| {"event": "hub", "repo": "AdaptiveChunking/ali_mem-shuffled_s1", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 2.1} |
| {"event": "start", "run": "ali_mem-shuffled_s1", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "shuffled", "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": 2.1} |
| {"event": "probe_schedule", "n": 45, "steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 200, 256, 400, 512, 600, 800, 1000, 1024, 1200, 1400, 1600, 1800, 2000, 2048, 2200, 2400, 2600, 2800, 3000, 3200, 3400, 3600, 3800, 4000, 4096, 4200, 4400, 4600, 4800, 5000, 5200, 5400, 5600, 5800, 6000], "t": 2.1} |
| {"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2674.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.153652, "cps": 20.890625, "mean_p": 0.478126, "boundary_entropy": 0.685438, "bpb": 8.041379}, "tr": {"chunks": 5645.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.311827, "cps": 44.101562, "mean_p": 0.486063, "boundary_entropy": 0.686375, "bpb": 8.01701}, "fi": {"chunks": 4732.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.251555, "cps": 36.96875, "mean_p": 0.48451, "boundary_entropy": 0.686595, "bpb": 8.075294}}, "bpb": {"en": 8.04138, "tr": 8.01701, "fi": 8.07529}, "gini_cpc": 0.147064, "gini_cps": 0.151764, "bpic": 4.346947, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.029439514461233, "fi": 1.6371767549398106}, "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": 3.4} |
| {"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.58509, "bpb": 8.05758, "chunk_loss": 0.030141, "F": 0.16724, "G": 0.47713, "chunks": 171.25, "bytes_per_chunk": 5.98, "boundary_entropy": 0.69022, "router_noise": 0.0, "grad_norm": 11.3982, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.007, "probe_sec": 0.0032, "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": 5434, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 3161, "admitted": 3161, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 3161, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.7190762758255005, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.003, "probe_overhead_seq_frac": 0.0, "util_loss": 0.0, "alpha": 0.03, "t": 4.3} |
| {"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2659.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.15279, "cps": 20.773438, "mean_p": 0.478104, "boundary_entropy": 0.685439, "bpb": 8.007687}, "tr": {"chunks": 5624.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.310667, "cps": 43.9375, "mean_p": 0.485999, "boundary_entropy": 0.686373, "bpb": 7.991801}, "fi": {"chunks": 4707.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.250226, "cps": 36.773438, "mean_p": 0.484401, "boundary_entropy": 0.686587, "bpb": 8.046083}}, "bpb": {"en": 8.00769, "tr": 7.9918, "fi": 8.04608}, "gini_cpc": 0.147476, "gini_cps": 0.152168, "bpic": 4.36736, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.0332957058204215, "fi": 1.6377141362380108}, "inventory": {"size": 3161, "admitted": 3161, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 3161, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.7190762758255005, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 4.9} |
| {"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2659.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.15279, "cps": 20.773438, "mean_p": 0.478078, "boundary_entropy": 0.685441, "bpb": 7.941635}, "tr": {"chunks": 5587.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.308623, "cps": 43.648438, "mean_p": 0.485898, "boundary_entropy": 0.686371, "bpb": 7.943341}, "fi": {"chunks": 4664.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.24794, "cps": 36.4375, "mean_p": 0.484225, "boundary_entropy": 0.686575, "bpb": 7.991336}}, "bpb": {"en": 7.94164, "tr": 7.94334, "fi": 7.99134}, "gini_cpc": 0.146456, "gini_cps": 0.151201, "bpic": 4.394423, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.0199187603873923, "fi": 1.6227530765698073}, "inventory": {"size": 5575, "admitted": 5575, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 5575, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.8444843292236328, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 6.6} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2615.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.150261, "cps": 20.429688, "mean_p": 0.478042, "boundary_entropy": 0.685455, "bpb": 7.725727}, "tr": {"chunks": 5457.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.301442, "cps": 42.632812, "mean_p": 0.485579, "boundary_entropy": 0.686365, "bpb": 7.788105}, "fi": {"chunks": 4538.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.241242, "cps": 35.453125, "mean_p": 0.483683, "boundary_entropy": 0.686539, "bpb": 7.812184}}, "bpb": {"en": 7.72573, "tr": 7.78811, "fi": 7.81218}, "gini_cpc": 0.145447, "gini_cps": 0.150251, "bpic": 4.498969, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.0061150191241555, "fi": 1.6054805002524357}, "inventory": {"size": 10483, "admitted": 10483, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 10483, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.9945626258850098, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 8.1} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2580.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.14825, "cps": 20.15625, "mean_p": 0.478076, "boundary_entropy": 0.685489, "bpb": 7.442497}, "tr": {"chunks": 5249.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.289952, "cps": 41.007812, "mean_p": 0.485094, "boundary_entropy": 0.68636, "bpb": 7.586648}, "fi": {"chunks": 4289.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228005, "cps": 33.507812, "mean_p": 0.482849, "boundary_entropy": 0.686485, "bpb": 7.574689}}, "bpb": {"en": 7.4425, "tr": 7.58665, "fi": 7.57469}, "gini_cpc": 0.141799, "gini_cps": 0.146834, "bpic": 4.681631, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.9558269925709346, "fi": 1.5379725247350324}, "inventory": {"size": 15095, "admitted": 15095, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 15095, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.039748191833496, "n_util_obs": 1, "util_count_spearman": NaN}, "t": 9.6} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2627.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.150951, "cps": 20.523438, "mean_p": 0.478865, "boundary_entropy": 0.685614, "bpb": 7.167441}, "tr": {"chunks": 5168.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.285478, "cps": 40.375, "mean_p": 0.484982, "boundary_entropy": 0.686393, "bpb": 7.395166}, "fi": {"chunks": 4121.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219074, "cps": 32.195312, "mean_p": 0.48238, "boundary_entropy": 0.68647, "bpb": 7.338518}}, "bpb": {"en": 7.16744, "tr": 7.39517, "fi": 7.33852}, "gini_cpc": 0.136818, "gini_cps": 0.142162, "bpic": 4.760994, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.8911936499387962, "fi": 1.451291923828595}, "inventory": {"size": 19584, "admitted": 19584, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 19584, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.0542280673980713, "n_util_obs": 2, "util_count_spearman": NaN}, "t": 11.2} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2608.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.149859, "cps": 20.375, "mean_p": 0.481164, "boundary_entropy": 0.686039, "bpb": 6.51484}, "tr": {"chunks": 4390.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.242501, "cps": 34.296875, "mean_p": 0.483003, "boundary_entropy": 0.686443, "bpb": 6.897431}, "fi": {"chunks": 2947.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.156664, "cps": 23.023438, "mean_p": 0.478364, "boundary_entropy": 0.686281, "bpb": 6.693122}}, "bpb": {"en": 6.51484, "tr": 6.89743, "fi": 6.69312}, "gini_cpc": 0.112493, "gini_cps": 0.119457, "bpic": 5.704575, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.6181936848076903, "fi": 1.0454055118187275}, "inventory": {"size": 38400, "admitted": 38400, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 38400, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.324791669845581, "n_util_obs": 3, "util_count_spearman": 0.5}, "t": 13.9} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3623.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208182, "cps": 28.304688, "mean_p": 0.490288, "boundary_entropy": 0.687162, "bpb": 6.161352}, "tr": {"chunks": 3898.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215323, "cps": 30.453125, "mean_p": 0.486103, "boundary_entropy": 0.687277, "bpb": 6.576846}, "fi": {"chunks": 3136.