic": 4.678928, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0128645321532095, "fi": 1.1066591382480764}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 39.65386962890625, "n_util_obs": 178, "util_count_spearman": 0.21421143036259982}, "t": 228.1} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.44304, "tr": 2.76709, "fi": 2.97605}, "bytes_per_chunk": {"en": 4.707, "tr": 5.14, "fi": 4.578}, "bpb_mean": 2.72873, "frozen_inventory": true, "t": 229.4} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.85821, "bpb": 2.68082, "chunk_loss": 0.030998, "F": 0.21579, "G": 0.25295, "chunks": 220.97, "bytes_per_chunk": 4.634, "boundary_entropy": 0.25056, "router_noise": 0.0, "grad_norm": 1.1988, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 0.31, "probe_sec": 0.1205, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11279, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22924, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.89676, "trunc_abs_frac": 6.5957, "mean_delta_full": 0.29403, "read_hit_rate": {"skel": 0.73393}, "mean_gate": {"skel": 0.20034}, "n_reads": 3870, "n_cand": 5273, "read_suppressed_frac": 0.101, "dropout_reads_logged": 3870, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 39.713836669921875, "n_util_obs": 178, "util_count_spearman": 0.21536425377039703}, "probe_overhead_sec_total": 38.571, "probe_overhead_seq_frac": 0.4982, "util_loss": 0.005778, "alpha": 0.03, "t": 230.6} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 242.8} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.80715, "bpb": 2.60716, "chunk_loss": 0.030809, "F": 0.21039, "G": 0.24518, "chunks": 215.44, "bytes_per_chunk": 4.753, "boundary_entropy": 0.24372, "router_noise": 0.0, "grad_norm": 1.1242, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.25, "probe_sec": 0.0432, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03481, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04654, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.78116, "trunc_abs_frac": 0.77685, "mean_delta_full": 0.62476, "read_hit_rate": {"skel": 0.74419}, "mean_gate": {"skel": 0.19626}, "n_reads": 4096, "n_cand": 5504, "read_suppressed_frac": 0.0942, "dropout_reads_logged": 4096, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 42.747039794921875, "n_util_obs": 184, "util_count_spearman": 0.23701454124815105}, "probe_overhead_sec_total": 41.631, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.005843, "alpha": 0.03, "t": 243.1} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 254.3} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61177, "bpb": 2.32529, "chunk_loss": 0.030786, "F": 0.2092, "G": 0.24639, "chunks": 214.22, "bytes_per_chunk": 4.78, "boundary_entropy": 0.23426, "router_noise": 0.0, "grad_norm": 1.4436, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.229, "probe_sec": 0.0503, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04316, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.25168, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.50542, "trunc_abs_frac": 0.72491, "mean_delta_full": 0.19141, "read_hit_rate": {"skel": 0.77869}, "mean_gate": {"skel": 0.19733}, "n_reads": 4268, "n_cand": 5481, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 4268, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 45.80859375, "n_util_obs": 190, "util_count_spearman": 0.22658848393985231}, "probe_overhead_sec_total": 44.178, "probe_overhead_seq_frac": 0.4984, "util_loss": 0.006844, "alpha": 0.03, "t": 254.5} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 266.8} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68465, "bpb": 2.43044, "chunk_loss": 0.030809, "F": 0.20999, "G": 0.24344, "chunks": 215.03, "bytes_per_chunk": 4.762, "boundary_entropy": 0.22972, "router_noise": 0.0, "grad_norm": 1.036, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.252, "probe_sec": 0.1817, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.16638, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08525, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.26862, "trunc_abs_frac": 0.83761, "mean_delta_full": 0.0437, "read_hit_rate": {"skel": 0.75333}, "mean_gate": {"skel": 0.19797}, "n_reads": 4016, "n_cand": 5331, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 4016, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 48.85585021972656, "n_util_obs": 209, "util_count_spearman": 0.19030587196089074}, "probe_overhead_sec_total": 47.094, "probe_overhead_seq_frac": 0.4985, "util_loss": 0.007328, "alpha": 0.03, "t": 267.2} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 279.9} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3736.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.214676, "cps": 29.1875, "mean_p": 0.249518, "boundary_entropy": 0.23337, "bpb": 2.332153}, "tr": {"chunks": 3924.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.21676, "cps": 30.65625, "mean_p": 0.229313, "boundary_entropy": 0.224392, "bpb": 2.633335}, "fi": {"chunks": 4439.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235979, "cps": 34.679688, "mean_p": 0.255231, "boundary_entropy": 0.236721, "bpb": 2.846476}}, "bpb": {"en": 2.33215, "tr": 2.63333, "fi": 2.84648}, "gini_cpc": 0.021279, "gini_cps": 0.038736, "bpic": 4.688983, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0097077737774838, "fi": 1.0992349144886182}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 51.809906005859375, "n_util_obs": 218, "util_count_spearman": 0.18985073555017842}, "t": 280.9} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.34183, "tr": 2.64197, "fi": 2.85153}, "bytes_per_chunk": {"en": 4.668, "tr": 5.19, "fi": 4.51}, "bpb_mean": 2.61178, "frozen_inventory": true, "t": 282.3} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.7349, "bpb": 2.50293, "chunk_loss": 0.030876, "F": 0.21387, "G": 0.24397, "chunks": 219.0, "bytes_per_chunk": 4.676, "boundary_entropy": 0.22173, "router_noise": 0.0, "grad_norm": 0.9934, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 0.329, "probe_sec": 0.0397, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02824, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05866, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -1.31074, "trunc_abs_frac": 0.58356, "mean_delta_full": 0.07095, "read_hit_rate": {"skel": 0.75591}, "mean_gate": {"skel": 0.19994}, "n_reads": 4187, "n_cand": 5539, "read_suppressed_frac": 0.0948, "dropout_reads_logged": 4187, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 51.87495422363281, "n_util_obs": 218, "util_count_spearman": 0.19072472883463318}, "probe_overhead_sec_total": 49.827, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.007707, "alpha": 0.03, "t": 283.6} |
| {"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3465.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199104, "cps": 27.070312, "mean_p": 0.238323, "boundary_entropy": 0.235486, "bpb": 2.331357}, "tr": {"chunks": 3779.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.20875, "cps": 29.523438, "mean_p": 0.224474, "boundary_entropy": 0.240147, "bpb": 2.638571}, "fi": {"chunks": 4225.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.224603, "cps": 33.007812, "mean_p": 0.250781, "boundary_entropy": 0.244382, "bpb": 2.832221}}, "bpb": {"en": 2.33136, "tr": 2.63857, "fi": 2.83222}, "gini_cpc": 0.026878, "gini_cps": 0.044177, "bpic": 4.946552, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0484487873981327, "fi": 1.1280691204671855}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 53.28179931640625, "n_util_obs": 223, "util_count_spearman": 0.22458347119129377}, "t": 290.7} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 298.6} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63574, "bpb": 2.35988, "chunk_loss": 0.030848, "F": 0.207, "G": 0.24022, "chunks": 211.97, "bytes_per_chunk": 4.831, "boundary_entropy": 0.2208, "router_noise": 0.0, "grad_norm": 1.1167, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.305, "probe_sec": 0.0319, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02031, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0233, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.10951, "trunc_abs_frac": 3.40352, "mean_delta_full": 0.30579, "read_hit_rate": {"skel": 0.74819}, "mean_gate": {"skel": 0.19901}, "n_reads": 3925, "n_cand": 5246, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 3925, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 54.94160461425781, "n_util_obs": 226, "util_count_spearman": 0.22735153580975861}, "probe_overhead_sec_total": 52.821, "probe_overhead_seq_frac": 0.4987, "util_loss": 0.008221, "alpha": 0.03, "t": 298.9} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 311.7} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.73806, "bpb": 2.5075, "chunk_loss": 0.030844, "F": 0.21207, "G": 0.24162, "chunks": 217.16, "bytes_per_chunk": 4.715, "boundary_entropy": 0.22192, "router_noise": 0.0, "grad_norm": 0.9485, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.0965, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09082, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10898, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 19.19395, "trunc_abs_frac": 0.99218, "mean_delta_full": 0.00749, "read_hit_rate": {"skel": 0.74651}, "mean_gate": {"skel": 0.19464}, "n_reads": 3952, "n_cand": 5294, "read_suppressed_frac": 0.0946, "dropout_reads_logged": 3952, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 57.997955322265625, "n_util_obs": 240, "util_count_spearman": 0.1803604143997524}, "probe_overhead_sec_total": 55.556, "probe_overhead_seq_frac": 0.4987, "util_loss": 0.007804, "alpha": 0.03, "t": 312.0} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 324.4} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63971, "bpb": 2.36559, "chunk_loss": 0.030702, "F": 0.20837, "G": 0.23843, "chunks": 213.38, "bytes_per_chunk": 4.799, "boundary_entropy": 0.21081, "router_noise": 0.0, "grad_norm": 0.8846, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.1553, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14789, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07155, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.09976, "trunc_abs_frac": 0.15098, "mean_delta_full": -0.25604, "read_hit_rate": {"skel": 0.77034}, "mean_gate": {"skel": 0.19547}, "n_reads": 4223, "n_cand": 5482, "read_suppressed_frac": 0.0954, "dropout_reads_logged": 4223, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 61.081939697265625, "n_util_obs": 253, "util_count_spearman": 0.17549295054508368}, "probe_overhead_sec_total": 58.516, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.008095, "alpha": 0.03, "t": 324.8} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 337.3} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3620.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20801, "cps": 28.28125, "mean_p": 0.240772, "boundary_entropy": 0.216438, "bpb": 2.264448}, "tr": {"chunks": 3777.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.208639, "cps": 29.507812, "mean_p": 0.217489, "boundary_entropy": 0.212031, "bpb": 2.553862}, "fi": {"chunks": 4385.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233108, "cps": 34.257812, "mean_p": 0.253899, "boundary_entropy": 0.218559, "bpb": 2.725659}}, "bpb": {"en": 2.26445, "tr": 2.55386, "fi": 2.72566}, "gini_cpc": 0.025751, "gini_cps": 0.043286, "bpic": 4.815142, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.003025520326749, "fi": 1.1206584339538022}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 64.05838012695312, "n_util_obs": 266, "util_count_spearman": 0.16752242090995378}, "t": 338.2} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.27312, "tr": 2.56271, "fi": 2.73323}, "bytes_per_chunk": {"en": 4.864, "tr": 5.305, "fi": 4.632}, "bpb_mean": 2.52302, "frozen_inventory": true, "t": 339.5} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61495, "bpb": 2.32988, "chunk_loss": 0.030634, "F": 0.2049, "G": 0.23643, "chunks": 209.81, "bytes_per_chunk": 4.881, "boundary_entropy": 0.21146, "router_noise": 0.0, "grad_norm": 0.8804, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 0.312, "probe_sec": 0.0283, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02164, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09731, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.85286, "trunc_abs_frac": 0.80666, "mean_delta_full": 0.06824, "read_hit_rate": {"skel": 0.76183}, "mean_gate": {"skel": 0.19812}, "n_reads": 4056, "n_cand": 5324, "read_suppressed_frac": 0.1048, "dropout_reads_logged": 4056, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 64.12042236328125, "n_util_obs": 266, "util_count_spearman": 0.16950836218186371}, "probe_overhead_sec_total": 61.61, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.008418, "alpha": 0.03, "t": 340.4} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 352.8} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63293, "bpb": 2.35581, "chunk_loss": 0.030614, "F": 0.20483, "G": 0.23385, "chunks": 209.75, "bytes_per_chunk": 4.882, "boundary_entropy": 0.20413, "router_noise": 0.0, "grad_norm": 0.9073, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.254, "probe_sec": 0.0382, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02945, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03589, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.63168, "trunc_abs_frac": 0.8977, "mean_delta_full": 0.10962, "read_hit_rate": {"skel": 0.75444}, "mean_gate": {"skel": 0.19911}, "n_reads": 3991, "n_cand": 5290, "read_suppressed_frac": 0.097, "dropout_reads_logged": 3991, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 67.21611022949219, "n_util_obs": 277, "util_count_spearman": 0.19456710576316508}, "probe_overhead_sec_total": 64.544, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.00822, "alpha": 0.03, "t": 353.0} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 366.1} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.66845, "bpb": 2.40707, "chunk_loss": 0.030978, "F": 0.21979, "G": 0.24332, "chunks": 225.06, "bytes_per_chunk": 4.55, "boundary_entropy": 0.20076, "router_noise": 0.0, "grad_norm": 0.8724, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.266, "probe_sec": 0.0518, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04553, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00643, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.59017, "trunc_abs_frac": 0.65477, "mean_delta_full": -0.10884, "read_hit_rate": {"skel": 0.77172}, "mean_gate": {"skel": 0.19604}, "n_reads": 4290, "n_cand": 5559, "read_suppressed_frac": 0.1105, "dropout_reads_logged": 4290, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 70.30226135253906, "n_util_obs": 286, "util_count_spearman": 0.17155136826757394}, "probe_overhead_sec_total": 67.765, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.009163, "alpha": 0.03, "t": 366.3} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 378.5} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59923, "bpb": 2.30719, "chunk_loss": 0.030685, "F": 0.2074, "G": 0.23448, "chunks": 212.38, "bytes_per_chunk": 4.822, "boundary_entropy": 0.2043, "router_noise": 0.0, "grad_norm": 0.9198, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.249, "probe_sec": 0.0312, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02464, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09729, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.60492, "trunc_abs_frac": 0.89318, "mean_delta_full": 0.51667, "read_hit_rate": {"skel": 0.74634}, "mean_gate": {"skel": 0.19871}, "n_reads": 3975, "n_cand": 5326, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 3975, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 73.37617492675781, "n_util_obs": 297, "util_count_spearman": 0.1527336455126636}, "probe_overhead_sec_total": 70.232, "probe_overhead_seq_frac": 0.499, "util_loss": 0.010219, "alpha": 0.03, "t": 378.8} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 392.3} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3634.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208815, "cps": 28.390625, "mean_p": 0.237953, "boundary_entropy": 0.203094, "bpb": 2.209689}, "tr": {"chunks": 3855.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.212948, "cps": 30.117188, "mean_p": 0.21596, "boundary_entropy": 0.191347, "bpb": 2.486538}, "fi": {"chunks": 4418.