| {"event": "hub", "repo": "AdaptiveChunking/ali_mem-random_s2", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 3.6} |
| {"event": "start", "run": "ali_mem-random_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": "random", "value_mode": "learned", "candidate_types": ["surface"], "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": 3.6} |
| {"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": 3.6} |
| {"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": 5.6} |
| {"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": 11.4934, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.019, "probe_sec": 0.0191, "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": {"surface": 0.0}, "mean_gate": {"surface": null}, "n_reads": 0, "n_cand": 19574, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 2006, "admitted": 2006, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 2006, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.757726669311523, "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": 6.9} |
| {"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10907.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.626731, "cps": 85.210938, "mean_p": 0.513143, "boundary_entropy": 0.686304, "bpb": 8.105516}, "tr": {"chunks": 8204.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.453185, "cps": 64.09375, "mean_p": 0.498082, "boundary_entropy": 0.686828, "bpb": 8.082919}, "fi": {"chunks": 13505.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.717931, "cps": 105.507812, "mean_p": 0.517039, "boundary_entropy": 0.686672, "bpb": 8.064982}}, "bpb": {"en": 8.10552, "tr": 8.08292, "fi": 8.06498}, "gini_cpc": 0.098172, "gini_cps": 0.108352, "bpic": 1.739392, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7230925837964345, "fi": 1.1455169299595527}, "inventory": {"size": 2006, "admitted": 2006, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 2006, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.757726669311523, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 7.8} |
| {"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10881.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.625237, "cps": 85.007812, "mean_p": 0.51299, "boundary_entropy": 0.686317, "bpb": 8.043342}, "tr": {"chunks": 8249.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.45567, "cps": 64.445312, "mean_p": 0.498258, "boundary_entropy": 0.686832, "bpb": 8.039123}, "fi": {"chunks": 13468.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.715964, "cps": 105.21875, "mean_p": 0.516808, "boundary_entropy": 0.686691, "bpb": 8.006897}}, "bpb": {"en": 8.04334, "tr": 8.03912, "fi": 8.0069}, "gini_cpc": 0.096573, "gini_cps": 0.106735, "bpic": 1.740352, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7287961371547589, "fi": 1.1451082252072065}, "inventory": {"size": 3536, "admitted": 3536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 3536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 10.45418643951416, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 9.6} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10720.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.615986, "cps": 83.75, "mean_p": 0.512136, "boundary_entropy": 0.686378, "bpb": 7.833662}, "tr": {"chunks": 8394.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.46368, "cps": 65.578125, "mean_p": 0.49871, "boundary_entropy": 0.686846, "bpb": 7.88636}, "fi": {"chunks": 13218.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.702674, "cps": 103.265625, "mean_p": 0.515691, "boundary_entropy": 0.686774, "bpb": 7.806576}}, "bpb": {"en": 7.83366, "tr": 7.88636, "fi": 7.80658}, "gini_cpc": 0.089393, "gini_cps": 0.099468, "bpic": 1.75467, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7527447726566308, "fi": 1.1407308818990478}, "inventory": {"size": 5987, "admitted": 5987, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 5987, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 12.38099193572998, "n_util_obs": 1, "util_count_spearman": NaN}, "t": 11.7} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10342.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.594265, "cps": 80.796875, "mean_p": 0.510427, "boundary_entropy": 0.68648, "bpb": 7.554479}, "tr": {"chunks": 8523.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.470806, "cps": 66.585938, "mean_p": 0.499245, "boundary_entropy": 0.686868, "bpb": 7.679644}, "fi": {"chunks": 12730.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.676732, "cps": 99.453125, "mean_p": 0.513623, "boundary_entropy": 0.686908, "bpb": 7.546268}}, "bpb": {"en": 7.55448, "tr": 7.67964, "fi": 7.54627}, "gini_cpc": 0.078817, "gini_cps": 0.088769, "bpic": 1.795601, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7922486791107757, "fi": 1.1387702346790944}, "inventory": {"size": 7708, "admitted": 7708, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 7708, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 14.460041999816895, "n_util_obs": 1, "util_count_spearman": NaN}, "t": 14.4} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 9680.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.556226, "cps": 75.625, "mean_p": 0.507432, "boundary_entropy": 0.686611, "bpb": 7.299168}, "tr": {"chunks": 8649.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.477766, "cps": 67.570312, "mean_p": 0.499662, "boundary_entropy": 0.6869, "bpb": 7.478937}, "fi": {"chunks": 11842.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.629525, "cps": 92.515625, "mean_p": 0.51022, "boundary_entropy": 0.687082, "bpb": 7.311924}}, "bpb": {"en": 7.29917, "tr": 7.47894, "fi": 7.31192}, "gini_cpc": 0.060819, "gini_cps": 0.070553, "bpic": 1.880349, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8589425329256875, "fi": 1.1317797943088515}, "inventory": {"size": 9781, "admitted": 9781, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 9781, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 14.897965431213379, "n_util_obs": 2, "util_count_spearman": NaN}, "t": 16.3} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3802.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.218468, "cps": 29.703125, "mean_p": 0.484743, "boundary_entropy": 0.686192, "bpb": 6.726197}, "tr": {"chunks": 7343.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.405623, "cps": 57.367188, "mean_p": 0.494694, "boundary_entropy": 0.687003, "bpb": 7.013662}, "fi": {"chunks": 4524.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.240498, "cps": 35.34375, "mean_p": 0.486317, "boundary_entropy": 0.686887, "bpb": 6.787457}}, "bpb": {"en": 6.7262, "tr": 7.01366, "fi": 6.78746}, "gini_cpc": 0.144312, "gini_cps": 0.150658, "bpic": 3.620652, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.8566711298951761, "fi": 1.1008362455725662}, "inventory": {"size": 23758, "admitted": 23758, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 23758, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.623495101928711, "n_util_obs": 4, "util_count_spearman": -0.8}, "t": 22.5} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2625.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.150836, "cps": 20.507812, "mean_p": 0.486862, "boundary_entropy": 0.686959, "bpb": 6.30563}, "tr": {"chunks": 4617.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.255041, "cps": 36.070312, "mean_p": 0.490024, "boundary_entropy": 0.687549, "bpb": 6.652271}, "fi": {"chunks": 1358.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.072192, "cps": 10.609375, "mean_p": 0.479003, "boundary_entropy": 0.686932, "bpb": 6.422055}}, "bpb": {"en": 6.30563, "tr": 6.65227, "fi": 6.42205}, "gini_cpc": 0.254983, "gini_cps": 0.252636, "bpic": 6.596744, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.690846315922381, "fi": 0.47861102546382434}, "inventory": {"size": 50409, "admitted": 50409, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 50409, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 6.429645538330078, "n_util_obs": 8, "util_count_spearman": 0.023809523809523808}, "t": 31.3} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 37.5} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.21302, "bpb": 6.07811, "chunk_loss": 0.034335, "F": 0.2276, "G": 0.48862, "chunks": 233.06, "bytes_per_chunk": 4.394, "boundary_entropy": 0.69161, "router_noise": 0.0, "grad_norm": 2.4797, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 0.62, "probe_sec": 0.0382, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03179, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00614, "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": 0.31027, "trunc_abs_frac": 0.85292, "mean_delta_full": -0.01823, "read_hit_rate": {"surface": 0.69483}, "mean_gate": {"surface": 0.11916}, "n_reads": 5157, "n_cand": 7422, "read_suppressed_frac": 0.103, "dropout_reads_logged": 5157, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 5.810028076171875, "n_util_obs": 11, "util_count_spearman": 0.045454545454545456}, "probe_overhead_sec_total": 2.629, "probe_overhead_seq_frac": 0.4828, "util_loss": 0.044625, "alpha": 0.03, "t": 37.9} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4917.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.282537, "cps": 38.414062, "mean_p": 0.485914, "boundary_entropy": 0.685574, "bpb": 5.552488}, "tr": {"chunks": 5557.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.306966, "cps": 43.414062, "mean_p": 0.487152, "boundary_entropy": 0.686519, "bpb": 5.954932}, "fi": {"chunks": 4780.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.254107, "cps": 37.34375, "mean_p": 0.48427, "boundary_entropy": 0.686341, "bpb": 5.718211}}, "bpb": {"en": 5.55249, "tr": 5.95493, "fi": 5.71821}, "gini_cpc": 0.041772, "gini_cps": 0.033958, "bpic": 3.719156, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.086460039169127, "fi": 0.8993731647869113}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 6.15966796875, "n_util_obs": 11, "util_count_spearman": -0.09090909090909091}, "t": 43.0} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 57.3} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.24952, "bpb": 4.68806, "chunk_loss": 0.029547, "F": 0.15915, "G": 0.45276, "chunks": 162.97, "bytes_per_chunk": 6.283, "boundary_entropy": 0.6821, "router_noise": 0.0, "grad_norm": 1.4846, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.399, "probe_sec": 0.2699, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.24187, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00362, "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": -0.0, "trunc_abs_frac": 2.96363, "mean_delta_full": -0.03174, "read_hit_rate": {"surface": 0.54816}, "mean_gate": {"surface": 0.11946}, "n_reads": 2840, "n_cand": 5181, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 2840, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 6.82940673828125, "n_util_obs": 15, "util_count_spearman": -0.2}, "probe_overhead_sec_total": 5.789, "probe_overhead_seq_frac": 0.4913, "util_loss": 0.000886, "alpha": 0.03, "t": 57.8} |
| {"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3577.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205539, "cps": 27.945312, "mean_p": 0.413038, "boundary_entropy": 0.643362, "bpb": 4.153193}, "tr": {"chunks": 4547.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.251174, "cps": 35.523438, "mean_p": 0.434195, "boundary_entropy": 0.659046, "bpb": 4.396497}, "fi": {"chunks": 3667.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.194939, "cps": 28.648438, "mean_p": 0.425315, "boundary_entropy": 0.659312, "bpb": 4.339411}}, "bpb": {"en": 4.15319, "tr": 4.3965, "fi": 4.33941}, "gini_cpc": 0.05753, "gini_cps": 0.054844, "bpic": 4.811466, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2220235708657528, "fi": 0.948427649719343}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 7.0527191162109375, "n_util_obs": 18, "util_count_spearman": -0.14551083591331268}, "t": 66.6} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 72.8} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.78174, "bpb": 4.0132, "chunk_loss": 0.031886, "F": 0.20572, "G": 0.38689, "chunks": 210.66, "bytes_per_chunk": 4.861, "boundary_entropy": 0.62214, "router_noise": 0.0, "grad_norm": 1.1144, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.306, "probe_sec": 0.0866, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07476, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01292, "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": -10.94444, "trunc_abs_frac": 1.19027, "mean_delta_full": -0.00037, "read_hit_rate": {"surface": 0.67462}, "mean_gate": {"surface": 0.12046}, "n_reads": 4526, "n_cand": 6709, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 4526, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 7.510528564453125, "n_util_obs": 20, "util_count_spearman": -0.2796992481203007}, "probe_overhead_sec_total": 8.387, "probe_overhead_seq_frac": 0.4942, "util_loss": 0.000427, "alpha": 0.03, "t": 73.1} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 86.1} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3779.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.217146, "cps": 29.523438, "mean_p": 0.3793, "boundary_entropy": 0.587912, "bpb": 3.630745}, "tr": {"chunks": 4478.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.247362, "cps": 34.984375, "mean_p": 0.379181, "boundary_entropy": 0.59889, "bpb": 3.813122}, "fi": {"chunks": 3522.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.187231, "cps": 27.515625, "mean_p": 0.389988, "boundary_entropy": 0.623406, "bpb": 3.864857}}, "bpb": {"en": 3.63074, "tr": 3.81312, "fi": 3.86486}, "gini_cpc": 0.061509, "gini_cps": 0.054108, "bpic": 4.816368, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1391496107576595, "fi": 0.8622331111996383}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 8.473342895507812, "n_util_obs": 24, "util_count_spearman": -0.09304347826086956}, "t": 86.9} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.63764, "tr": 3.82357, "fi": 3.86315}, "bytes_per_chunk": {"en": 5.202, "tr": 5.06, "fi": 6.381}, "bpb_mean": 3.77479, "frozen_inventory": true, "t": 87.9} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.52957, "bpb": 3.64939, "chunk_loss": 0.030833, "F": 0.18518, "G": 0.36903, "chunks": 189.62, "bytes_per_chunk": 5.4, "boundary_entropy": 0.5899, "router_noise": 0.0, "grad_norm": 1.2896, "lr": 0.000201, "router_trainable": true, "sec_per_step": 0.317, "probe_sec": 0.0306, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02049, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0136, "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": -0.26905, "trunc_abs_frac": 0.28268, "mean_delta_full": 0.12653, "read_hit_rate": {"surface": 0.62326}, "mean_gate": {"surface": 0.12296}, "n_reads": 3762, "n_cand": 6036, "read_suppressed_frac": 0.096, "dropout_reads_logged": 3762, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 8.505355834960938, "n_util_obs": 25, "util_count_spearman": -0.003076923076923077}, "probe_overhead_sec_total": 10.766, "probe_overhead_seq_frac": 0.4956, "util_loss": 0.000443, "alpha": 0.03, "t": 88.9} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 101.2} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.38506, "bpb": 3.44091, "chunk_loss": 0.030845, "F": 0.18744, "G": 0.34692, "chunks": 191.94, "bytes_per_chunk": 5.335, "boundary_entropy": 0.54734, "router_noise": 0.0, "grad_norm": 2.235, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.248, "probe_sec": 0.0292, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02418, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02319, "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": -12.12803, "trunc_abs_frac": 1.7441, "mean_delta_full": -0.00629, "read_hit_rate": {"surface": 0.63191}, "mean_gate": {"surface": 0.13161}, "n_reads": 3861, "n_cand": 6110, "read_suppressed_frac": 0.0912, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.26519775390625, "n_util_obs": 34, "util_count_spearman": 0.07776928953399542}, "probe_overhead_sec_total": 13.597, "probe_overhead_seq_frac": 0.4965, "util_loss": 0.00081, "alpha": 0.03, "t": 101.4} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3132.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.179969, "cps": 24.46875, "mean_p": 0.34636, "boundary_entropy": 0.541865, "bpb": 3.344389}, "tr": {"chunks": 4532.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.250345, "cps": 35.40625, "mean_p": 0.330798, "boundary_entropy": 0.526875, "bpb": 3.504978}, "fi": {"chunks": 4804.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.255382, "cps": 37.53125, "mean_p": 0.374047, "boundary_entropy": 0.585009, "bpb": 3.651423}}, "bpb": {"en": 3.34439, "tr": 3.50498, "fi": 3.65142}, "gini_cpc": 0.073321, "gini_cps": 0.089402, "bpic": 4.550209, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.3910467200986776, "fi": 1.419036224626864}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.479171752929688, "n_util_obs": 36, "util_count_spearman": 0.008751608751608752}, "t": 103.5} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 116.0} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.27649, "bpb": 3.28429, "chunk_loss": 0.030916, "F": 0.19452, "G": 0.31532, "chunks": 199.19, "bytes_per_chunk": 5.141, "boundary_entropy": 0.4862, "router_noise": 0.0, "grad_norm": 1.2175, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.298, "probe_sec": 0.0295, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02392, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01201, "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": -0.04294, "trunc_abs_frac": 0.60016, "mean_delta_full": -0.22955, "read_hit_rate": {"surface": 0.60754}, "mean_gate": {"surface": 0.14759}, "n_reads": 3853, "n_cand": 6342, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 3853, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 10.012680053710938, "n_util_obs": 43, "util_count_spearman": -0.05421322863183328}, "probe_overhead_sec_total": 16.214, "probe_overhead_seq_frac": 0.4971, "util_loss": 0.000993, "alpha": 0.03, "t": 116.3} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 129.0} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.16117, "bpb": 3.11791, "chunk_loss": 0.031146, "F": 0.20587, "G": 0.29325, "chunks": 210.81, "bytes_per_chunk": 4.857, "boundary_entropy": 0.43192, "router_noise": 0.0, "grad_norm": 1.0805, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.258, "probe_sec": 0.0287, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02384, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00258, "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": 0.57988, "trunc_abs_frac": 0.57988, "mean_delta_full": 0.27124, "read_hit_rate": {"surface": 0.66309}, "mean_gate": {"surface": 0.17468}, "n_reads": 4452, "n_cand": 6714, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 4452, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 11.686294555664062, "n_util_obs": 53, "util_count_spearman": -0.12360909530720851}, "probe_overhead_sec_total": 18.924, "probe_overhead_seq_frac": 0.4975, "util_loss": 0.002497, "alpha": 0.03, "t": 129.2} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 140.4} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3075.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.176694, "cps": 24.023438, "mean_p": 0.268252, "boundary_entropy": 0.393501, "bpb": 2.904928}, "tr": {"chunks": 3522.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194553, "cps": 27.515625, "mean_p": 0.274207, "boundary_entropy": 0.429776, "bpb": 3.170221}, "fi": {"chunks": 4139.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220031, "cps": 32.335938, "mean_p": 0.28611, "boundary_entropy": 0.449395, "bpb": 3.382524}}, "bpb": {"en": 2.90493, "tr": 3.17022, "fi": 3.38252}, "gini_cpc": 0.048863, "gini_cps": 0.066071, "bpic": 5.284277, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1010772773142303, "fi": 1.2452671827978181}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 13.232025146484375, "n_util_obs": 65, "util_count_spearman": -0.013636363636363636}, "t": 141.0} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.91712, "tr": 3.17813, "fi": 3.37572}, "bytes_per_chunk": {"en": 5.353, "tr": 5.194, "fi": 4.902}, "bpb_mean": 3.15699, "frozen_inventory": true, "t": 142.1} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.08143, "bpb": 3.00287, "chunk_loss": 0.030557, "F": 0.18927, "G": 0.27351, "chunks": 193.81, "bytes_per_chunk": 5.283, "boundary_entropy": 0.39687, "router_noise": 0.0, "grad_norm": 1.4327, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 0.278, "probe_sec": 0.0614, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05342, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11144, "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": -1.8519, "trunc_abs_frac": 3.65634, "mean_delta_full": -0.02246, "read_hit_rate": {"surface": 0.61394}, "mean_gate": {"surface": 0.20839}, "n_reads": 3788, "n_cand": 6170, "read_suppressed_frac": 0.1059, "dropout_reads_logged": 3788, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 13.249267578125, "n_util_obs": 65, "util_count_spearman": -0.01215034965034965}, "probe_overhead_sec_total": 21.592, "probe_overhead_seq_frac": 0.4978, "util_loss": 0.003808, "alpha": 0.03, "t": 143.1} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 158.0} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.98721, "bpb": 2.86694, "chunk_loss": 0.030796, "F": 0.20078, "G": 0.26924, "chunks": 205.59, "bytes_per_chunk": 4.981, "boundary_entropy": 0.36125, "router_noise": 0.0, "grad_norm": 1.4385, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 0.304, "probe_sec": 0.0368, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02829, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00794, "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": 0.88868, "trunc_abs_frac": 1.10757, "mean_delta_full": -0.20231, "read_hit_rate": {"surface": 0.65618}, "mean_gate": {"surface": 0.23064}, "n_reads": 4296, "n_cand": 6547, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 4296, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 14.778533935546875, "n_util_obs": 77, "util_count_spearman": -0.01682527998317472}, "probe_overhead_sec_total": 24.308, "probe_overhead_seq_frac": 0.4981, "util_loss": 0.005054, "alpha": 0.03, "t": 158.3} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 173.7} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.90074, "bpb": 2.74219, "chunk_loss": 0.031387, "F": 0.22305, "G": 0.27492, "chunks": 228.41, "bytes_per_chunk": 4.483, "boundary_entropy": 0.3243, "router_noise": 0.0, "grad_norm": 1.4361, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.314, "probe_sec": 0.1147, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10671, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12938, "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": 1.00122, "trunc_abs_frac": 0.90987, "mean_delta_full": 0.37443, "read_hit_rate": {"surface": 0.69878}, "mean_gate": {"surface": 0.25226}, "n_reads": 5085, "n_cand": 7277, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 5085, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 16.320571899414062, "n_util_obs": 80, "util_count_spearman": -0.02184716361931552}, "probe_overhead_sec_total": 28.14, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.009973, "alpha": 0.03, "t": 174.0} |