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.166711, "cps": 24.5, "mean_p": 0.484989, "boundary_entropy": 0.687295, "bpb": 6.341366}}, "bpb": {"en": 6.16135, "tr": 6.57685, "fi": 6.34137}, "gini_cpc": 0.054909, "gini_cps": 0.047668, "bpic": 5.323449, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0343012975601982, "fi": 0.800792425747099}, "inventory": {"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": 2.5770263671875, "n_util_obs": 4, "util_count_spearman": 0.4}, "t": 18.6} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 23.6} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.09838, "bpb": 5.91271, "chunk_loss": 0.030281, "F": 0.17038, "G": 0.4861, "chunks": 174.47, "bytes_per_chunk": 5.869, "boundary_entropy": 0.69163, "router_noise": 0.0, "grad_norm": 2.3619, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 0.39, "probe_sec": 0.0313, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02626, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0078, "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": -0.38346, "trunc_abs_frac": 1.08171, "mean_delta_full": -0.01562, "read_hit_rate": {"surface": 0.61362}, "mean_gate": {"surface": 0.11915}, "n_reads": 3405, "n_cand": 5549, "read_suppressed_frac": 0.1069, "dropout_reads_logged": 3405, "inventory": {"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": 3.6025848388671875, "n_util_obs": 9, "util_count_spearman": 0.4166666666666667}, "probe_overhead_sec_total": 2.406, "probe_overhead_seq_frac": 0.4902, "util_loss": 0.002381, "alpha": 0.03, "t": 23.8} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3764.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.216285, "cps": 29.40625, "mean_p": 0.484615, "boundary_entropy": 0.686192, "bpb": 5.428941}, "tr": {"chunks": 4516.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.249461, "cps": 35.28125, "mean_p": 0.481778, "boundary_entropy": 0.68615, "bpb": 5.921189}, "fi": {"chunks": 3986.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.211897, "cps": 31.140625, "mean_p": 0.482677, "boundary_entropy": 0.686558, "bpb": 5.653718}}, "bpb": {"en": 5.42894, "tr": 5.92119, "fi": 5.65372}, "gini_cpc": 0.036956, "gini_cps": 0.040872, "bpic": 4.625143, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.15339452957903, "fi": 0.9797153586227245}, "inventory": {"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": 4.289794921875, "n_util_obs": 11, "util_count_spearman": 0.2636363636363636}, "t": 27.3} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 35.9} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.20877, "bpb": 4.62928, "chunk_loss": 0.033023, "F": 0.21375, "G": 0.46259, "chunks": 218.88, "bytes_per_chunk": 4.678, "boundary_entropy": 0.68389, "router_noise": 0.0, "grad_norm": 1.7744, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.245, "probe_sec": 0.0265, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0199, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00207, "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": -0.08545, "trunc_abs_frac": 0.09178, "mean_delta_full": 0.02337, "read_hit_rate": {"surface": 0.71343}, "mean_gate": {"surface": 0.11944}, "n_reads": 4974, "n_cand": 6972, "read_suppressed_frac": 0.1041, "dropout_reads_logged": 4974, "inventory": {"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": 6.5779876708984375, "n_util_obs": 23, "util_count_spearman": 0.18774703557312253}, "probe_overhead_sec_total": 4.736, "probe_overhead_seq_frac": 0.495, "util_loss": 0.000518, "alpha": 0.03, "t": 36.0} |
| {"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3620.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20801, "cps": 28.28125, "mean_p": 0.423396, "boundary_entropy": 0.61858, "bpb": 4.116156}, "tr": {"chunks": 3950.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.218196, "cps": 30.859375, "mean_p": 0.405891, "boundary_entropy": 0.639531, "bpb": 4.363962}, "fi": {"chunks": 4053.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.215459, "cps": 31.664062, "mean_p": 0.412502, "boundary_entropy": 0.642374, "bpb": 4.320276}}, "bpb": {"en": 4.11616, "tr": 4.36396, "fi": 4.32028}, "gini_cpc": 0.010583, "gini_cps": 0.024836, "bpic": 4.881012, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0489676476808734, "fi": 1.035810406571211}, "inventory": {"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": 8.726455688476562, "n_util_obs": 31, "util_count_spearman": 0.36088709677419356}, "t": 42.4} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 48.6} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.75152, "bpb": 3.9696, "chunk_loss": 0.033097, "F": 0.22641, "G": 0.40434, "chunks": 231.84, "bytes_per_chunk": 4.417, "boundary_entropy": 0.58713, "router_noise": 0.0, "grad_norm": 2.0043, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.256, "probe_sec": 0.0488, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04142, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00984, "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": 0.03122, "trunc_abs_frac": 5.15566, "mean_delta_full": -0.0769, "read_hit_rate": {"surface": 0.81522}, "mean_gate": {"surface": 0.11927}, "n_reads": 6022, "n_cand": 7387, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 6022, "inventory": {"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": 10.62451171875, "n_util_obs": 35, "util_count_spearman": 0.40476190476190477}, "probe_overhead_sec_total": 7.302, "probe_overhead_seq_frac": 0.4967, "util_loss": 0.000322, "alpha": 0.03, "t": 48.8} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 58.7} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3215.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.184738, "cps": 25.117188, "mean_p": 0.319757, "boundary_entropy": 0.45895, "bpb": 3.54874}, "tr": {"chunks": 4266.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.235652, "cps": 33.328125, "mean_p": 0.303305, "boundary_entropy": 0.48058, "bpb": 3.783523}, "fi": {"chunks": 3826.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.203392, "cps": 29.890625, "mean_p": 0.334326, "boundary_entropy": 0.529073, "bpb": 3.887811}}, "bpb": {"en": 3.54874, "tr": 3.78352, "fi": 3.88781}, "gini_cpc": 0.054414, "gini_cps": 0.061968, "bpic": 5.017423, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.275596863257587, "fi": 1.1009718760068332}, "inventory": {"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": 14.924911499023438, "n_util_obs": 49, "util_count_spearman": 0.20795918367346938}, "t": 59.4} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.55671, "tr": 3.79276, "fi": 3.88588}, "bytes_per_chunk": {"en": 5.183, "tr": 4.565, "fi": 5.2}, "bpb_mean": 3.74511, "frozen_inventory": true, "t": 60.5} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.48725, "bpb": 3.58834, "chunk_loss": 0.030984, "F": 0.1958, "G": 0.32309, "chunks": 200.5, "bytes_per_chunk": 5.107, "boundary_entropy": 0.47178, "router_noise": 0.0, "grad_norm": 1.3624, "lr": 0.000201, "router_trainable": true, "sec_per_step": 0.251, "probe_sec": 0.0443, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03555, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00157, "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": 0.07056, "trunc_abs_frac": 9.26961, "mean_delta_full": -0.05017, "read_hit_rate": {"surface": 0.80529}, "mean_gate": {"surface": 0.12105}, "n_reads": 5141, "n_cand": 6384, "read_suppressed_frac": 0.1019, "dropout_reads_logged": 5141, "inventory": {"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": 15.004898071289062, "n_util_obs": 49, "util_count_spearman": 0.22969387755102041}, "probe_overhead_sec_total": 9.857, "probe_overhead_seq_frac": 0.4975, "util_loss": 0.000405, "alpha": 0.03, "t": 61.4} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 71.2} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.36087, "bpb": 3.40602, "chunk_loss": 0.030662, "F": 0.18948, "G": 0.29699, "chunks": 194.03, "bytes_per_chunk": 5.278, "boundary_entropy": 0.40464, "router_noise": 0.0, "grad_norm": 1.4303, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.202, "probe_sec": 0.1294, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11316, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05445, "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": 0.21288, "trunc_abs_frac": 2.81297, "mean_delta_full": -0.14852, "read_hit_rate": {"surface": 0.79602}, "mean_gate": {"surface": 0.12848}, "n_reads": 4917, "n_cand": 6177, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 4917, "inventory": {"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": 18.928680419921875, "n_util_obs": 62, "util_count_spearman": 0.21175492936465967}, "probe_overhead_sec_total": 12.523, "probe_overhead_seq_frac": 0.498, "util_loss": 0.000683, "alpha": 0.03, "t": 71.4} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3142.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.180544, "cps": 24.546875, "mean_p": 0.285243, "boundary_entropy": 0.384021, "bpb": 3.266573}, "tr": {"chunks": 3811.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.210518, "cps": 29.773438, "mean_p": 0.288167, "boundary_entropy": 0.418133, "bpb": 3.505176}, "fi": {"chunks": 3338.