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234863, "cps": 34.515625, "mean_p": 0.247712, "boundary_entropy": 0.204702, "bpb": 2.637292}}, "bpb": {"en": 2.20969, "tr": 2.48654, "fi": 2.63729}, "gini_cpc": 0.026446, "gini_cps": 0.043896, "bpic": 4.764592, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0197953519259981, "fi": 1.124742291356534}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 76.4287109375, "n_util_obs": 310, "util_count_spearman": 0.16556237838269708}, "t": 393.1} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.21923, "tr": 2.49451, "fi": 2.64473}, "bytes_per_chunk": {"en": 4.862, "tr": 5.208, "fi": 4.638}, "bpb_mean": 2.45282, "frozen_inventory": true, "t": 394.5} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59611, "bpb": 2.3027, "chunk_loss": 0.030612, "F": 0.20395, "G": 0.23073, "chunks": 208.84, "bytes_per_chunk": 4.903, "boundary_entropy": 0.19523, "router_noise": 0.0, "grad_norm": 0.8153, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 0.333, "probe_sec": 0.0371, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02661, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07573, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -2.99674, "trunc_abs_frac": 2.18047, "mean_delta_full": 0.01777, "read_hit_rate": {"skel": 0.747}, "mean_gate": {"skel": 0.2014}, "n_reads": 3924, "n_cand": 5253, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 3924, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 76.48847961425781, "n_util_obs": 310, "util_count_spearman": 0.16524561782295139}, "probe_overhead_sec_total": 73.534, "probe_overhead_seq_frac": 0.499, "util_loss": 0.00871, "alpha": 0.03, "t": 395.4} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 409.3} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64306, "bpb": 2.37043, "chunk_loss": 0.030547, "F": 0.20392, "G": 0.22635, "chunks": 208.81, "bytes_per_chunk": 4.904, "boundary_entropy": 0.19161, "router_noise": 0.0, "grad_norm": 0.8665, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.285, "probe_sec": 0.1777, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.17044, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07832, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 3.38355, "trunc_abs_frac": 0.38738, "mean_delta_full": 0.05049, "read_hit_rate": {"skel": 0.74475}, "mean_gate": {"skel": 0.20453}, "n_reads": 3904, "n_cand": 5242, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 3904, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 79.5738525390625, "n_util_obs": 322, "util_count_spearman": 0.15659834682347892}, "probe_overhead_sec_total": 76.713, "probe_overhead_seq_frac": 0.499, "util_loss": 0.008664, "alpha": 0.03, "t": 409.6} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 423.1} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.66273, "bpb": 2.39881, "chunk_loss": 0.030873, "F": 0.21426, "G": 0.23763, "chunks": 219.41, "bytes_per_chunk": 4.667, "boundary_entropy": 0.1906, "router_noise": 0.0, "grad_norm": 0.8574, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.276, "probe_sec": 0.1233, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10986, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18983, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.96556, "trunc_abs_frac": 0.91614, "mean_delta_full": 0.56921, "read_hit_rate": {"skel": 0.75302}, "mean_gate": {"skel": 0.20317}, "n_reads": 3988, "n_cand": 5296, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 3988, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 82.64486694335938, "n_util_obs": 330, "util_count_spearman": 0.1693947025295475}, "probe_overhead_sec_total": 80.117, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.009828, "alpha": 0.03, "t": 423.4} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 440.1} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5831, "bpb": 2.28393, "chunk_loss": 0.030751, "F": 0.20929, "G": 0.23278, "chunks": 214.31, "bytes_per_chunk": 4.778, "boundary_entropy": 0.19073, "router_noise": 0.0, "grad_norm": 1.0097, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.337, "probe_sec": 0.0354, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02925, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20995, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.60987, "trunc_abs_frac": 0.60987, "mean_delta_full": -0.33887, "read_hit_rate": {"skel": 0.76985}, "mean_gate": {"skel": 0.20119}, "n_reads": 4188, "n_cand": 5440, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 4188, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 85.72128295898438, "n_util_obs": 347, "util_count_spearman": 0.16845271808000722}, "probe_overhead_sec_total": 83.258, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.009813, "alpha": 0.03, "t": 440.3} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 453.9} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3584.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205942, "cps": 28.0, "mean_p": 0.233714, "boundary_entropy": 0.19304, "bpb": 2.169207}, "tr": {"chunks": 3840.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.21212, "cps": 30.0, "mean_p": 0.211952, "boundary_entropy": 0.181993, "bpb": 2.420317}, "fi": {"chunks": 4486.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238477, "cps": 35.046875, "mean_p": 0.247117, "boundary_entropy": 0.186909, "bpb": 2.587439}}, "bpb": {"en": 2.16921, "tr": 2.42032, "fi": 2.58744}, "gini_cpc": 0.033038, "gini_cps": 0.05049, "bpic": 4.763392, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0299989741242572, "fi": 1.1579865231304631}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 88.73696899414062, "n_util_obs": 357, "util_count_spearman": 0.16657306004508926}, "t": 454.7} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.18013, "tr": 2.42989, "fi": 2.59286}, "bytes_per_chunk": {"en": 4.909, "tr": 5.254, "fi": 4.56}, "bpb_mean": 2.40096, "frozen_inventory": true, "t": 456.2} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51854, "bpb": 2.19079, "chunk_loss": 0.030543, "F": 0.20294, "G": 0.23081, "chunks": 207.81, "bytes_per_chunk": 4.928, "boundary_entropy": 0.18924, "router_noise": 0.0, "grad_norm": 0.9138, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.337, "probe_sec": 0.0524, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04473, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12149, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.39153, "trunc_abs_frac": 1.07658, "mean_delta_full": 0.10382, "read_hit_rate": {"skel": 0.78328}, "mean_gate": {"skel": 0.19278}, "n_reads": 4225, "n_cand": 5394, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 4225, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 88.80168151855469, "n_util_obs": 357, "util_count_spearman": 0.16670532692668852}, "probe_overhead_sec_total": 86.443, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.009497, "alpha": 0.03, "t": 457.2} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 471.5} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57683, "bpb": 2.27489, "chunk_loss": 0.030713, "F": 0.21121, "G": 0.23542, "chunks": 216.28, "bytes_per_chunk": 4.735, "boundary_entropy": 0.18689, "router_noise": 0.0, "grad_norm": 0.9056, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.292, "probe_sec": 0.0309, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02033, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02262, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.70915, "trunc_abs_frac": 1.73113, "mean_delta_full": 0.12625, "read_hit_rate": {"skel": 0.7725}, "mean_gate": {"skel": 0.19486}, "n_reads": 4241, "n_cand": 5490, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 4241, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 91.91867065429688, "n_util_obs": 368, "util_count_spearman": 0.19340168681671407}, "probe_overhead_sec_total": 90.199, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.00975, "alpha": 0.03, "t": 471.7} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 485.3} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53801, "bpb": 2.21888, "chunk_loss": 0.030568, "F": 0.20499, "G": 0.22458, "chunks": 209.91, "bytes_per_chunk": 4.878, "boundary_entropy": 0.1809, "router_noise": 0.0, "grad_norm": 0.8265, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.275, "probe_sec": 0.0337, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02814, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06397, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.81857, "trunc_abs_frac": 2.32444, "mean_delta_full": 0.25441, "read_hit_rate": {"skel": 0.75446}, "mean_gate": {"skel": 0.2016}, "n_reads": 3976, "n_cand": 5270, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 3976, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 94.98248291015625, "n_util_obs": 376, "util_count_spearman": 0.1866093421212733}, "probe_overhead_sec_total": 92.635, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.009336, "alpha": 0.03, "t": 485.5} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 497.9} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56573, "bpb": 2.25887, "chunk_loss": 0.030754, "F": 0.21326, "G": 0.23285, "chunks": 218.38, "bytes_per_chunk": 4.689, "boundary_entropy": 0.17491, "router_noise": 0.0, "grad_norm": 0.8131, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.252, "probe_sec": 0.0484, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04085, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09185, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.96167, "trunc_abs_frac": 0.8273, "mean_delta_full": -0.12817, "read_hit_rate": {"skel": 0.77225}, "mean_gate": {"skel": 0.19943}, "n_reads": 4235, "n_cand": 5484, "read_suppressed_frac": 0.1048, "dropout_reads_logged": 4235, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 98.04681396484375, "n_util_obs": 380, "util_count_spearman": 0.17358442494985074}, "probe_overhead_sec_total": 95.212, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.009995, "alpha": 0.03, "t": 498.1} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 514.5} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3566.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204907, "cps": 27.859375, "mean_p": 0.230802, "boundary_entropy": 0.18701, "bpb": 2.134578}, "tr": {"chunks": 3890.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.214882, "cps": 30.390625, "mean_p": 0.213102, "boundary_entropy": 0.171575, "bpb": 2.353274}, "fi": {"chunks": 4622.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.245707, "cps": 36.109375, "mean_p": 0.248899, "boundary_entropy": 0.178957, "bpb": 2.524087}}, "bpb": {"en": 2.13458, "tr": 2.35327, "fi": 2.52409}, "gini_cpc": 0.040872, "gini_cps": 0.058288, "bpic": 4.697135, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0486772131390363, "fi": 1.1991150092875635}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 101.061279296875, "n_util_obs": 397, "util_count_spearman": 0.1525489406755254}, "t": 515.3} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.14881, "tr": 2.36104, "fi": 2.52864}, "bytes_per_chunk": {"en": 4.926, "tr": 5.218, "fi": 4.418}, "bpb_mean": 2.34617, "frozen_inventory": true, "t": 516.6} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53689, "bpb": 2.21727, "chunk_loss": 0.030445, "F": 0.19885, "G": 0.2217, "chunks": 203.62, "bytes_per_chunk": 5.029, "boundary_entropy": 0.17874, "router_noise": 0.0, "grad_norm": 0.8042, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.393, "probe_sec": 0.0824, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07427, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08073, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.19336, "trunc_abs_frac": 0.92883, "mean_delta_full": 0.2077, "read_hit_rate": {"skel": 0.74654}, "mean_gate": {"skel": 0.19576}, "n_reads": 3879, "n_cand": 5196, "read_suppressed_frac": 0.098, "dropout_reads_logged": 3879, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 101.12022399902344, "n_util_obs": 398, "util_count_spearman": 0.1540163197474639}, "probe_overhead_sec_total": 98.349, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.008455, "alpha": 0.03, "t": 517.7} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 533.2} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55037, "bpb": 2.23672, "chunk_loss": 0.030681, "F": 0.2084, "G": 0.23094, "chunks": 213.41, "bytes_per_chunk": 4.798, "boundary_entropy": 0.1725, "router_noise": 0.0, "grad_norm": 0.7387, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.314, "probe_sec": 0.0398, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02716, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08524, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.45673, "trunc_abs_frac": 6.6428, "mean_delta_full": 0.11707, "read_hit_rate": {"skel": 0.76249}, "mean_gate": {"skel": 0.19624}, "n_reads": 4090, "n_cand": 5364, "read_suppressed_frac": 0.0936, "dropout_reads_logged": 4090, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 104.19384765625, "n_util_obs": 414, "util_count_spearman": 0.15044123233459117}, "probe_overhead_sec_total": 101.251, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.008836, "alpha": 0.03, "t": 533.4} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 561.0} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52991, "bpb": 2.20719, "chunk_loss": 0.030823, "F": 0.21118, "G": 0.23111, "chunks": 216.25, "bytes_per_chunk": 4.735, "boundary_entropy": 0.17782, "router_noise": 0.0, "grad_norm": 0.8716, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.57, "probe_sec": 0.0469, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02342, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11265, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.48979, "trunc_abs_frac": 0.39249, "mean_delta_full": 0.84241, "read_hit_rate": {"skel": 0.76359}, "mean_gate": {"skel": 0.20247}, "n_reads": 4102, "n_cand": 5372, "read_suppressed_frac": 0.0943, "dropout_reads_logged": 4102, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 107.2825927734375, "n_util_obs": 421, "util_count_spearman": 0.1489713006283724}, "probe_overhead_sec_total": 104.15, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.010275, "alpha": 0.03, "t": 561.9} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 578.0} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48106, "bpb": 2.13672, "chunk_loss": 0.030635, "F": 0.20468, "G": 0.22786, "chunks": 209.59, "bytes_per_chunk": 4.886, "boundary_entropy": 0.16971, "router_noise": 0.0, "grad_norm": 0.7276, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.326, "probe_sec": 0.0495, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04141, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19776, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.58404, "trunc_abs_frac": 0.67451, "mean_delta_full": 0.25271, "read_hit_rate": {"skel": 0.76363}, "mean_gate": {"skel": 0.20525}, "n_reads": 3964, "n_cand": 5191, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 3964, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 110.3746337890625, "n_util_obs": 429, "util_count_spearman": 0.17001352132099784}, "probe_overhead_sec_total": 106.864, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.009021, "alpha": 0.03, "t": 578.2} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 593.4} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3609.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207378, "cps": 28.195312, "mean_p": 0.230964, "boundary_entropy": 0.175984, "bpb": 2.107104}, "tr": {"chunks": 3782.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.208916, "cps": 29.546875, "mean_p": 0.207951, "boundary_entropy": 0.171987, "bpb": 2.304475}, "fi": {"chunks": 4474.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.23784, "cps": 34.953125, "mean_p": 0.244795, "boundary_entropy": 0.169122, "bpb": 2.476941}}, "bpb": {"en": 2.1071, "tr": 2.30447, "fi": 2.47694}, "gini_cpc": 0.031045, "gini_cps": 0.048602, "bpic": 4.781458, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0074145318542749, "fi": 1.1468888602833285}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 113.38240051269531, "n_util_obs": 440, "util_count_spearman": 0.1802468083608188}, "t": 594.2} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.11986, "tr": 2.31516, "fi": 2.48649}, "bytes_per_chunk": {"en": 4.883, "tr": 5.328, "fi": 4.614}, "bpb_mean": 2.30717, "frozen_inventory": true, "t": 595.5} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64385, "bpb": 2.37158, "chunk_loss": 0.030849, "F": 0.21405, "G": 0.23522, "chunks": 219.19, "bytes_per_chunk": 4.672, "boundary_entropy": 0.17269, "router_noise": 0.0, "grad_norm": 0.8491, "lr": 0.000245, "router_trainable": true, "sec_per_step": 0.385, "probe_sec": 0.0604, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03361, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13755, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.03469, "trunc_abs_frac": 0.21631, "mean_delta_full": 0.05048, "read_hit_rate": {"skel": 0.7468}, "mean_gate": {"skel": 0.19541}, "n_reads": 3973, "n_cand": 5320, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 3973, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 113.4422607421875, "n_util_obs": 440, "util_count_spearman": 0.18050669242191425}, "probe_overhead_sec_total": 109.85, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.010179, "alpha": 0.03, "t": 597.5} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3613.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207608, "cps": 28.226562, "mean_p": 0.228524, "boundary_entropy": 0.17322, "bpb": 2.097364}, "tr": {"chunks": 3881.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.214384, "cps": 30.320312, "mean_p": 0.21246, "boundary_entropy": 0.157586, "bpb": 2.301996}, "fi": {"chunks": 4598.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.244431, "cps": 35.921875, "mean_p": 0.248203, "boundary_entropy": 0.166108, "bpb": 2.465116}}, "bpb": {"en": 2.09736, "tr": 2.302, "fi": 2.46512}, "gini_cpc": 0.036837, "gini_cps": 0.054306, "bpic": 4.691697, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0326407281126928, "fi": 1.1773707497496142}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 116.329833984375, "n_util_obs": 449, "util_count_spearman": 0.17982290995016437}, "t": 614.8} |