| {"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3625.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208297, "cps": 28.320312, "mean_p": 0.265868, "boundary_entropy": 0.324312, "bpb": 2.684575}, "tr": {"chunks": 3328.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.183837, "cps": 26.0, "mean_p": 0.260954, "boundary_entropy": 0.366682, "bpb": 3.023136}, "fi": {"chunks": 4055.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.215565, "cps": 31.679688, "mean_p": 0.267655, "boundary_entropy": 0.373051, "bpb": 3.219019}}, "bpb": {"en": 2.68458, "tr": 3.02314, "fi": 3.21902}, "gini_cpc": 0.034807, "gini_cps": 0.044029, "bpic": 5.153706, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.882569419814205, "fi": 1.0348921302466092}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 16.685806274414062, "n_util_obs": 80, "util_count_spearman": -0.04474917955930614}, "t": 177.9} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 190.1} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.89737, "bpb": 2.73733, "chunk_loss": 0.030947, "F": 0.20685, "G": 0.26843, "chunks": 211.81, "bytes_per_chunk": 4.834, "boundary_entropy": 0.32251, "router_noise": 0.0, "grad_norm": 1.5599, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 0.327, "probe_sec": 0.0341, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02755, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.30105, "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": 1.00726, "trunc_abs_frac": 0.63734, "mean_delta_full": 0.30115, "read_hit_rate": {"surface": 0.66884}, "mean_gate": {"surface": 0.26591}, "n_reads": 4512, "n_cand": 6746, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 4512, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 17.941162109375, "n_util_obs": 93, "util_count_spearman": -0.10799479273594549}, "probe_overhead_sec_total": 31.212, "probe_overhead_seq_frac": 0.4984, "util_loss": 0.009195, "alpha": 0.03, "t": 190.3} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 203.4} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3598.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206746, "cps": 28.109375, "mean_p": 0.257612, "boundary_entropy": 0.297234, "bpb": 2.571573}, "tr": {"chunks": 3371.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.186212, "cps": 26.335938, "mean_p": 0.258763, "boundary_entropy": 0.334698, "bpb": 2.922527}, "fi": {"chunks": 4288.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.227952, "cps": 33.5, "mean_p": 0.275247, "boundary_entropy": 0.339764, "bpb": 3.12653}}, "bpb": {"en": 2.57157, "tr": 2.92253, "fi": 3.12653}, "gini_cpc": 0.044815, "gini_cps": 0.054307, "bpic": 5.039709, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.900681333793222, "fi": 1.1025691950078644}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 19.156036376953125, "n_util_obs": 104, "util_count_spearman": -0.19604339094347026}, "t": 204.1} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.58395, "tr": 2.93018, "fi": 3.13017}, "bytes_per_chunk": {"en": 4.591, "tr": 5.954, "fi": 4.656}, "bpb_mean": 2.88143, "frozen_inventory": true, "t": 205.2} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.87919, "bpb": 2.71109, "chunk_loss": 0.031023, "F": 0.2132, "G": 0.26307, "chunks": 218.31, "bytes_per_chunk": 4.691, "boundary_entropy": 0.2985, "router_noise": 0.0, "grad_norm": 1.1374, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 0.317, "probe_sec": 0.0284, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02222, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12114, "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": 0.83083, "trunc_abs_frac": 2.07235, "mean_delta_full": -0.07796, "read_hit_rate": {"surface": 0.68047}, "mean_gate": {"surface": 0.26959}, "n_reads": 4732, "n_cand": 6954, "read_suppressed_frac": 0.1033, "dropout_reads_logged": 4732, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 19.196136474609375, "n_util_obs": 104, "util_count_spearman": -0.19009922116718234}, "probe_overhead_sec_total": 33.52, "probe_overhead_seq_frac": 0.4985, "util_loss": 0.010359, "alpha": 0.03, "t": 206.2} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 220.0} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.83093, "bpb": 2.64148, "chunk_loss": 0.030811, "F": 0.20349, "G": 0.2589, "chunks": 208.38, "bytes_per_chunk": 4.914, "boundary_entropy": 0.29251, "router_noise": 0.0, "grad_norm": 1.2409, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.281, "probe_sec": 0.101, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09338, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.26995, "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": -0.00533, "trunc_abs_frac": 0.00533, "mean_delta_full": 0.16028, "read_hit_rate": {"surface": 0.67119}, "mean_gate": {"surface": 0.26769}, "n_reads": 4454, "n_cand": 6636, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 4454, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 20.752792358398438, "n_util_obs": 109, "util_count_spearman": -0.14051524418496897}, "probe_overhead_sec_total": 36.154, "probe_overhead_seq_frac": 0.4987, "util_loss": 0.020738, "alpha": 0.03, "t": 220.2} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 233.4} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.97781, "bpb": 2.85338, "chunk_loss": 0.031279, "F": 0.21396, "G": 0.27068, "chunks": 219.09, "bytes_per_chunk": 4.674, "boundary_entropy": 0.29802, "router_noise": 0.0, "grad_norm": 1.1335, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 0.269, "probe_sec": 0.0456, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03952, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.08159, "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": 0.60577, "trunc_abs_frac": 2.01719, "mean_delta_full": -0.28219, "read_hit_rate": {"surface": 0.65411}, "mean_gate": {"surface": 0.27425}, "n_reads": 4565, "n_cand": 6979, "read_suppressed_frac": 0.0942, "dropout_reads_logged": 4565, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 22.607025146484375, "n_util_obs": 117, "util_count_spearman": -0.09759624750857948}, "probe_overhead_sec_total": 38.644, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.018109, "alpha": 0.03, "t": 233.7} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 247.0} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.77312, "bpb": 2.55807, "chunk_loss": 0.031127, "F": 0.21869, "G": 0.26409, "chunks": 223.94, "bytes_per_chunk": 4.573, "boundary_entropy": 0.28119, "router_noise": 0.0, "grad_norm": 1.2042, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.274, "probe_sec": 0.1111, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1019, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.317, "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": 1.34911, "trunc_abs_frac": 3.59361, "mean_delta_full": 0.26908, "read_hit_rate": {"surface": 0.68881}, "mean_gate": {"surface": 0.27086}, "n_reads": 4914, "n_cand": 7134, "read_suppressed_frac": 0.103, "dropout_reads_logged": 4914, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 24.376235961914062, "n_util_obs": 124, "util_count_spearman": -0.05171990558615264}, "probe_overhead_sec_total": 41.184, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.019441, "alpha": 0.03, "t": 247.4} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 260.4} |
| {"event": "probe", "step": 800, "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": 0.249314, "boundary_entropy": 0.268115, "bpb": 2.427222}, "tr": {"chunks": 3458.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.191018, "cps": 27.015625, "mean_p": 0.255127, "boundary_entropy": 0.312535, "bpb": 2.766895}, "fi": {"chunks": 4191.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.222795, "cps": 32.742188, "mean_p": 0.264454, "boundary_entropy": 0.301888, "bpb": 2.97797}}, "bpb": {"en": 2.42722, "tr": 2.76689, "fi": 2.97797}, "gini_cpc": 0.034103, "gini_cps": 0.043406, "bpic": 5.039261, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9211103784114443, "fi": 1.0743431411371098}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 26.225616455078125, "n_util_obs": 128, "util_count_spearman": -0.05352736095638282}, "t": 261.2} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.43774, "tr": 2.77429, "fi": 2.98027}, "bytes_per_chunk": {"en": 4.574, "tr": 5.566, "fi": 4.758}, "bpb_mean": 2.73077, "frozen_inventory": true, "t": 262.5} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.8599, "bpb": 2.68327, "chunk_loss": 0.031191, "F": 0.21866, "G": 0.26338, "chunks": 223.91, "bytes_per_chunk": 4.573, "boundary_entropy": 0.27454, "router_noise": 0.0, "grad_norm": 1.3663, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 0.325, "probe_sec": 0.0322, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02201, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13486, "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": 1.14554, "trunc_abs_frac": 0.78127, "mean_delta_full": 0.25403, "read_hit_rate": {"surface": 0.66662}, "mean_gate": {"surface": 0.26259}, "n_reads": 4755, "n_cand": 7133, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 4755, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 26.27764892578125, "n_util_obs": 128, "util_count_spearman": -0.051656061160959534}, "probe_overhead_sec_total": 43.985, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.018789, "alpha": 0.03, "t": 263.6} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 276.5} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.79877, "bpb": 2.59507, "chunk_loss": 0.030937, "F": 0.21219, "G": 0.25792, "chunks": 217.28, "bytes_per_chunk": 4.713, "boundary_entropy": 0.27092, "router_noise": 0.0, "grad_norm": 0.9999, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.264, "probe_sec": 0.0599, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05127, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16643, "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": 0.40664, "trunc_abs_frac": 0.40664, "mean_delta_full": -0.93652, "read_hit_rate": {"surface": 0.66002}, "mean_gate": {"surface": 0.26592}, "n_reads": 4568, "n_cand": 6921, "read_suppressed_frac": 0.0924, "dropout_reads_logged": 4568, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 28.0220947265625, "n_util_obs": 132, "util_count_spearman": -0.041476055000550544}, "probe_overhead_sec_total": 46.858, "probe_overhead_seq_frac": 0.499, "util_loss": 0.01929, "alpha": 0.03, "t": 276.8} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 289.6} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60444, "bpb": 2.31472, "chunk_loss": 0.031106, "F": 0.21945, "G": 0.25744, "chunks": 224.72, "bytes_per_chunk": 4.557, "boundary_entropy": 0.25102, "router_noise": 0.0, "grad_norm": 1.2563, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.26, "probe_sec": 0.0613, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05114, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.10596, "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": 0.59943, "trunc_abs_frac": 0.59943, "mean_delta_full": 0.5813, "read_hit_rate": {"surface": 0.72021}, "mean_gate": {"surface": 0.28127}, "n_reads": 5156, "n_cand": 7159, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 5156, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 29.721328735351562, "n_util_obs": 136, "util_count_spearman": -0.13974795989725342}, "probe_overhead_sec_total": 49.188, "probe_overhead_seq_frac": 0.499, "util_loss": 0.01982, "alpha": 0.03, "t": 289.8} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 303.3} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.67678, "bpb": 2.41908, "chunk_loss": 0.031053, "F": 0.22067, "G": 0.25042, "chunks": 225.97, "bytes_per_chunk": 4.532, "boundary_entropy": 0.24191, "router_noise": 0.0, "grad_norm": 1.2754, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.277, "probe_sec": 0.0256, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01872, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01675, "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": 0.77344, "trunc_abs_frac": 0.77344, "mean_delta_full": 0.40625, "read_hit_rate": {"surface": 0.70288}, "mean_gate": {"surface": 0.27557}, "n_reads": 5060, "n_cand": 7199, "read_suppressed_frac": 0.0945, "dropout_reads_logged": 5060, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 31.56396484375, "n_util_obs": 141, "util_count_spearman": -0.0885525921486365}, "probe_overhead_sec_total": 52.155, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.02022, "alpha": 0.03, "t": 303.7} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 316.4} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3652.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209849, "cps": 28.53125, "mean_p": 0.240358, "boundary_entropy": 0.239302, "bpb": 2.321958}, "tr": {"chunks": 3611.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19947, "cps": 28.210938, "mean_p": 0.245223, "boundary_entropy": 0.290396, "bpb": 2.637957}, "fi": {"chunks": 4255.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.226197, "cps": 33.242188, "mean_p": 0.260192, "boundary_entropy": 0.272902, "bpb": 2.846832}}, "bpb": {"en": 2.32196, "tr": 2.63796, "fi": 2.84683}, "gini_cpc": 0.028038, "gini_cps": 0.037275, "bpic": 4.925508, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9505397615530796, "fi": 1.077906354808712}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 33.195556640625, "n_util_obs": 144, "util_count_spearman": -0.0762580907044597}, "t": 317.1} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.33154, "tr": 2.64522, "fi": 2.85369}, "bytes_per_chunk": {"en": 4.563, "tr": 5.463, "fi": 4.714}, "bpb_mean": 2.61015, "frozen_inventory": true, "t": 318.3} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.73552, "bpb": 2.50383, "chunk_loss": 0.030933, "F": 0.2153, "G": 0.25074, "chunks": 220.47, "bytes_per_chunk": 4.645, "boundary_entropy": 0.24859, "router_noise": 0.0, "grad_norm": 1.006, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 0.334, "probe_sec": 0.1046, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08763, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04527, "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": 0.70515, "trunc_abs_frac": 0.97592, "mean_delta_full": 0.30725, "read_hit_rate": {"surface": 0.66239}, "mean_gate": {"surface": 0.26915}, "n_reads": 4652, "n_cand": 7023, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 4652, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 33.2279052734375, "n_util_obs": 144, "util_count_spearman": -0.07580178442247408}, "probe_overhead_sec_total": 55.053, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.024884, "alpha": 0.03, "t": 320.4} |
| {"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3516.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202034, "cps": 27.46875, "mean_p": 0.23594, "boundary_entropy": 0.245324, "bpb": 2.310457}, "tr": {"chunks": 3591.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198365, "cps": 28.054688, "mean_p": 0.246363, "boundary_entropy": 0.292782, "bpb": 2.630793}, "fi": {"chunks": 4268.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.226889, "cps": 33.34375, "mean_p": 0.260818, "boundary_entropy": 0.271266, "bpb": 2.829863}}, "bpb": {"en": 2.31046, "tr": 2.63079, "fi": 2.82986}, "gini_cpc": 0.030314, "gini_cps": 0.044073, "bpic": 4.987429, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9818386125355121, "fi": 1.1230207508750145}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 33.88233947753906, "n_util_obs": 146, "util_count_spearman": -0.061170564015880675}, "t": 327.2} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 335.3} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63391, "bpb": 2.35723, "chunk_loss": 0.030971, "F": 0.21442, "G": 0.24576, "chunks": 219.56, "bytes_per_chunk": 4.664, "boundary_entropy": 0.23837, "router_noise": 0.0, "grad_norm": 1.1338, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.305, "probe_sec": 0.1346, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12144, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01646, "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": 6.88871, "trunc_abs_frac": 1.21211, "mean_delta_full": 0.01113, "read_hit_rate": {"surface": 0.69073}, "mean_gate": {"surface": 0.26705}, "n_reads": 4831, "n_cand": 6994, "read_suppressed_frac": 0.0956, "dropout_reads_logged": 4831, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 34.5003662109375, "n_util_obs": 146, "util_count_spearman": -0.06644950637650353}, "probe_overhead_sec_total": 58.439, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.024295, "alpha": 0.03, "t": 335.6} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 348.2} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.73765, "bpb": 2.50689, "chunk_loss": 0.0309, "F": 0.21109, "G": 0.24686, "chunks": 216.16, "bytes_per_chunk": 4.737, "boundary_entropy": 0.24129, "router_noise": 0.0, "grad_norm": 0.987, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.26, "probe_sec": 0.0651, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05698, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1732, "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": 1.73066, "trunc_abs_frac": 0.85526, "mean_delta_full": -0.05933, "read_hit_rate": {"surface": 0.6658}, "mean_gate": {"surface": 0.26734}, "n_reads": 4584, "n_cand": 6885, "read_suppressed_frac": 0.0951, "dropout_reads_logged": 4584, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 36.026702880859375, "n_util_obs": 152, "util_count_spearman": -0.04554094610146799}, "probe_overhead_sec_total": 61.467, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.02126, "alpha": 0.03, "t": 348.6} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 360.9} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63847, "bpb": 2.36381, "chunk_loss": 0.0307, "F": 0.20502, "G": 0.24334, "chunks": 209.94, "bytes_per_chunk": 4.878, "boundary_entropy": 0.22995, "router_noise": 0.0, "grad_norm": 0.8486, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.256, "probe_sec": 0.3123, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07308, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16552, "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": 0.16124, "trunc_abs_frac": 0.17491, "mean_delta_full": -0.43536, "read_hit_rate": {"surface": 0.656}, "mean_gate": {"surface": 0.26653}, "n_reads": 4386, "n_cand": 6686, "read_suppressed_frac": 0.0953, "dropout_reads_logged": 4386, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 37.41435241699219, "n_util_obs": 152, "util_count_spearman": -0.05296646368550906}, "probe_overhead_sec_total": 64.532, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.021724, "alpha": 0.03, "t": 361.4} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 375.0} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3426.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196863, "cps": 26.765625, "mean_p": 0.23193, "boundary_entropy": 0.222453, "bpb": 2.254014}, "tr": {"chunks": 3557.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196487, "cps": 27.789062, "mean_p": 0.240218, "boundary_entropy": 0.278123, "bpb": 2.552033}, "fi": {"chunks": 4349.