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.177449, "cps": 26.078125, "mean_p": 0.292116, "boundary_entropy": 0.454689, "bpb": 3.660912}}, "bpb": {"en": 3.26657, "tr": 3.50518, "fi": 3.66091}, "gini_cpc": 0.038778, "gini_cps": 0.043339, "bpic": 5.512778, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.166020905271072, "fi": 0.9828616468667637}, "inventory": {"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": 19.31549072265625, "n_util_obs": 62, "util_count_spearman": 0.23736795056086382}, "t": 73.2} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 82.0} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.21394, "bpb": 3.19403, "chunk_loss": 0.03053, "F": 0.18649, "G": 0.28891, "chunks": 190.97, "bytes_per_chunk": 5.362, "boundary_entropy": 0.37424, "router_noise": 0.0, "grad_norm": 1.73, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.217, "probe_sec": 0.0657, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05381, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06339, "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": 0.09115, "trunc_abs_frac": 0.61953, "mean_delta_full": 0.02885, "read_hit_rate": {"surface": 0.79026}, "mean_gate": {"surface": 0.1501}, "n_reads": 4804, "n_cand": 6079, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 4804, "inventory": {"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": 22.758956909179688, "n_util_obs": 75, "util_count_spearman": 0.2532859174964438}, "probe_overhead_sec_total": 14.849, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.001008, "alpha": 0.03, "t": 82.3} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 91.8} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.191, "bpb": 3.16095, "chunk_loss": 0.030469, "F": 0.18683, "G": 0.27486, "chunks": 191.31, "bytes_per_chunk": 5.352, "boundary_entropy": 0.34499, "router_noise": 0.0, "grad_norm": 1.528, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.193, "probe_sec": 0.0389, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02924, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01942, "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": 0.38269, "trunc_abs_frac": 1.06947, "mean_delta_full": -0.12371, "read_hit_rate": {"surface": 0.77274}, "mean_gate": {"surface": 0.15404}, "n_reads": 4706, "n_cand": 6090, "read_suppressed_frac": 0.096, "dropout_reads_logged": 4706, "inventory": {"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": 26.606948852539062, "n_util_obs": 84, "util_count_spearman": 0.23620532550369544}, "probe_overhead_sec_total": 17.293, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.001628, "alpha": 0.03, "t": 92.0} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 102.2} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3320.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190772, "cps": 25.9375, "mean_p": 0.255319, "boundary_entropy": 0.283697, "bpb": 2.859028}, "tr": {"chunks": 3478.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.192123, "cps": 27.171875, "mean_p": 0.297601, "boundary_entropy": 0.404862, "bpb": 3.130164}, "fi": {"chunks": 3288.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.174791, "cps": 25.6875, "mean_p": 0.28225, "boundary_entropy": 0.41121, "bpb": 3.365287}}, "bpb": {"en": 2.85903, "tr": 3.13016, "fi": 3.36529}, "gini_cpc": 0.020718, "gini_cps": 0.012559, "bpic": 5.624826, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0070825310855085, "fi": 0.9162330679370503}, "inventory": {"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": 30.513702392578125, "n_util_obs": 88, "util_count_spearman": 0.2942129548096228}, "t": 102.8} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.87207, "tr": 3.13748, "fi": 3.35992}, "bytes_per_chunk": {"en": 4.926, "tr": 5.667, "fi": 5.839}, "bpb_mean": 3.12316, "frozen_inventory": true, "t": 103.7} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.02149, "bpb": 2.91639, "chunk_loss": 0.030615, "F": 0.19678, "G": 0.2636, "chunks": 201.5, "bytes_per_chunk": 5.082, "boundary_entropy": 0.29134, "router_noise": 0.0, "grad_norm": 1.0672, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 0.249, "probe_sec": 0.0308, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0251, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0787, "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": 0.19474, "trunc_abs_frac": 0.42528, "mean_delta_full": 0.27377, "read_hit_rate": {"surface": 0.83151}, "mean_gate": {"surface": 0.16978}, "n_reads": 5335, "n_cand": 6416, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 5335, "inventory": {"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": 30.597000122070312, "n_util_obs": 88, "util_count_spearman": 0.30312422950935153}, "probe_overhead_sec_total": 19.859, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.002126, "alpha": 0.03, "t": 104.4} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 114.6} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.05967, "bpb": 2.97147, "chunk_loss": 0.030591, "F": 0.1962, "G": 0.25425, "chunks": 200.91, "bytes_per_chunk": 5.097, "boundary_entropy": 0.26793, "router_noise": 0.0, "grad_norm": 1.1657, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 0.209, "probe_sec": 0.0355, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02473, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03228, "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": 0.53315, "trunc_abs_frac": 0.64638, "mean_delta_full": -0.07642, "read_hit_rate": {"surface": 0.8171}, "mean_gate": {"surface": 0.18517}, "n_reads": 5227, "n_cand": 6397, "read_suppressed_frac": 0.101, "dropout_reads_logged": 5227, "inventory": {"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": 34.64588928222656, "n_util_obs": 98, "util_count_spearman": 0.22426027580666755}, "probe_overhead_sec_total": 22.636, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.002651, "alpha": 0.03, "t": 114.9} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 124.7} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.02189, "bpb": 2.91697, "chunk_loss": 0.030446, "F": 0.1918, "G": 0.25225, "chunks": 196.41, "bytes_per_chunk": 5.214, "boundary_entropy": 0.26999, "router_noise": 0.0, "grad_norm": 1.1803, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.201, "probe_sec": 0.1012, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09484, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01881, "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": 0.34725, "trunc_abs_frac": 0.52461, "mean_delta_full": -0.25517, "read_hit_rate": {"surface": 0.79098}, "mean_gate": {"surface": 0.20822}, "n_reads": 4946, "n_cand": 6253, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 4946, "inventory": {"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": 38.69267272949219, "n_util_obs": 111, "util_count_spearman": 0.1906019656019656}, "probe_overhead_sec_total": 25.131, "probe_overhead_seq_frac": 0.499, "util_loss": 0.003, "alpha": 0.03, "t": 124.9} |
| {"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3490.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20054, "cps": 27.265625, "mean_p": 0.248249, "boundary_entropy": 0.24377, "bpb": 2.676047}, "tr": {"chunks": 3271.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.180688, "cps": 25.554688, "mean_p": 0.291211, "boundary_entropy": 0.387796, "bpb": 2.970911}, "fi": {"chunks": 3392.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.18032, "cps": 26.5, "mean_p": 0.262444, "boundary_entropy": 0.361862, "bpb": 3.225536}}, "bpb": {"en": 2.67605, "tr": 2.97091, "fi": 3.22554}, "gini_cpc": 0.024005, "gini_cps": 0.01438, "bpic": 5.587708, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9010080806312557, "fi": 0.8991717490177896}, "inventory": {"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": 39.59417724609375, "n_util_obs": 116, "util_count_spearman": 0.17354817504390466}, "t": 127.6} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 136.5} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.94719, "bpb": 2.8092, "chunk_loss": 0.030809, "F": 0.20685, "G": 0.25509, "chunks": 211.81, "bytes_per_chunk": 4.834, "boundary_entropy": 0.24357, "router_noise": 0.0, "grad_norm": 1.3103, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 0.237, "probe_sec": 0.0568, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04891, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04401, "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": 2.25302, "trunc_abs_frac": 0.5577, "mean_delta_full": 0.02524, "read_hit_rate": {"surface": 0.8362}, "mean_gate": {"surface": 0.22828}, "n_reads": 5641, "n_cand": 6746, "read_suppressed_frac": 0.1053, "dropout_reads_logged": 5641, "inventory": {"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": 42.75679016113281, "n_util_obs": 126, "util_count_spearman": 0.2177071131717682}, "probe_overhead_sec_total": 28.063, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.003858, "alpha": 0.03, "t": 136.8} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 146.2} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3470.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199391, "cps": 27.109375, "mean_p": 0.23869, "boundary_entropy": 0.217999, "bpb": 2.573903}, "tr": {"chunks": 3311.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.182898, "cps": 25.867188, "mean_p": 0.293918, "boundary_entropy": 0.384464, "bpb": 2.882513}, "fi": {"chunks": 3395.