| {"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 618.4} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.47958, "bpb": 2.13458, "chunk_loss": 0.030579, "F": 0.19965, "G": 0.22049, "chunks": 204.44, "bytes_per_chunk": 5.009, "boundary_entropy": 0.16943, "router_noise": 0.0, "grad_norm": 0.8348, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 0.431, "probe_sec": 0.0819, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06871, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15708, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -3.11008, "trunc_abs_frac": 0.48368, "mean_delta_full": 0.01248, "read_hit_rate": {"skel": 0.75735}, "mean_gate": {"skel": 0.20016}, "n_reads": 3939, "n_cand": 5201, "read_suppressed_frac": 0.0947, "dropout_reads_logged": 3939, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 116.51237487792969, "n_util_obs": 449, "util_count_spearman": 0.17990753787818448}, "probe_overhead_sec_total": 112.764, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.009602, "alpha": 0.03, "t": 619.0} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 637.3} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54124, "bpb": 2.22354, "chunk_loss": 0.030522, "F": 0.20377, "G": 0.22293, "chunks": 208.66, "bytes_per_chunk": 4.908, "boundary_entropy": 0.1654, "router_noise": 0.0, "grad_norm": 0.8077, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 0.371, "probe_sec": 0.0341, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02148, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16002, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.54231, "trunc_abs_frac": 0.54231, "mean_delta_full": -0.05518, "read_hit_rate": {"skel": 0.75238}, "mean_gate": {"skel": 0.20325}, "n_reads": 3947, "n_cand": 5246, "read_suppressed_frac": 0.096, "dropout_reads_logged": 3947, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 119.58438110351562, "n_util_obs": 458, "util_count_spearman": 0.14680801566085191}, "probe_overhead_sec_total": 115.476, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.009508, "alpha": 0.03, "t": 637.6} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 651.4} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49898, "bpb": 2.16257, "chunk_loss": 0.030582, "F": 0.20691, "G": 0.22601, "chunks": 211.88, "bytes_per_chunk": 4.833, "boundary_entropy": 0.16755, "router_noise": 0.0, "grad_norm": 0.7429, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.279, "probe_sec": 0.0462, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03942, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16879, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.90497, "trunc_abs_frac": 1.09712, "mean_delta_full": 0.06873, "read_hit_rate": {"skel": 0.75521}, "mean_gate": {"skel": 0.20062}, "n_reads": 4026, "n_cand": 5331, "read_suppressed_frac": 0.1051, "dropout_reads_logged": 4026, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 122.69609069824219, "n_util_obs": 470, "util_count_spearman": 0.14071019694316098}, "probe_overhead_sec_total": 118.129, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.011021, "alpha": 0.03, "t": 651.6} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 666.2} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3615.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207723, "cps": 28.242188, "mean_p": 0.227403, "boundary_entropy": 0.166239, "bpb": 2.082792}, "tr": {"chunks": 3903.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.2156, "cps": 30.492188, "mean_p": 0.20881, "boundary_entropy": 0.144791, "bpb": 2.283716}, "fi": {"chunks": 4496.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.239009, "cps": 35.125, "mean_p": 0.243419, "boundary_entropy": 0.161603, "bpb": 2.441267}}, "bpb": {"en": 2.08279, "tr": 2.28372, "fi": 2.44127}, "gini_cpc": 0.031491, "gini_cps": 0.048887, "bpic": 4.722157, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0379198514356416, "fi": 1.1506155465229468}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 125.72370910644531, "n_util_obs": 485, "util_count_spearman": 0.15957062431906607}, "t": 667.1} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.09368, "tr": 2.29181, "fi": 2.44949}, "bytes_per_chunk": {"en": 4.899, "tr": 5.203, "fi": 4.566}, "bpb_mean": 2.27833, "frozen_inventory": true, "t": 668.5} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.47398, "bpb": 2.1265, "chunk_loss": 0.030527, "F": 0.20511, "G": 0.224, "chunks": 210.03, "bytes_per_chunk": 4.875, "boundary_entropy": 0.16024, "router_noise": 0.0, "grad_norm": 0.6934, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 0.364, "probe_sec": 0.132, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12309, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13699, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.10969, "trunc_abs_frac": 0.58144, "mean_delta_full": -0.21777, "read_hit_rate": {"skel": 0.76975}, "mean_gate": {"skel": 0.19934}, "n_reads": 4122, "n_cand": 5355, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 4122, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 125.78704833984375, "n_util_obs": 486, "util_count_spearman": 0.1593334172077217}, "probe_overhead_sec_total": 120.575, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.011705, "alpha": 0.03, "t": 669.8} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 682.9} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49839, "bpb": 2.16171, "chunk_loss": 0.030552, "F": 0.20514, "G": 0.22302, "chunks": 210.06, "bytes_per_chunk": 4.875, "boundary_entropy": 0.15495, "router_noise": 0.0, "grad_norm": 0.8001, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.271, "probe_sec": 0.0435, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03837, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08275, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.28439, "trunc_abs_frac": 0.429, "mean_delta_full": 0.13384, "read_hit_rate": {"skel": 0.75248}, "mean_gate": {"skel": 0.19667}, "n_reads": 4028, "n_cand": 5353, "read_suppressed_frac": 0.1038, "dropout_reads_logged": 4028, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 128.85975646972656, "n_util_obs": 494, "util_count_spearman": 0.15989813564814487}, "probe_overhead_sec_total": 123.157, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.009873, "alpha": 0.03, "t": 683.3} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 696.0} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51245, "bpb": 2.18201, "chunk_loss": 0.030473, "F": 0.20251, "G": 0.22204, "chunks": 207.38, "bytes_per_chunk": 4.938, "boundary_entropy": 0.16092, "router_noise": 0.0, "grad_norm": 0.7821, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.258, "probe_sec": 0.0311, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02555, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16879, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.73291, "trunc_abs_frac": 0.73536, "mean_delta_full": 0.38962, "read_hit_rate": {"skel": 0.75613}, "mean_gate": {"skel": 0.19363}, "n_reads": 4009, "n_cand": 5302, "read_suppressed_frac": 0.0925, "dropout_reads_logged": 4009, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 131.94479370117188, "n_util_obs": 511, "util_count_spearman": 0.16000367647158598}, "probe_overhead_sec_total": 125.75, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.011684, "alpha": 0.03, "t": 696.2} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 709.4} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56003, "bpb": 2.25064, "chunk_loss": 0.030596, "F": 0.20926, "G": 0.22789, "chunks": 214.28, "bytes_per_chunk": 4.779, "boundary_entropy": 0.15888, "router_noise": 0.0, "grad_norm": 0.7154, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.268, "probe_sec": 0.0283, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02108, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13697, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.37156, "trunc_abs_frac": 0.97875, "mean_delta_full": -0.27338, "read_hit_rate": {"skel": 0.74791}, "mean_gate": {"skel": 0.19934}, "n_reads": 4035, "n_cand": 5395, "read_suppressed_frac": 0.1046, "dropout_reads_logged": 4035, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 135.04067993164062, "n_util_obs": 523, "util_count_spearman": 0.1418209175731959}, "probe_overhead_sec_total": 128.558, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.010238, "alpha": 0.03, "t": 709.6} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 723.2} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3587.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206114, "cps": 28.023438, "mean_p": 0.225349, "boundary_entropy": 0.162923, "bpb": 2.061035}, "tr": {"chunks": 3941.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.217699, "cps": 30.789062, "mean_p": 0.210608, "boundary_entropy": 0.144968, "bpb": 2.241363}, "fi": {"chunks": 4669.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.248206, "cps": 36.476562, "mean_p": 0.246781, "boundary_entropy": 0.147879, "bpb": 2.387578}}, "bpb": {"en": 2.06103, "tr": 2.24136, "fi": 2.38758}, "gini_cpc": 0.041757, "gini_cps": 0.05914, "bpic": 4.651308, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0562059919771727, "fi": 1.2042169449371491}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 138.07272338867188, "n_util_obs": 533, "util_count_spearman": 0.12939049786848125}, "t": 724.0} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.07291, "tr": 2.24993, "fi": 2.39755}, "bytes_per_chunk": {"en": 4.873, "tr": 5.193, "fi": 4.402}, "bpb_mean": 2.24013, "frozen_inventory": true, "t": 725.5} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 726.5} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49826, "bpb": 2.16154, "chunk_loss": 0.030736, "F": 0.21533, "G": 0.22913, "chunks": 220.5, "bytes_per_chunk": 4.644, "boundary_entropy": 0.15094, "router_noise": 0.0, "grad_norm": 0.7555, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.344, "probe_sec": 0.0485, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04202, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03473, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.41836, "trunc_abs_frac": 0.52661, "mean_delta_full": 0.22504, "read_hit_rate": {"skel": 0.79016}, "mean_gate": {"skel": 0.19816}, "n_reads": 4466, "n_cand": 5652, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 4466, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 138.1415557861328, "n_util_obs": 533, "util_count_spearman": 0.129440814161223}, "probe_overhead_sec_total": 131.355, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.01102, "alpha": 0.03, "t": 726.8} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 739.2} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45922, "bpb": 2.10521, "chunk_loss": 0.030494, "F": 0.20319, "G": 0.21985, "chunks": 208.06, "bytes_per_chunk": 4.922, "boundary_entropy": 0.15005, "router_noise": 0.0, "grad_norm": 0.6918, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.256, "probe_sec": 0.0547, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0453, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0946, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.38058, "trunc_abs_frac": 4.26054, "mean_delta_full": 0.26947, "read_hit_rate": {"skel": 0.76903}, "mean_gate": {"skel": 0.19883}, "n_reads": 4132, "n_cand": 5373, "read_suppressed_frac": 0.1053, "dropout_reads_logged": 4132, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 141.2508087158203, "n_util_obs": 542, "util_count_spearman": 0.12245293101071349}, "probe_overhead_sec_total": 134.022, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.011569, "alpha": 0.03, "t": 739.6} |
| {"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 752.3} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43663, "bpb": 2.07261, "chunk_loss": 0.03059, "F": 0.20822, "G": 0.22648, "chunks": 213.22, "bytes_per_chunk": 4.803, "boundary_entropy": 0.1551, "router_noise": 0.0, "grad_norm": 0.7662, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.26, "probe_sec": 0.123, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11288, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09244, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.48746, "trunc_abs_frac": 1.01881, "mean_delta_full": -0.0578, "read_hit_rate": {"skel": 0.77376}, "mean_gate": {"skel": 0.19183}, "n_reads": 4241, "n_cand": 5481, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 4241, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 144.3369140625, "n_util_obs": 550, "util_count_spearman": 0.13320272428948615}, "probe_overhead_sec_total": 136.861, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.011645, "alpha": 0.03, "t": 752.6} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 764.8} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46417, "bpb": 2.11235, "chunk_loss": 0.030509, "F": 0.20541, "G": 0.22092, "chunks": 210.34, "bytes_per_chunk": 4.868, "boundary_entropy": 0.14545, "router_noise": 0.0, "grad_norm": 0.7581, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.1039, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09835, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16121, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.4789, "trunc_abs_frac": 1.38236, "mean_delta_full": -0.05111, "read_hit_rate": {"skel": 0.77244}, "mean_gate": {"skel": 0.19674}, "n_reads": 4226, "n_cand": 5471, "read_suppressed_frac": 0.1058, "dropout_reads_logged": 4226, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 147.40785217285156, "n_util_obs": 562, "util_count_spearman": 0.14508107519769037}, "probe_overhead_sec_total": 139.553, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.011482, "alpha": 0.03, "t": 765.2} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 778.8} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3649.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209676, "cps": 28.507812, "mean_p": 0.225381, "boundary_entropy": 0.155446, "bpb": 2.040573}, "tr": {"chunks": 3844.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.21234, "cps": 30.03125, "mean_p": 0.204826, "boundary_entropy": 0.145056, "bpb": 2.232979}, "fi": {"chunks": 4390.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233374, "cps": 34.296875, "mean_p": 0.238874, "boundary_entropy": 0.143376, "bpb": 2.371184}}, "bpb": {"en": 2.04057, "tr": 2.23298, "fi": 2.37118}, "gini_cpc": 0.024105, "gini_cps": 0.041572, "bpic": 4.774215, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0127053035803604, "fi": 1.1130198086607839}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 150.4307861328125, "n_util_obs": 579, "util_count_spearman": 0.17232866185134155}, "t": 779.6} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.05483, "tr": 2.24293, "fi": 2.37882}, "bytes_per_chunk": {"en": 4.876, "tr": 5.315, "fi": 4.708}, "bpb_mean": 2.22552, "frozen_inventory": true, "t": 780.9} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41818, "bpb": 2.04601, "chunk_loss": 0.03051, "F": 0.2056, "G": 0.22141, "chunks": 210.53, "bytes_per_chunk": 4.864, "boundary_entropy": 0.15166, "router_noise": 0.0, "grad_norm": 0.7483, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 0.338, "probe_sec": 0.0298, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02476, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12704, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.04709, "trunc_abs_frac": 1.1619, "mean_delta_full": 0.09009, "read_hit_rate": {"skel": 0.76859}, "mean_gate": {"skel": 0.19041}, "n_reads": 4062, "n_cand": 5285, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 4062, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 150.49317932128906, "n_util_obs": 580, "util_count_spearman": 0.17406815489055977}, "probe_overhead_sec_total": 142.124, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.012943, "alpha": 0.03, "t": 782.1} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 794.8} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47081, "bpb": 2.12194, "chunk_loss": 0.030349, "F": 0.1958, "G": 0.21285, "chunks": 200.5, "bytes_per_chunk": 5.107, "boundary_entropy": 0.14665, "router_noise": 0.0, "grad_norm": 0.7247, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 0.259, "probe_sec": 0.0721, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06581, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.14945, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 13.2739, "trunc_abs_frac": 1.81611, "mean_delta_full": -0.00862, "read_hit_rate": {"skel": 0.73674}, "mean_gate": {"skel": 0.19857}, "n_reads": 3848, "n_cand": 5223, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 3848, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 153.59259033203125, "n_util_obs": 592, "util_count_spearman": 0.16564695832052687}, "probe_overhead_sec_total": 145.101, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.012422, "alpha": 0.03, "t": 795.1} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 808.4} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44735, "bpb": 2.08809, "chunk_loss": 0.030508, "F": 0.2034, "G": 0.22044, "chunks": 208.28, "bytes_per_chunk": 4.916, "boundary_entropy": 0.14738, "router_noise": 0.0, "grad_norm": 0.6949, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.271, "probe_sec": 0.0396, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03135, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07301, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.38297, "trunc_abs_frac": 0.53189, "mean_delta_full": 0.38228, "read_hit_rate": {"skel": 0.75368}, "mean_gate": {"skel": 0.19246}, "n_reads": 3999, "n_cand": 5306, "read_suppressed_frac": 0.0923, "dropout_reads_logged": 3999, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 156.69674682617188, "n_util_obs": 608, "util_count_spearman": 0.15285337452592046}, "probe_overhead_sec_total": 148.185, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.011599, "alpha": 0.03, "t": 808.6} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 821.8} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45351, "bpb": 2.09698, "chunk_loss": 0.030657, "F": 0.21039, "G": 0.2248, "chunks": 215.44, "bytes_per_chunk": 4.753, "boundary_entropy": 0.15241, "router_noise": 0.0, "grad_norm": 0.6906, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.268, "probe_sec": 0.0861, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08062, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08602, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": -0.89223, "trunc_abs_frac": 0.5314, "mean_delta_full": -0.1212, "read_hit_rate": {"skel": 0.77659}, "mean_gate": {"skel": 0.18907}, "n_reads": 4147, "n_cand": 5340, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 4147, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 159.81748962402344, "n_util_obs": 620, "util_count_spearman": 0.1433517489739654}, "probe_overhead_sec_total": 150.736, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.011686, "alpha": 0.03, "t": 822.0} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 835.9} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3619.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207953, "cps": 28.273438, "mean_p": 0.225305, "boundary_entropy": 0.155114, "bpb": 2.022556}, "tr": {"chunks": 3885.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.214605, "cps": 30.351562, "mean_p": 0.205739, "boundary_entropy": 0.133238, "bpb": 2.182364}, "fi": {"chunks": 4383.