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231195, "cps": 33.976562, "mean_p": 0.260055, "boundary_entropy": 0.250424, "bpb": 2.728483}}, "bpb": {"en": 2.25401, "tr": 2.55203, "fi": 2.72848}, "gini_cpc": 0.037049, "gini_cps": 0.054301, "bpic": 5.006354, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9980908525674208, "fi": 1.174395249416702}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 38.09742736816406, "n_util_obs": 148, "util_count_spearman": -0.03749767478880359}, "t": 375.8} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.26496, "tr": 2.56116, "fi": 2.73752}, "bytes_per_chunk": {"en": 4.754, "tr": 5.547, "fi": 4.536}, "bpb_mean": 2.52121, "frozen_inventory": true, "t": 377.0} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61549, "bpb": 2.33066, "chunk_loss": 0.03072, "F": 0.20715, "G": 0.24453, "chunks": 212.12, "bytes_per_chunk": 4.827, "boundary_entropy": 0.23203, "router_noise": 0.0, "grad_norm": 0.8816, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 0.337, "probe_sec": 0.0285, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02054, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05091, "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": 0.9281, "trunc_abs_frac": 1.64877, "mean_delta_full": 0.53412, "read_hit_rate": {"surface": 0.66533}, "mean_gate": {"surface": 0.28349}, "n_reads": 4495, "n_cand": 6756, "read_suppressed_frac": 0.105, "dropout_reads_logged": 4495, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 38.14251708984375, "n_util_obs": 148, "util_count_spearman": -0.038804420353781885}, "probe_overhead_sec_total": 67.432, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.021866, "alpha": 0.03, "t": 378.2} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 395.4} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63474, "bpb": 2.35844, "chunk_loss": 0.030459, "F": 0.19312, "G": 0.23649, "chunks": 197.75, "bytes_per_chunk": 5.178, "boundary_entropy": 0.23075, "router_noise": 0.0, "grad_norm": 0.8634, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.348, "probe_sec": 0.0228, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01744, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.1013, "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": 1.81853, "trunc_abs_frac": 3.3625, "mean_delta_full": 0.08016, "read_hit_rate": {"surface": 0.63787}, "mean_gate": {"surface": 0.25896}, "n_reads": 4016, "n_cand": 6296, "read_suppressed_frac": 0.0969, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 39.2579345703125, "n_util_obs": 146, "util_count_spearman": 0.04876755590562252}, "probe_overhead_sec_total": 70.329, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.023701, "alpha": 0.03, "t": 395.6} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 408.7} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.67031, "bpb": 2.40975, "chunk_loss": 0.0308, "F": 0.20926, "G": 0.24257, "chunks": 214.28, "bytes_per_chunk": 4.779, "boundary_entropy": 0.21794, "router_noise": 0.0, "grad_norm": 0.9012, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.267, "probe_sec": 0.083, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07441, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.28119, "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": 0.78478, "trunc_abs_frac": 0.86948, "mean_delta_full": 0.36151, "read_hit_rate": {"surface": 0.64938}, "mean_gate": {"surface": 0.26554}, "n_reads": 4432, "n_cand": 6825, "read_suppressed_frac": 0.1085, "dropout_reads_logged": 4432, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 40.70817565917969, "n_util_obs": 143, "util_count_spearman": 0.07790390360812896}, "probe_overhead_sec_total": 73.393, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.022322, "alpha": 0.03, "t": 409.0} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 424.4} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59277, "bpb": 2.29789, "chunk_loss": 0.030701, "F": 0.2067, "G": 0.23803, "chunks": 211.66, "bytes_per_chunk": 4.838, "boundary_entropy": 0.2144, "router_noise": 0.0, "grad_norm": 0.8646, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.313, "probe_sec": 0.0409, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0334, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06257, "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": 0.76128, "trunc_abs_frac": 0.83117, "mean_delta_full": -0.1915, "read_hit_rate": {"surface": 0.664}, "mean_gate": {"surface": 0.27076}, "n_reads": 4476, "n_cand": 6741, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 4476, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 42.10643005371094, "n_util_obs": 141, "util_count_spearman": 0.05078989981369814}, "probe_overhead_sec_total": 75.809, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.020293, "alpha": 0.03, "t": 424.7} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 447.9} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3381.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194277, "cps": 26.414062, "mean_p": 0.224349, "boundary_entropy": 0.201716, "bpb": 2.205884}, "tr": {"chunks": 3367.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.185991, "cps": 26.304688, "mean_p": 0.234606, "boundary_entropy": 0.268456, "bpb": 2.496275}, "fi": {"chunks": 4321.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229706, "cps": 33.757812, "mean_p": 0.252242, "boundary_entropy": 0.233281, "bpb": 2.630767}}, "bpb": {"en": 2.20588, "tr": 2.49627, "fi": 2.63077}, "gini_cpc": 0.047778, "gini_cps": 0.057458, "bpic": 5.125305, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9573517034855415, "fi": 1.1823643654289198}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 43.06074523925781, "n_util_obs": 143, "util_count_spearman": 0.057823631767293736}, "t": 448.7} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.21733, "tr": 2.50471, "fi": 2.63805}, "bytes_per_chunk": {"en": 4.877, "tr": 5.982, "fi": 4.635}, "bpb_mean": 2.45336, "frozen_inventory": true, "t": 450.1} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59637, "bpb": 2.30308, "chunk_loss": 0.030545, "F": 0.19647, "G": 0.23466, "chunks": 201.19, "bytes_per_chunk": 5.09, "boundary_entropy": 0.21623, "router_noise": 0.0, "grad_norm": 0.8861, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 0.579, "probe_sec": 0.1211, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06355, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17879, "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": 0.899, "trunc_abs_frac": 1.09533, "mean_delta_full": -0.10258, "read_hit_rate": {"surface": 0.64377}, "mean_gate": {"surface": 0.26203}, "n_reads": 4124, "n_cand": 6406, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 4124, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 43.09120178222656, "n_util_obs": 143, "util_count_spearman": 0.05644472897994025}, "probe_overhead_sec_total": 78.787, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.018287, "alpha": 0.03, "t": 453.6} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 470.1} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64766, "bpb": 2.37707, "chunk_loss": 0.030579, "F": 0.20395, "G": 0.23239, "chunks": 208.84, "bytes_per_chunk": 4.903, "boundary_entropy": 0.20662, "router_noise": 0.0, "grad_norm": 0.8613, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.337, "probe_sec": 0.054, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04265, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06632, "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": 8, "trunc_ratio_sum": 1.8734, "trunc_abs_frac": 1.08941, "mean_delta_full": -0.04541, "read_hit_rate": {"surface": 0.64652}, "mean_gate": {"surface": 0.25876}, "n_reads": 4300, "n_cand": 6651, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 4300, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 44.57928466796875, "n_util_obs": 146, "util_count_spearman": 0.03426296356993435}, "probe_overhead_sec_total": 81.417, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.017248, "alpha": 0.03, "t": 470.5} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 484.8} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.66332, "bpb": 2.39967, "chunk_loss": 0.030721, "F": 0.20679, "G": 0.2382, "chunks": 211.75, "bytes_per_chunk": 4.836, "boundary_entropy": 0.21112, "router_noise": 0.0, "grad_norm": 0.8728, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.293, "probe_sec": 0.1587, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14563, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12373, "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": -0.23075, "trunc_abs_frac": 2.40665, "mean_delta_full": 0.10632, "read_hit_rate": {"surface": 0.64294}, "mean_gate": {"surface": 0.25624}, "n_reads": 4336, "n_cand": 6744, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 4336, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 45.271575927734375, "n_util_obs": 144, "util_count_spearman": -0.015205786518159872}, "probe_overhead_sec_total": 84.226, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.017833, "alpha": 0.03, "t": 485.1} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 503.6} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59069, "bpb": 2.29488, "chunk_loss": 0.030605, "F": 0.20093, "G": 0.23107, "chunks": 205.75, "bytes_per_chunk": 4.977, "boundary_entropy": 0.19999, "router_noise": 0.0, "grad_norm": 0.8925, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.375, "probe_sec": 0.0627, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04981, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01001, "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": 1.12381, "trunc_abs_frac": 0.79236, "mean_delta_full": 0.34644, "read_hit_rate": {"surface": 0.64606}, "mean_gate": {"surface": 0.26173}, "n_reads": 4233, "n_cand": 6552, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 4233, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 46.807952880859375, "n_util_obs": 147, "util_count_spearman": -0.05994378499270867}, "probe_overhead_sec_total": 87.437, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.02409, "alpha": 0.03, "t": 503.9} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 518.6} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3276.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188243, "cps": 25.59375, "mean_p": 0.217597, "boundary_entropy": 0.183142, "bpb": 2.163054}, "tr": {"chunks": 3640.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.201072, "cps": 28.4375, "mean_p": 0.225539, "boundary_entropy": 0.2365, "bpb": 2.426111}, "fi": {"chunks": 4303.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228749, "cps": 33.617188, "mean_p": 0.250404, "boundary_entropy": 0.213337, "bpb": 2.573508}}, "bpb": {"en": 2.16305, "tr": 2.42611, "fi": 2.57351}, "gini_cpc": 0.043691, "gini_cps": 0.061027, "bpic": 5.056779, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0681470842770076, "fi": 1.2151774164559237}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 47.33027648925781, "n_util_obs": 144, "util_count_spearman": -0.06024837231733784}, "t": 519.3} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.17424, "tr": 2.435, "fi": 2.58125}, "bytes_per_chunk": {"en": 4.981, "tr": 5.339, "fi": 4.643}, "bpb_mean": 2.39683, "frozen_inventory": true, "t": 520.4} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5115, "bpb": 2.18064, "chunk_loss": 0.030535, "F": 0.20197, "G": 0.23147, "chunks": 206.81, "bytes_per_chunk": 4.951, "boundary_entropy": 0.19153, "router_noise": 0.0, "grad_norm": 0.8105, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.35, "probe_sec": 0.0306, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02051, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06763, "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": 0.91926, "trunc_abs_frac": 0.91926, "mean_delta_full": 1.87329, "read_hit_rate": {"surface": 0.66338}, "mean_gate": {"surface": 0.25895}, "n_reads": 4369, "n_cand": 6586, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 4369, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 47.366241455078125, "n_util_obs": 144, "util_count_spearman": -0.061180773249738765}, "probe_overhead_sec_total": 90.281, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.024819, "alpha": 0.03, "t": 521.4} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 542.7} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57386, "bpb": 2.2706, "chunk_loss": 0.030482, "F": 0.19965, "G": 0.22792, "chunks": 204.44, "bytes_per_chunk": 5.009, "boundary_entropy": 0.19119, "router_noise": 0.0, "grad_norm": 0.8716, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.429, "probe_sec": 0.0262, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02109, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0051, "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": -0.74496, "trunc_abs_frac": 1.68726, "mean_delta_full": -0.10757, "read_hit_rate": {"surface": 0.64531}, "mean_gate": {"surface": 0.251}, "n_reads": 4201, "n_cand": 6510, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 4201, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 48.54042053222656, "n_util_obs": 144, "util_count_spearman": -0.08798328108672936}, "probe_overhead_sec_total": 93.019, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.018399, "alpha": 0.03, "t": 542.8} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 557.5} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53654, "bpb": 2.21676, "chunk_loss": 0.030443, "F": 0.19681, "G": 0.2254, "chunks": 201.53, "bytes_per_chunk": 5.081, "boundary_entropy": 0.18805, "router_noise": 0.0, "grad_norm": 0.7876, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.296, "probe_sec": 0.0261, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02002, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13757, "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": 0.94007, "trunc_abs_frac": 0.94007, "mean_delta_full": 0.25867, "read_hit_rate": {"surface": 0.6461}, "mean_gate": {"surface": 0.24018}, "n_reads": 4146, "n_cand": 6417, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 4146, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 49.85569763183594, "n_util_obs": 139, "util_count_spearman": -0.03126145046695563}, "probe_overhead_sec_total": 96.243, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.021667, "alpha": 0.03, "t": 557.7} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 571.8} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56801, "bpb": 2.26216, "chunk_loss": 0.030592, "F": 0.20511, "G": 0.2308, "chunks": 210.03, "bytes_per_chunk": 4.875, "boundary_entropy": 0.18591, "router_noise": 0.0, "grad_norm": 0.829, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.289, "probe_sec": 0.0327, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02507, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22593, "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": 0.66514, "trunc_abs_frac": 0.44081, "mean_delta_full": 0.70243, "read_hit_rate": {"surface": 0.64972}, "mean_gate": {"surface": 0.25314}, "n_reads": 4346, "n_cand": 6689, "read_suppressed_frac": 0.1045, "dropout_reads_logged": 4346, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 50.9254150390625, "n_util_obs": 138, "util_count_spearman": 0.005518403430223134}, "probe_overhead_sec_total": 98.912, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.0187, "alpha": 0.03, "t": 572.1} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 584.2} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3362.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.193185, "cps": 26.265625, "mean_p": 0.214675, "boundary_entropy": 0.16639, "bpb": 2.128512}, "tr": {"chunks": 3512.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194001, "cps": 27.4375, "mean_p": 0.219, "boundary_entropy": 0.226261, "bpb": 2.367999}, "fi": {"chunks": 4421.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235022, "cps": 34.539062, "mean_p": 0.24806, "boundary_entropy": 0.194875, "bpb": 2.526206}}, "bpb": {"en": 2.12851, "tr": 2.368, "fi": 2.52621}, "gini_cpc": 0.044826, "gini_cps": 0.062506, "bpic": 5.022753, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0042234693583478, "fi": 1.2165642288292404}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 52.237579345703125, "n_util_obs": 139, "util_count_spearman": 0.006380982170785111}, "t": 585.0} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.14151, "tr": 2.37751, "fi": 2.53139}, "bytes_per_chunk": {"en": 4.895, "tr": 5.606, "fi": 4.518}, "bpb_mean": 2.35014, "frozen_inventory": true, "t": 586.0} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54503, "bpb": 2.22901, "chunk_loss": 0.030471, "F": 0.19861, "G": 0.22346, "chunks": 203.38, "bytes_per_chunk": 5.035, "boundary_entropy": 0.17933, "router_noise": 0.0, "grad_norm": 0.8044, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.301, "probe_sec": 0.0686, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06199, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01658, "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": 2.9964, "trunc_abs_frac": 1.32466, "mean_delta_full": -0.03394, "read_hit_rate": {"surface": 0.64515}, "mean_gate": {"surface": 0.25132}, "n_reads": 4178, "n_cand": 6476, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 4178, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 52.279510498046875, "n_util_obs": 139, "util_count_spearman": 0.005518566513249028}, "probe_overhead_sec_total": 101.62, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.019846, "alpha": 0.03, "t": 587.2} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 599.9} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55184, "bpb": 2.23884, "chunk_loss": 0.030594, "F": 0.20322, "G": 0.22581, "chunks": 208.09, "bytes_per_chunk": 4.921, "boundary_entropy": 0.17112, "router_noise": 0.0, "grad_norm": 0.7811, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.26, "probe_sec": 0.0302, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02071, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05051, "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": 0.38757, "trunc_abs_frac": 0.38757, "mean_delta_full": 0.39575, "read_hit_rate": {"surface": 0.65339}, "mean_gate": {"surface": 0.24372}, "n_reads": 4330, "n_cand": 6627, "read_suppressed_frac": 0.0921, "dropout_reads_logged": 4330, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 53.73707580566406, "n_util_obs": 140, "util_count_spearman": 0.02782801125722526}, "probe_overhead_sec_total": 104.545, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.019298, "alpha": 0.03, "t": 600.1} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 611.5} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53318, "bpb": 2.21191, "chunk_loss": 0.030614, "F": 0.20224, "G": 0.22581, "chunks": 207.09, "bytes_per_chunk": 4.945, "boundary_entropy": 0.1761, "router_noise": 0.0, "grad_norm": 0.8578, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.232, "probe_sec": 0.0298, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02387, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0222, "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": 1.42752, "trunc_abs_frac": 4.22041, "mean_delta_full": 0.11593, "read_hit_rate": {"surface": 0.64746}, "mean_gate": {"surface": 0.24743}, "n_reads": 4270, "n_cand": 6595, "read_suppressed_frac": 0.0951, "dropout_reads_logged": 4270, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 55.26654052734375, "n_util_obs": 140, "util_count_spearman": 0.045177232074522754}, "probe_overhead_sec_total": 106.895, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.01904, "alpha": 0.03, "t": 611.7} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 623.8} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48355, "bpb": 2.1403, "chunk_loss": 0.030544, "F": 0.20203, "G": 0.22419, "chunks": 206.88, "bytes_per_chunk": 4.95, "boundary_entropy": 0.17353, "router_noise": 0.0, "grad_norm": 0.746, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.246, "probe_sec": 0.031, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0248, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09274, "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": 0.22685, "trunc_abs_frac": 0.22685, "mean_delta_full": 0.76733, "read_hit_rate": {"surface": 0.6741}, "mean_gate": {"surface": 0.25537}, "n_reads": 4441, "n_cand": 6588, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 4441, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 56.5186767578125, "n_util_obs": 140, "util_count_spearman": -0.048483523211825676}, "probe_overhead_sec_total": 109.98, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017518, "alpha": 0.03, "t": 624.0} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 636.3} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3361.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.193128, "cps": 26.257812, "mean_p": 0.215969, "boundary_entropy": 0.160665, "bpb": 2.104037}, "tr": {"chunks": 3451.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.190631, "cps": 26.960938, "mean_p": 0.216956, "boundary_entropy": 0.216656, "bpb": 2.319371}, "fi": {"chunks": 4407.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234278, "cps": 34.429688, "mean_p": 0.248156, "boundary_entropy": 0.187392, "bpb": 2.477544}}, "bpb": {"en": 2.10404, "tr": 2.31937, "fi": 2.47754}, "gini_cpc": 0.047081, "gini_cps": 0.062156, "bpic": 5.056779, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9870746887997492, "fi": 1.213072548298329}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 57.98323059082031, "n_util_obs": 141, "util_count_spearman": -0.07974656449334304}, "t": 636.8} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.11492, "tr": 2.3301, "fi": 2.48792}, "bytes_per_chunk": {"en": 4.935, "tr": 5.833, "fi": 4.553}, "bpb_mean": 2.31098, "frozen_inventory": true, "t": 638.0} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64222, "bpb": 2.36923, "chunk_loss": 0.030776, "F": 0.2103, "G": 0.23495, "chunks": 215.34, "bytes_per_chunk": 4.755, "boundary_entropy": 0.1762, "router_noise": 0.0, "grad_norm": 0.8228, "lr": 0.000245, "router_trainable": true, "sec_per_step": 0.304, "probe_sec": 0.0277, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02195, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01238, "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": 0.40647, "trunc_abs_frac": 1.74109, "mean_delta_full": 0.35783, "read_hit_rate": {"surface": 0.64062}, "mean_gate": {"surface": 0.23888}, "n_reads": 4394, "n_cand": 6859, "read_suppressed_frac": 0.099, "dropout_reads_logged": 4394, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 58.009063720703125, "n_util_obs": 141, "util_count_spearman": -0.0798921186694636}, "probe_overhead_sec_total": 112.626, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.019137, "alpha": 0.03, "t": 639.2} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3351.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.192553, "cps": 26.179688, "mean_p": 0.212084, "boundary_entropy": 0.155003, "bpb": 2.090527}, "tr": {"chunks": 3503.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193504, "cps": 27.367188, "mean_p": 0.212643, "boundary_entropy": 0.211046, "bpb": 2.311049}, "fi": {"chunks": 4361.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231832, "cps": 34.070312, "mean_p": 0.243519, "boundary_entropy": 0.184251, "bpb": 2.457992}}, "bpb": {"en": 2.09053, "tr": 2.31105, "fi": 2.45799}, "gini_cpc": 0.04238, "gini_cps": 0.060039, "bpic": 5.058582, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0049380222620476, "fi": 1.2039928160664486}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 59.29449462890625, "n_util_obs": 144, "util_count_spearman": -0.06710875331564986}, "t": 652.8} |