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.18048, "cps": 26.523438, "mean_p": 0.260445, "boundary_entropy": 0.349136, "bpb": 3.155848}}, "bpb": {"en": 2.5739, "tr": 2.88251, "fi": 3.15585}, "gini_cpc": 0.022403, "gini_cps": 0.010417, "bpic": 5.575079, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9172828519991512, "fi": 0.9051541367741635}, "inventory": {"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": 46.790863037109375, "n_util_obs": 132, "util_count_spearman": 0.15300589216708319}, "t": 146.8} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.58809, "tr": 2.89318, "fi": 3.15192}, "bytes_per_chunk": {"en": 4.783, "tr": 5.844, "fi": 5.729}, "bpb_mean": 2.87773, "frozen_inventory": true, "t": 147.6} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.92214, "bpb": 2.77306, "chunk_loss": 0.030596, "F": 0.19821, "G": 0.25616, "chunks": 202.97, "bytes_per_chunk": 5.045, "boundary_entropy": 0.25369, "router_noise": 0.0, "grad_norm": 1.3613, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 0.236, "probe_sec": 0.0305, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02364, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21196, "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": 0.87097, "trunc_abs_frac": 0.57422, "mean_delta_full": 0.46167, "read_hit_rate": {"surface": 0.82145}, "mean_gate": {"surface": 0.23014}, "n_reads": 5309, "n_cand": 6463, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 5309, "inventory": {"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": 46.8702392578125, "n_util_obs": 132, "util_count_spearman": 0.15985703291167222}, "probe_overhead_sec_total": 30.456, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.004861, "alpha": 0.03, "t": 148.6} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 158.7} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.85599, "bpb": 2.67763, "chunk_loss": 0.030504, "F": 0.19601, "G": 0.25336, "chunks": 200.72, "bytes_per_chunk": 5.102, "boundary_entropy": 0.24424, "router_noise": 0.0, "grad_norm": 1.3266, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.209, "probe_sec": 0.0577, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05037, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.08762, "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": -0.11696, "trunc_abs_frac": 0.43877, "mean_delta_full": -0.21379, "read_hit_rate": {"surface": 0.82131}, "mean_gate": {"surface": 0.23967}, "n_reads": 5249, "n_cand": 6391, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 5249, "inventory": {"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": 50.99969482421875, "n_util_obs": 147, "util_count_spearman": 0.14775153925568857}, "probe_overhead_sec_total": 33.251, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.005377, "alpha": 0.03, "t": 159.0} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 169.0} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.87314, "bpb": 2.70237, "chunk_loss": 0.030353, "F": 0.18811, "G": 0.26252, "chunks": 192.62, "bytes_per_chunk": 5.316, "boundary_entropy": 0.27603, "router_noise": 0.0, "grad_norm": 1.0732, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 0.202, "probe_sec": 0.0265, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01953, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0802, "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": 0.76294, "trunc_abs_frac": 0.559, "mean_delta_full": 0.29248, "read_hit_rate": {"surface": 0.77299}, "mean_gate": {"surface": 0.24614}, "n_reads": 4740, "n_cand": 6132, "read_suppressed_frac": 0.1061, "dropout_reads_logged": 4740, "inventory": {"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": 55.14935302734375, "n_util_obs": 160, "util_count_spearman": 0.08622559930976129}, "probe_overhead_sec_total": 35.909, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.006084, "alpha": 0.03, "t": 169.1} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 178.9} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.77219, "bpb": 2.55673, "chunk_loss": 0.030484, "F": 0.19733, "G": 0.24738, "chunks": 202.06, "bytes_per_chunk": 5.068, "boundary_entropy": 0.22289, "router_noise": 0.0, "grad_norm": 0.9365, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.198, "probe_sec": 0.0295, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02285, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04316, "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": -0.19814, "trunc_abs_frac": 0.05538, "mean_delta_full": 0.02295, "read_hit_rate": {"surface": 0.83043}, "mean_gate": {"surface": 0.25608}, "n_reads": 5343, "n_cand": 6434, "read_suppressed_frac": 0.1059, "dropout_reads_logged": 5343, "inventory": {"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": 59.334503173828125, "n_util_obs": 172, "util_count_spearman": 0.12616775868749802}, "probe_overhead_sec_total": 38.336, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.0069, "alpha": 0.03, "t": 179.0} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 189.4} |
| {"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3542.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203528, "cps": 27.671875, "mean_p": 0.230872, "boundary_entropy": 0.176602, "bpb": 2.404116}, "tr": {"chunks": 3377.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.186544, "cps": 26.382812, "mean_p": 0.28053, "boundary_entropy": 0.35093, "bpb": 2.699936}, "fi": {"chunks": 3497.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.185902, "cps": 27.320312, "mean_p": 0.256931, "boundary_entropy": 0.309278, "bpb": 2.970504}}, "bpb": {"en": 2.40412, "tr": 2.69994, "fi": 2.9705}, "gini_cpc": 0.020402, "gini_cps": 0.010561, "bpic": 5.446621, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9165498117569477, "fi": 0.9133964350003531}, "inventory": {"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": 63.380859375, "n_util_obs": 184, "util_count_spearman": 0.1692427127610663}, "t": 189.9} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.42098, "tr": 2.71205, "fi": 2.9699}, "bytes_per_chunk": {"en": 4.676, "tr": 5.695, "fi": 5.631}, "bpb_mean": 2.70098, "frozen_inventory": true, "t": 191.0} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.78192, "bpb": 2.57077, "chunk_loss": 0.030591, "F": 0.19989, "G": 0.25121, "chunks": 204.69, "bytes_per_chunk": 5.003, "boundary_entropy": 0.23042, "router_noise": 0.0, "grad_norm": 1.013, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.0275, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02137, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.27873, "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": 0.6727, "trunc_abs_frac": 0.74428, "mean_delta_full": 0.63005, "read_hit_rate": {"surface": 0.80868}, "mean_gate": {"surface": 0.26639}, "n_reads": 5271, "n_cand": 6518, "read_suppressed_frac": 0.104, "dropout_reads_logged": 5271, "inventory": {"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": 63.461639404296875, "n_util_obs": 184, "util_count_spearman": 0.17057849010806958}, "probe_overhead_sec_total": 40.712, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.008039, "alpha": 0.03, "t": 191.8} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 201.4} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.71929, "bpb": 2.48041, "chunk_loss": 0.030543, "F": 0.20364, "G": 0.23331, "chunks": 208.53, "bytes_per_chunk": 4.911, "boundary_entropy": 0.17692, "router_noise": 0.0, "grad_norm": 1.2114, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.196, "probe_sec": 0.0282, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02281, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22078, "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": 0.85981}, "mean_gate": {"surface": 0.27477}, "n_reads": 5710, "n_cand": 6641, "read_suppressed_frac": 0.0956, "dropout_reads_logged": 5710, "inventory": {"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": 67.60855102539062, "n_util_obs": 193, "util_count_spearman": 0.15741867098458473}, "probe_overhead_sec_total": 43.34, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.008706, "alpha": 0.03, "t": 201.5} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 211.5} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.69495, "bpb": 2.4453, "chunk_loss": 0.030465, "F": 0.19962, "G": 0.23984, "chunks": 204.41, "bytes_per_chunk": 5.01, "boundary_entropy": 0.19061, "router_noise": 0.0, "grad_norm": 0.9657, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.202, "probe_sec": 0.0402, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03448, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04572, "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": 0.91963, "trunc_abs_frac": 3.12268, "mean_delta_full": -0.02057, "read_hit_rate": {"surface": 0.84683}, "mean_gate": {"surface": 0.27674}, "n_reads": 5512, "n_cand": 6509, "read_suppressed_frac": 0.103, "dropout_reads_logged": 5512, "inventory": {"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": 71.80767822265625, "n_util_obs": 208, "util_count_spearman": 0.187061728263397}, "probe_overhead_sec_total": 46.032, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.009165, "alpha": 0.03, "t": 211.6} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 221.6} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.69803, "bpb": 2.44974, "chunk_loss": 0.030499, "F": 0.20139, "G": 0.23513, "chunks": 206.22, "bytes_per_chunk": 4.966, "boundary_entropy": 0.17477, "router_noise": 0.0, "grad_norm": 1.0232, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.201, "probe_sec": 0.0385, "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": 0.01497, "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": 0.14605, "trunc_abs_frac": 0.34398, "mean_delta_full": -0.19973, "read_hit_rate": {"surface": 0.85595}, "mean_gate": {"surface": 0.28384}, "n_reads": 5621, "n_cand": 6567, "read_suppressed_frac": 0.1028, "dropout_reads_logged": 5621, "inventory": {"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": 75.98173522949219, "n_util_obs": 218, "util_count_spearman": 0.19583614221150605}, "probe_overhead_sec_total": 48.842, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.0105, "alpha": 0.03, "t": 221.7} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 231.4} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3491.