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233002, "cps": 34.242188, "mean_p": 0.237242, "boundary_entropy": 0.142812, "bpb": 2.330233}}, "bpb": {"en": 2.02256, "tr": 2.18236, "fi": 2.33023}, "gini_cpc": 0.025474, "gini_cps": 0.042848, "bpic": 4.772609, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0319912351960643, "fi": 1.1204568197172118}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 162.8459014892578, "n_util_obs": 623, "util_count_spearman": 0.1417811950128607}, "t": 836.8} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.03507, "tr": 2.19287, "fi": 2.33595}, "bytes_per_chunk": {"en": 4.867, "tr": 5.234, "fi": 4.701}, "bpb_mean": 2.18796, "frozen_inventory": true, "t": 837.8} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40927, "bpb": 2.03315, "chunk_loss": 0.030535, "F": 0.20609, "G": 0.22146, "chunks": 211.03, "bytes_per_chunk": 4.852, "boundary_entropy": 0.14651, "router_noise": 0.0, "grad_norm": 0.7245, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 0.342, "probe_sec": 0.0274, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0217, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.14506, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.20931, "trunc_abs_frac": 0.1285, "mean_delta_full": 0.39862, "read_hit_rate": {"skel": 0.766}, "mean_gate": {"skel": 0.19499}, "n_reads": 4118, "n_cand": 5376, "read_suppressed_frac": 0.0947, "dropout_reads_logged": 4118, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 162.90940856933594, "n_util_obs": 623, "util_count_spearman": 0.1414882092724801}, "probe_overhead_sec_total": 153.582, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.011927, "alpha": 0.03, "t": 839.1} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 851.6} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45941, "bpb": 2.10549, "chunk_loss": 0.030653, "F": 0.21057, "G": 0.22781, "chunks": 215.62, "bytes_per_chunk": 4.749, "boundary_entropy": 0.15274, "router_noise": 0.0, "grad_norm": 0.7707, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.256, "probe_sec": 0.0314, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02506, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02513, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.88034, "trunc_abs_frac": 2.52766, "mean_delta_full": -0.27218, "read_hit_rate": {"skel": 0.77361}, "mean_gate": {"skel": 0.1784}, "n_reads": 4268, "n_cand": 5517, "read_suppressed_frac": 0.0923, "dropout_reads_logged": 4268, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 165.99363708496094, "n_util_obs": 635, "util_count_spearman": 0.14734953489469582}, "probe_overhead_sec_total": 156.115, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.012124, "alpha": 0.03, "t": 851.9} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 864.9} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39067, "bpb": 2.00631, "chunk_loss": 0.030323, "F": 0.19385, "G": 0.21255, "chunks": 198.5, "bytes_per_chunk": 5.159, "boundary_entropy": 0.15079, "router_noise": 0.0, "grad_norm": 0.6619, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.268, "probe_sec": 0.0785, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07121, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20375, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.89899, "trunc_abs_frac": 0.95824, "mean_delta_full": 0.63724, "read_hit_rate": {"skel": 0.75005}, "mean_gate": {"skel": 0.19243}, "n_reads": 3835, "n_cand": 5113, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 3835, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 169.1114501953125, "n_util_obs": 643, "util_count_spearman": 0.15074292397658204}, "probe_overhead_sec_total": 158.996, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.012012, "alpha": 0.03, "t": 865.3} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 878.4} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54072, "bpb": 2.22279, "chunk_loss": 0.030629, "F": 0.21008, "G": 0.22281, "chunks": 215.12, "bytes_per_chunk": 4.76, "boundary_entropy": 0.14143, "router_noise": 0.0, "grad_norm": 0.7037, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.269, "probe_sec": 0.0571, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0442, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01746, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.41354, "trunc_abs_frac": 1.15736, "mean_delta_full": -0.20667, "read_hit_rate": {"skel": 0.7555}, "mean_gate": {"skel": 0.19258}, "n_reads": 4017, "n_cand": 5317, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 4017, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 172.22381591796875, "n_util_obs": 655, "util_count_spearman": 0.15818934377177488}, "probe_overhead_sec_total": 161.84, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.011628, "alpha": 0.03, "t": 878.8} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 891.7} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3586.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206056, "cps": 28.015625, "mean_p": 0.220916, "boundary_entropy": 0.148567, "bpb": 2.002432}, "tr": {"chunks": 3882.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.21444, "cps": 30.328125, "mean_p": 0.206832, "boundary_entropy": 0.12862, "bpb": 2.166177}, "fi": {"chunks": 4420.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234969, "cps": 34.53125, "mean_p": 0.238222, "boundary_entropy": 0.136732, "bpb": 2.311414}}, "bpb": {"en": 2.00243, "tr": 2.16618, "fi": 2.31141}, "gini_cpc": 0.029407, "gini_cps": 0.04677, "bpic": 4.772207, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0406838491486359, "fi": 1.1403133818030726}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 175.25225830078125, "n_util_obs": 667, "util_count_spearman": 0.16578800583994316}, "t": 892.5} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.01643, "tr": 2.1764, "fi": 2.31525}, "bytes_per_chunk": {"en": 4.93, "tr": 5.245, "fi": 4.654}, "bpb_mean": 2.16936, "frozen_inventory": true, "t": 894.1} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48561, "bpb": 2.14328, "chunk_loss": 0.030473, "F": 0.20056, "G": 0.21463, "chunks": 205.38, "bytes_per_chunk": 4.986, "boundary_entropy": 0.13858, "router_noise": 0.0, "grad_norm": 0.7068, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.334, "probe_sec": 0.0276, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02153, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18043, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.09695, "trunc_abs_frac": 1.49501, "mean_delta_full": 0.34775, "read_hit_rate": {"skel": 0.7548}, "mean_gate": {"skel": 0.18634}, "n_reads": 3974, "n_cand": 5265, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 3974, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 175.3124542236328, "n_util_obs": 667, "util_count_spearman": 0.1661378269824852}, "probe_overhead_sec_total": 164.272, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.011717, "alpha": 0.03, "t": 895.5} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 908.8} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43695, "bpb": 2.07308, "chunk_loss": 0.030631, "F": 0.20889, "G": 0.22175, "chunks": 213.91, "bytes_per_chunk": 4.787, "boundary_entropy": 0.14178, "router_noise": 0.0, "grad_norm": 0.6721, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.272, "probe_sec": 0.0455, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03033, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17572, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.91663, "trunc_abs_frac": 0.91663, "mean_delta_full": -0.26648, "read_hit_rate": {"skel": 0.75784}, "mean_gate": {"skel": 0.18457}, "n_reads": 4034, "n_cand": 5323, "read_suppressed_frac": 0.0967, "dropout_reads_logged": 4034, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 178.3768310546875, "n_util_obs": 674, "util_count_spearman": 0.1669549321947302}, "probe_overhead_sec_total": 166.521, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.011824, "alpha": 0.03, "t": 909.1} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 922.0} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48777, "bpb": 2.1464, "chunk_loss": 0.030447, "F": 0.20261, "G": 0.21568, "chunks": 207.47, "bytes_per_chunk": 4.936, "boundary_entropy": 0.13605, "router_noise": 0.0, "grad_norm": 0.7181, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.0352, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02924, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18646, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 2.03319, "trunc_abs_frac": 0.71108, "mean_delta_full": -0.04919, "read_hit_rate": {"skel": 0.75842}, "mean_gate": {"skel": 0.19773}, "n_reads": 4006, "n_cand": 5282, "read_suppressed_frac": 0.0946, "dropout_reads_logged": 4006, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 181.44898986816406, "n_util_obs": 683, "util_count_spearman": 0.16342028511221562}, "probe_overhead_sec_total": 169.249, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.011096, "alpha": 0.03, "t": 922.2} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 935.1} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44709, "bpb": 2.0877, "chunk_loss": 0.03046, "F": 0.20117, "G": 0.2175, "chunks": 206.0, "bytes_per_chunk": 4.971, "boundary_entropy": 0.14932, "router_noise": 0.0, "grad_norm": 0.7134, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.261, "probe_sec": 0.0356, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02954, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0091, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.53607, "trunc_abs_frac": 0.58947, "mean_delta_full": -0.16467, "read_hit_rate": {"skel": 0.7499}, "mean_gate": {"skel": 0.18357}, "n_reads": 3940, "n_cand": 5254, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 3940, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 184.53067016601562, "n_util_obs": 695, "util_count_spearman": 0.16897672864525634}, "probe_overhead_sec_total": 171.722, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.011513, "alpha": 0.03, "t": 935.3} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 948.0} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3587.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206114, "cps": 28.023438, "mean_p": 0.221665, "boundary_entropy": 0.142672, "bpb": 1.984981}, "tr": {"chunks": 3834.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.211788, "cps": 29.953125, "mean_p": 0.202971, "boundary_entropy": 0.130608, "bpb": 2.1391}, "fi": {"chunks": 4346.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231035, "cps": 33.953125, "mean_p": 0.236544, "boundary_entropy": 0.137453, "bpb": 2.283274}}, "bpb": {"en": 1.98498, "tr": 2.1391, "fi": 2.28327}, "gini_cpc": 0.025602, "gini_cps": 0.043002, "bpic": 4.82128, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0275295034865464, "fi": 1.120909582929289}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 187.54727172851562, "n_util_obs": 712, "util_count_spearman": 0.17623625571356263}, "t": 948.8} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 1.99781, "tr": 2.14831, "fi": 2.28953}, "bytes_per_chunk": {"en": 4.903, "tr": 5.309, "fi": 4.693}, "bpb_mean": 2.14521, "frozen_inventory": true, "t": 950.3} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43385, "bpb": 2.06861, "chunk_loss": 0.030431, "F": 0.20139, "G": 0.21657, "chunks": 206.22, "bytes_per_chunk": 4.966, "boundary_entropy": 0.13774, "router_noise": 0.0, "grad_norm": 0.6224, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.332, "probe_sec": 0.1434, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13717, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20428, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.47026, "trunc_abs_frac": 0.73995, "mean_delta_full": 0.58718, "read_hit_rate": {"skel": 0.75324}, "mean_gate": {"skel": 0.18429}, "n_reads": 3950, "n_cand": 5244, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 3950, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 187.6077117919922, "n_util_obs": 712, "util_count_spearman": 0.17425036177704342}, "probe_overhead_sec_total": 174.772, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.011118, "alpha": 0.03, "t": 951.9} |
| {"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 965.5} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43361, "bpb": 2.06826, "chunk_loss": 0.030528, "F": 0.20493, "G": 0.2168, "chunks": 209.84, "bytes_per_chunk": 4.88, "boundary_entropy": 0.13342, "router_noise": 0.0, "grad_norm": 0.7117, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.277, "probe_sec": 0.066, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05338, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02658, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.38756, "trunc_abs_frac": 1.01036, "mean_delta_full": -0.05923, "read_hit_rate": {"skel": 0.7603}, "mean_gate": {"skel": 0.18887}, "n_reads": 3987, "n_cand": 5244, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 3987, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 190.68997192382812, "n_util_obs": 725, "util_count_spearman": 0.18179313894039853}, "probe_overhead_sec_total": 177.575, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.013808, "alpha": 0.03, "t": 965.7} |
| {"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 978.8} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40865, "bpb": 2.03225, "chunk_loss": 0.030621, "F": 0.20953, "G": 0.22397, "chunks": 214.56, "bytes_per_chunk": 4.773, "boundary_entropy": 0.14235, "router_noise": 0.0, "grad_norm": 0.6503, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.266, "probe_sec": 0.0493, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03853, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1011, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -3.18889, "trunc_abs_frac": 2.86489, "mean_delta_full": -0.09631, "read_hit_rate": {"skel": 0.76442}, "mean_gate": {"skel": 0.18335}, "n_reads": 4121, "n_cand": 5391, "read_suppressed_frac": 0.1019, "dropout_reads_logged": 4121, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 193.79904174804688, "n_util_obs": 731, "util_count_spearman": 0.18445317439301437}, "probe_overhead_sec_total": 180.446, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.013068, "alpha": 0.03, "t": 979.0} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 994.4} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35299, "bpb": 1.95195, "chunk_loss": 0.030481, "F": 0.20383, "G": 0.2172, "chunks": 208.72, "bytes_per_chunk": 4.906, "boundary_entropy": 0.13413, "router_noise": 0.0, "grad_norm": 0.7054, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.326, "probe_sec": 0.0591, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03102, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10826, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 11.93441, "trunc_abs_frac": 0.90545, "mean_delta_full": 0.01644, "read_hit_rate": {"skel": 0.76242}, "mean_gate": {"skel": 0.18466}, "n_reads": 4050, "n_cand": 5312, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 4050, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 196.9247589111328, "n_util_obs": 741, "util_count_spearman": 0.18419695208718767}, "probe_overhead_sec_total": 183.181, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.012766, "alpha": 0.03, "t": 995.3} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1010.4} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3627.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208412, "cps": 28.335938, "mean_p": 0.221753, "boundary_entropy": 0.141422, "bpb": 1.973255}, "tr": {"chunks": 3882.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.21444, "cps": 30.328125, "mean_p": 0.204275, "boundary_entropy": 0.122926, "bpb": 2.13997}, "fi": {"chunks": 4411.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.23449, "cps": 34.460938, "mean_p": 0.236733, "boundary_entropy": 0.128886, "bpb": 2.25406}}, "bpb": {"en": 1.97326, "tr": 2.13997, "fi": 2.25406}, "gini_cpc": 0.026448, "gini_cps": 0.043848, "bpic": 4.759396, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0289198464425169, "fi": 1.125127498081895}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 199.9427947998047, "n_util_obs": 753, "util_count_spearman": 0.17533846436159412}, "t": 1011.3} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.9864, "tr": 2.15141, "fi": 2.26212}, "bytes_per_chunk": {"en": 4.847, "tr": 5.255, "fi": 4.7}, "bpb_mean": 2.13331, "frozen_inventory": true, "t": 1012.7} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41695, "bpb": 2.04423, "chunk_loss": 0.030489, "F": 0.20529, "G": 0.2176, "chunks": 210.22, "bytes_per_chunk": 4.871, "boundary_entropy": 0.13375, "router_noise": 0.0, "grad_norm": 0.7513, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 0.367, "probe_sec": 0.0361, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.023, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23315, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.0594, "trunc_abs_frac": 0.59852, "mean_delta_full": -0.35449, "read_hit_rate": {"skel": 0.76354}, "mean_gate": {"skel": 0.18415}, "n_reads": 4075, "n_cand": 5337, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 4075, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 200.00564575195312, "n_util_obs": 754, "util_count_spearman": 0.17060432055648395}, "probe_overhead_sec_total": 186.282, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.010959, "alpha": 0.03, "t": 1013.7} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1027.9} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38559, "bpb": 1.99899, "chunk_loss": 0.030469, "F": 0.20413, "G": 0.21553, "chunks": 209.03, "bytes_per_chunk": 4.899, "boundary_entropy": 0.13122, "router_noise": 0.0, "grad_norm": 0.769, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.289, "probe_sec": 0.0329, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02709, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08018, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.07295, "trunc_abs_frac": 5.13099, "mean_delta_full": 0.07498, "read_hit_rate": {"skel": 0.77203}, "mean_gate": {"skel": 0.18748}, "n_reads": 4162, "n_cand": 5391, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 4162, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 203.0942840576172, "n_util_obs": 765, "util_count_spearman": 0.17954411386377264}, "probe_overhead_sec_total": 189.094, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.011674, "alpha": 0.03, "t": 1028.1} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1043.7} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36518, "bpb": 1.96954, "chunk_loss": 0.030405, "F": 0.20111, "G": 0.2145, "chunks": 205.94, "bytes_per_chunk": 4.972, "boundary_entropy": 0.13563, "router_noise": 0.0, "grad_norm": 0.6861, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.317, "probe_sec": 0.0312, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.021, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0366, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.80657, "trunc_abs_frac": 1.39524, "mean_delta_full": -0.11975, "read_hit_rate": {"skel": 0.76643}, "mean_gate": {"skel": 0.18619}, "n_reads": 4059, "n_cand": 5296, "read_suppressed_frac": 0.0951, "dropout_reads_logged": 4059, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 206.18223571777344, "n_util_obs": 782, "util_count_spearman": 0.18488767278943588}, "probe_overhead_sec_total": 191.418, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.012757, "alpha": 0.03, "t": 1044.0} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1060.9} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40059, "bpb": 2.02062, "chunk_loss": 0.030414, "F": 0.20197, "G": 0.21246, "chunks": 206.81, "bytes_per_chunk": 4.951, "boundary_entropy": 0.12556, "router_noise": 0.0, "grad_norm": 0.6663, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.341, "probe_sec": 0.0564, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04684, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19938, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.0736, "trunc_abs_frac": 0.82264, "mean_delta_full": 1.82245, "read_hit_rate": {"skel": 0.75654}, "mean_gate": {"skel": 0.18731}, "n_reads": 4021, "n_cand": 5315, "read_suppressed_frac": 0.0915, "dropout_reads_logged": 4021, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 209.27828979492188, "n_util_obs": 792, "util_count_spearman": 0.18743855164138765}, "probe_overhead_sec_total": 194.365, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.012195, "alpha": 0.03, "t": 1061.1} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1090.2} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3653.