| {"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 654.7} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48091, "bpb": 2.13649, "chunk_loss": 0.030415, "F": 0.19205, "G": 0.21521, "chunks": 196.66, "bytes_per_chunk": 5.207, "boundary_entropy": 0.1662, "router_noise": 0.0, "grad_norm": 0.8067, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 0.316, "probe_sec": 0.0236, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01811, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01283, "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": 0.09846, "trunc_abs_frac": 0.65546, "mean_delta_full": 0.08699, "read_hit_rate": {"surface": 0.64015}, "mean_gate": {"surface": 0.24474}, "n_reads": 4008, "n_cand": 6261, "read_suppressed_frac": 0.0941, "dropout_reads_logged": 4008, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 59.41160583496094, "n_util_obs": 144, "util_count_spearman": -0.0670484687726067}, "probe_overhead_sec_total": 115.054, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017714, "alpha": 0.03, "t": 655.0} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 667.2} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54189, "bpb": 2.22447, "chunk_loss": 0.030441, "F": 0.19873, "G": 0.22037, "chunks": 203.5, "bytes_per_chunk": 5.032, "boundary_entropy": 0.16605, "router_noise": 0.0, "grad_norm": 0.7621, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 0.247, "probe_sec": 0.0249, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01875, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15654, "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": 0.86443, "trunc_abs_frac": 4.58718, "mean_delta_full": 0.50009, "read_hit_rate": {"surface": 0.64151}, "mean_gate": {"surface": 0.23831}, "n_reads": 4157, "n_cand": 6480, "read_suppressed_frac": 0.0969, "dropout_reads_logged": 4157, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 60.772705078125, "n_util_obs": 149, "util_count_spearman": -0.043859967349900233}, "probe_overhead_sec_total": 117.522, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017417, "alpha": 0.03, "t": 667.4} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 679.6} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49589, "bpb": 2.15811, "chunk_loss": 0.030564, "F": 0.20505, "G": 0.22537, "chunks": 209.97, "bytes_per_chunk": 4.877, "boundary_entropy": 0.16482, "router_noise": 0.0, "grad_norm": 0.6972, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.249, "probe_sec": 0.0387, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03343, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1462, "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": -2.01933, "trunc_abs_frac": 1.25463, "mean_delta_full": -0.10422, "read_hit_rate": {"surface": 0.66607}, "mean_gate": {"surface": 0.24606}, "n_reads": 4454, "n_cand": 6687, "read_suppressed_frac": 0.1075, "dropout_reads_logged": 4454, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 62.1319580078125, "n_util_obs": 149, "util_count_spearman": -0.033488149363472296}, "probe_overhead_sec_total": 119.634, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.018757, "alpha": 0.03, "t": 679.8} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 691.7} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3258.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187209, "cps": 25.453125, "mean_p": 0.208873, "boundary_entropy": 0.152213, "bpb": 2.074065}, "tr": {"chunks": 3616.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.199746, "cps": 28.25, "mean_p": 0.21294, "boundary_entropy": 0.197264, "bpb": 2.283389}, "fi": {"chunks": 4277.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.227367, "cps": 33.414062, "mean_p": 0.24096, "boundary_entropy": 0.177203, "bpb": 2.431684}}, "bpb": {"en": 2.07407, "tr": 2.28339, "fi": 2.43168}, "gini_cpc": 0.04358, "gini_cps": 0.060921, "bpic": 5.087615, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.066966811255707, "fi": 1.2145080760218958}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 63.26771545410156, "n_util_obs": 144, "util_count_spearman": -0.045414355759183345}, "t": 692.4} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.08542, "tr": 2.29182, "fi": 2.43973}, "bytes_per_chunk": {"en": 4.967, "tr": 5.278, "fi": 4.655}, "bpb_mean": 2.27232, "frozen_inventory": true, "t": 693.6} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.4754, "bpb": 2.12855, "chunk_loss": 0.030422, "F": 0.19934, "G": 0.21969, "chunks": 204.12, "bytes_per_chunk": 5.017, "boundary_entropy": 0.15867, "router_noise": 0.0, "grad_norm": 0.6811, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 0.298, "probe_sec": 0.0667, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06125, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05426, "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": 0.04594, "trunc_abs_frac": 1.23134, "mean_delta_full": -0.16772, "read_hit_rate": {"surface": 0.65908}, "mean_gate": {"surface": 0.25164}, "n_reads": 4284, "n_cand": 6500, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 4284, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 63.308013916015625, "n_util_obs": 144, "util_count_spearman": -0.044144361385740696}, "probe_overhead_sec_total": 122.827, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.019513, "alpha": 0.03, "t": 694.7} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 707.1} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49777, "bpb": 2.16083, "chunk_loss": 0.030494, "F": 0.20163, "G": 0.22052, "chunks": 206.47, "bytes_per_chunk": 4.96, "boundary_entropy": 0.15545, "router_noise": 0.0, "grad_norm": 0.7577, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.251, "probe_sec": 0.0275, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02106, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.15472, "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": 0.76513, "trunc_abs_frac": 1.1131, "mean_delta_full": -0.07515, "read_hit_rate": {"surface": 0.64669}, "mean_gate": {"surface": 0.24064}, "n_reads": 4252, "n_cand": 6575, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 4252, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 64.00083923339844, "n_util_obs": 144, "util_count_spearman": -0.03631942769873804}, "probe_overhead_sec_total": 126.071, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.018851, "alpha": 0.03, "t": 707.3} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 719.8} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51177, "bpb": 2.18102, "chunk_loss": 0.030451, "F": 0.20157, "G": 0.21847, "chunks": 206.41, "bytes_per_chunk": 4.961, "boundary_entropy": 0.14888, "router_noise": 0.0, "grad_norm": 0.7354, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.0588, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05129, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01201, "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": 0.27059, "trunc_abs_frac": 0.94535, "mean_delta_full": 0.5488, "read_hit_rate": {"surface": 0.64461}, "mean_gate": {"surface": 0.23844}, "n_reads": 4237, "n_cand": 6573, "read_suppressed_frac": 0.0923, "dropout_reads_logged": 4237, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 65.51689147949219, "n_util_obs": 141, "util_count_spearman": -0.021019735433879874}, "probe_overhead_sec_total": 128.739, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.018969, "alpha": 0.03, "t": 720.0} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 732.1} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56102, "bpb": 2.25208, "chunk_loss": 0.030521, "F": 0.20599, "G": 0.22229, "chunks": 210.94, "bytes_per_chunk": 4.855, "boundary_entropy": 0.15148, "router_noise": 0.0, "grad_norm": 0.7272, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.246, "probe_sec": 0.0279, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02261, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13358, "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": 0.27505, "trunc_abs_frac": 0.29085, "mean_delta_full": -0.19389, "read_hit_rate": {"surface": 0.64483}, "mean_gate": {"surface": 0.23641}, "n_reads": 4332, "n_cand": 6718, "read_suppressed_frac": 0.1043, "dropout_reads_logged": 4332, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 66.70097351074219, "n_util_obs": 139, "util_count_spearman": -0.06783591760132267}, "probe_overhead_sec_total": 131.148, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.021093, "alpha": 0.03, "t": 732.3} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 745.0} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3329.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.191289, "cps": 26.007812, "mean_p": 0.205788, "boundary_entropy": 0.136198, "bpb": 2.055112}, "tr": {"chunks": 3516.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194222, "cps": 27.46875, "mean_p": 0.206188, "boundary_entropy": 0.189603, "bpb": 2.255381}, "fi": {"chunks": 4331.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230238, "cps": 33.835938, "mean_p": 0.239063, "boundary_entropy": 0.164812, "bpb": 2.389514}}, "bpb": {"en": 2.05511, "tr": 2.25538, "fi": 2.38951}, "gini_cpc": 0.04217, "gini_cps": 0.059771, "bpic": 5.076235, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0153333233662247, "fi": 1.2036123224079454}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 67.62295532226562, "n_util_obs": 135, "util_count_spearman": -0.02972880694566384}, "t": 745.7} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.06738, "tr": 2.26326, "fi": 2.39868}, "bytes_per_chunk": {"en": 4.932, "tr": 5.613, "fi": 4.573}, "bpb_mean": 2.24311, "frozen_inventory": true, "t": 746.7} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 747.8} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49973, "bpb": 2.16365, "chunk_loss": 0.030542, "F": 0.20529, "G": 0.22173, "chunks": 210.22, "bytes_per_chunk": 4.871, "boundary_entropy": 0.14752, "router_noise": 0.0, "grad_norm": 0.7486, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.313, "probe_sec": 0.034, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02719, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1312, "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": 0.95455, "trunc_abs_frac": 6.65133, "mean_delta_full": 0.16113, "read_hit_rate": {"surface": 0.65228}, "mean_gate": {"surface": 0.23013}, "n_reads": 4367, "n_cand": 6695, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 4367, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 67.6514892578125, "n_util_obs": 135, "util_count_spearman": -0.03011901277924105}, "probe_overhead_sec_total": 133.843, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.019029, "alpha": 0.03, "t": 748.0} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 760.5} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46648, "bpb": 2.11568, "chunk_loss": 0.030426, "F": 0.19836, "G": 0.21799, "chunks": 203.12, "bytes_per_chunk": 5.041, "boundary_entropy": 0.15336, "router_noise": 0.0, "grad_norm": 0.7404, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.0285, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0224, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23516, "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": 2.06175, "trunc_abs_frac": 49.47537, "mean_delta_full": 0.17273, "read_hit_rate": {"surface": 0.64904}, "mean_gate": {"surface": 0.2326}, "n_reads": 4198, "n_cand": 6468, "read_suppressed_frac": 0.1048, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 69.19511413574219, "n_util_obs": 138, "util_count_spearman": -0.051441989915293855}, "probe_overhead_sec_total": 136.873, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.018784, "alpha": 0.03, "t": 760.8} |
| {"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 773.5} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43557, "bpb": 2.07108, "chunk_loss": 0.030551, "F": 0.20734, "G": 0.22233, "chunks": 212.31, "bytes_per_chunk": 4.823, "boundary_entropy": 0.14085, "router_noise": 0.0, "grad_norm": 0.7139, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.259, "probe_sec": 0.0466, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04114, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.40011, "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": 1.62474, "trunc_abs_frac": 12.9959, "mean_delta_full": 0.18741, "read_hit_rate": {"surface": 0.6711}, "mean_gate": {"surface": 0.23316}, "n_reads": 4538, "n_cand": 6762, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 4538, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 70.31878662109375, "n_util_obs": 137, "util_count_spearman": -0.06865234192693262}, "probe_overhead_sec_total": 139.185, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018274, "alpha": 0.03, "t": 773.7} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 785.7} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46647, "bpb": 2.11567, "chunk_loss": 0.030456, "F": 0.20081, "G": 0.21591, "chunks": 205.62, "bytes_per_chunk": 4.98, "boundary_entropy": 0.14029, "router_noise": 0.0, "grad_norm": 0.7486, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.243, "probe_sec": 0.0312, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02487, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1545, "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": 0.5836, "trunc_abs_frac": 0.5424, "mean_delta_full": 0.37503, "read_hit_rate": {"surface": 0.65822}, "mean_gate": {"surface": 0.22909}, "n_reads": 4310, "n_cand": 6548, "read_suppressed_frac": 0.1058, "dropout_reads_logged": 4310, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 71.64910888671875, "n_util_obs": 141, "util_count_spearman": -0.021441437391849894}, "probe_overhead_sec_total": 141.748, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016624, "alpha": 0.03, "t": 785.9} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 797.9} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3359.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.193013, "cps": 26.242188, "mean_p": 0.205815, "boundary_entropy": 0.127556, "bpb": 2.032145}, "tr": {"chunks": 3547.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195934, "cps": 27.710938, "mean_p": 0.204485, "boundary_entropy": 0.176374, "bpb": 2.239669}, "fi": {"chunks": 4355.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231513, "cps": 34.023438, "mean_p": 0.23963, "boundary_entropy": 0.155664, "bpb": 2.369502}}, "bpb": {"en": 2.03214, "tr": 2.23967, "fi": 2.3695}, "gini_cpc": 0.041368, "gini_cps": 0.058965, "bpic": 5.037918, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.015137224512501, "fi": 1.1994727673623533}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 73.02827453613281, "n_util_obs": 138, "util_count_spearman": -0.0125505435068004}, "t": 798.6} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.04536, "tr": 2.25031, "fi": 2.37634}, "bytes_per_chunk": {"en": 4.905, "tr": 5.687, "fi": 4.581}, "bpb_mean": 2.224, "frozen_inventory": true, "t": 799.7} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41587, "bpb": 2.04267, "chunk_loss": 0.030387, "F": 0.20004, "G": 0.21386, "chunks": 204.84, "bytes_per_chunk": 4.999, "boundary_entropy": 0.13363, "router_noise": 0.0, "grad_norm": 0.7395, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 0.3, "probe_sec": 0.0315, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02585, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00941, "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": 0.81448, "trunc_abs_frac": 1.04919, "mean_delta_full": -0.16943, "read_hit_rate": {"surface": 0.65629}, "mean_gate": {"surface": 0.23323}, "n_reads": 4281, "n_cand": 6523, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 4281, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 73.06935119628906, "n_util_obs": 138, "util_count_spearman": -0.012534575679506601}, "probe_overhead_sec_total": 144.281, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016024, "alpha": 0.03, "t": 800.9} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 813.5} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46057, "bpb": 2.10716, "chunk_loss": 0.030349, "F": 0.19583, "G": 0.21104, "chunks": 200.53, "bytes_per_chunk": 5.106, "boundary_entropy": 0.13858, "router_noise": 0.0, "grad_norm": 0.678, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 0.258, "probe_sec": 0.0943, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08246, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06129, "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": 0.61105, "trunc_abs_frac": 0.49797, "mean_delta_full": -0.22599, "read_hit_rate": {"surface": 0.64103}, "mean_gate": {"surface": 0.23394}, "n_reads": 4093, "n_cand": 6385, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 4093, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 74.42611694335938, "n_util_obs": 143, "util_count_spearman": 0.04444708916813583}, "probe_overhead_sec_total": 147.012, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01816, "alpha": 0.03, "t": 813.7} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 826.7} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44857, "bpb": 2.08984, "chunk_loss": 0.030478, "F": 0.20364, "G": 0.21694, "chunks": 208.53, "bytes_per_chunk": 4.911, "boundary_entropy": 0.13649, "router_noise": 0.0, "grad_norm": 0.6778, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.264, "probe_sec": 0.1071, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09824, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.10352, "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": 12.93056, "trunc_abs_frac": 5.58647, "mean_delta_full": -0.01025, "read_hit_rate": {"surface": 0.64463}, "mean_gate": {"surface": 0.22606}, "n_reads": 4281, "n_cand": 6641, "read_suppressed_frac": 0.0946, "dropout_reads_logged": 4281, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 75.46063232421875, "n_util_obs": 141, "util_count_spearman": 0.08893360160965795}, "probe_overhead_sec_total": 149.496, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.017882, "alpha": 0.03, "t": 826.9} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 839.4} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45519, "bpb": 2.09939, "chunk_loss": 0.030511, "F": 0.20407, "G": 0.21682, "chunks": 208.97, "bytes_per_chunk": 4.9, "boundary_entropy": 0.13259, "router_noise": 0.0, "grad_norm": 0.673, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.1311, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12597, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.27348, "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": 0.84368, "trunc_abs_frac": 1.69928, "mean_delta_full": 0.42126, "read_hit_rate": {"surface": 0.66236}, "mean_gate": {"surface": 0.22325}, "n_reads": 4408, "n_cand": 6655, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 4408, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 76.96115112304688, "n_util_obs": 143, "util_count_spearman": 0.09328523589086969}, "probe_overhead_sec_total": 152.106, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.017077, "alpha": 0.03, "t": 839.7} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 852.1} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3280.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188473, "cps": 25.625, "mean_p": 0.201144, "boundary_entropy": 0.117182, "bpb": 2.016283}, "tr": {"chunks": 3560.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196652, "cps": 27.8125, "mean_p": 0.202923, "boundary_entropy": 0.172352, "bpb": 2.192416}, "fi": {"chunks": 4360.