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200598, "cps": 27.273438, "mean_p": 0.220994, "boundary_entropy": 0.15161, "bpb": 2.307995}, "tr": {"chunks": 3233.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.178589, "cps": 25.257812, "mean_p": 0.268546, "boundary_entropy": 0.330342, "bpb": 2.618252}, "fi": {"chunks": 3382.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.179788, "cps": 26.421875, "mean_p": 0.243506, "boundary_entropy": 0.27295, "bpb": 2.849546}}, "bpb": {"en": 2.308, "tr": 2.61825, "fi": 2.84955}, "gini_cpc": 0.026249, "gini_cps": 0.01702, "bpic": 5.613695, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8902857551191089, "fi": 0.896264079716761}, "inventory": {"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": 80.08824157714844, "n_util_obs": 225, "util_count_spearman": 0.17052842948191865}, "t": 232.0} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.32228, "tr": 2.62725, "fi": 2.85324}, "bytes_per_chunk": {"en": 4.739, "tr": 6.056, "fi": 5.738}, "bpb_mean": 2.60093, "frozen_inventory": true, "t": 232.9} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.70758, "bpb": 2.46351, "chunk_loss": 0.030471, "F": 0.19995, "G": 0.23691, "chunks": 204.75, "bytes_per_chunk": 5.001, "boundary_entropy": 0.18332, "router_noise": 0.0, "grad_norm": 0.8857, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 0.245, "probe_sec": 0.0288, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02081, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0925, "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": 1.23118, "trunc_abs_frac": 1.06326, "mean_delta_full": 0.30273, "read_hit_rate": {"surface": 0.83129}, "mean_gate": {"surface": 0.29079}, "n_reads": 5420, "n_cand": 6520, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 5420, "inventory": {"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": 80.1722412109375, "n_util_obs": 225, "util_count_spearman": 0.17114333830089581}, "probe_overhead_sec_total": 51.435, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.0117, "alpha": 0.03, "t": 233.9} |
| {"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3470.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199391, "cps": 27.109375, "mean_p": 0.220532, "boundary_entropy": 0.149772, "bpb": 2.303791}, "tr": {"chunks": 3129.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.172844, "cps": 24.445312, "mean_p": 0.260827, "boundary_entropy": 0.32367, "bpb": 2.58962}, "fi": {"chunks": 3391.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.180267, "cps": 26.492188, "mean_p": 0.24209, "boundary_entropy": 0.273303, "bpb": 2.820313}}, "bpb": {"en": 2.30379, "tr": 2.58962, "fi": 2.82031}, "gini_cpc": 0.032032, "gini_cps": 0.022756, "bpic": 5.678879, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8668613844474008, "fi": 0.9040876812374634}, "inventory": {"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": 82.09236145019531, "n_util_obs": 234, "util_count_spearman": 0.1432306349405714}, "t": 239.4} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 245.1} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.72675, "bpb": 2.49118, "chunk_loss": 0.030376, "F": 0.19501, "G": 0.23368, "chunks": 199.69, "bytes_per_chunk": 5.128, "boundary_entropy": 0.18748, "router_noise": 0.0, "grad_norm": 0.9691, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.227, "probe_sec": 0.0274, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02173, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01848, "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": -0.76247, "trunc_abs_frac": 1.6419, "mean_delta_full": 0.09482, "read_hit_rate": {"surface": 0.81409}, "mean_gate": {"surface": 0.28154}, "n_reads": 5176, "n_cand": 6358, "read_suppressed_frac": 0.096, "dropout_reads_logged": 5176, "inventory": {"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": 84.34916687011719, "n_util_obs": 241, "util_count_spearman": 0.1465005257452407}, "probe_overhead_sec_total": 54.047, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.013257, "alpha": 0.03, "t": 245.3} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 255.2} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.71033, "bpb": 2.46749, "chunk_loss": 0.030519, "F": 0.20383, "G": 0.23315, "chunks": 208.72, "bytes_per_chunk": 4.906, "boundary_entropy": 0.1655, "router_noise": 0.0, "grad_norm": 1.0199, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.202, "probe_sec": 0.0256, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01935, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16335, "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": 0.41441, "trunc_abs_frac": 0.60772, "mean_delta_full": 0.2061, "read_hit_rate": {"surface": 0.83842}, "mean_gate": {"surface": 0.28549}, "n_reads": 5573, "n_cand": 6647, "read_suppressed_frac": 0.0942, "dropout_reads_logged": 5573, "inventory": {"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": 88.56027221679688, "n_util_obs": 248, "util_count_spearman": 0.16492992861874048}, "probe_overhead_sec_total": 56.478, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.012702, "alpha": 0.03, "t": 255.4} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 265.5} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.71882, "bpb": 2.47974, "chunk_loss": 0.030522, "F": 0.19897, "G": 0.24324, "chunks": 203.75, "bytes_per_chunk": 5.026, "boundary_entropy": 0.20105, "router_noise": 0.0, "grad_norm": 0.8682, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.205, "probe_sec": 0.0363, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03091, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03625, "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": 0.6898, "trunc_abs_frac": 0.93177, "mean_delta_full": -0.44425, "read_hit_rate": {"surface": 0.80672}, "mean_gate": {"surface": 0.29366}, "n_reads": 5234, "n_cand": 6488, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 5234, "inventory": {"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": 92.7939453125, "n_util_obs": 257, "util_count_spearman": 0.1621039534402932}, "probe_overhead_sec_total": 59.018, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.013146, "alpha": 0.03, "t": 265.6} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 274.9} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3491.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200598, "cps": 27.273438, "mean_p": 0.21631, "boundary_entropy": 0.135428, "bpb": 2.246193}, "tr": {"chunks": 3195.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.17649, "cps": 24.960938, "mean_p": 0.255212, "boundary_entropy": 0.302304, "bpb": 2.534878}, "fi": {"chunks": 3346.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.177875, "cps": 26.140625, "mean_p": 0.232279, "boundary_entropy": 0.245938, "bpb": 2.738661}}, "bpb": {"en": 2.24619, "tr": 2.53488, "fi": 2.73866}, "gini_cpc": 0.02896, "gini_cps": 0.01967, "bpic": 5.655104, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.879821524158847, "fi": 0.8867237169521828}, "inventory": {"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": 96.89503479003906, "n_util_obs": 269, "util_count_spearman": 0.1771681176423166}, "t": 275.5} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.26019, "tr": 2.54542, "fi": 2.74417}, "bytes_per_chunk": {"en": 4.738, "tr": 6.089, "fi": 5.747}, "bpb_mean": 2.51659, "frozen_inventory": true, "t": 276.4} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62281, "bpb": 2.34122, "chunk_loss": 0.030431, "F": 0.19827, "G": 0.23449, "chunks": 203.03, "bytes_per_chunk": 5.044, "boundary_entropy": 0.17438, "router_noise": 0.0, "grad_norm": 0.8998, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 0.236, "probe_sec": 0.0801, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07292, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01247, "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": 0.51082, "trunc_abs_frac": 0.70522, "mean_delta_full": -0.69373, "read_hit_rate": {"surface": 0.84114}, "mean_gate": {"surface": 0.30184}, "n_reads": 5438, "n_cand": 6465, "read_suppressed_frac": 0.1041, "dropout_reads_logged": 5438, "inventory": {"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": 96.97927856445312, "n_util_obs": 270, "util_count_spearman": 0.17759229520817363}, "probe_overhead_sec_total": 61.7, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.012178, "alpha": 0.03, "t": 277.4} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 287.0} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.66336, "bpb": 