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209906, "cps": 28.539062, "mean_p": 0.222029, "boundary_entropy": 0.136573, "bpb": 1.960161}, "tr": {"chunks": 3863.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.21339, "cps": 30.179688, "mean_p": 0.203357, "boundary_entropy": 0.11648, "bpb": 2.103673}, "fi": {"chunks": 4345.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230982, "cps": 33.945312, "mean_p": 0.234181, "boundary_entropy": 0.129705, "bpb": 2.233721}}, "bpb": {"en": 1.96016, "tr": 2.10367, "fi": 2.23372}, "gini_cpc": 0.021475, "gini_cps": 0.038895, "bpic": 4.783071, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.016596487244969, "fi": 1.100404468598213}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 212.33184814453125, "n_util_obs": 800, "util_count_spearman": 0.18376779748019276}, "t": 1091.0} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.97328, "tr": 2.11206, "fi": 2.24012}, "bytes_per_chunk": {"en": 4.852, "tr": 5.261, "fi": 4.745}, "bpb_mean": 2.10849, "frozen_inventory": true, "t": 1092.4} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40682, "bpb": 2.02962, "chunk_loss": 0.03052, "F": 0.20584, "G": 0.21807, "chunks": 210.78, "bytes_per_chunk": 4.858, "boundary_entropy": 0.12882, "router_noise": 0.0, "grad_norm": 0.6275, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.663, "probe_sec": 0.0399, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03407, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0041, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.29336, "trunc_abs_frac": 0.73846, "mean_delta_full": 0.22928, "read_hit_rate": {"skel": 0.77173}, "mean_gate": {"skel": 0.17503}, "n_reads": 4199, "n_cand": 5441, "read_suppressed_frac": 0.1048, "dropout_reads_logged": 4199, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 212.39630126953125, "n_util_obs": 800, "util_count_spearman": 0.18393655841948744}, "probe_overhead_sec_total": 197.51, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.012646, "alpha": 0.03, "t": 1094.2} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1108.9} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43867, "bpb": 2.07557, "chunk_loss": 0.030629, "F": 0.20956, "G": 0.22069, "chunks": 214.59, "bytes_per_chunk": 4.772, "boundary_entropy": 0.13052, "router_noise": 0.0, "grad_norm": 0.6923, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.3, "probe_sec": 0.0675, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05999, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06524, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.81512, "trunc_abs_frac": 2.02762, "mean_delta_full": 0.05649, "read_hit_rate": {"skel": 0.76287}, "mean_gate": {"skel": 0.17596}, "n_reads": 4150, "n_cand": 5440, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 4150, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 215.5127410888672, "n_util_obs": 814, "util_count_spearman": 0.18506694561165482}, "probe_overhead_sec_total": 200.296, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.011557, "alpha": 0.03, "t": 1109.2} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1127.9} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38676, "bpb": 2.00067, "chunk_loss": 0.030494, "F": 0.20645, "G": 0.21669, "chunks": 211.41, "bytes_per_chunk": 4.844, "boundary_entropy": 0.12912, "router_noise": 0.0, "grad_norm": 0.649, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.393, "probe_sec": 0.063, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0328, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10603, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 22.33333, "trunc_abs_frac": 1.60795, "mean_delta_full": 0.00632, "read_hit_rate": {"skel": 0.76365}, "mean_gate": {"skel": 0.18447}, "n_reads": 4029, "n_cand": 5276, "read_suppressed_frac": 0.1052, "dropout_reads_logged": 4029, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 218.6145477294922, "n_util_obs": 822, "util_count_spearman": 0.18880891236564304}, "probe_overhead_sec_total": 203.782, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.011763, "alpha": 0.03, "t": 1128.8} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1148.8} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37768, "bpb": 1.98757, "chunk_loss": 0.030408, "F": 0.2012, "G": 0.21249, "chunks": 206.03, "bytes_per_chunk": 4.97, "boundary_entropy": 0.12713, "router_noise": 0.0, "grad_norm": 0.6793, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.413, "probe_sec": 0.0358, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02333, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.13674, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.25014, "trunc_abs_frac": 1.14474, "mean_delta_full": -0.11005, "read_hit_rate": {"skel": 0.77435}, "mean_gate": {"skel": 0.17552}, "n_reads": 4111, "n_cand": 5309, "read_suppressed_frac": 0.1024, "dropout_reads_logged": 4111, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 221.73965454101562, "n_util_obs": 832, "util_count_spearman": 0.20250776052158875}, "probe_overhead_sec_total": 206.292, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.011615, "alpha": 0.03, "t": 1149.5} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1170.8} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3619.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207953, "cps": 28.273438, "mean_p": 0.220042, "boundary_entropy": 0.138036, "bpb": 1.947511}, "tr": {"chunks": 3888.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.214771, "cps": 30.375, "mean_p": 0.203876, "boundary_entropy": 0.111746, "bpb": 2.09363}, "fi": {"chunks": 4374.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232524, "cps": 34.171875, "mean_p": 0.23505, "boundary_entropy": 0.124604, "bpb": 2.213582}}, "bpb": {"en": 1.94751, "tr": 2.09363, "fi": 2.21358}, "gini_cpc": 0.024999, "gini_cps": 0.042365, "bpic": 4.775019, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.032788139624787, "fi": 1.118156087027854}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 224.77125549316406, "n_util_obs": 842, "util_count_spearman": 0.19767725384679113}, "t": 1171.6} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.96082, "tr": 2.10105, "fi": 2.22042}, "bytes_per_chunk": {"en": 4.895, "tr": 5.221, "fi": 4.73}, "bpb_mean": 2.0941, "frozen_inventory": true, "t": 1173.0} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44246, "bpb": 2.08103, "chunk_loss": 0.030469, "F": 0.20383, "G": 0.21587, "chunks": 208.72, "bytes_per_chunk": 4.906, "boundary_entropy": 0.12704, "router_noise": 0.0, "grad_norm": 0.7106, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.498, "probe_sec": 0.0384, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03074, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13746, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.26809, "trunc_abs_frac": 0.89147, "mean_delta_full": 0.38688, "read_hit_rate": {"skel": 0.76069}, "mean_gate": {"skel": 0.17353}, "n_reads": 4129, "n_cand": 5428, "read_suppressed_frac": 0.1068, "dropout_reads_logged": 4129, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 224.83412170410156, "n_util_obs": 842, "util_count_spearman": 0.19767310692981735}, "probe_overhead_sec_total": 209.304, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.01265, "alpha": 0.03, "t": 1174.4} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1189.4} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.53861, "bpb": 2.21975, "chunk_loss": 0.030841, "F": 0.2099, "G": 0.21928, "chunks": 214.94, "bytes_per_chunk": 4.764, "boundary_entropy": 0.12057, "router_noise": 0.0, "grad_norm": 0.7226, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.306, "probe_sec": 0.0306, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02452, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02645, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -6.30112, "trunc_abs_frac": 1.15955, "mean_delta_full": -0.05615, "read_hit_rate": {"skel": 0.73715}, "mean_gate": {"skel": 0.18515}, "n_reads": 3814, "n_cand": 5174, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 3814, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 227.92428588867188, "n_util_obs": 856, "util_count_spearman": 0.21248558881252202}, "probe_overhead_sec_total": 211.588, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.011203, "alpha": 0.03, "t": 1189.7} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1203.3} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4313, "bpb": 2.06493, "chunk_loss": 0.030533, "F": 0.20633, "G": 0.21659, "chunks": 211.28, "bytes_per_chunk": 4.847, "boundary_entropy": 0.12926, "router_noise": 0.0, "grad_norm": 0.7276, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.277, "probe_sec": 0.0366, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.029, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09002, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.47964, "trunc_abs_frac": 0.48804, "mean_delta_full": -0.43083, "read_hit_rate": {"skel": 0.76545}, "mean_gate": {"skel": 0.17181}, "n_reads": 4112, "n_cand": 5372, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 4112, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 231.0196533203125, "n_util_obs": 867, "util_count_spearman": 0.20808912570833868}, "probe_overhead_sec_total": 215.144, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.012247, "alpha": 0.03, "t": 1203.5} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1216.3} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48676, "bpb": 2.14494, "chunk_loss": 0.030501, "F": 0.20343, "G": 0.2131, "chunks": 208.31, "bytes_per_chunk": 4.916, "boundary_entropy": 0.12452, "router_noise": 0.0, "grad_norm": 0.7722, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.26, "probe_sec": 0.0345, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02842, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10031, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.31272, "trunc_abs_frac": 1.18374, "mean_delta_full": -0.24487, "read_hit_rate": {"skel": 0.75273}, "mean_gate": {"skel": 0.17449}, "n_reads": 3991, "n_cand": 5302, "read_suppressed_frac": 0.1037, "dropout_reads_logged": 3991, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 234.1436309814453, "n_util_obs": 883, "util_count_spearman": 0.2135734237354656}, "probe_overhead_sec_total": 218.034, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.01143, "alpha": 0.03, "t": 1216.5} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1229.1} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3615.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207723, "cps": 28.242188, "mean_p": 0.219672, "boundary_entropy": 0.137854, "bpb": 1.935712}, "tr": {"chunks": 3858.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.213114, "cps": 30.140625, "mean_p": 0.201906, "boundary_entropy": 0.113322, "bpb": 2.080461}, "fi": {"chunks": 4349.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231195, "cps": 33.976562, "mean_p": 0.234636, "boundary_entropy": 0.121598, "bpb": 2.196431}}, "bpb": {"en": 1.93571, "tr": 2.08046, "fi": 2.19643}, "gini_cpc": 0.023999, "gini_cps": 0.041392, "bpic": 4.79885, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0259530583752767, "fi": 1.1129953318123433}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 237.18675231933594, "n_util_obs": 890, "util_count_spearman": 0.219463071874216}, "t": 1230.0} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.94855, "tr": 2.08751, "fi": 2.2026}, "bytes_per_chunk": {"en": 4.887, "tr": 5.271, "fi": 4.756}, "bpb_mean": 2.07955, "frozen_inventory": true, "t": 1231.4} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40877, "bpb": 2.03243, "chunk_loss": 0.03048, "F": 0.20322, "G": 0.21475, "chunks": 208.09, "bytes_per_chunk": 4.921, "boundary_entropy": 0.13041, "router_noise": 0.0, "grad_norm": 0.6766, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.326, "probe_sec": 0.067, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06118, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.13371, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.04056, "trunc_abs_frac": 0.8825, "mean_delta_full": -0.18124, "read_hit_rate": {"skel": 0.75512}, "mean_gate": {"skel": 0.18023}, "n_reads": 4021, "n_cand": 5325, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 4021, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 237.2474822998047, "n_util_obs": 890, "util_count_spearman": 0.21839198029303153}, "probe_overhead_sec_total": 220.859, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.011593, "alpha": 0.03, "t": 1232.8} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1245.7} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42326, "bpb": 2.05333, "chunk_loss": 0.030477, "F": 0.20541, "G": 0.21525, "chunks": 210.34, "bytes_per_chunk": 4.868, "boundary_entropy": 0.11992, "router_noise": 0.0, "grad_norm": 0.7154, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.262, "probe_sec": 0.0545, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04514, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0412, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.32916, "trunc_abs_frac": 1.99404, "mean_delta_full": 0.06091, "read_hit_rate": {"skel": 0.76013}, "mean_gate": {"skel": 0.18372}, "n_reads": 4015, "n_cand": 5282, "read_suppressed_frac": 0.1131, "dropout_reads_logged": 4015, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 240.36964416503906, "n_util_obs": 900, "util_count_spearman": 0.22278969217770805}, "probe_overhead_sec_total": 223.746, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011493, "alpha": 0.03, "t": 1246.0} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3627.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208412, "cps": 28.335938, "mean_p": 0.219758, "boundary_entropy": 0.129456, "bpb": 1.929186}, "tr": {"chunks": 3893.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215047, "cps": 30.414062, "mean_p": 0.204083, "boundary_entropy": 0.108773, "bpb": 2.069403}, "fi": {"chunks": 4417.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234809, "cps": 34.507812, "mean_p": 0.235569, "boundary_entropy": 0.121031, "bpb": 2.189988}}, "bpb": {"en": 1.92919, "tr": 2.0694, "fi": 2.18999}, "gini_cpc": 0.026734, "gini_cps": 0.044121, "bpic": 4.752618, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0318353843896748, "fi": 1.126657936755323}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 243.14122009277344, "n_util_obs": 913, "util_count_spearman": 0.22330474962109512}, "t": 1258.5} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1262.4} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35473, "bpb": 1.95446, "chunk_loss": 0.030536, "F": 0.20609, "G": 0.21765, "chunks": 211.03, "bytes_per_chunk": 4.852, "boundary_entropy": 0.12418, "router_noise": 0.0, "grad_norm": 0.7083, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.337, "probe_sec": 0.102, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08761, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.08077, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.02726, "trunc_abs_frac": 1.6766, "mean_delta_full": -0.14383, "read_hit_rate": {"skel": 0.77062}, "mean_gate": {"skel": 0.17467}, "n_reads": 4129, "n_cand": 5358, "read_suppressed_frac": 0.1075, "dropout_reads_logged": 4129, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 243.4461212158203, "n_util_obs": 913, "util_count_spearman": 0.2230393040030261}, "probe_overhead_sec_total": 226.231, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011639, "alpha": 0.03, "t": 1262.8} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1275.1} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42014, "bpb": 2.04883, "chunk_loss": 0.030551, "F": 0.20721, "G": 0.21786, "chunks": 212.19, "bytes_per_chunk": 4.826, "boundary_entropy": 0.12456, "router_noise": 0.0, "grad_norm": 0.7418, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.251, "probe_sec": 0.0322, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02607, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05916, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.88297, "trunc_abs_frac": 0.37642, "mean_delta_full": -0.31128, "read_hit_rate": {"skel": 0.75897}, "mean_gate": {"skel": 0.17073}, "n_reads": 4144, "n_cand": 5460, "read_suppressed_frac": 0.0951, "dropout_reads_logged": 4144, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 246.53143310546875, "n_util_obs": 922, "util_count_spearman": 0.22505974325121392}, "probe_overhead_sec_total": 229.252, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012324, "alpha": 0.03, "t": 1275.3} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1289.2} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3637.