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231779, "cps": 34.0625, "mean_p": 0.235385, "boundary_entropy": 0.14484, "bpb": 2.329963}}, "bpb": {"en": 2.01628, "tr": 2.19242, "fi": 2.32996}, "gini_cpc": 0.046799, "gini_cps": 0.064286, "bpic": 5.065357, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.043397334763272, "fi": 1.2297727977013968}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 78.21176147460938, "n_util_obs": 143, "util_count_spearman": 0.06871121835910568}, "t": 852.8} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.02842, "tr": 2.20196, "fi": 2.33664}, "bytes_per_chunk": {"en": 4.956, "tr": 5.628, "fi": 4.598}, "bpb_mean": 2.18901, "frozen_inventory": true, "t": 853.9} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40841, "bpb": 2.03191, "chunk_loss": 0.030382, "F": 0.2002, "G": 0.21171, "chunks": 205.0, "bytes_per_chunk": 4.995, "boundary_entropy": 0.12391, "router_noise": 0.0, "grad_norm": 0.6835, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 0.312, "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": 0.06787, "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": -0.18832, "trunc_abs_frac": 0.18832, "mean_delta_full": 0.14001, "read_hit_rate": {"surface": 0.66207}, "mean_gate": {"surface": 0.23151}, "n_reads": 4322, "n_cand": 6528, "read_suppressed_frac": 0.0942, "dropout_reads_logged": 4322, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 78.25422668457031, "n_util_obs": 143, "util_count_spearman": 0.06924061853639318}, "probe_overhead_sec_total": 154.943, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.029081, "alpha": 0.03, "t": 855.3} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 867.5} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45962, "bpb": 2.10579, "chunk_loss": 0.030568, "F": 0.20963, "G": 0.22094, "chunks": 214.66, "bytes_per_chunk": 4.77, "boundary_entropy": 0.12254, "router_noise": 0.0, "grad_norm": 0.7321, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.249, "probe_sec": 0.0345, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02747, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03431, "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": 7.26503, "trunc_abs_frac": 0.79917, "mean_delta_full": -0.02576, "read_hit_rate": {"surface": 0.66345}, "mean_gate": {"surface": 0.21556}, "n_reads": 4536, "n_cand": 6837, "read_suppressed_frac": 0.0915, "dropout_reads_logged": 4536, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 79.73579406738281, "n_util_obs": 145, "util_count_spearman": 0.09605967564163125}, "probe_overhead_sec_total": 157.322, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.017718, "alpha": 0.03, "t": 867.7} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 879.6} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38705, "bpb": 2.0011, "chunk_loss": 0.030342, "F": 0.19849, "G": 0.20948, "chunks": 203.25, "bytes_per_chunk": 5.038, "boundary_entropy": 0.12705, "router_noise": 0.0, "grad_norm": 0.6645, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.243, "probe_sec": 0.0238, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01862, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09409, "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": 0.49987, "trunc_abs_frac": 0.97231, "mean_delta_full": 0.21204, "read_hit_rate": {"surface": 0.66564}, "mean_gate": {"surface": 0.22459}, "n_reads": 4308, "n_cand": 6472, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 4308, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 81.34403991699219, "n_util_obs": 145, "util_count_spearman": 0.12682648401826485}, "probe_overhead_sec_total": 159.417, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016007, "alpha": 0.03, "t": 879.9} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 892.0} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54066, "bpb": 2.22271, "chunk_loss": 0.030636, "F": 0.20724, "G": 0.21995, "chunks": 212.22, "bytes_per_chunk": 4.825, "boundary_entropy": 0.12652, "router_noise": 0.0, "grad_norm": 0.7062, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.248, "probe_sec": 0.0739, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06235, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.26476, "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": -0.76328, "trunc_abs_frac": 0.65861, "mean_delta_full": -0.12866, "read_hit_rate": {"surface": 0.63856}, "mean_gate": {"surface": 0.22181}, "n_reads": 4316, "n_cand": 6759, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 4316, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 82.80735778808594, "n_util_obs": 150, "util_count_spearman": 0.11961242721898752}, "probe_overhead_sec_total": 161.499, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016692, "alpha": 0.03, "t": 892.2} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 909.1} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3238.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18606, "cps": 25.296875, "mean_p": 0.197921, "boundary_entropy": 0.117177, "bpb": 1.999746}, "tr": {"chunks": 3547.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195934, "cps": 27.710938, "mean_p": 0.202303, "boundary_entropy": 0.16387, "bpb": 2.173817}, "fi": {"chunks": 4380.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232842, "cps": 34.21875, "mean_p": 0.235666, "boundary_entropy": 0.139578, "bpb": 2.312045}}, "bpb": {"en": 1.99975, "tr": 2.17382, "fi": 2.31204}, "gini_cpc": 0.050726, "gini_cps": 0.068189, "bpic": 5.081236, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0530716297521592, "fi": 1.2514384743737887}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 84.08148193359375, "n_util_obs": 147, "util_count_spearman": 0.09651376285427166}, "t": 909.8} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.01376, "tr": 2.18444, "fi": 2.31619}, "bytes_per_chunk": {"en": 5.017, "tr": 5.692, "fi": 4.567}, "bpb_mean": 2.17147, "frozen_inventory": true, "t": 911.0} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48627, "bpb": 2.14424, "chunk_loss": 0.030384, "F": 0.19757, "G": 0.2083, "chunks": 202.31, "bytes_per_chunk": 5.061, "boundary_entropy": 0.12478, "router_noise": 0.0, "grad_norm": 0.6968, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.408, "probe_sec": 0.0672, "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": 0.12069, "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": 0.43617, "trunc_abs_frac": 0.43617, "mean_delta_full": -0.32227, "read_hit_rate": {"surface": 0.63148}, "mean_gate": {"surface": 0.21665}, "n_reads": 4068, "n_cand": 6442, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 4068, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 84.11186218261719, "n_util_obs": 147, "util_count_spearman": 0.0982931491737754}, "probe_overhead_sec_total": 164.23, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.02005, "alpha": 0.03, "t": 912.7} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 925.0} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4356, "bpb": 2.07113, "chunk_loss": 0.030456, "F": 0.20139, "G": 0.21247, "chunks": 206.22, "bytes_per_chunk": 4.966, "boundary_entropy": 0.11925, "router_noise": 0.0, "grad_norm": 0.6645, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0484, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04179, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02963, "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": 0.66882, "trunc_abs_frac": 0.53416, "mean_delta_full": 0.27838, "read_hit_rate": {"surface": 0.64413}, "mean_gate": {"surface": 0.21544}, "n_reads": 4230, "n_cand": 6567, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 4230, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 85.64016723632812, "n_util_obs": 152, "util_count_spearman": 0.10606619782803327}, "probe_overhead_sec_total": 167.135, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01649, "alpha": 0.03, "t": 925.3} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 937.7} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48599, "bpb": 2.14383, "chunk_loss": 0.030347, "F": 0.19662, "G": 0.20751, "chunks": 201.34, "bytes_per_chunk": 5.086, "boundary_entropy": 0.11749, "router_noise": 0.0, "grad_norm": 0.6854, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0281, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02196, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.10371, "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": 0.37015, "trunc_abs_frac": 0.30903, "mean_delta_full": -0.05188, "read_hit_rate": {"surface": 0.63609}, "mean_gate": {"surface": 0.22756}, "n_reads": 4078, "n_cand": 6411, "read_suppressed_frac": 0.0937, "dropout_reads_logged": 4078, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 86.82142639160156, "n_util_obs": 153, "util_count_spearman": 0.09421280473912053}, "probe_overhead_sec_total": 169.678, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.015673, "alpha": 0.03, "t": 938.0} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 952.2} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44632, "bpb": 2.0866, "chunk_loss": 0.030399, "F": 0.19989, "G": 0.21017, "chunks": 204.69, "bytes_per_chunk": 5.003, "boundary_entropy": 0.1157, "router_noise": 0.0, "grad_norm": 0.7014, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.289, "probe_sec": 0.0445, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03841, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09424, "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": 0.69011, "trunc_abs_frac": 1.03418, "mean_delta_full": -0.18145, "read_hit_rate": {"surface": 0.63946}, "mean_gate": {"surface": 0.21233}, "n_reads": 4168, "n_cand": 6518, "read_suppressed_frac": 0.0969, "dropout_reads_logged": 4168, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 88.14271545410156, "n_util_obs": 153, "util_count_spearman": 0.08552806274384035}, "probe_overhead_sec_total": 172.669, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.017479, "alpha": 0.03, "t": 952.4} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 969.1} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3340.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.191921, "cps": 26.09375, "mean_p": 0.201055, "boundary_entropy": 0.105162, "bpb": 1.980329}, "tr": {"chunks": 3558.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196542, "cps": 27.796875, "mean_p": 0.198727, "boundary_entropy": 0.148456, "bpb": 2.147093}, "fi": {"chunks": 4430.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235501, "cps": 34.609375, "mean_p": 0.237976, "boundary_entropy": 0.132332, "bpb": 2.281504}}, "bpb": {"en": 1.98033, "tr": 2.14709, "fi": 2.2815}, "gini_cpc": 0.046562, "gini_cps": 0.064148, "bpic": 5.008121, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0240780219376746, "fi": 1.2270704457864348}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 89.48426818847656, "n_util_obs": 157, "util_count_spearman": 0.03476826326136977}, "t": 969.7} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 1.99349, "tr": 2.15526, "fi": 2.28897}, "bytes_per_chunk": {"en": 4.875, "tr": 5.596, "fi": 4.482}, "bpb_mean": 2.14591, "frozen_inventory": true, "t": 970.8} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.432, "bpb": 2.06594, "chunk_loss": 0.030458, "F": 0.20535, "G": 0.21387, "chunks": 210.28, "bytes_per_chunk": 4.87, "boundary_entropy": 0.1097, "router_noise": 0.0, "grad_norm": 0.6186, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.426, "probe_sec": 0.106, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06088, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0718, "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": 20.25893, "trunc_abs_frac": 0.86985, "mean_delta_full": -0.00342, "read_hit_rate": {"surface": 0.66134}, "mean_gate": {"surface": 0.21899}, "n_reads": 4429, "n_cand": 6697, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 4429, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 89.51434326171875, "n_util_obs": 157, "util_count_spearman": 0.03338212219128127}, "probe_overhead_sec_total": 175.142, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014713, "alpha": 0.03, "t": 973.7} |
| {"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 998.9} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43068, "bpb": 2.06403, "chunk_loss": 0.030471, "F": 0.20306, "G": 0.21153, "chunks": 207.94, "bytes_per_chunk": 4.925, "boundary_entropy": 0.11571, "router_noise": 0.0, "grad_norm": 0.7217, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.509, "probe_sec": 0.0286, "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": 0.00709, "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": -5.40846, "trunc_abs_frac": 0.62449, "mean_delta_full": 0.01519, "read_hit_rate": {"surface": 0.65751}, "mean_gate": {"surface": 0.21752}, "n_reads": 4354, "n_cand": 6622, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 4354, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 90.92829895019531, "n_util_obs": 155, "util_count_spearman": 0.0016290174575570508}, "probe_overhead_sec_total": 178.088, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.015115, "alpha": 0.03, "t": 999.1} |
| {"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1014.3} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41304, "bpb": 2.03858, "chunk_loss": 0.03058, "F": 0.20932, "G": 0.21755, "chunks": 214.34, "bytes_per_chunk": 4.777, "boundary_entropy": 0.10835, "router_noise": 0.0, "grad_norm": 0.6849, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.308, "probe_sec": 0.0718, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06569, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19579, "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": -1.44269, "trunc_abs_frac": 1.43127, "mean_delta_full": -0.04706, "read_hit_rate": {"surface": 0.67233}, "mean_gate": {"surface": 0.20678}, "n_reads": 4590, "n_cand": 6827, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 4590, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 92.4805908203125, "n_util_obs": 153, "util_count_spearman": -0.03919561537167768}, "probe_overhead_sec_total": 180.599, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.013455, "alpha": 0.03, "t": 1014.6} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1028.0} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35452, "bpb": 1.95415, "chunk_loss": 0.030494, "F": 0.20682, "G": 0.21447, "chunks": 211.78, "bytes_per_chunk": 4.835, "boundary_entropy": 0.10743, "router_noise": 0.0, "grad_norm": 0.6888, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.273, "probe_sec": 0.0295, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02345, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06587, "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": 0, "read_hit_rate": {"surface": 0.68154}, "mean_gate": {"surface": 0.20235}, "n_reads": 4597, "n_cand": 6745, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 4597, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 93.81379699707031, "n_util_obs": 152, "util_count_spearman": -0.04359997983857058}, "probe_overhead_sec_total": 183.047, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.013282, "alpha": 0.03, "t": 1028.2} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1043.4} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3318.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190657, "cps": 25.921875, "mean_p": 0.199116, "boundary_entropy": 0.100704, "bpb": 1.971192}, "tr": {"chunks": 3574.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197426, "cps": 27.921875, "mean_p": 0.198559, "boundary_entropy": 0.148511, "bpb": 2.147735}, "fi": {"chunks": 4451.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.236617, "cps": 34.773438, "mean_p": 0.237248, "boundary_entropy": 0.128648, "bpb": 2.256307}}, "bpb": {"en": 1.97119, "tr": 2.14774, "fi": 2.25631}, "gini_cpc": 0.049048, "gini_cps": 0.06659, "bpic": 5.001499, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.03550389128554, "fi": 1.241061917621014}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 95.40769958496094, "n_util_obs": 153, "util_count_spearman": -0.06691031297356108}, "t": 1043.9} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.984, "tr": 2.15689, "fi": 2.2611}, "bytes_per_chunk": {"en": 4.894, "tr": 5.588, "fi": 4.491}, "bpb_mean": 2.134, "frozen_inventory": true, "t": 1044.9} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41862, "bpb": 2.04664, "chunk_loss": 0.030398, "F": 0.19949, "G": 0.20865, "chunks": 204.28, "bytes_per_chunk": 5.013, "boundary_entropy": 0.10966, "router_noise": 0.0, "grad_norm": 0.77, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 0.366, "probe_sec": 0.0366, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02785, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.07473, "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": 1.63104, "trunc_abs_frac": 1.47204, "mean_delta_full": 0.10697, "read_hit_rate": {"surface": 0.64904}, "mean_gate": {"surface": 0.21198}, "n_reads": 4222, "n_cand": 6505, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 4222, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 95.44277954101562, "n_util_obs": 153, "util_count_spearman": -0.06588329111548925}, "probe_overhead_sec_total": 185.714, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016286, "alpha": 0.03, "t": 1046.5} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1069.2} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38219, "bpb": 1.99408, "chunk_loss": 0.030399, "F": 0.20123, "G": 0.21011, "chunks": 206.06, "bytes_per_chunk": 4.969, "boundary_entropy": 0.11219, "router_noise": 0.0, "grad_norm": 0.7703, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.46, "probe_sec": 0.1166, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10631, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03219, "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": 0.76127, "trunc_abs_frac": 0.7347, "mean_delta_full": 0.3291, "read_hit_rate": {"surface": 0.65712}, "mean_gate": {"surface": 0.20763}, "n_reads": 4312, "n_cand": 6562, "read_suppressed_frac": 0.102, "dropout_reads_logged": 4312, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 96.62039184570312, "n_util_obs": 152, "util_count_spearman": -0.06343332034800675}, "probe_overhead_sec_total": 188.676, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01594, "alpha": 0.03, "t": 1069.5} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1083.7} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36173, "bpb": 1.96456, "chunk_loss": 0.030336, "F": 0.19711, "G": 0.20517, "chunks": 201.84, "bytes_per_chunk": 5.073, "boundary_entropy": 0.09789, "router_noise": 0.0, "grad_norm": 0.6572, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.287, "probe_sec": 0.0413, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03464, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09486, "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": -0.28103, "trunc_abs_frac": 0.42135, "mean_delta_full": 0.21191, "read_hit_rate": {"surface": 0.66734}, "mean_gate": {"surface": 0.21147}, "n_reads": 4289, "n_cand": 6427, "read_suppressed_frac": 0.097, "dropout_reads_logged": 4289, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 96.86039733886719, "n_util_obs": 153, "util_count_spearman": 0.010040148868287926}, "probe_overhead_sec_total": 191.392, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014584, "alpha": 0.03, "t": 1083.9} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1101.8} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40415, "bpb": 2.02576, "chunk_loss": 0.030435, "F": 0.20212, "G": 0.21107, "chunks": 206.97, "bytes_per_chunk": 4.948, "boundary_entropy": 0.10837, "router_noise": 0.0, "grad_norm": 0.666, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.362, "probe_sec": 0.0321, "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": 0.01151, "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": -0.36145, "trunc_abs_frac": 1.43639, "mean_delta_full": -0.11603, "read_hit_rate": {"surface": 0.65514}, "mean_gate": {"surface": 0.20677}, "n_reads": 4318, "n_cand": 6591, "read_suppressed_frac": 0.0908, "dropout_reads_logged": 4318, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 97.4351806640625, "n_util_obs": 150, "util_count_spearman": 0.010882261433841503}, "probe_overhead_sec_total": 193.843, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.012571, "alpha": 0.03, "t": 1102.0} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1114.8} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3315.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190484, "cps": 25.898438, "mean_p": 0.198572, "boundary_entropy": 0.097332, "bpb": 1.955155}, "tr": {"chunks": 3546.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195879, "cps": 27.703125, "mean_p": 0.195466, "boundary_entropy": 0.133403, "bpb": 2.111496}, "fi": {"chunks": 4384.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233055, "cps": 34.25, "mean_p": 0.234304, "boundary_entropy": 0.123797, "bpb": 2.234667}}, "bpb": {"en": 1.95516, "tr": 2.1115, "fi": 2.23467}, "gini_cpc": 0.045818, "gini_cps": 0.063376, "bpic": 5.045087, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0283211477410683, "fi": 1.2234866910210185}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 98.82356262207031, "n_util_obs": 148, "util_count_spearman": -0.027108467185169006}, "t": 1115.6} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.96909, "tr": 2.12011, "fi": 2.2394}, "bytes_per_chunk": {"en": 4.916, "tr": 5.678, "fi": 4.574}, "bpb_mean": 2.10954, "frozen_inventory": true, "t": 1116.6} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40133, "bpb": 2.02169, "chunk_loss": 0.03039, "F": 0.20078, "G": 0.20925, "chunks": 205.59, "bytes_per_chunk": 4.981, "boundary_entropy": 0.1039, "router_noise": 0.0, "grad_norm": 0.6434, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.32, "probe_sec": 0.035, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0216, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04043, "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": -1.18614, "trunc_abs_frac": 0.75051, "mean_delta_full": -0.05835, "read_hit_rate": {"surface": 0.64931}, "mean_gate": {"surface": 0.20571}, "n_reads": 4251, "n_cand": 6547, "read_suppressed_frac": 0.1044, "dropout_reads_logged": 4251, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 98.85865783691406, "n_util_obs": 148, "util_count_spearman": -0.026801215679451536}, "probe_overhead_sec_total": 196.025, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015563, "alpha": 0.03, "t": 1118.0} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1135.3} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44003, "bpb": 2.07753, "chunk_loss": 0.030488, "F": 0.20374, "G": 0.21186, "chunks": 208.62, "bytes_per_chunk": 4.908, "boundary_entropy": 0.10613, "router_noise": 0.0, "grad_norm": 0.714, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.351, "probe_sec": 0.0498, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03849, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0211, "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": 0.13089, "trunc_abs_frac": 0.5433, "mean_delta_full": -0.20815, "read_hit_rate": {"surface": 0.64916}, "mean_gate": {"surface": 0.19879}, "n_reads": 4313, "n_cand": 6644, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 4313, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 100.39369201660156, "n_util_obs": 148, "util_count_spearman": -0.033009176825092454}, "probe_overhead_sec_total": 199.258, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017317, "alpha": 0.03, "t": 1135.5} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1150.3} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38535, "bpb": 1.99863, "chunk_loss": 0.030458, "F": 0.2056, "G": 0.21319, "chunks": 210.53, "bytes_per_chunk": 4.864, "boundary_entropy": 0.10109, "router_noise": 0.0, "grad_norm": 0.6311, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.299, "probe_sec": 0.0274, "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": 0.11398, "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": -0.19181, "trunc_abs_frac": 2.30939, "mean_delta_full": -0.10762, "read_hit_rate": {"surface": 0.66637}, "mean_gate": {"surface": 0.207}, "n_reads": 4468, "n_cand": 6705, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 4468, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 101.59033203125, "n_util_obs": 148, "util_count_spearman": -0.011201723569892314}, "probe_overhead_sec_total": 201.953, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016306, "alpha": 0.03, "t": 1150.5} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1161.7} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37386, "bpb": 1.98206, "chunk_loss": 0.030373, "F": 0.20029, "G": 0.20814, "chunks": 205.09, "bytes_per_chunk": 4.993, "boundary_entropy": 0.10356, "router_noise": 0.0, "grad_norm": 0.6514, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.228, "probe_sec": 0.0316, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02098, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04686, "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": -0.1131, "trunc_abs_frac": 0.1131, "mean_delta_full": -0.43604, "read_hit_rate": {"surface": 0.65916}, "mean_gate": {"surface": 0.20446}, "n_reads": 4305, "n_cand": 6531, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 4305, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 102.71818542480469, "n_util_obs": 150, "util_count_spearman": -0.021723632161429397}, "probe_overhead_sec_total": 204.516, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015268, "alpha": 0.03, "t": 1161.9} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1171.7} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3321.