2.39972, "chunk_loss": 0.030417, "F": 0.1965, "G": 0.23405, "chunks": 201.22, "bytes_per_chunk": 5.089, "boundary_entropy": 0.17557, "router_noise": 0.0, "grad_norm": 0.9195, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.197, "probe_sec": 0.0839, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07881, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18748, "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": 1.10087, "trunc_abs_frac": 0.75057, "mean_delta_full": 0.755, "read_hit_rate": {"surface": 0.81848}, "mean_gate": {"surface": 0.29887}, "n_reads": 5244, "n_cand": 6407, "read_suppressed_frac": 0.1053, "dropout_reads_logged": 5244, "inventory": {"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": 101.15817260742188, "n_util_obs": 273, "util_count_spearman": 0.1896654505638215}, "probe_overhead_sec_total": 64.333, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.013399, "alpha": 0.03, "t": 287.3} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 297.1} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58315, "bpb": 2.284, "chunk_loss": 0.030486, "F": 0.2034, "G": 0.23038, "chunks": 208.28, "bytes_per_chunk": 4.916, "boundary_entropy": 0.14554, "router_noise": 0.0, "grad_norm": 0.8364, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.199, "probe_sec": 0.0319, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02538, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.31778, "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": 1.61443, "trunc_abs_frac": 6.69021, "mean_delta_full": 0.31942, "read_hit_rate": {"surface": 0.8616}, "mean_gate": {"surface": 0.31054}, "n_reads": 5715, "n_cand": 6633, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 5715, "inventory": {"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": 105.36965942382812, "n_util_obs": 283, "util_count_spearman": 0.165557572918586}, "probe_overhead_sec_total": 67.201, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.012948, "alpha": 0.03, "t": 297.3} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 307.0} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6033, "bpb": 2.31307, "chunk_loss": 0.030495, "F": 0.2048, "G": 0.22434, "chunks": 209.72, "bytes_per_chunk": 4.883, "boundary_entropy": 0.13087, "router_noise": 0.0, "grad_norm": 0.9127, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.198, "probe_sec": 0.0284, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01854, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00876, "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": -0.90802, "trunc_abs_frac": 0.92335, "mean_delta_full": -0.19626, "read_hit_rate": {"surface": 0.8657}, "mean_gate": {"surface": 0.30945}, "n_reads": 5782, "n_cand": 6679, "read_suppressed_frac": 0.1019, "dropout_reads_logged": 5782, "inventory": {"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": 109.62045288085938, "n_util_obs": 294, "util_count_spearman": 0.18321762759920848}, "probe_overhead_sec_total": 69.854, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.018817, "alpha": 0.03, "t": 307.2} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 317.1} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3524.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202494, "cps": 27.53125, "mean_p": 0.214785, "boundary_entropy": 0.118086, "bpb": 2.187496}, "tr": {"chunks": 3242.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.179086, "cps": 25.328125, "mean_p": 0.254145, "boundary_entropy": 0.294876, "bpb": 2.463513}, "fi": {"chunks": 3558.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.189145, "cps": 27.796875, "mean_p": 0.230879, "boundary_entropy": 0.224354, "bpb": 2.664038}}, "bpb": {"en": 2.1875, "tr": 2.46351, "fi": 2.66404}, "gini_cpc": 0.027342, "gini_cps": 0.020406, "bpic": 5.495157, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8844039621178923, "fi": 0.9340761446182111}, "inventory": {"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": 113.75341796875, "n_util_obs": 305, "util_count_spearman": 0.199113374914741}, "t": 317.7} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.20168, "tr": 2.47273, "fi": 2.66895}, "bytes_per_chunk": {"en": 4.702, "tr": 5.957, "fi": 5.486}, "bpb_mean": 2.44779, "frozen_inventory": true, "t": 318.6} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57461, "bpb": 2.27168, "chunk_loss": 0.030442, "F": 0.20264, "G": 0.22788, "chunks": 207.5, "bytes_per_chunk": 4.935, "boundary_entropy": 0.1486, "router_noise": 0.0, "grad_norm": 0.8975, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 0.251, "probe_sec": 0.0248, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01836, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03188, "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": 0.99971, "trunc_abs_frac": 0.94818, "mean_delta_full": 0.36292, "read_hit_rate": {"surface": 0.85775}, "mean_gate": {"surface": 0.30547}, "n_reads": 5668, "n_cand": 6608, "read_suppressed_frac": 0.0942, "dropout_reads_logged": 5668, "inventory": {"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": 113.84120178222656, "n_util_obs": 305, "util_count_spearman": 0.19948673165057765}, "probe_overhead_sec_total": 72.618, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.018363, "alpha": 0.03, "t": 319.7} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 331.1} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55143, "bpb": 2.23824, "chunk_loss": 0.030497, "F": 0.20212, "G": 0.23332, "chunks": 206.97, "bytes_per_chunk": 4.948, "boundary_entropy": 0.157, "router_noise": 0.0, "grad_norm": 0.8449, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.232, "probe_sec": 0.0421, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03465, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23557, "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": 1.84018, "trunc_abs_frac": 0.9615, "mean_delta_full": 0.13908, "read_hit_rate": {"surface": 0.85693}, "mean_gate": {"surface": 0.31265}, "n_reads": 5648, "n_cand": 6591, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 5648, "inventory": {"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": 118.03877258300781, "n_util_obs": 318, "util_count_spearman": 0.21068163219155522}, "probe_overhead_sec_total": 74.899, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01649, "alpha": 0.03, "t": 331.3} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 345.8} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54633, "bpb": 2.23088, "chunk_loss": 0.03038, "F": 0.19879, "G": 0.22115, "chunks": 203.56, "bytes_per_chunk": 5.03, "boundary_entropy": 0.13483, "router_noise": 0.0, "grad_norm": 1.0168, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.308, "probe_sec": 0.034, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02376, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00408, "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": 0.17956, "trunc_abs_frac": 0.17956, "mean_delta_full": -0.44849, "read_hit_rate": {"surface": 0.8556}, "mean_gate": {"surface": 0.30219}, "n_reads": 5546, "n_cand": 6482, "read_suppressed_frac": 0.0893, "dropout_reads_logged": 5546, "inventory": {"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": 122.24969482421875, "n_util_obs": 325, "util_count_spearman": 0.2092119838201237}, "probe_overhead_sec_total": 77.496, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.015736, "alpha": 0.03, "t": 346.7} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 363.5} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53638, "bpb": 2.21652, "chunk_loss": 0.030517, "F": 0.20514, "G": 0.22766, "chunks": 210.06, "bytes_per_chunk": 4.875, "boundary_entropy": 0.13548, "router_noise": 0.0, "grad_norm": 0.8211, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.341, "probe_sec": 0.0334, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02596, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.38, "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": 0.65307, "trunc_abs_frac": 0.87503, "mean_delta_full": -0.39535, "read_hit_rate": {"surface": 0.8562}, "mean_gate": {"surface": 0.31456}, "n_reads": 5728, "n_cand": 6690, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 5728, "inventory": {"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": 126.51556396484375, "n_util_obs": 336, "util_count_spearman": 0.17826872432081614}, "probe_overhead_sec_total": 80.049, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01624, "alpha": 0.03, "t": 363.8} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 381.8} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3461.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198874, "cps": 27.039062, "mean_p": 0.210302, "boundary_entropy": 0.106734, "bpb": 2.152123}, "tr": {"chunks": 3188.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.176103, "cps": 24.90625, "mean_p": 0.244475, "boundary_entropy": 0.278853, "bpb": 2.415213}, "fi": {"chunks": 3450.