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208987, "cps": 28.414062, "mean_p": 0.219294, "boundary_entropy": 0.127985, "bpb": 1.925996}, "tr": {"chunks": 3900.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215434, "cps": 30.46875, "mean_p": 0.204035, "boundary_entropy": 0.106678, "bpb": 2.06118}, "fi": {"chunks": 4385.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233108, "cps": 34.257812, "mean_p": 0.234676, "boundary_entropy": 0.119906, "bpb": 2.185384}}, "bpb": {"en": 1.926, "tr": 2.06118, "fi": 2.18538}, "gini_cpc": 0.024457, "gini_cps": 0.041827, "bpic": 4.758598, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0308485745978269, "fi": 1.1154202724533306}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 249.58123779296875, "n_util_obs": 931, "util_count_spearman": 0.2265875927588894}, "t": 1290.0} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.94001, "tr": 2.06852, "fi": 2.19243}, "bytes_per_chunk": {"en": 4.879, "tr": 5.24, "fi": 4.731}, "bpb_mean": 2.06699, "frozen_inventory": true, "t": 1291.5} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 1292.3} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35047, "bpb": 1.94832, "chunk_loss": 0.030348, "F": 0.19846, "G": 0.20922, "chunks": 203.22, "bytes_per_chunk": 5.039, "boundary_entropy": 0.11705, "router_noise": 0.0, "grad_norm": 0.7231, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.345, "probe_sec": 0.1379, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12953, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13144, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.22245, "trunc_abs_frac": 1.1241, "mean_delta_full": 0.15621, "read_hit_rate": {"skel": 0.76472}, "mean_gate": {"skel": 0.17695}, "n_reads": 4027, "n_cand": 5266, "read_suppressed_frac": 0.1028, "dropout_reads_logged": 4027, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 249.64247131347656, "n_util_obs": 932, "util_count_spearman": 0.2273019311418266}, "probe_overhead_sec_total": 232.883, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01233, "alpha": 0.03, "t": 1292.6} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1305.9} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37908, "bpb": 1.98959, "chunk_loss": 0.030497, "F": 0.20697, "G": 0.21719, "chunks": 211.94, "bytes_per_chunk": 4.832, "boundary_entropy": 0.12014, "router_noise": 0.0, "grad_norm": 0.6628, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.272, "probe_sec": 0.0673, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05749, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.035, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.2469, "trunc_abs_frac": 0.60281, "mean_delta_full": -0.28394, "read_hit_rate": {"skel": 0.76063}, "mean_gate": {"skel": 0.17798}, "n_reads": 4115, "n_cand": 5410, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 4115, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 252.72695922851562, "n_util_obs": 942, "util_count_spearman": 0.2317034035706154}, "probe_overhead_sec_total": 235.868, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012295, "alpha": 0.03, "t": 1306.2} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1319.6} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37805, "bpb": 1.98811, "chunk_loss": 0.030454, "F": 0.20383, "G": 0.21308, "chunks": 208.72, "bytes_per_chunk": 4.906, "boundary_entropy": 0.1211, "router_noise": 0.0, "grad_norm": 0.7503, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.274, "probe_sec": 0.062, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04991, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04119, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.92397, "trunc_abs_frac": 0.86842, "mean_delta_full": -0.12297, "read_hit_rate": {"skel": 0.7588}, "mean_gate": {"skel": 0.17192}, "n_reads": 4011, "n_cand": 5286, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 4011, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 255.84689331054688, "n_util_obs": 950, "util_count_spearman": 0.2294831435314289}, "probe_overhead_sec_total": 238.551, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012324, "alpha": 0.03, "t": 1319.9} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1332.7} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40243, "bpb": 2.02328, "chunk_loss": 0.030488, "F": 0.20551, "G": 0.2143, "chunks": 210.44, "bytes_per_chunk": 4.866, "boundary_entropy": 0.12035, "router_noise": 0.0, "grad_norm": 0.6669, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.089, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08061, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.08708, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.53253, "trunc_abs_frac": 0.69959, "mean_delta_full": -0.25427, "read_hit_rate": {"skel": 0.76617}, "mean_gate": {"skel": 0.16814}, "n_reads": 4135, "n_cand": 5397, "read_suppressed_frac": 0.0902, "dropout_reads_logged": 4135, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 258.94049072265625, "n_util_obs": 964, "util_count_spearman": 0.22261385233268022}, "probe_overhead_sec_total": 241.593, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012298, "alpha": 0.03, "t": 1333.1} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1346.0} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3636.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208929, "cps": 28.40625, "mean_p": 0.218163, "boundary_entropy": 0.131203, "bpb": 1.91553}, "tr": {"chunks": 3928.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.216981, "cps": 30.6875, "mean_p": 0.204304, "boundary_entropy": 0.106319, "bpb": 2.049124}, "fi": {"chunks": 4486.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238477, "cps": 35.046875, "mean_p": 0.237455, "boundary_entropy": 0.120364, "bpb": 2.167233}}, "bpb": {"en": 1.91553, "tr": 2.04912, "fi": 2.16723}, "gini_cpc": 0.029649, "gini_cps": 0.047026, "bpic": 4.70805, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0385350859010112, "fi": 1.1414256597633607}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 261.9915771484375, "n_util_obs": 974, "util_count_spearman": 0.22629913434376972}, "t": 1346.9} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.92957, "tr": 2.05685, "fi": 2.17408}, "bytes_per_chunk": {"en": 4.867, "tr": 5.235, "fi": 4.632}, "bpb_mean": 2.0535, "frozen_inventory": true, "t": 1348.4} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35613, "bpb": 1.95648, "chunk_loss": 0.030475, "F": 0.207, "G": 0.21584, "chunks": 211.97, "bytes_per_chunk": 4.831, "boundary_entropy": 0.12598, "router_noise": 0.0, "grad_norm": 0.6357, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.331, "probe_sec": 0.0315, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02504, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02766, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.13225, "trunc_abs_frac": 0.13225, "mean_delta_full": 1.01477, "read_hit_rate": {"skel": 0.77103}, "mean_gate": {"skel": 0.17466}, "n_reads": 4179, "n_cand": 5420, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 4179, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 262.0552673339844, "n_util_obs": 974, "util_count_spearman": 0.22667484554610942}, "probe_overhead_sec_total": 244.81, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012082, "alpha": 0.03, "t": 1349.6} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1362.5} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35204, "bpb": 1.95059, "chunk_loss": 0.030399, "F": 0.20016, "G": 0.20929, "chunks": 204.97, "bytes_per_chunk": 4.996, "boundary_entropy": 0.12238, "router_noise": 0.0, "grad_norm": 0.6948, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.26, "probe_sec": 0.031, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02475, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05494, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.0488, "trunc_abs_frac": 0.76193, "mean_delta_full": -0.38507, "read_hit_rate": {"skel": 0.75883}, "mean_gate": {"skel": 0.17655}, "n_reads": 3930, "n_cand": 5179, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 3930, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 265.16339111328125, "n_util_obs": 980, "util_count_spearman": 0.22328444387740645}, "probe_overhead_sec_total": 246.854, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011239, "alpha": 0.03, "t": 1362.6} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1375.9} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32942, "bpb": 1.91795, "chunk_loss": 0.030475, "F": 0.20593, "G": 0.21541, "chunks": 210.88, "bytes_per_chunk": 4.856, "boundary_entropy": 0.11958, "router_noise": 0.0, "grad_norm": 0.6468, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.271, "probe_sec": 0.073, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06722, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17292, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.50637, "trunc_abs_frac": 19.86105, "mean_delta_full": 0.25137, "read_hit_rate": {"skel": 0.77554}, "mean_gate": {"skel": 0.16983}, "n_reads": 4198, "n_cand": 5413, "read_suppressed_frac": 0.0917, "dropout_reads_logged": 4198, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 268.2623291015625, "n_util_obs": 989, "util_count_spearman": 0.22889934846449406}, "probe_overhead_sec_total": 249.329, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012092, "alpha": 0.03, "t": 1376.2} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1388.4} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36179, "bpb": 1.96465, "chunk_loss": 0.030372, "F": 0.20038, "G": 0.20966, "chunks": 205.19, "bytes_per_chunk": 4.991, "boundary_entropy": 0.11624, "router_noise": 0.0, "grad_norm": 0.6554, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.25, "probe_sec": 0.0818, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07549, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19885, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.14764, "trunc_abs_frac": 0.06487, "mean_delta_full": -0.18976, "read_hit_rate": {"skel": 0.75445}, "mean_gate": {"skel": 0.17166}, "n_reads": 3988, "n_cand": 5286, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 3988, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 271.38458251953125, "n_util_obs": 1002, "util_count_spearman": 0.2317795702860477}, "probe_overhead_sec_total": 251.896, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012564, "alpha": 0.03, "t": 1388.7} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1401.2} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3669.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210826, "cps": 28.664062, "mean_p": 0.219455, "boundary_entropy": 0.12615, "bpb": 1.90685}, "tr": {"chunks": 3901.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215489, "cps": 30.476562, "mean_p": 0.203629, "boundary_entropy": 0.104242, "bpb": 2.03884}, "fi": {"chunks": 4424.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235182, "cps": 34.5625, "mean_p": 0.235589, "boundary_entropy": 0.114349, "bpb": 2.156673}}, "bpb": {"en": 1.90685, "tr": 2.03884, "fi": 2.15667}, "gini_cpc": 0.024546, "gini_cps": 0.041965, "bpic": 4.730032, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0221198141695682, "fi": 1.1155258620491055}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 274.442626953125, "n_util_obs": 1010, "util_count_spearman": 0.23254622306734735}, "t": 1402.0} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.92093, "tr": 2.04658, "fi": 2.16205}, "bytes_per_chunk": {"en": 4.841, "tr": 5.236, "fi": 4.719}, "bpb_mean": 2.04319, "frozen_inventory": true, "t": 1403.4} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36237, "bpb": 1.96549, "chunk_loss": 0.030622, "F": 0.2132, "G": 0.22242, "chunks": 218.31, "bytes_per_chunk": 4.691, "boundary_entropy": 0.12247, "router_noise": 0.0, "grad_norm": 0.6973, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.32, "probe_sec": 0.0364, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02451, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21814, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.81741, "trunc_abs_frac": 1.17086, "mean_delta_full": 0.08698, "read_hit_rate": {"skel": 0.77532}, "mean_gate": {"skel": 0.17027}, "n_reads": 4241, "n_cand": 5470, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 4241, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 274.50775146484375, "n_util_obs": 1010, "util_count_spearman": 0.23281754376993843}, "probe_overhead_sec_total": 254.683, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011567, "alpha": 0.03, "t": 1404.7} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1416.8} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38786, "bpb": 2.00226, "chunk_loss": 0.030536, "F": 0.20798, "G": 0.2165, "chunks": 212.97, "bytes_per_chunk": 4.808, "boundary_entropy": 0.11321, "router_noise": 0.0, "grad_norm": 0.6614, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.247, "probe_sec": 0.031, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02398, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.13554, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.80574, "trunc_abs_frac": 0.77447, "mean_delta_full": -0.32593, "read_hit_rate": {"skel": 0.77053}, "mean_gate": {"skel": 0.17379}, "n_reads": 4120, "n_cand": 5347, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 4120, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 277.6174011230469, "n_util_obs": 1023, "util_count_spearman": 0.2394441061755452}, "probe_overhead_sec_total": 257.911, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01122, "alpha": 0.03, "t": 1417.0} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1428.9} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41724, "bpb": 2.04464, "chunk_loss": 0.030697, "F": 0.21426, "G": 0.22353, "chunks": 219.41, "bytes_per_chunk": 4.667, "boundary_entropy": 0.12228, "router_noise": 0.0, "grad_norm": 0.6845, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.245, "probe_sec": 0.0397, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03108, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0208, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.6592, "trunc_abs_frac": 1.57617, "mean_delta_full": 0.02655, "read_hit_rate": {"skel": 0.77319}, "mean_gate": {"skel": 0.16247}, "n_reads": 4193, "n_cand": 5423, "read_suppressed_frac": 0.0878, "dropout_reads_logged": 4193, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 280.7204895019531, "n_util_obs": 1035, "util_count_spearman": 0.23843726570040377}, "probe_overhead_sec_total": 259.928, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.013185, "alpha": 0.03, "t": 1429.3} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1441.7} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40571, "bpb": 2.02802, "chunk_loss": 0.030494, "F": 0.20712, "G": 0.21643, "chunks": 212.09, "bytes_per_chunk": 4.828, "boundary_entropy": 0.11715, "router_noise": 0.0, "grad_norm": 0.6797, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.252, "probe_sec": 0.0299, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02237, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.12651, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.67882, "trunc_abs_frac": 1.54625, "mean_delta_full": -0.3114, "read_hit_rate": {"skel": 0.7611}, "mean_gate": {"skel": 0.172}, "n_reads": 4062, "n_cand": 5337, "read_suppressed_frac": 0.1054, "dropout_reads_logged": 4062, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 283.81402587890625, "n_util_obs": 1047, "util_count_spearman": 0.23723037026063118}, "probe_overhead_sec_total": 262.69, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011242, "alpha": 0.03, "t": 1441.9} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1455.1} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3623.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208182, "cps": 28.304688, "mean_p": 0.217333, "boundary_entropy": 0.124345, "bpb": 1.898155}, "tr": {"chunks": 3863.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.21339, "cps": 30.179688, "mean_p": 0.202141, "boundary_entropy": 0.105223, "bpb": 2.02807}, "fi": {"chunks": 4410.