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190829, "cps": 25.945312, "mean_p": 0.197852, "boundary_entropy": 0.092452, "bpb": 1.942613}, "tr": {"chunks": 3565.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196929, "cps": 27.851562, "mean_p": 0.194464, "boundary_entropy": 0.131685, "bpb": 2.101524}, "fi": {"chunks": 4402.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234012, "cps": 34.390625, "mean_p": 0.234151, "boundary_entropy": 0.115263, "bpb": 2.214067}}, "bpb": {"en": 1.94261, "tr": 2.10152, "fi": 2.21407}, "gini_cpc": 0.046301, "gini_cps": 0.063844, "bpic": 5.025868, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0319632399586807, "fi": 1.2262906004035674}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 104.06507873535156, "n_util_obs": 153, "util_count_spearman": 0.0671096189672041}, "t": 1172.3} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.95746, "tr": 2.10998, "fi": 2.22025}, "bytes_per_chunk": {"en": 4.905, "tr": 5.655, "fi": 4.531}, "bpb_mean": 2.0959, "frozen_inventory": true, "t": 1173.1} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4456, "bpb": 2.08557, "chunk_loss": 0.030459, "F": 0.20477, "G": 0.2121, "chunks": 209.69, "bytes_per_chunk": 4.883, "boundary_entropy": 0.10073, "router_noise": 0.0, "grad_norm": 0.6994, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.243, "probe_sec": 0.0256, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01859, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0512, "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": -0.42589, "trunc_abs_frac": 1.23754, "mean_delta_full": -0.17, "read_hit_rate": {"surface": 0.6442}, "mean_gate": {"surface": 0.19641}, "n_reads": 4302, "n_cand": 6678, "read_suppressed_frac": 0.1076, "dropout_reads_logged": 4302, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 104.09576416015625, "n_util_obs": 153, "util_count_spearman": 0.06688188868217453}, "probe_overhead_sec_total": 206.601, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.013555, "alpha": 0.03, "t": 1174.0} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1186.7} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54543, "bpb": 2.22959, "chunk_loss": 0.03063, "F": 0.20108, "G": 0.21022, "chunks": 205.91, "bytes_per_chunk": 4.973, "boundary_entropy": 0.10252, "router_noise": 0.0, "grad_norm": 0.705, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.257, "probe_sec": 0.0236, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01759, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22635, "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": 0.59372, "trunc_abs_frac": 0.77767, "mean_delta_full": 0.25704, "read_hit_rate": {"surface": 0.62208}, "mean_gate": {"surface": 0.20245}, "n_reads": 4079, "n_cand": 6557, "read_suppressed_frac": 0.099, "dropout_reads_logged": 4079, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 104.60420227050781, "n_util_obs": 149, "util_count_spearman": 0.09270814438599674}, "probe_overhead_sec_total": 209.226, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.013557, "alpha": 0.03, "t": 1186.8} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1199.2} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43374, "bpb": 2.06846, "chunk_loss": 0.030507, "F": 0.2049, "G": 0.21208, "chunks": 209.81, "bytes_per_chunk": 4.881, "boundary_entropy": 0.10187, "router_noise": 0.0, "grad_norm": 0.7534, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.251, "probe_sec": 0.0281, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02261, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23743, "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": 1.15773, "trunc_abs_frac": 1.10398, "mean_delta_full": 1.40656, "read_hit_rate": {"surface": 0.64397}, "mean_gate": {"surface": 0.18871}, "n_reads": 4303, "n_cand": 6682, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 4303, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 105.67982482910156, "n_util_obs": 150, "util_count_spearman": 0.11495799813325036}, "probe_overhead_sec_total": 211.678, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017129, "alpha": 0.03, "t": 1199.4} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1211.2} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48796, "bpb": 2.14667, "chunk_loss": 0.030386, "F": 0.19794, "G": 0.2052, "chunks": 202.69, "bytes_per_chunk": 5.052, "boundary_entropy": 0.09919, "router_noise": 0.0, "grad_norm": 0.7699, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.239, "probe_sec": 0.028, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0224, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12553, "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": 1.1894, "trunc_abs_frac": 0.96143, "mean_delta_full": 0.18506, "read_hit_rate": {"surface": 0.62705}, "mean_gate": {"surface": 0.19497}, "n_reads": 4047, "n_cand": 6454, "read_suppressed_frac": 0.1038, "dropout_reads_logged": 4047, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 107.14553833007812, "n_util_obs": 150, "util_count_spearman": 0.14443486377172318}, "probe_overhead_sec_total": 214.664, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018575, "alpha": 0.03, "t": 1211.4} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1223.5} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3305.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18991, "cps": 25.820312, "mean_p": 0.196796, "boundary_entropy": 0.090017, "bpb": 1.929835}, "tr": {"chunks": 3550.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.1961, "cps": 27.734375, "mean_p": 0.19371, "boundary_entropy": 0.125429, "bpb": 2.086015}, "fi": {"chunks": 4397.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233746, "cps": 34.351562, "mean_p": 0.234138, "boundary_entropy": 0.112031, "bpb": 2.194241}}, "bpb": {"en": 1.92983, "tr": 2.08601, "fi": 2.19424}, "gini_cpc": 0.047154, "gini_cps": 0.0647, "bpic": 5.041948, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0325960466762598, "fi": 1.2308276348108356}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 108.13334655761719, "n_util_obs": 151, "util_count_spearman": 0.12465318926455211}, "t": 1224.1} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.94494, "tr": 2.09486, "fi": 2.1993}, "bytes_per_chunk": {"en": 4.919, "tr": 5.627, "fi": 4.565}, "bpb_mean": 2.0797, "frozen_inventory": true, "t": 1225.2} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41038, "bpb": 2.03475, "chunk_loss": 0.030456, "F": 0.20197, "G": 0.20888, "chunks": 206.81, "bytes_per_chunk": 4.951, "boundary_entropy": 0.09699, "router_noise": 0.0, "grad_norm": 0.6717, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.3, "probe_sec": 0.0395, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02054, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03288, "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": -0.10103, "trunc_abs_frac": 1.71682, "mean_delta_full": 0.28577, "read_hit_rate": {"surface": 0.65578}, "mean_gate": {"surface": 0.19836}, "n_reads": 4319, "n_cand": 6586, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 4319, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 108.16534423828125, "n_util_obs": 151, "util_count_spearman": 0.12465318926455211}, "probe_overhead_sec_total": 217.048, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017847, "alpha": 0.03, "t": 1226.4} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1238.1} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4221, "bpb": 2.05166, "chunk_loss": 0.030462, "F": 0.20233, "G": 0.21159, "chunks": 207.19, "bytes_per_chunk": 4.942, "boundary_entropy": 0.10263, "router_noise": 0.0, "grad_norm": 0.7106, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.241, "probe_sec": 0.0703, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06311, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00147, "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": 1.30311, "trunc_abs_frac": 1.30311, "mean_delta_full": -0.23157, "read_hit_rate": {"surface": 0.6411}, "mean_gate": {"surface": 0.2013}, "n_reads": 4230, "n_cand": 6598, "read_suppressed_frac": 0.1111, "dropout_reads_logged": 4230, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 109.38363647460938, "n_util_obs": 151, "util_count_spearman": 0.0728999651446497}, "probe_overhead_sec_total": 219.243, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016671, "alpha": 0.03, "t": 1238.4} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3286.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188818, "cps": 25.671875, "mean_p": 0.197395, "boundary_entropy": 0.089438, "bpb": 1.925484}, "tr": {"chunks": 3585.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198033, "cps": 28.007812, "mean_p": 0.192864, "boundary_entropy": 0.125212, "bpb": 2.081885}, "fi": {"chunks": 4353.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231407, "cps": 34.007812, "mean_p": 0.233712, "boundary_entropy": 0.112692, "bpb": 2.185725}}, "bpb": {"en": 1.92548, "tr": 2.08189, "fi": 2.18573}, "gini_cpc": 0.045924, "gini_cps": 0.063376, "bpic": 5.054526, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.048806015648133, "fi": 1.2255565201503258}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 110.55694580078125, "n_util_obs": 152, "util_count_spearman": 0.09295101798126013}, "t": 1250.2} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1252.8} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35983, "bpb": 1.96183, "chunk_loss": 0.030432, "F": 0.20319, "G": 0.20946, "chunks": 208.06, "bytes_per_chunk": 4.922, "boundary_entropy": 0.09174, "router_noise": 0.0, "grad_norm": 0.6897, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.29, "probe_sec": 0.0275, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01912, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09181, "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": 1.52745, "trunc_abs_frac": 1.15284, "mean_delta_full": 0.09448, "read_hit_rate": {"surface": 0.66707}, "mean_gate": {"surface": 0.20039}, "n_reads": 4420, "n_cand": 6626, "read_suppressed_frac": 0.1063, "dropout_reads_logged": 4420, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 110.62440490722656, "n_util_obs": 151, "util_count_spearman": 0.08453475473578909}, "probe_overhead_sec_total": 222.256, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020882, "alpha": 0.03, "t": 1252.9} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1263.7} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41949, "bpb": 2.0479, "chunk_loss": 0.030574, "F": 0.20886, "G": 0.21513, "chunks": 213.88, "bytes_per_chunk": 4.788, "boundary_entropy": 0.09526, "router_noise": 0.0, "grad_norm": 0.7404, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.222, "probe_sec": 0.0319, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02316, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13263, "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": 1.1447, "trunc_abs_frac": 4.97336, "mean_delta_full": 0.28583, "read_hit_rate": {"surface": 0.65722}, "mean_gate": {"surface": 0.18912}, "n_reads": 4477, "n_cand": 6812, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 4477, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 111.43666076660156, "n_util_obs": 144, "util_count_spearman": 0.09172494172494172}, "probe_overhead_sec_total": 224.688, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017807, "alpha": 0.03, "t": 1264.0} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1275.5} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3291.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189105, "cps": 25.710938, "mean_p": 0.19617, "boundary_entropy": 0.087156, "bpb": 1.924919}, "tr": {"chunks": 3575.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197481, "cps": 27.929688, "mean_p": 0.193713, "boundary_entropy": 0.122872, "bpb": 2.071891}, "fi": {"chunks": 4376.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.23263, "cps": 34.1875, "mean_p": 0.232727, "boundary_entropy": 0.110308, "bpb": 2.182806}}, "bpb": {"en": 1.92492, "tr": 2.07189, "fi": 2.18281}, "gini_cpc": 0.04686, "gini_cps": 0.064342, "bpic": 5.046433, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.044291474828706, "fi": 1.2301601881829165}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 112.47434997558594, "n_util_obs": 143, "util_count_spearman": 0.1231778784595686}, "t": 1276.2} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.93827, "tr": 2.08054, "fi": 2.1902}, "bytes_per_chunk": {"en": 4.939, "tr": 5.59, "fi": 4.577}, "bpb_mean": 2.06967, "frozen_inventory": true, "t": 1277.2} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 1278.2} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35327, "bpb": 1.95236, "chunk_loss": 0.030301, "F": 0.19693, "G": 0.20358, "chunks": 201.66, "bytes_per_chunk": 5.078, "boundary_entropy": 0.09295, "router_noise": 0.0, "grad_norm": 0.6982, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.29, "probe_sec": 0.03, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02381, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03792, "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": 2.33131, "trunc_abs_frac": 0.5372, "mean_delta_full": 0.04514, "read_hit_rate": {"surface": 0.64928}, "mean_gate": {"surface": 0.20113}, "n_reads": 4169, "n_cand": 6421, "read_suppressed_frac": 0.1039, "dropout_reads_logged": 4169, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 112.50831604003906, "n_util_obs": 143, "util_count_spearman": 0.12309580091270232}, "probe_overhead_sec_total": 227.405, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016668, "alpha": 0.03, "t": 1278.5} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1290.5} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38241, "bpb": 1.9944, "chunk_loss": 0.030474, "F": 0.20572, "G": 0.21124, "chunks": 210.66, "bytes_per_chunk": 4.861, "boundary_entropy": 0.091, "router_noise": 0.0, "grad_norm": 0.6521, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.246, "probe_sec": 0.04, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02978, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10574, "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": 0.84882, "trunc_abs_frac": 1.28429, "mean_delta_full": 0.25888, "read_hit_rate": {"surface": 0.65882}, "mean_gate": {"surface": 0.19418}, "n_reads": 4420, "n_cand": 6709, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 4420, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 114.05320739746094, "n_util_obs": 146, "util_count_spearman": 0.1545911137246585}, "probe_overhead_sec_total": 229.739, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018777, "alpha": 0.03, "t": 1290.8} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1302.4} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38329, "bpb": 1.99566, "chunk_loss": 0.03031, "F": 0.19617, "G": 0.2035, "chunks": 200.88, "bytes_per_chunk": 5.098, "boundary_entropy": 0.09606, "router_noise": 0.0, "grad_norm": 0.7241, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.241, "probe_sec": 0.1132, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09481, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09422, "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": -2.2386, "trunc_abs_frac": 2.08531, "mean_delta_full": -0.03613, "read_hit_rate": {"surface": 0.64181}, "mean_gate": {"surface": 0.1911}, "n_reads": 4105, "n_cand": 6396, "read_suppressed_frac": 0.097, "dropout_reads_logged": 4105, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 115.23536682128906, "n_util_obs": 145, "util_count_spearman": 0.20183435679420564}, "probe_overhead_sec_total": 232.672, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018377, "alpha": 0.03, "t": 1302.8} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1314.2} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40591, "bpb": 2.02829, "chunk_loss": 0.030383, "F": 0.20135, "G": 0.20808, "chunks": 206.19, "bytes_per_chunk": 4.966, "boundary_entropy": 0.09005, "router_noise": 0.0, "grad_norm": 0.6591, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.231, "probe_sec": 0.0303, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02047, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01215, "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": 1.38979, "trunc_abs_frac": 1.07876, "mean_delta_full": 0.11884, "read_hit_rate": {"surface": 0.64545}, "mean_gate": {"surface": 0.18905}, "n_reads": 4238, "n_cand": 6566, "read_suppressed_frac": 0.0897, "dropout_reads_logged": 4238, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 116.79554748535156, "n_util_obs": 149, "util_count_spearman": 0.17653018005646665}, "probe_overhead_sec_total": 235.192, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015546, "alpha": 0.03, "t": 1314.4} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1325.9} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3309.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.19014, "cps": 25.851562, "mean_p": 0.196915, "boundary_entropy": 0.085883, "bpb": 1.911992}, "tr": {"chunks": 3563.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196818, "cps": 27.835938, "mean_p": 0.19235, "boundary_entropy": 0.115342, "bpb": 2.058012}, "fi": {"chunks": 4399.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233853, "cps": 34.367188, "mean_p": 0.234231, "boundary_entropy": 0.1088, "bpb": 2.164496}}, "bpb": {"en": 1.91199, "tr": 2.05801, "fi": 2.1645}, "gini_cpc": 0.046942, "gini_cps": 0.064472, "bpic": 5.033449, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.035124586023294, "fi": 1.2298989523741268}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 118.0699462890625, "n_util_obs": 149, "util_count_spearman": 0.23323417377108652}, "t": 1326.4} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.92629, "tr": 2.0658, "fi": 2.17138}, "bytes_per_chunk": {"en": 4.938, "tr": 5.616, "fi": 4.57}, "bpb_mean": 2.05449, "frozen_inventory": true, "t": 1327.4} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35838, "bpb": 1.95973, "chunk_loss": 0.030301, "F": 0.19821, "G": 0.20442, "chunks": 202.97, "bytes_per_chunk": 5.045, "boundary_entropy": 0.08796, "router_noise": 0.0, "grad_norm": 0.6212, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.285, "probe_sec": 0.082, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07636, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16973, "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": 0.56702, "trunc_abs_frac": 0.59196, "mean_delta_full": 0.6163, "read_hit_rate": {"surface": 0.64862}, "mean_gate": {"surface": 0.19126}, "n_reads": 4192, "n_cand": 6463, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 4192, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 118.09320068359375, "n_util_obs": 149, "util_count_spearman": 0.23315638777117134}, "probe_overhead_sec_total": 237.691, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014304, "alpha": 0.03, "t": 1328.6} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1339.5} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35318, "bpb": 1.95223, "chunk_loss": 0.03028, "F": 0.19385, "G": 0.20064, "chunks": 198.5, "bytes_per_chunk": 5.159, "boundary_entropy": 0.09093, "router_noise": 0.0, "grad_norm": 0.6901, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.221, "probe_sec": 0.0254, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01875, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15737, "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": 0.20205, "trunc_abs_frac": 1.38076, "mean_delta_full": -0.11179, "read_hit_rate": {"surface": 0.64615}, "mean_gate": {"surface": 0.19272}, "n_reads": 4083, "n_cand": 6319, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 4083, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 119.57337951660156, "n_util_obs": 152, "util_count_spearman": 0.17195989584401206}, "probe_overhead_sec_total": 240.212, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014528, "alpha": 0.03, "t": 1339.7} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1351.1} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33177, "bpb": 1.92134, "chunk_loss": 0.030368, "F": 0.20157, "G": 0.20785, "chunks": 206.41, "bytes_per_chunk": 4.961, "boundary_entropy": 0.08933, "router_noise": 0.0, "grad_norm": 0.6636, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.232, "probe_sec": 0.0273, "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": -0.01043, "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": 0.08932, "trunc_abs_frac": 0.66311, "mean_delta_full": -0.06287, "read_hit_rate": {"surface": 0.6513}, "mean_gate": {"surface": 0.18141}, "n_reads": 4281, "n_cand": 6573, "read_suppressed_frac": 0.0913, "dropout_reads_logged": 4281, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 120.62283325195312, "n_util_obs": 154, "util_count_spearman": 0.19018868544485243}, "probe_overhead_sec_total": 242.418, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015222, "alpha": 0.03, "t": 1351.3} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1362.9} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36077, "bpb": 1.96317, "chunk_loss": 0.030289, "F": 0.19562, "G": 0.20277, "chunks": 200.31, "bytes_per_chunk": 5.112, "boundary_entropy": 0.08868, "router_noise": 0.0, "grad_norm": 0.6769, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.237, "probe_sec": 0.0315, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02456, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21947, "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": 1.12252, "trunc_abs_frac": 0.7925, "mean_delta_full": 0.2607, "read_hit_rate": {"surface": 0.63515}, "mean_gate": {"surface": 0.18529}, "n_reads": 4051, "n_cand": 6378, "read_suppressed_frac": 0.1024, "dropout_reads_logged": 4051, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 121.8265380859375, "n_util_obs": 153, "util_count_spearman": 0.19624931312237814}, "probe_overhead_sec_total": 245.031, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.013827, "alpha": 0.03, "t": 1363.1} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1374.2} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3330.