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.183403, "cps": 26.953125, "mean_p": 0.228823, "boundary_entropy": 0.219325, "bpb": 2.599096}}, "bpb": {"en": 2.15212, "tr": 2.41521, "fi": 2.5991}, "gini_cpc": 0.027186, "gini_cps": 0.018022, "bpic": 5.617586, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8855035002044466, "fi": 0.9222097990179566}, "inventory": {"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": 130.66452026367188, "n_util_obs": 343, "util_count_spearman": 0.1875240826333593}, "t": 382.3} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.16788, "tr": 2.42685, "fi": 2.6078}, "bytes_per_chunk": {"en": 4.778, "tr": 6.057, "fi": 5.628}, "bpb_mean": 2.40084, "frozen_inventory": true, "t": 383.1} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51685, "bpb": 2.18835, "chunk_loss": 0.030347, "F": 0.19583, "G": 0.22673, "chunks": 200.53, "bytes_per_chunk": 5.106, "boundary_entropy": 0.15131, "router_noise": 0.0, "grad_norm": 0.8425, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.416, "probe_sec": 0.0359, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02289, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04539, "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": -3.53589, "trunc_abs_frac": 1.06136, "mean_delta_full": -0.05896, "read_hit_rate": {"surface": 0.84558}, "mean_gate": {"surface": 0.30751}, "n_reads": 5399, "n_cand": 6385, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 5399, "inventory": {"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": 130.7479705810547, "n_util_obs": 343, "util_count_spearman": 0.1872861974969932}, "probe_overhead_sec_total": 82.744, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014756, "alpha": 0.03, "t": 384.6} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 395.3} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56828, "bpb": 2.26255, "chunk_loss": 0.030423, "F": 0.202, "G": 0.22317, "chunks": 206.84, "bytes_per_chunk": 4.951, "boundary_entropy": 0.13376, "router_noise": 0.0, "grad_norm": 0.769, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.218, "probe_sec": 0.0346, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02808, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.33831, "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": 0.22586, "trunc_abs_frac": 1.35353, "mean_delta_full": -0.18014, "read_hit_rate": {"surface": 0.84894}, "mean_gate": {"surface": 0.30765}, "n_reads": 5592, "n_cand": 6587, "read_suppressed_frac": 0.1055, "dropout_reads_logged": 5592, "inventory": {"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": 134.96925354003906, "n_util_obs": 355, "util_count_spearman": 0.1804203736542037}, "probe_overhead_sec_total": 85.57, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.020007, "alpha": 0.03, "t": 395.5} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 405.9} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56507, "bpb": 2.25792, "chunk_loss": 0.030346, "F": 0.19617, "G": 0.22361, "chunks": 200.88, "bytes_per_chunk": 5.098, "boundary_entropy": 0.14357, "router_noise": 0.0, "grad_norm": 0.8844, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.214, "probe_sec": 0.0538, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04599, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02422, "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": 0.69371, "trunc_abs_frac": 41.70243, "mean_delta_full": -0.22759, "read_hit_rate": {"surface": 0.8324}, "mean_gate": {"surface": 0.31209}, "n_reads": 5324, "n_cand": 6396, "read_suppressed_frac": 0.0913, "dropout_reads_logged": 5324, "inventory": {"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": 139.21914672851562, "n_util_obs": 366, "util_count_spearman": 0.19038283180980817}, "probe_overhead_sec_total": 88.085, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018407, "alpha": 0.03, "t": 406.2} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 423.8} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5351, "bpb": 2.21468, "chunk_loss": 0.03049, "F": 0.20456, "G": 0.22516, "chunks": 209.47, "bytes_per_chunk": 4.889, "boundary_entropy": 0.12209, "router_noise": 0.0, "grad_norm": 0.8094, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.359, "probe_sec": 0.036, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02362, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.38706, "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": 0.72505, "trunc_abs_frac": 1.06231, "mean_delta_full": 0.53076, "read_hit_rate": {"surface": 0.86329}, "mean_gate": {"surface": 0.31969}, "n_reads": 5759, "n_cand": 6671, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 5759, "inventory": {"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": 143.4713592529297, "n_util_obs": 377, "util_count_spearman": 0.1895888016790364}, "probe_overhead_sec_total": 90.891, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016578, "alpha": 0.03, "t": 424.1} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 441.6} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3478.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199851, "cps": 27.171875, "mean_p": 0.209396, "boundary_entropy": 0.102249, "bpb": 2.11251}, "tr": {"chunks": 3545.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195824, "cps": 27.695312, "mean_p": 0.248665, "boundary_entropy": 0.26335, "bpb": 2.357348}, "fi": {"chunks": 3477.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.184839, "cps": 27.164062, "mean_p": 0.223035, "boundary_entropy": 0.200911, "bpb": 2.566116}}, "bpb": {"en": 2.11251, "tr": 2.35735, "fi": 2.56612}, "gini_cpc": 0.01724, "gini_cps": 0.004317, "bpic": 5.403048, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9798514296681404, "fi": 0.9248841774610423}, "inventory": {"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": 147.60162353515625, "n_util_obs": 388, "util_count_spearman": 0.17108895830772541}, "t": 442.1} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.12926, "tr": 2.36705, "fi": 2.57468}, "bytes_per_chunk": {"en": 4.765, "tr": 5.581, "fi": 5.566}, "bpb_mean": 2.357, "frozen_inventory": true, "t": 443.2} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58589, "bpb": 2.28796, "chunk_loss": 0.030478, "F": 0.20261, "G": 0.22886, "chunks": 207.47, "bytes_per_chunk": 4.936, "boundary_entropy": 0.1439, "router_noise": 0.0, "grad_norm": 0.8326, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.402, "probe_sec": 0.0385, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03196, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.40439, "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": 0.82428}, "mean_gate": {"surface": 0.30973}, "n_reads": 5446, "n_cand": 6607, "read_suppressed_frac": 0.1039, "dropout_reads_logged": 5446, "inventory": {"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": 147.68092346191406, "n_util_obs": 389, "util_count_spearman": 0.1666829840920313}, "probe_overhead_sec_total": 93.457, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.022601, "alpha": 0.03, "t": 444.2} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 455.7} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52928, "bpb": 2.20628, "chunk_loss": 0.030449, "F": 0.20197, "G": 0.22416, "chunks": 206.81, "bytes_per_chunk": 4.951, "boundary_entropy": 0.13243, "router_noise": 0.0, "grad_norm": 0.7485, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.235, "probe_sec": 0.0894, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08359, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24873, "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": 1.03678, "trunc_abs_frac": 1.03678, "mean_delta_full": 1.15491, "read_hit_rate": {"surface": 0.84634}, "mean_gate": {"surface": 0.31028}, "n_reads": 5574, "n_cand": 6586, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 5574, "inventory": {"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": 151.8953857421875, "n_util_obs": 405, "util_count_spearman": 0.17026368781791926}, "probe_overhead_sec_total": 96.298, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018048, "alpha": 0.03, "t": 456.0} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 465.9} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57685, "bpb": 2.27491, "chunk_loss": 0.030433, "F": 0.20245, "G": 0.22171, "chunks": 207.31, "bytes_per_chunk": 4.939, "boundary_entropy": 0.12277, "router_noise": 0.0, "grad_norm": 0.7466, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.205, "probe_sec": 0.0811, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07136, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.11755, "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": 0.00513, "trunc_abs_frac": 0.72557, "mean_delta_full": -0.11309, "read_hit_rate": {"surface": 0.84444}, "mean_gate": {"surface": 0.30702}, "n_reads": 5575, "n_cand": 6602, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 5575, "inventory": {"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": 156.1011505126953, "n_util_obs": 414, "util_count_spearman": 0.17909335764986903}, "probe_overhead_sec_total": 98.867, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.017021, "alpha": 0.03, "t": 466.2} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 475.9} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53733, "bpb": 2.21789, "chunk_loss": 0.030429, "F": 0.20111, "G": 0.21956, "chunks": 205.94, "bytes_per_chunk": 4.972, "boundary_entropy": 0.11994, "router_noise": 0.0, "grad_norm": 0.7707, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.199, "probe_sec": 0.0551, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04575, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04729, "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": -24.725, "trunc_abs_frac": 16.08096, "mean_delta_full": 0.00081, "read_hit_rate": {"surface": 0.84324}, "mean_gate": {"surface": 0.31383}, "n_reads": 5530, "n_cand": 6558, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 5530, "inventory": {"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": 160.34532165527344, "n_util_obs": 425, "util_count_spearman": 0.18756877873739605}, "probe_overhead_sec_total": 101.394, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019753, "alpha": 0.03, "t": 476.1} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 485.6} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3464.