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234437, "cps": 34.453125, "mean_p": 0.234287, "boundary_entropy": 0.115134, "bpb": 2.145756}}, "bpb": {"en": 1.89815, "tr": 2.02807, "fi": 2.14576}, "gini_cpc": 0.026681, "gini_cps": 0.044104, "bpic": 4.768998, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.025014343887903, "fi": 1.126114348706858}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 286.85260009765625, "n_util_obs": 1059, "util_count_spearman": 0.23839678605148765}, "t": 1456.0} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.91206, "tr": 2.03558, "fi": 2.14903}, "bytes_per_chunk": {"en": 4.892, "tr": 5.285, "fi": 4.719}, "bpb_mean": 2.03222, "frozen_inventory": true, "t": 1457.5} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37433, "bpb": 1.98274, "chunk_loss": 0.030527, "F": 0.20724, "G": 0.21652, "chunks": 212.22, "bytes_per_chunk": 4.825, "boundary_entropy": 0.12013, "router_noise": 0.0, "grad_norm": 0.6682, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.34, "probe_sec": 0.0465, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04057, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10813, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.19031, "trunc_abs_frac": 0.48085, "mean_delta_full": -0.15667, "read_hit_rate": {"skel": 0.76333}, "mean_gate": {"skel": 0.15987}, "n_reads": 4067, "n_cand": 5328, "read_suppressed_frac": 0.1047, "dropout_reads_logged": 4067, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 286.91455078125, "n_util_obs": 1059, "util_count_spearman": 0.23836505718008177}, "probe_overhead_sec_total": 264.948, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012033, "alpha": 0.03, "t": 1458.9} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1471.3} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35186, "bpb": 1.95032, "chunk_loss": 0.03049, "F": 0.20569, "G": 0.21387, "chunks": 210.62, "bytes_per_chunk": 4.862, "boundary_entropy": 0.11258, "router_noise": 0.0, "grad_norm": 0.6844, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0255, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0197, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.2318, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.93282, "trunc_abs_frac": 0.61265, "mean_delta_full": 0.72729, "read_hit_rate": {"skel": 0.76904}, "mean_gate": {"skel": 0.17188}, "n_reads": 4069, "n_cand": 5291, "read_suppressed_frac": 0.104, "dropout_reads_logged": 4069, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 290.0338134765625, "n_util_obs": 1075, "util_count_spearman": 0.23813081692751803}, "probe_overhead_sec_total": 267.386, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011868, "alpha": 0.03, "t": 1471.5} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1484.7} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38076, "bpb": 1.99201, "chunk_loss": 0.030434, "F": 0.20169, "G": 0.20973, "chunks": 206.53, "bytes_per_chunk": 4.958, "boundary_entropy": 0.1112, "router_noise": 0.0, "grad_norm": 0.6767, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.268, "probe_sec": 0.0315, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02533, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03392, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.1543, "trunc_abs_frac": 1.1543, "mean_delta_full": -0.86353, "read_hit_rate": {"skel": 0.75955}, "mean_gate": {"skel": 0.16948}, "n_reads": 4018, "n_cand": 5290, "read_suppressed_frac": 0.0951, "dropout_reads_logged": 4018, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 293.14105224609375, "n_util_obs": 1087, "util_count_spearman": 0.2341884933995732}, "probe_overhead_sec_total": 269.745, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01366, "alpha": 0.03, "t": 1484.9} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1498.5} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3401, "bpb": 1.93335, "chunk_loss": 0.030394, "F": 0.19949, "G": 0.20864, "chunks": 204.28, "bytes_per_chunk": 5.013, "boundary_entropy": 0.11574, "router_noise": 0.0, "grad_norm": 0.6654, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.276, "probe_sec": 0.0654, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05978, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11248, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.08047, "trunc_abs_frac": 1.72353, "mean_delta_full": 0.37775, "read_hit_rate": {"skel": 0.75358}, "mean_gate": {"skel": 0.16886}, "n_reads": 3890, "n_cand": 5162, "read_suppressed_frac": 0.0946, "dropout_reads_logged": 3890, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 296.23333740234375, "n_util_obs": 1097, "util_count_spearman": 0.23012490442962327}, "probe_overhead_sec_total": 272.624, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012166, "alpha": 0.03, "t": 1498.7} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1511.1} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3654.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209964, "cps": 28.546875, "mean_p": 0.217647, "boundary_entropy": 0.123243, "bpb": 1.890948}, "tr": {"chunks": 3887.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.214716, "cps": 30.367188, "mean_p": 0.202508, "boundary_entropy": 0.10171, "bpb": 2.017893}, "fi": {"chunks": 4456.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.236883, "cps": 34.8125, "mean_p": 0.23544, "boundary_entropy": 0.110354, "bpb": 2.132409}}, "bpb": {"en": 1.89095, "tr": 2.01789, "fi": 2.13241}, "gini_cpc": 0.027127, "gini_cps": 0.044567, "bpic": 4.728849, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0226324425085538, "fi": 1.1282072231956655}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 299.262451171875, "n_util_obs": 1108, "util_count_spearman": 0.2308215480586193}, "t": 1511.8} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.90604, "tr": 2.02562, "fi": 2.13759}, "bytes_per_chunk": {"en": 4.871, "tr": 5.257, "fi": 4.68}, "bpb_mean": 2.02308, "frozen_inventory": true, "t": 1513.0} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32649, "bpb": 1.91371, "chunk_loss": 0.030454, "F": 0.2049, "G": 0.21394, "chunks": 209.81, "bytes_per_chunk": 4.881, "boundary_entropy": 0.11975, "router_noise": 0.0, "grad_norm": 0.6433, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.327, "probe_sec": 0.0716, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03657, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02182, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.50827, "trunc_abs_frac": 0.74325, "mean_delta_full": -0.26016, "read_hit_rate": {"skel": 0.77252}, "mean_gate": {"skel": 0.15765}, "n_reads": 4082, "n_cand": 5284, "read_suppressed_frac": 0.1036, "dropout_reads_logged": 4082, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 299.3251647949219, "n_util_obs": 1108, "util_count_spearman": 0.23034125674063818}, "probe_overhead_sec_total": 275.168, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.010905, "alpha": 0.03, "t": 1515.1} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1531.4} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34941, "bpb": 1.94678, "chunk_loss": 0.030512, "F": 0.20593, "G": 0.21599, "chunks": 210.88, "bytes_per_chunk": 4.856, "boundary_entropy": 0.11995, "router_noise": 0.0, "grad_norm": 0.7026, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.333, "probe_sec": 0.0351, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02542, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02092, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.13155, "trunc_abs_frac": 0.7168, "mean_delta_full": -0.23871, "read_hit_rate": {"skel": 0.7623}, "mean_gate": {"skel": 0.16143}, "n_reads": 4089, "n_cand": 5364, "read_suppressed_frac": 0.0932, "dropout_reads_logged": 4089, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 302.43707275390625, "n_util_obs": 1115, "util_count_spearman": 0.22553660357880012}, "probe_overhead_sec_total": 278.164, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.010855, "alpha": 0.03, "t": 1531.8} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1544.2} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37408, "bpb": 1.98237, "chunk_loss": 0.030569, "F": 0.20923, "G": 0.21802, "chunks": 214.25, "bytes_per_chunk": 4.779, "boundary_entropy": 0.11172, "router_noise": 0.0, "grad_norm": 0.7393, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.254, "probe_sec": 0.107, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09966, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04183, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 3.25033, "trunc_abs_frac": 0.48185, "mean_delta_full": 0.01831, "read_hit_rate": {"skel": 0.76064}, "mean_gate": {"skel": 0.16635}, "n_reads": 4004, "n_cand": 5264, "read_suppressed_frac": 0.1059, "dropout_reads_logged": 4004, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 305.54241943359375, "n_util_obs": 1126, "util_count_spearman": 0.22870189349704417}, "probe_overhead_sec_total": 280.876, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.013815, "alpha": 0.03, "t": 1544.4} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1557.9} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36433, "bpb": 1.96831, "chunk_loss": 0.030604, "F": 0.21121, "G": 0.21832, "chunks": 216.28, "bytes_per_chunk": 4.735, "boundary_entropy": 0.11414, "router_noise": 0.0, "grad_norm": 0.6799, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.274, "probe_sec": 0.0382, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0305, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15504, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 9, "trunc_ratio_sum": 1.22583, "trunc_abs_frac": 1.94845, "mean_delta_full": 0.21497, "read_hit_rate": {"skel": 0.77175}, "mean_gate": {"skel": 0.16181}, "n_reads": 4081, "n_cand": 5288, "read_suppressed_frac": 0.0953, "dropout_reads_logged": 4081, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 308.657470703125, "n_util_obs": 1133, "util_count_spearman": 0.22586687763542068}, "probe_overhead_sec_total": 284.064, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01541, "alpha": 0.03, "t": 1558.1} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1570.8} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3637.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208987, "cps": 28.414062, "mean_p": 0.217244, "boundary_entropy": 0.122732, "bpb": 1.884353}, "tr": {"chunks": 3892.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.214992, "cps": 30.40625, "mean_p": 0.203074, "boundary_entropy": 0.101718, "bpb": 2.010994}, "fi": {"chunks": 4392.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.23348, "cps": 34.3125, "mean_p": 0.234305, "boundary_entropy": 0.111088, "bpb": 2.123685}}, "bpb": {"en": 1.88435, "tr": 2.01099, "fi": 2.12368}, "gini_cpc": 0.024836, "gini_cps": 0.042222, "bpic": 4.758997, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0287340134191647, "fi": 1.1172008749407134}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 311.689453125, "n_util_obs": 1145, "util_count_spearman": 0.22046978444419377}, "t": 1571.7} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.89944, "tr": 2.01861, "fi": 2.12898}, "bytes_per_chunk": {"en": 4.873, "tr": 5.235, "fi": 4.737}, "bpb_mean": 2.01568, "frozen_inventory": true, "t": 1573.1} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32339, "bpb": 1.90924, "chunk_loss": 0.030342, "F": 0.19986, "G": 0.20843, "chunks": 204.66, "bytes_per_chunk": 5.004, "boundary_entropy": 0.11194, "router_noise": 0.0, "grad_norm": 0.661, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.342, "probe_sec": 0.0563, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02727, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04204, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.4167, "trunc_abs_frac": 0.40377, "mean_delta_full": -0.28918, "read_hit_rate": {"skel": 0.76925}, "mean_gate": {"skel": 0.15538}, "n_reads": 4137, "n_cand": 5378, "read_suppressed_frac": 0.1047, "dropout_reads_logged": 4137, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 311.7529296875, "n_util_obs": 1145, "util_count_spearman": 0.22042179685244023}, "probe_overhead_sec_total": 286.548, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012503, "alpha": 0.03, "t": 1575.2} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1593.2} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32128, "bpb": 1.90621, "chunk_loss": 0.030429, "F": 0.20291, "G": 0.21094, "chunks": 207.78, "bytes_per_chunk": 4.928, "boundary_entropy": 0.11682, "router_noise": 0.0, "grad_norm": 0.6594, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.364, "probe_sec": 0.0412, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03157, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06219, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.11033, "trunc_abs_frac": 1.23711, "mean_delta_full": 0.15914, "read_hit_rate": {"skel": 0.76015}, "mean_gate": {"skel": 0.1583}, "n_reads": 4063, "n_cand": 5345, "read_suppressed_frac": 0.1036, "dropout_reads_logged": 4063, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 314.84857177734375, "n_util_obs": 1156, "util_count_spearman": 0.22179505882145062}, "probe_overhead_sec_total": 288.985, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011855, "alpha": 0.03, "t": 1593.4} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1612.7} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37343, "bpb": 1.98144, "chunk_loss": 0.030541, "F": 0.20944, "G": 0.21683, "chunks": 214.47, "bytes_per_chunk": 4.775, "boundary_entropy": 0.1094, "router_noise": 0.0, "grad_norm": 0.6947, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.391, "probe_sec": 0.0348, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02442, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04151, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.13538, "trunc_abs_frac": 0.58218, "mean_delta_full": 0.07881, "read_hit_rate": {"skel": 0.78404}, "mean_gate": {"skel": 0.1625}, "n_reads": 4353, "n_cand": 5552, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 4353, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 317.93829345703125, "n_util_obs": 1164, "util_count_spearman": 0.22176976491499464}, "probe_overhead_sec_total": 291.602, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012015, "alpha": 0.03, "t": 1613.0} |
| {"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1638.1} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30797, "bpb": 1.88699, "chunk_loss": 0.030333, "F": 0.1994, "G": 0.20805, "chunks": 204.19, "bytes_per_chunk": 5.015, "boundary_entropy": 0.11406, "router_noise": 0.0, "grad_norm": 0.6979, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.508, "probe_sec": 0.0575, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04408, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15175, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.38357, "trunc_abs_frac": 1.04541, "mean_delta_full": 0.38513, "read_hit_rate": {"skel": 0.77197}, "mean_gate": {"skel": 0.1576}, "n_reads": 4120, "n_cand": 5337, "read_suppressed_frac": 0.1036, "dropout_reads_logged": 4120, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 321.03045654296875, "n_util_obs": 1178, "util_count_spearman": 0.22316985611746923}, "probe_overhead_sec_total": 294.634, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012446, "alpha": 0.03, "t": 1638.4} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1653.6} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3653.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209906, "cps": 28.539062, "mean_p": 0.218392, "boundary_entropy": 0.123235, "bpb": 1.879043}, "tr": {"chunks": 3904.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215655, "cps": 30.5, "mean_p": 0.203556, "boundary_entropy": 0.099952, "bpb": 2.003394}, "fi": {"chunks": 4436.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235819, "cps": 34.65625, "mean_p": 0.234618, "boundary_entropy": 0.110867, "bpb": 2.117311}}, "bpb": {"en": 1.87904, "tr": 2.00339, "fi": 2.11731}, "gini_cpc": 0.02612, "gini_cps": 0.043525, "bpic": 4.730426, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0273861470888839, "fi": 1.1234509143156899}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 324.0805969238281, "n_util_obs": 1189, "util_count_spearman": 0.2351359924347411}, "t": 1654.4} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.89337, "tr": 2.01208, "fi": 2.12302}, "bytes_per_chunk": {"en": 4.853, "tr": 5.207, "fi": 4.691}, "bpb_mean": 2.00949, "frozen_inventory": true, "t": 1655.8} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38168, "bpb": 1.99335, "chunk_loss": 0.030628, "F": 0.20963, "G": 0.21906, "chunks": 214.66, "bytes_per_chunk": 4.77, "boundary_entropy": 0.11498, "router_noise": 0.0, "grad_norm": 0.7083, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.377, "probe_sec": 0.0739, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.069, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04689, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.75095, "trunc_abs_frac": 0.51052, "mean_delta_full": 0.06443, "read_hit_rate": {"skel": 0.76392}, "mean_gate": {"skel": 0.15632}, "n_reads": 4171, "n_cand": 5460, "read_suppressed_frac": 0.0928, "dropout_reads_logged": 4171, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 324.1427307128906, "n_util_obs": 1190, "util_count_spearman": 0.23501088381005567}, "probe_overhead_sec_total": 297.107, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012236, "alpha": 0.03, "t": 1657.2} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1681.1} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31428, "bpb": 1.89611, "chunk_loss": 0.030619, "F": 0.21298, "G": 0.21923, "chunks": 218.09, "bytes_per_chunk": 4.695, "boundary_entropy": 0.11247, "router_noise": 0.0, "grad_norm": 0.6897, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.487, "probe_sec": 0.0477, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02661, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00477, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 4.98697, "trunc_abs_frac": 1.08053, "mean_delta_full": 0.0203, "read_hit_rate": {"skel": 0.77817}, "mean_gate": {"skel": 0.16267}, "n_reads": 4206, "n_cand": 5405, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 4206, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 327.24810791015625, "n_util_obs": 1194, "util_count_spearman": 0.23704316066538453}, "probe_overhead_sec_total": 299.886, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.013091, "alpha": 0.03, "t": 1681.6} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1704.8} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30314, "bpb": 1.88003, "chunk_loss": 0.030531, "F": 0.20755, "G": 0.21623, "chunks": 212.53, "bytes_per_chunk": 4.818, "boundary_entropy": 0.11447, "router_noise": 0.0, "grad_norm": 0.6722, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.471, "probe_sec": 0.0698, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05817, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13085, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.59876, "trunc_abs_frac": 0.84182, "mean_delta_full": 0.16748, "read_hit_rate": {"skel": 0.76677}, "mean_gate": {"skel": 0.15475}, "n_reads": 4116, "n_cand": 5368, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 4116, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 330.36663818359375, "n_util_obs": 1206, "util_count_spearman": 0.24029129394306292}, "probe_overhead_sec_total": 302.946, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011915, "alpha": 0.03, "t": 1705.1} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1720.7} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35691, "bpb": 1.9576, "chunk_loss": 0.030479, "F": 0.20596, "G": 0.21296, "chunks": 210.91, "bytes_per_chunk": 4.855, "boundary_entropy": 0.11262, "router_noise": 0.0, "grad_norm": 0.6951, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.316, "probe_sec": 0.0673, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.056, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01424, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.66941, "trunc_abs_frac": 0.90877, "mean_delta_full": -0.16082, "read_hit_rate": {"skel": 0.75731}, "mean_gate": {"skel": 0.1624}, "n_reads": 3935, "n_cand": 5196, "read_suppressed_frac": 0.1055, "dropout_reads_logged": 3935, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 333.481689453125, "n_util_obs": 1216, "util_count_spearman": 0.24975202557247977}, "probe_overhead_sec_total": 305.937, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011791, "alpha": 0.03, "t": 1720.9} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1736.7} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3670.