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.191346, "cps": 26.015625, "mean_p": 0.197293, "boundary_entropy": 0.082676, "bpb": 1.902561}, "tr": {"chunks": 3582.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197868, "cps": 27.984375, "mean_p": 0.193693, "boundary_entropy": 0.116014, "bpb": 2.048946}, "fi": {"chunks": 4416.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234756, "cps": 34.5, "mean_p": 0.234106, "boundary_entropy": 0.103861, "bpb": 2.154561}}, "bpb": {"en": 1.90256, "tr": 2.04895, "fi": 2.15456}, "gini_cpc": 0.04638, "gini_cps": 0.063912, "bpic": 5.008121, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0340818529406057, "fi": 1.2268658217517927}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 123.28262329101562, "n_util_obs": 160, "util_count_spearman": 0.1800714871674675}, "t": 1374.9} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.91649, "tr": 2.05747, "fi": 2.16046}, "bytes_per_chunk": {"en": 4.906, "tr": 5.614, "fi": 4.533}, "bpb_mean": 2.04481, "frozen_inventory": true, "t": 1375.9} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35896, "bpb": 1.96056, "chunk_loss": 0.030501, "F": 0.20804, "G": 0.2147, "chunks": 213.03, "bytes_per_chunk": 4.807, "boundary_entropy": 0.08929, "router_noise": 0.0, "grad_norm": 0.671, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.283, "probe_sec": 0.0526, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03917, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01933, "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": -0.95237, "trunc_abs_frac": 1.01354, "mean_delta_full": 0.0496, "read_hit_rate": {"surface": 0.66691}, "mean_gate": {"surface": 0.18813}, "n_reads": 4525, "n_cand": 6785, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 4525, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 123.32551574707031, "n_util_obs": 160, "util_count_spearman": 0.17992499707019805}, "probe_overhead_sec_total": 247.515, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010706, "alpha": 0.03, "t": 1377.3} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1388.7} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38758, "bpb": 2.00185, "chunk_loss": 0.030394, "F": 0.20084, "G": 0.20818, "chunks": 205.66, "bytes_per_chunk": 4.979, "boundary_entropy": 0.09214, "router_noise": 0.0, "grad_norm": 0.68, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.233, "probe_sec": 0.0585, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05116, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08445, "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": 3.76942, "trunc_abs_frac": 0.64715, "mean_delta_full": 0.0673, "read_hit_rate": {"surface": 0.63781}, "mean_gate": {"surface": 0.18829}, "n_reads": 4177, "n_cand": 6549, "read_suppressed_frac": 0.101, "dropout_reads_logged": 4177, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 124.73970031738281, "n_util_obs": 160, "util_count_spearman": 0.17636102217939673}, "probe_overhead_sec_total": 250.15, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010826, "alpha": 0.03, "t": 1388.9} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1400.7} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41763, "bpb": 2.04521, "chunk_loss": 0.030542, "F": 0.20782, "G": 0.21326, "chunks": 212.81, "bytes_per_chunk": 4.812, "boundary_entropy": 0.08694, "router_noise": 0.0, "grad_norm": 0.6539, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.242, "probe_sec": 0.0302, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02066, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20347, "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": 0.0789, "trunc_abs_frac": 0.78111, "mean_delta_full": -0.07349, "read_hit_rate": {"surface": 0.65403}, "mean_gate": {"surface": 0.18219}, "n_reads": 4433, "n_cand": 6778, "read_suppressed_frac": 0.0891, "dropout_reads_logged": 4433, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 126.032958984375, "n_util_obs": 157, "util_count_spearman": 0.1752336593504053}, "probe_overhead_sec_total": 252.576, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010095, "alpha": 0.03, "t": 1401.0} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1414.9} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.4027, "bpb": 2.02367, "chunk_loss": 0.030456, "F": 0.20609, "G": 0.21155, "chunks": 211.03, "bytes_per_chunk": 4.852, "boundary_entropy": 0.08903, "router_noise": 0.0, "grad_norm": 0.7164, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.281, "probe_sec": 0.0394, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03336, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00768, "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": 0.68376, "trunc_abs_frac": 0.59457, "mean_delta_full": 0.13867, "read_hit_rate": {"surface": 0.65362}, "mean_gate": {"surface": 0.18612}, "n_reads": 4393, "n_cand": 6721, "read_suppressed_frac": 0.1056, "dropout_reads_logged": 4393, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 126.69905090332031, "n_util_obs": 152, "util_count_spearman": 0.18932264435924248}, "probe_overhead_sec_total": 255.089, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014365, "alpha": 0.03, "t": 1415.1} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1429.5} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3303.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189795, "cps": 25.804688, "mean_p": 0.195899, "boundary_entropy": 0.081419, "bpb": 1.894863}, "tr": {"chunks": 3531.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195051, "cps": 27.585938, "mean_p": 0.190312, "boundary_entropy": 0.11247, "bpb": 2.036772}, "fi": {"chunks": 4402.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234012, "cps": 34.390625, "mean_p": 0.232761, "boundary_entropy": 0.100987, "bpb": 2.140706}}, "bpb": {"en": 1.89486, "tr": 2.03677, "fi": 2.14071}, "gini_cpc": 0.047633, "gini_cps": 0.065207, "bpic": 5.049128, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0276913772703307, "fi": 1.2329733829670746}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 127.70579528808594, "n_util_obs": 149, "util_count_spearman": 0.11589334300743696}, "t": 1430.1} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.9097, "tr": 2.04473, "fi": 2.14547}, "bytes_per_chunk": {"en": 4.937, "tr": 5.657, "fi": 4.557}, "bpb_mean": 2.0333, "frozen_inventory": true, "t": 1431.2} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37569, "bpb": 1.9847, "chunk_loss": 0.030455, "F": 0.20651, "G": 0.21149, "chunks": 211.47, "bytes_per_chunk": 4.842, "boundary_entropy": 0.08435, "router_noise": 0.0, "grad_norm": 0.6482, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.348, "probe_sec": 0.0229, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01781, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06581, "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": 0.45266, "trunc_abs_frac": 1.59077, "mean_delta_full": 0.06656, "read_hit_rate": {"surface": 0.65598}, "mean_gate": {"surface": 0.17908}, "n_reads": 4418, "n_cand": 6735, "read_suppressed_frac": 0.1055, "dropout_reads_logged": 4418, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 127.74649047851562, "n_util_obs": 149, "util_count_spearman": 0.11931264222083833}, "probe_overhead_sec_total": 257.339, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010311, "alpha": 0.03, "t": 1432.5} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1443.6} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35146, "bpb": 1.94974, "chunk_loss": 0.030387, "F": 0.19958, "G": 0.20643, "chunks": 204.38, "bytes_per_chunk": 5.01, "boundary_entropy": 0.09107, "router_noise": 0.0, "grad_norm": 0.6891, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.228, "probe_sec": 0.0556, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04796, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.07292, "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": 0.21711, "trunc_abs_frac": 0.22861, "mean_delta_full": -0.22223, "read_hit_rate": {"surface": 0.65012}, "mean_gate": {"surface": 0.18418}, "n_reads": 4231, "n_cand": 6508, "read_suppressed_frac": 0.1047, "dropout_reads_logged": 4231, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 129.1665496826172, "n_util_obs": 149, "util_count_spearman": 0.13091601668782876}, "probe_overhead_sec_total": 259.722, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011712, "alpha": 0.03, "t": 1443.9} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1455.1} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38008, "bpb": 1.99103, "chunk_loss": 0.030426, "F": 0.19815, "G": 0.20409, "chunks": 202.91, "bytes_per_chunk": 5.047, "boundary_entropy": 0.08918, "router_noise": 0.0, "grad_norm": 0.6714, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.229, "probe_sec": 0.0556, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04712, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03353, "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": -9.79648, "trunc_abs_frac": 2.89116, "mean_delta_full": -0.01353, "read_hit_rate": {"surface": 0.6403}, "mean_gate": {"surface": 0.17709}, "n_reads": 4137, "n_cand": 6461, "read_suppressed_frac": 0.095, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 130.4761962890625, "n_util_obs": 145, "util_count_spearman": 0.10183435679420563}, "probe_overhead_sec_total": 262.178, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011757, "alpha": 0.03, "t": 1455.3} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1475.7} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33611, "bpb": 1.9276, "chunk_loss": 0.030382, "F": 0.20041, "G": 0.20588, "chunks": 205.22, "bytes_per_chunk": 4.99, "boundary_entropy": 0.08816, "router_noise": 0.0, "grad_norm": 0.6583, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.411, "probe_sec": 0.0265, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01991, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14286, "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": 1.15046, "trunc_abs_frac": 1.40736, "mean_delta_full": 0.18774, "read_hit_rate": {"surface": 0.65815}, "mean_gate": {"surface": 0.18687}, "n_reads": 4301, "n_cand": 6535, "read_suppressed_frac": 0.0956, "dropout_reads_logged": 4301, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 131.7606201171875, "n_util_obs": 148, "util_count_spearman": 0.18150049789551229}, "probe_overhead_sec_total": 264.914, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011398, "alpha": 0.03, "t": 1475.9} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1497.2} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3287.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188875, "cps": 25.679688, "mean_p": 0.19494, "boundary_entropy": 0.081196, "bpb": 1.887973}, "tr": {"chunks": 3555.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196376, "cps": 27.773438, "mean_p": 0.190906, "boundary_entropy": 0.112574, "bpb": 2.028411}, "fi": {"chunks": 4377.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232683, "cps": 34.195312, "mean_p": 0.231761, "boundary_entropy": 0.10076, "bpb": 2.129551}}, "bpb": {"en": 1.88797, "tr": 2.02841, "fi": 2.12955}, "gini_cpc": 0.047262, "gini_cps": 0.064771, "bpic": 5.056779, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0397129893958885, "fi": 1.2319386460261836}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 132.8737030029297, "n_util_obs": 145, "util_count_spearman": 0.13036529680365297}, "t": 1497.7} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.9034, "tr": 2.03745, "fi": 2.13518}, "bytes_per_chunk": {"en": 4.958, "tr": 5.592, "fi": 4.579}, "bpb_mean": 2.02534, "frozen_inventory": true, "t": 1498.7} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32886, "bpb": 1.91714, "chunk_loss": 0.030373, "F": 0.2012, "G": 0.20694, "chunks": 206.03, "bytes_per_chunk": 4.97, "boundary_entropy": 0.08303, "router_noise": 0.0, "grad_norm": 0.62, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.485, "probe_sec": 0.0987, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08418, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09311, "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": 0.11532, "trunc_abs_frac": 0.59922, "mean_delta_full": 0.08699, "read_hit_rate": {"surface": 0.66621}, "mean_gate": {"surface": 0.17671}, "n_reads": 4371, "n_cand": 6561, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 4371, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 132.90582275390625, "n_util_obs": 145, "util_count_spearman": 0.13270744764604}, "probe_overhead_sec_total": 267.782, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.013803, "alpha": 0.03, "t": 1500.1} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1513.6} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34756, "bpb": 1.94411, "chunk_loss": 0.030434, "F": 0.20514, "G": 0.21032, "chunks": 210.06, "bytes_per_chunk": 4.875, "boundary_entropy": 0.08352, "router_noise": 0.0, "grad_norm": 0.6917, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.272, "probe_sec": 0.0266, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02068, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20113, "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": 1.42956, "trunc_abs_frac": 0.5901, "mean_delta_full": 0.15177, "read_hit_rate": {"surface": 0.64993}, "mean_gate": {"surface": 0.176}, "n_reads": 4348, "n_cand": 6690, "read_suppressed_frac": 0.0943, "dropout_reads_logged": 4348, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 134.3872528076172, "n_util_obs": 143, "util_count_spearman": 0.06461144489313503}, "probe_overhead_sec_total": 270.522, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011185, "alpha": 0.03, "t": 1513.7} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1531.2} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37711, "bpb": 1.98676, "chunk_loss": 0.030464, "F": 0.2038, "G": 0.20932, "chunks": 208.69, "bytes_per_chunk": 4.907, "boundary_entropy": 0.08853, "router_noise": 0.0, "grad_norm": 0.7005, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.356, "probe_sec": 0.0333, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02439, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.2797, "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": 3.50137, "trunc_abs_frac": 1.93752, "mean_delta_full": 0.11688, "read_hit_rate": {"surface": 0.64385}, "mean_gate": {"surface": 0.17573}, "n_reads": 4279, "n_cand": 6646, "read_suppressed_frac": 0.1047, "dropout_reads_logged": 4279, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 135.5848846435547, "n_util_obs": 140, "util_count_spearman": 0.08668080732982003}, "probe_overhead_sec_total": 273.037, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010571, "alpha": 0.03, "t": 1531.5} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1547.9} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37036, "bpb": 1.97702, "chunk_loss": 0.030561, "F": 0.2103, "G": 0.21482, "chunks": 215.34, "bytes_per_chunk": 4.755, "boundary_entropy": 0.08531, "router_noise": 0.0, "grad_norm": 0.6862, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.333, "probe_sec": 0.0297, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0216, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21002, "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": 0.47528, "trunc_abs_frac": 0.79655, "mean_delta_full": 0.32174, "read_hit_rate": {"surface": 0.67415}, "mean_gate": {"surface": 0.17717}, "n_reads": 4624, "n_cand": 6859, "read_suppressed_frac": 0.0954, "dropout_reads_logged": 4624, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 137.19180297851562, "n_util_obs": 143, "util_count_spearman": 0.15258215962441316}, "probe_overhead_sec_total": 275.719, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010888, "alpha": 0.03, "t": 1548.1} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1564.0} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3307.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190025, "cps": 25.835938, "mean_p": 0.1953, "boundary_entropy": 0.079288, "bpb": 1.881535}, "tr": {"chunks": 3595.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198586, "cps": 28.085938, "mean_p": 0.192289, "boundary_entropy": 0.107422, "bpb": 2.021549}, "fi": {"chunks": 4392.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.23348, "cps": 34.3125, "mean_p": 0.232315, "boundary_entropy": 0.098046, "bpb": 2.121222}}, "bpb": {"en": 1.88154, "tr": 2.02155, "fi": 2.12122}, "gini_cpc": 0.04657, "gini_cps": 0.064046, "bpic": 5.023198, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0450528851461318, "fi": 1.2286844820560552}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 138.78082275390625, "n_util_obs": 148, "util_count_spearman": 0.18809344887964255}, "t": 1564.7} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.89719, "tr": 2.0309, "fi": 2.12653}, "bytes_per_chunk": {"en": 4.932, "tr": 5.526, "fi": 4.564}, "bpb_mean": 2.01821, "frozen_inventory": true, "t": 1565.5} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32058, "bpb": 1.90519, "chunk_loss": 0.030341, "F": 0.19952, "G": 0.20539, "chunks": 204.31, "bytes_per_chunk": 5.012, "boundary_entropy": 0.08434, "router_noise": 0.0, "grad_norm": 0.6564, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.371, "probe_sec": 0.0725, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06164, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19865, "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": 0.36843, "trunc_abs_frac": 1.04215, "mean_delta_full": 0.33775, "read_hit_rate": {"surface": 0.65186}, "mean_gate": {"surface": 0.17133}, "n_reads": 4241, "n_cand": 6506, "read_suppressed_frac": 0.1052, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 138.8199462890625, "n_util_obs": 148, "util_count_spearman": 0.18813416895871354}, "probe_overhead_sec_total": 278.025, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010716, "alpha": 0.03, "t": 1566.7} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1578.8} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32232, "bpb": 1.9077, "chunk_loss": 0.03042, "F": 0.20255, "G": 0.20796, "chunks": 207.41, "bytes_per_chunk": 4.937, "boundary_entropy": 0.08276, "router_noise": 0.0, "grad_norm": 0.6441, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.248, "probe_sec": 0.0332, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02731, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01829, "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": -0.21107, "trunc_abs_frac": 1.36102, "mean_delta_full": 0.02783, "read_hit_rate": {"surface": 0.66798}, "mean_gate": {"surface": 0.17826}, "n_reads": 4412, "n_cand": 6605, "read_suppressed_frac": 0.1049, "dropout_reads_logged": 4412, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 140.03363037109375, "n_util_obs": 149, "util_count_spearman": 0.21602031561763105}, "probe_overhead_sec_total": 280.652, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01076, "alpha": 0.03, "t": 1579.1} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1591.3} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37801, "bpb": 1.98805, "chunk_loss": 0.0304, "F": 0.20245, "G": 0.20769, "chunks": 207.31, "bytes_per_chunk": 4.939, "boundary_entropy": 0.08367, "router_noise": 0.0, "grad_norm": 0.689, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.25, "probe_sec": 0.047, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04081, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19328, "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": -3.55847, "trunc_abs_frac": 0.40315, "mean_delta_full": 0.02557, "read_hit_rate": {"surface": 0.64783}, "mean_gate": {"surface": 0.17578}, "n_reads": 4277, "n_cand": 6602, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 4277, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 141.34742736816406, "n_util_obs": 148, "util_count_spearman": 0.19580805295091008}, "probe_overhead_sec_total": 283.312, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010076, "alpha": 0.03, "t": 1591.6} |
| {"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1604.1} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30674, "bpb": 1.88523, "chunk_loss": 0.030286, "F": 0.19754, "G": 0.20287, "chunks": 202.28, "bytes_per_chunk": 5.062, "boundary_entropy": 0.08271, "router_noise": 0.0, "grad_norm": 0.6785, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.0397, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03408, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12406, "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": 1.70235, "trunc_abs_frac": 1.72747, "mean_delta_full": 0.21456, "read_hit_rate": {"surface": 0.653}, "mean_gate": {"surface": 0.17234}, "n_reads": 4206, "n_cand": 6441, "read_suppressed_frac": 0.1044, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 142.4797821044922, "n_util_obs": 147, "util_count_spearman": 0.18728513249061193}, "probe_overhead_sec_total": 285.594, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01638, "alpha": 0.03, "t": 1604.3} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1616.0} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3324.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.191002, "cps": 25.96875, "mean_p": 0.196396, "boundary_entropy": 0.079101, "bpb": 1.87621}, "tr": {"chunks": 3545.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195824, "cps": 27.695312, "mean_p": 0.190831, "boundary_entropy": 0.112634, "bpb": 2.013035}, "fi": {"chunks": 4432.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235607, "cps": 34.625, "mean_p": 0.232711, "boundary_entropy": 0.095213, "bpb": 2.113978}}, "bpb": {"en": 1.87621, "tr": 2.01304, "fi": 2.11398}, "gini_cpc": 0.047775, "gini_cps": 0.065363, "bpic": 5.020087, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0252476752063155, "fi": 1.2335335707830524}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 144.11544799804688, "n_util_obs": 151, "util_count_spearman": 0.2040885325897525}, "t": 1616.7} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.89189, "tr": 2.02366, "fi": 2.12024}, "bytes_per_chunk": {"en": 4.908, "tr": 5.588, "fi": 4.525}, "bpb_mean": 2.01193, "frozen_inventory": true, "t": 1617.8} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37917, "bpb": 1.98972, "chunk_loss": 0.03074, "F": 0.21271, "G": 0.21724, "chunks": 217.81, "bytes_per_chunk": 4.701, "boundary_entropy": 0.08579, "router_noise": 0.0, "grad_norm": 0.7199, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.295, "probe_sec": 0.0482, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04136, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01107, "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": -0.29888, "trunc_abs_frac": 0.89963, "mean_delta_full": 0.16357, "read_hit_rate": {"surface": 0.66316}, "mean_gate": {"surface": 0.16602}, "n_reads": 4601, "n_cand": 6938, "read_suppressed_frac": 0.0935, "dropout_reads_logged": 4601, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 144.15623474121094, "n_util_obs": 151, "util_count_spearman": 0.20441983654576995}, "probe_overhead_sec_total": 287.884, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.024187, "alpha": 0.03, "t": 1619.1} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1630.5} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31519, "bpb": 1.89742, "chunk_loss": 0.030576, "F": 0.20938, "G": 0.21384, "chunks": 214.41, "bytes_per_chunk": 4.776, "boundary_entropy": 0.08336, "router_noise": 0.0, "grad_norm": 0.6757, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.233, "probe_sec": 0.0422, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03229, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.25164, "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": 1.60282, "trunc_abs_frac": 6.78087, "mean_delta_full": 0.26819, "read_hit_rate": {"surface": 0.67477}, "mean_gate": {"surface": 0.17692}, "n_reads": 4608, "n_cand": 6829, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 4608, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 145.23809814453125, "n_util_obs": 150, "util_count_spearman": 0.16015822925463355}, "probe_overhead_sec_total": 290.488, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.023462, "alpha": 0.03, "t": 1630.7} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1642.2} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30592, "bpb": 1.88405, "chunk_loss": 0.030393, "F": 0.20303, "G": 0.20822, "chunks": 207.91, "bytes_per_chunk": 4.925, "boundary_entropy": 0.08299, "router_noise": 0.0, "grad_norm": 0.6558, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.232, "probe_sec": 0.0284, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02216, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09361, "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": 0.53117, "trunc_abs_frac": 8.97618, "mean_delta_full": -0.05048, "read_hit_rate": {"surface": 0.65594}, "mean_gate": {"surface": 0.17231}, "n_reads": 4343, "n_cand": 6621, "read_suppressed_frac": 0.0967, "dropout_reads_logged": 4343, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 146.70651245117188, "n_util_obs": 148, "util_count_spearman": 0.13483898910552794}, "probe_overhead_sec_total": 293.089, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020855, "alpha": 0.03, "t": 1642.3} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1653.9} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35886, "bpb": 1.96042, "chunk_loss": 0.030322, "F": 0.19748, "G": 0.20291, "chunks": 202.22, "bytes_per_chunk": 5.064, "boundary_entropy": 0.08309, "router_noise": 0.0, "grad_norm": 0.7137, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.235, "probe_sec": 0.0364, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03098, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03892, "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": 0.7991, "trunc_abs_frac": 4.39368, "mean_delta_full": 0.22406, "read_hit_rate": {"surface": 0.65398}, "mean_gate": {"surface": 0.17676}, "n_reads": 4211, "n_cand": 6439, "read_suppressed_frac": 0.1047, "dropout_reads_logged": 4211, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 147.7847137451172, "n_util_obs": 147, "util_count_spearman": 0.14297817038496502}, "probe_overhead_sec_total": 295.206, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.013054, "alpha": 0.03, "t": 1654.1} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1665.6} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3302.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189737, "cps": 25.796875, "mean_p": 0.195172, "boundary_entropy": 0.078674, "bpb": 1.870459}, "tr": {"chunks": 3574.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197426, "cps": 27.921875, "mean_p": 0.190274, "boundary_entropy": 0.105584, "bpb": 2.007339}, "fi": {"chunks": 4387.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233215, "cps": 34.273438, "mean_p": 0.231642, "boundary_entropy": 0.096074, "bpb": 2.10578}}, "bpb": {"en": 1.87046, "tr": 2.00734, "fi": 2.10578}, "gini_cpc": 0.046721, "gini_cps": 0.064222, "bpic": 5.037024, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.040521475253005, "fi": 1.2291441039578856}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 149.3180389404297, "n_util_obs": 147, "util_count_spearman": 0.15294033200099735}, "t": 1666.3} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.88695, "tr": 2.01716, "fi": 2.11231}, "bytes_per_chunk": {"en": 4.933, "tr": 5.551, "fi": 4.559}, "bpb_mean": 2.00547, "frozen_inventory": true, "t": 1667.1} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34833, "bpb": 1.94522, "chunk_loss": 0.030385, "F": 0.20099, "G": 0.20679, "chunks": 205.81, "bytes_per_chunk": 4.975, "boundary_entropy": 0.08601, "router_noise": 0.0, "grad_norm": 0.7309, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.289, "probe_sec": 0.039, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03133, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.12342, "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": 0.42578, "trunc_abs_frac": 0.42578, "mean_delta_full": -1.3103, "read_hit_rate": {"surface": 0.64205}, "mean_gate": {"surface": 0.17102}, "n_reads": 4208, "n_cand": 6554, "read_suppressed_frac": 0.0941, "dropout_reads_logged": 4208, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 149.34140014648438, "n_util_obs": 147, "util_count_spearman": 0.1589358438673507}, "probe_overhead_sec_total": 297.51, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.012028, "alpha": 0.03, "t": 1668.6} |
| {"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1681.0} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33411, "bpb": 1.92471, "chunk_loss": 0.030521, "F": 0.20822, "G": 0.21288, "chunks": 213.22, "bytes_per_chunk": 4.803, "boundary_entropy": 0.08162, "router_noise": 0.0, "grad_norm": 0.7195, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.026, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01958, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02514, "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": 0.93977, "trunc_abs_frac": 0.94857, "mean_delta_full": 0.13578, "read_hit_rate": {"surface": 0.67545}, "mean_gate": {"surface": 0.16616}, "n_reads": 4587, "n_cand": 6791, "read_suppressed_frac": 0.0931, "dropout_reads_logged": 4587, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 150.50067138671875, "n_util_obs": 144, "util_count_spearman": 0.13490876939152802}, "probe_overhead_sec_total": 299.944, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010325, "alpha": 0.03, "t": 1681.2} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1693.0} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34089, "bpb": 1.93449, "chunk_loss": 0.030329, "F": 0.19897, "G": 0.20441, "chunks": 203.75, "bytes_per_chunk": 5.026, "boundary_entropy": 0.08273, "router_noise": 0.0, "grad_norm": 0.7213, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.239, "probe_sec": 0.0319, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02501, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.14511, "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": 2.26764, "trunc_abs_frac": 0.87383, "mean_delta_full": -0.09511, "read_hit_rate": {"surface": 0.64704}, "mean_gate": {"surface": 0.16712}, "n_reads": 4198, "n_cand": 6488, "read_suppressed_frac": 0.0958, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 151.68907165527344, "n_util_obs": 147, "util_count_spearman": 0.1781766390376958}, "probe_overhead_sec_total": 302.583, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010081, "alpha": 0.03, "t": 1693.2} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1705.4} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39419, "bpb": 2.01139, "chunk_loss": 0.030557, "F": 0.20844, "G": 0.21346, "chunks": 213.44, "bytes_per_chunk": 4.798, "boundary_entropy": 0.08237, "router_noise": 0.0, "grad_norm": 0.7423, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.252, "probe_sec": 0.0444, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03252, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11262, "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": 1.41021, "trunc_abs_frac": 1.41021, "mean_delta_full": 0.27258, "read_hit_rate": {"surface": 0.65137}, "mean_gate": {"surface": 0.16665}, "n_reads": 4428, "n_cand": 6798, "read_suppressed_frac": 0.098, "dropout_reads_logged": 4428, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 153.21575927734375, "n_util_obs": 149, "util_count_spearman": 0.21174315254852166}, "probe_overhead_sec_total": 304.963, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008787, "alpha": 0.03, "t": 1705.8} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1716.9} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3308.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190082, "cps": 25.84375, "mean_p": 0.195432, "boundary_entropy": 0.077287, "bpb": 1.868406}, "tr": {"chunks": 3536.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195327, "cps": 27.625, "mean_p": 0.189048, "boundary_entropy": 0.107976, "bpb": 2.004869}, "fi": {"chunks": 4396.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233693, "cps": 34.34375, "mean_p": 0.231893, "boundary_entropy": 0.094195, "bpb": 2.102028}}, "bpb": {"en": 1.86841, "tr": 2.00487, "fi": 2.10203}, "gini_cpc": 0.046961, "gini_cps": 0.064531, "bpic": 5.047331, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0275910764822094, "fi": 1.2294317360555393}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 154.74093627929688, "n_util_obs": 150, "util_count_spearman": 0.16067736343837505}, "t": 1717.4} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.88391, "tr": 2.01349, "fi": 2.10857}, "bytes_per_chunk": {"en": 4.929, "tr": 5.625, "fi": 4.561}, "bpb_mean": 2.00199, "frozen_inventory": true, "t": 1718.4} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31689, "bpb": 1.89988, "chunk_loss": 0.030496, "F": 0.20468, "G": 0.20911, "chunks": 209.59, "bytes_per_chunk": 4.886, "boundary_entropy": 0.08405, "router_noise": 0.0, "grad_norm": 0.6974, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.281, "probe_sec": 0.0335, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02003, "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": 0.48313, "trunc_abs_frac": 0.58692, "mean_delta_full": -0.24934, "read_hit_rate": {"surface": 0.66262}, "mean_gate": {"surface": 0.17228}, "n_reads": 4423, "n_cand": 6675, "read_suppressed_frac": 0.1056, "dropout_reads_logged": 4423, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 154.76939392089844, "n_util_obs": 150, "util_count_spearman": 0.16052091204053515}, "probe_overhead_sec_total": 307.342, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009518, "alpha": 0.03, "t": 1719.8} |
| {"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1731.2} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34118, "bpb": 1.93491, "chunk_loss": 0.030463, "F": 0.20633, "G": 0.21162, "chunks": 211.28, "bytes_per_chunk": 4.847, "boundary_entropy": 0.08041, "router_noise": 0.0, "grad_norm": 0.729, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.236, "probe_sec": 0.048, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04196, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18712, "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": 3.1707, "trunc_abs_frac": 1.96664, "mean_delta_full": 0.0782, "read_hit_rate": {"surface": 0.64408}, "mean_gate": {"surface": 0.16257}, "n_reads": 4334, "n_cand": 6729, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 4334, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 156.30287170410156, "n_util_obs": 150, "util_count_spearman": 0.13402373438819504}, "probe_overhead_sec_total": 310.052, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010533, "alpha": 0.03, "t": 1731.6} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1742.6} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27452, "bpb": 1.83874, "chunk_loss": 0.030324, "F": 0.1983, "G": 0.20396, "chunks": 203.06, "bytes_per_chunk": 5.043, "boundary_entropy": 0.08212, "router_noise": 0.0, "grad_norm": 0.6891, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.225, "probe_sec": 0.036, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02971, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09023, "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": 0.74797, "trunc_abs_frac": 0.50974, "mean_delta_full": -0.18672, "read_hit_rate": {"surface": 0.6545}, "mean_gate": {"surface": 0.16951}, "n_reads": 4232, "n_cand": 6466, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 4232, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 157.99464416503906, "n_util_obs": 152, "util_count_spearman": 0.10638399661014633}, "probe_overhead_sec_total": 312.677, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.012447, "alpha": 0.03, "t": 1742.9} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1754.8} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37817, "bpb": 1.98829, "chunk_loss": 0.030418, "F": 0.20374, "G": 0.2085, "chunks": 208.62, "bytes_per_chunk": 4.908, "boundary_entropy": 0.08413, "router_noise": 0.0, "grad_norm": 0.7315, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.243, "probe_sec": 0.0409, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03479, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01851, "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": 0.77955, "trunc_abs_frac": 0.31257, "mean_delta_full": -0.12805, "read_hit_rate": {"surface": 0.63621}, "mean_gate": {"surface": 0.16825}, "n_reads": 4227, "n_cand": 6644, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 4227, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 159.3297882080078, "n_util_obs": 152, "util_count_spearman": 0.10634137256731696}, "probe_overhead_sec_total": 315.39, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.013328, "alpha": 0.03, "t": 1755.0} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1766.6} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37872, "bpb": 1.98907, "chunk_loss": 0.03052, "F": 0.20599, "G": 0.21014, "chunks": 210.94, "bytes_per_chunk": 4.855, "boundary_entropy": 0.07815, "router_noise": 0.0, "grad_norm": 0.7147, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.231, "probe_sec": 0.0553, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04303, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16844, "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": -0.29305, "trunc_abs_frac": 0.3127, "mean_delta_full": -0.20276, "read_hit_rate": {"surface": 0.65585}, "mean_gate": {"surface": 0.16976}, "n_reads": 4406, "n_cand": 6718, "read_suppressed_frac": 0.1044, "dropout_reads_logged": 4406, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 160.794677734375, "n_util_obs": 153, "util_count_spearman": 0.15476370971192566}, "probe_overhead_sec_total": 317.644, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.012842, "alpha": 0.03, "t": 1766.6} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3301.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18968, "cps": 25.789062, "mean_p": 0.195122, "boundary_entropy": 0.075917, "bpb": 1.865938}, "tr": {"chunks": 3552.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196211, "cps": 27.75, "mean_p": 0.18935, "boundary_entropy": 0.105107, "bpb": 2.000132}, "fi": {"chunks": 4379.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232789, "cps": 34.210938, "mean_p": 0.231144, "boundary_entropy": 0.094825, "bpb": 2.101119}}, "bpb": {"en": 1.86594, "tr": 2.00013, "fi": 2.10112}, "gini_cpc": 0.046453, "gini_cps": 0.063984, "bpic": 5.050926, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0344297482631741, "fi": 1.2272743501497243}, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 160.794677734375, "n_util_obs": 153, "util_count_spearman": 0.15476370971192566}, "t": 1767.3} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.88124, "tr": 2.00924, "fi": 2.10687}, "bytes_per_chunk": {"en": 4.927, "tr": 5.597, "fi": 4.581}, "bpb_mean": 1.99912, "frozen_inventory": true, "t": 1768.1} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_mem-random_s2/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-random_s2/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-random_s2/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-random_s2/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-random_s2/probe/reps/th0.70"], "t": 1770.0} |
| {"event": "final", "run": "ali_mem-random_s2", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 2, "per_lang": {"en": {"cpc": 0.18967994024018847, "cps": 25.7890625, "bpb": 1.887210698121041, "bytes_per_chunk": 4.876054425564763, "reads_per_chunk": 0.6769993396761623}, "tr": {"cpc": 0.19621057283323207, "cps": 27.75, "bpb": 2.0005761688738923, "bytes_per_chunk": 5.610838809100833, "reads_per_chunk": 0.5789259647695897}, "fi": {"cpc": 0.23278932539471586, "cps": 34.2109375, "bpb": 2.1247721718224195, "bytes_per_chunk": 4.562160092585342, "reads_per_chunk": 0.5796068603649464}}, "bpb_mean": 2.004186346272451, "gini_cpc": 0.04645308981488383, "gini_cps": 0.06398385565052234, "bpic": 5.050925925925926, "mean_boundary_entropy": 0.0781458169221878, "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": 317.6440327167511, "probe_overhead_seq_frac": 0.49985416666666665, "wall_sec": 1772.23020529747, "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": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 160.794677734375, "n_util_obs": 153, "util_count_spearman": 0.15476370971192566}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "random", "value_mode": "learned", "candidate_types": ["surface"], "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": 0.15476370971192566, "paired_vs_dropout": {"n": 188, "pearson": 0.2360564433337164, "spearman": 0.1246345088647967, "mean_abs_discrepancy": 0.44449761509895325, "mean_paired": 0.08757366985082626, "mean_dropout": 0.0988227054476738}, "predictor_calibration": {"n": 273, "all": {"n": 273, "pearson": 0.2399563679231443, "spearman": 0.30818419450879736, "mae": 0.1049981415271759, "reliability": [{"pred": 0.015780987218022346, "obs": 0.010552691295742989, "n": 28}, {"pred": 0.03746779263019562, "obs": -0.0008808771963231266, "n": 27}, {"pred": 0.048890046775341034, "obs": 0.07203880697488785, "n": 27}, {"pred": 0.05779610201716423, "obs": 0.06596147269010544, "n": 27}, {"pred": 0.06574138253927231, "obs": 0.09322180598974228, "n": 27}, {"pred": 0.0741511806845665, "obs": 0.12963199615478516, "n": 28}, {"pred": 0.08890796452760696, "obs": 0.053287651389837265, "n": 27}, {"pred": 0.10822872072458267, "obs": 0.0980893149971962, "n": 27}, {"pred": 0.1182013601064682, "obs": 0.12967272102832794, "n": 27}, {"pred": 0.12872345745563507, "obs": 0.1480996310710907, "n": 28}]}, "admitted": {"n": 273, "pearson": 0.2399563679231443, "spearman": 0.30818419450879736, "mae": 0.1049981415271759, "reliability": [{"pred": 0.015780987218022346, "obs": 0.010552691295742989, "n": 28}, {"pred": 0.03746779263019562, "obs": -0.0008808771963231266, "n": 27}, {"pred": 0.048890046775341034, "obs": 0.07203880697488785, "n": 27}, {"pred": 0.05779610201716423, "obs": 0.06596147269010544, "n": 27}, {"pred": 0.06574138253927231, "obs": 0.09322180598974228, "n": 27}, {"pred": 0.0741511806845665, "obs": 0.12963199615478516, "n": 28}, {"pred": 0.08890796452760696, "obs": 0.053287651389837265, "n": 27}, {"pred": 0.10822872072458267, "obs": 0.0980893149971962, "n": 27}, {"pred": 0.1182013601064682, "obs": 0.12967272102832794, "n": 27}, {"pred": 0.12872345745563507, "obs": 0.1480996310710907, "n": 28}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 1772.2} |
| {"event": "hub_done", "repo": "AdaptiveChunking/ali_mem-random_s2", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 1798.0} |
|
|