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210883, "cps": 28.671875, "mean_p": 0.219819, "boundary_entropy": 0.122961, "bpb": 1.874436}, "tr": {"chunks": 3915.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.216262, "cps": 30.585938, "mean_p": 0.203347, "boundary_entropy": 0.096981, "bpb": 1.997952}, "fi": {"chunks": 4462.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.237202, "cps": 34.859375, "mean_p": 0.235674, "boundary_entropy": 0.109018, "bpb": 2.108888}}, "bpb": {"en": 1.87444, "tr": 1.99795, "fi": 2.10889}, "gini_cpc": 0.02641, "gini_cps": 0.043828, "bpic": 4.709222, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0255085153814814, "fi": 1.124801115701767}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 336.5386962890625, "n_util_obs": 1226, "util_count_spearman": 0.2525199111584296}, "t": 1737.7} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.88989, "tr": 2.00609, "fi": 2.11446}, "bytes_per_chunk": {"en": 4.833, "tr": 5.213, "fi": 4.678}, "bpb_mean": 2.00348, "frozen_inventory": true, "t": 1739.1} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34471, "bpb": 1.94001, "chunk_loss": 0.030416, "F": 0.20264, "G": 0.21126, "chunks": 207.5, "bytes_per_chunk": 4.935, "boundary_entropy": 0.1122, "router_noise": 0.0, "grad_norm": 0.7329, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.391, "probe_sec": 0.0285, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02179, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02004, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 11.0049, "trunc_abs_frac": 1.30165, "mean_delta_full": 0.01169, "read_hit_rate": {"skel": 0.77861}, "mean_gate": {"skel": 0.1602}, "n_reads": 4259, "n_cand": 5470, "read_suppressed_frac": 0.0939, "dropout_reads_logged": 4259, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 336.6036376953125, "n_util_obs": 1226, "util_count_spearman": 0.25270333819514207}, "probe_overhead_sec_total": 308.8, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012639, "alpha": 0.03, "t": 1740.5} |
| {"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1753.3} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33374, "bpb": 1.92418, "chunk_loss": 0.030563, "F": 0.20923, "G": 0.21742, "chunks": 214.25, "bytes_per_chunk": 4.779, "boundary_entropy": 0.11494, "router_noise": 0.0, "grad_norm": 0.7429, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.1139, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1018, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21175, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.44982, "trunc_abs_frac": 0.772, "mean_delta_full": 0.11677, "read_hit_rate": {"skel": 0.77456}, "mean_gate": {"skel": 0.15184}, "n_reads": 4195, "n_cand": 5416, "read_suppressed_frac": 0.0925, "dropout_reads_logged": 4195, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 339.703369140625, "n_util_obs": 1233, "util_count_spearman": 0.2505985188583773}, "probe_overhead_sec_total": 311.362, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.012877, "alpha": 0.03, "t": 1753.7} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1766.6} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34364, "bpb": 1.93847, "chunk_loss": 0.030369, "F": 0.20056, "G": 0.20952, "chunks": 205.38, "bytes_per_chunk": 4.986, "boundary_entropy": 0.11343, "router_noise": 0.0, "grad_norm": 0.7176, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.265, "probe_sec": 0.0672, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05191, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04507, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.24487, "trunc_abs_frac": 1.24487, "mean_delta_full": -0.07727, "read_hit_rate": {"skel": 0.74953}, "mean_gate": {"skel": 0.15884}, "n_reads": 3947, "n_cand": 5266, "read_suppressed_frac": 0.095, "dropout_reads_logged": 3947, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 342.81219482421875, "n_util_obs": 1247, "util_count_spearman": 0.25925482316865284}, "probe_overhead_sec_total": 314.162, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.012032, "alpha": 0.03, "t": 1766.9} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1779.9} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39191, "bpb": 2.0081, "chunk_loss": 0.030553, "F": 0.20959, "G": 0.21728, "chunks": 214.62, "bytes_per_chunk": 4.771, "boundary_entropy": 0.11033, "router_noise": 0.0, "grad_norm": 0.726, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.276, "probe_sec": 0.0414, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02558, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13596, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.70608, "trunc_abs_frac": 1.67634, "mean_delta_full": -0.10923, "read_hit_rate": {"skel": 0.76904}, "mean_gate": {"skel": 0.15295}, "n_reads": 4139, "n_cand": 5382, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 4139, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 345.9118957519531, "n_util_obs": 1258, "util_count_spearman": 0.2563553456554022}, "probe_overhead_sec_total": 316.971, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018178, "alpha": 0.03, "t": 1780.7} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1792.9} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3646.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209504, "cps": 28.484375, "mean_p": 0.217231, "boundary_entropy": 0.119841, "bpb": 1.872288}, "tr": {"chunks": 3877.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.214163, "cps": 30.289062, "mean_p": 0.201703, "boundary_entropy": 0.09944, "bpb": 1.994349}, "fi": {"chunks": 4409.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234384, "cps": 34.445312, "mean_p": 0.233724, "boundary_entropy": 0.107824, "bpb": 2.105884}}, "bpb": {"en": 1.87229, "tr": 1.99435, "fi": 2.10588}, "gini_cpc": 0.025206, "gini_cps": 0.04263, "bpic": 4.754609, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.022239610852376, "fi": 1.118756756770943}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 348.95794677734375, "n_util_obs": 1271, "util_count_spearman": 0.25322582402979243}, "t": 1793.8} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.88709, "tr": 2.00182, "fi": 2.11144}, "bytes_per_chunk": {"en": 4.858, "tr": 5.256, "fi": 4.728}, "bpb_mean": 2.00012, "frozen_inventory": true, "t": 1794.9} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31701, "bpb": 1.90005, "chunk_loss": 0.030588, "F": 0.20609, "G": 0.21331, "chunks": 211.03, "bytes_per_chunk": 4.852, "boundary_entropy": 0.11298, "router_noise": 0.0, "grad_norm": 0.7014, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.311, "probe_sec": 0.0267, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02027, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0157, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -5.3922, "trunc_abs_frac": 1.28317, "mean_delta_full": 0.0224, "read_hit_rate": {"skel": 0.77304}, "mean_gate": {"skel": 0.15768}, "n_reads": 4101, "n_cand": 5305, "read_suppressed_frac": 0.1051, "dropout_reads_logged": 4101, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 349.0212707519531, "n_util_obs": 1271, "util_count_spearman": 0.2532871978890859}, "probe_overhead_sec_total": 319.752, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018021, "alpha": 0.03, "t": 1796.3} |
| {"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1809.3} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34062, "bpb": 1.9341, "chunk_loss": 0.030491, "F": 0.2056, "G": 0.21497, "chunks": 210.53, "bytes_per_chunk": 4.864, "boundary_entropy": 0.11197, "router_noise": 0.0, "grad_norm": 0.7381, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.268, "probe_sec": 0.0465, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03973, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10976, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.01968, "trunc_abs_frac": 1.23597, "mean_delta_full": 0.08259, "read_hit_rate": {"skel": 0.77046}, "mean_gate": {"skel": 0.14763}, "n_reads": 4226, "n_cand": 5485, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 4226, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 352.1033020019531, "n_util_obs": 1280, "util_count_spearman": 0.25261112882212977}, "probe_overhead_sec_total": 322.327, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.012246, "alpha": 0.03, "t": 1809.7} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1823.3} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27374, "bpb": 1.83762, "chunk_loss": 0.030357, "F": 0.20032, "G": 0.20833, "chunks": 205.12, "bytes_per_chunk": 4.992, "boundary_entropy": 0.10945, "router_noise": 0.0, "grad_norm": 0.6985, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.277, "probe_sec": 0.0316, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02409, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13587, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.45149, "trunc_abs_frac": 1.4305, "mean_delta_full": 0.28406, "read_hit_rate": {"skel": 0.77311}, "mean_gate": {"skel": 0.15375}, "n_reads": 4089, "n_cand": 5289, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 4089, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 355.2392272949219, "n_util_obs": 1287, "util_count_spearman": 0.2534453981788791}, "probe_overhead_sec_total": 325.44, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.012511, "alpha": 0.03, "t": 1823.5} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1836.6} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37624, "bpb": 1.9855, "chunk_loss": 0.030419, "F": 0.20319, "G": 0.21191, "chunks": 208.06, "bytes_per_chunk": 4.922, "boundary_entropy": 0.11136, "router_noise": 0.0, "grad_norm": 0.7298, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.266, "probe_sec": 0.0257, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08106, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.38012, "trunc_abs_frac": 0.68946, "mean_delta_full": 0.11243, "read_hit_rate": {"skel": 0.76606}, "mean_gate": {"skel": 0.15662}, "n_reads": 4126, "n_cand": 5386, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 4126, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 358.34332275390625, "n_util_obs": 1293, "util_count_spearman": 0.2550584247881943}, "probe_overhead_sec_total": 328.491, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011673, "alpha": 0.03, "t": 1836.8} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1849.5} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38139, "bpb": 1.99293, "chunk_loss": 0.030558, "F": 0.20773, "G": 0.2147, "chunks": 212.72, "bytes_per_chunk": 4.814, "boundary_entropy": 0.11171, "router_noise": 0.0, "grad_norm": 0.6995, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.0367, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02484, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04637, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 32.73571, "trunc_abs_frac": 0.31298, "mean_delta_full": -0.00107, "read_hit_rate": {"skel": 0.75577}, "mean_gate": {"skel": 0.15579}, "n_reads": 3964, "n_cand": 5245, "read_suppressed_frac": 0.107, "dropout_reads_logged": 3964, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 361.39111328125, "n_util_obs": 1300, "util_count_spearman": 0.2596489470935999}, "probe_overhead_sec_total": 331.373, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011859, "alpha": 0.03, "t": 1849.5} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3636.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208929, "cps": 28.40625, "mean_p": 0.216399, "boundary_entropy": 0.119409, "bpb": 1.869837}, "tr": {"chunks": 3900.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215434, "cps": 30.46875, "mean_p": 0.202367, "boundary_entropy": 0.099521, "bpb": 1.989811}, "fi": {"chunks": 4382.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232949, "cps": 34.234375, "mean_p": 0.23301, "boundary_entropy": 0.106587, "bpb": 2.104174}}, "bpb": {"en": 1.86984, "tr": 1.98981, "fi": 2.10417}, "gini_cpc": 0.024361, "gini_cps": 0.04173, "bpic": 4.760195, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0311320863070121, "fi": 1.1149637184759356}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 361.39111328125, "n_util_obs": 1300, "util_count_spearman": 0.2596489470935999}, "t": 1850.3} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.88486, "tr": 1.99774, "fi": 2.10921}, "bytes_per_chunk": {"en": 4.873, "tr": 5.239, "fi": 4.739}, "bpb_mean": 1.99727, "frozen_inventory": true, "t": 1851.6} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_mem-skel_s2/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-skel_s2/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-skel_s2/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-skel_s2/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-skel_s2/probe/reps/th0.70"], "t": 1854.3} |
| {"event": "final", "run": "ali_mem-skel_s2", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 2, "per_lang": {"en": {"cpc": 0.2089294949146699, "cps": 28.40625, "bpb": 1.8891463736054175, "bytes_per_chunk": 4.814840664891174, "reads_per_chunk": 0.5980989989898062}, "tr": {"cpc": 0.21543390598243384, "cps": 30.46875, "bpb": 1.9904279993433465, "bytes_per_chunk": 5.251176859437912, "reads_per_chunk": 0.5799963943029988}, "fi": {"cpc": 0.23294880654935943, "cps": 34.234375, "bpb": 2.126474477598491, "bytes_per_chunk": 4.7085956514320095, "reads_per_chunk": 0.6858918515990552}}, "bpb_mean": 2.002016283515752, "gini_cpc": 0.024361139564611722, "gini_cps": 0.04172959668848253, "bpic": 4.76019466353415, "mean_boundary_entropy": 0.11170578002929688, "dechunk": "ema", "router_noise": 0.0, "transitions": [{"step": 0, "phase": 1, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0}, {"step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0}, {"step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0}], "probe_overhead_sec_total": 331.3731119632721, "probe_overhead_seq_frac": 0.499765625, "wall_sec": 1857.3115632534027, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 361.39111328125, "n_util_obs": 1300, "util_count_spearman": 0.2596489470935999}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["skel"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "util_vs_frequency_spearman": 0.2596489470935999, "paired_vs_dropout": {"n": 2730, "pearson": -0.006880562740271094, "spearman": -0.002931798169149572, "mean_abs_discrepancy": 1.059653639793396, "mean_paired": 0.044733691960573196, "mean_dropout": 0.055564913898706436}, "predictor_calibration": {"n": 6592, "all": {"n": 6592, "pearson": 0.1363131604910194, "spearman": 0.1383622981261577, "mae": 0.0987764224410057, "reliability": [{"pred": 0.002392061287537217, "obs": 0.010740666650235653, "n": 660}, {"pred": 0.012135072611272335, "obs": 0.018849022686481476, "n": 659}, {"pred": 0.014819816686213017, "obs": 0.02018786035478115, "n": 659}, {"pred": 0.01694437488913536, "obs": 0.022722948342561722, "n": 659}, {"pred": 0.018824057653546333, "obs": 0.021593667566776276, "n": 659}, {"pred": 0.02109544351696968, "obs": 0.03046458587050438, "n": 659}, {"pred": 0.024069782346487045, "obs": 0.036752745509147644, "n": 659}, {"pred": 0.02879488095641136, "obs": 0.026973681524395943, "n": 659}, {"pred": 0.036858052015304565, "obs": 0.0521981343626976, "n": 659}, {"pred": 0.06617799401283264, "obs": 0.08191295713186264, "n": 660}]}, "admitted": {"n": 6592, "pearson": 0.1363131604910194, "spearman": 0.1383622981261577, "mae": 0.0987764224410057, "reliability": [{"pred": 0.002392061287537217, "obs": 0.010740666650235653, "n": 660}, {"pred": 0.012135072611272335, "obs": 0.018849022686481476, "n": 659}, {"pred": 0.014819816686213017, "obs": 0.02018786035478115, "n": 659}, {"pred": 0.01694437488913536, "obs": 0.022722948342561722, "n": 659}, {"pred": 0.018824057653546333, "obs": 0.021593667566776276, "n": 659}, {"pred": 0.02109544351696968, "obs": 0.03046458587050438, "n": 659}, {"pred": 0.024069782346487045, "obs": 0.036752745509147644, "n": 659}, {"pred": 0.02879488095641136, "obs": 0.026973681524395943, "n": 659}, {"pred": 0.036858052015304565, "obs": 0.0521981343626976, "n": 659}, {"pred": 0.06617799401283264, "obs": 0.08191295713186264, "n": 660}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 1857.4} |
| {"event": "hub_done", "repo": "AdaptiveChunking/ali_mem-skel_s2", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 1873.0} |
|
|