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199046, "cps": 27.0625, "mean_p": 0.206511, "boundary_entropy": 0.094461, "bpb": 2.085289}, "tr": {"chunks": 3245.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.179252, "cps": 25.351562, "mean_p": 0.239096, "boundary_entropy": 0.255623, "bpb": 2.334918}, "fi": {"chunks": 3454.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.183616, "cps": 26.984375, "mean_p": 0.22082, "boundary_entropy": 0.195428, "bpb": 2.50192}}, "bpb": {"en": 2.08529, "tr": 2.33492, "fi": 2.50192}, "gini_cpc": 0.023484, "gini_cps": 0.014366, "bpic": 5.58221, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9005553001244226, "fi": 0.9224794212221412}, "inventory": {"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": 164.48231506347656, "n_util_obs": 437, "util_count_spearman": 0.19204450179089735}, "t": 486.2} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.10266, "tr": 2.34271, "fi": 2.51159}, "bytes_per_chunk": {"en": 4.774, "tr": 5.984, "fi": 5.573}, "bpb_mean": 2.31899, "frozen_inventory": true, "t": 487.2} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55195, "bpb": 2.239, "chunk_loss": 0.030372, "F": 0.19864, "G": 0.22084, "chunks": 203.41, "bytes_per_chunk": 5.034, "boundary_entropy": 0.12811, "router_noise": 0.0, "grad_norm": 0.8018, "lr": 0.000245, "router_trainable": true, "sec_per_step": 0.243, "probe_sec": 0.1604, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1541, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.4835, "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": 1.05207, "trunc_abs_frac": 1.08964, "mean_delta_full": 0.66583, "read_hit_rate": {"surface": 0.83341}, "mean_gate": {"surface": 0.31798}, "n_reads": 5398, "n_cand": 6477, "read_suppressed_frac": 0.1058, "dropout_reads_logged": 5398, "inventory": {"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": 164.56556701660156, "n_util_obs": 437, "util_count_spearman": 0.1919753952065513}, "probe_overhead_sec_total": 104.142, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.021903, "alpha": 0.03, "t": 488.3} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3463.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198989, "cps": 27.054688, "mean_p": 0.207443, "boundary_entropy": 0.09235, "bpb": 2.082028}, "tr": {"chunks": 3240.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.178976, "cps": 25.3125, "mean_p": 0.237854, "boundary_entropy": 0.252543, "bpb": 2.33113}, "fi": {"chunks": 3447.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.183244, "cps": 26.929688, "mean_p": 0.218903, "boundary_entropy": 0.189439, "bpb": 2.494654}}, "bpb": {"en": 2.08203, "tr": 2.33113, "fi": 2.49465}, "gini_cpc": 0.023773, "gini_cps": 0.014647, "bpic": 5.58936, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8994273455823719, "fi": 0.9208757331408853}, "inventory": {"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": 168.53228759765625, "n_util_obs": 446, "util_count_spearman": 0.19249764447941123}, "t": 498.3} |
| {"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 500.5} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60612, "bpb": 2.31713, "chunk_loss": 0.030357, "F": 0.19614, "G": 0.22265, "chunks": 200.84, "bytes_per_chunk": 5.098, "boundary_entropy": 0.13563, "router_noise": 0.0, "grad_norm": 0.8739, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 0.248, "probe_sec": 0.0447, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03698, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04283, "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": -0.80442, "trunc_abs_frac": 0.80442, "mean_delta_full": -0.01794, "read_hit_rate": {"surface": 0.81579}, "mean_gate": {"surface": 0.30684}, "n_reads": 5217, "n_cand": 6395, "read_suppressed_frac": 0.105, "dropout_reads_logged": 5217, "inventory": {"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": 168.7769317626953, "n_util_obs": 447, "util_count_spearman": 0.1962229839156009}, "probe_overhead_sec_total": 106.754, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.02563, "alpha": 0.03, "t": 500.7} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 510.5} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49425, "bpb": 2.15575, "chunk_loss": 0.030442, "F": 0.20264, "G": 0.2243, "chunks": 207.5, "bytes_per_chunk": 4.935, "boundary_entropy": 0.12359, "router_noise": 0.0, "grad_norm": 0.7385, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 0.2, "probe_sec": 0.0274, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02011, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07918, "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": -0.99873, "trunc_abs_frac": 1.01491, "mean_delta_full": -0.21704, "read_hit_rate": {"surface": 0.85427}, "mean_gate": {"surface": 0.31539}, "n_reads": 5645, "n_cand": 6608, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 5645, "inventory": {"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": 172.9785919189453, "n_util_obs": 458, "util_count_spearman": 0.19270341105227148}, "probe_overhead_sec_total": 109.388, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019902, "alpha": 0.03, "t": 510.7} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 520.8} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51227, "bpb": 2.18174, "chunk_loss": 0.030318, "F": 0.19623, "G": 0.21546, "chunks": 200.94, "bytes_per_chunk": 5.096, "boundary_entropy": 0.11524, "router_noise": 0.0, "grad_norm": 0.7261, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.209, "probe_sec": 0.0671, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05798, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24291, "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": 0.86069, "trunc_abs_frac": 0.85844, "mean_delta_full": 0.72729, "read_hit_rate": {"surface": 0.83823}, "mean_gate": {"surface": 0.30763}, "n_reads": 5363, "n_cand": 6398, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 5363, "inventory": {"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": 177.18597412109375, "n_util_obs": 474, "util_count_spearman": 0.20292087776648884}, "probe_overhead_sec_total": 112.406, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020822, "alpha": 0.03, "t": 521.1} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 530.8} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3471.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199448, "cps": 27.117188, "mean_p": 0.206327, "boundary_entropy": 0.087669, "bpb": 2.055152}, "tr": {"chunks": 3276.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.180964, "cps": 25.59375, "mean_p": 0.232794, "boundary_entropy": 0.238232, "bpb": 2.286534}, "fi": {"chunks": 3328.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.176918, "cps": 26.0, "mean_p": 0.207877, "boundary_entropy": 0.168197, "bpb": 2.46831}}, "bpb": {"en": 2.05515, "tr": 2.28653, "fi": 2.46831}, "gini_cpc": 0.026951, "gini_cps": 0.012903, "bpic": 5.630968, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.90732493900384, "fi": 0.8870353767428693}, "inventory": {"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": 181.32159423828125, "n_util_obs": 481, "util_count_spearman": 0.19464255218027865}, "t": 531.5} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.07311, "tr": 2.29563, "fi": 2.47806}, "bytes_per_chunk": {"en": 4.772, "tr": 5.875, "fi": 5.714}, "bpb_mean": 2.28227, "frozen_inventory": true, "t": 532.5} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.4795, "bpb": 2.13447, "chunk_loss": 0.030416, "F": 0.20242, "G": 0.21797, "chunks": 207.28, "bytes_per_chunk": 4.94, "boundary_entropy": 0.1054, "router_noise": 0.0, "grad_norm": 0.6908, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 0.248, "probe_sec": 0.0463, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03984, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06589, "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": 0.73251, "trunc_abs_frac": 1.51006, "mean_delta_full": 0.36169, "read_hit_rate": {"surface": 0.85426}, "mean_gate": {"surface": 0.31491}, "n_reads": 5639, "n_cand": 6601, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 5639, "inventory": {"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": 181.40859985351562, "n_util_obs": 482, "util_count_spearman": 0.19327932156832411}, "probe_overhead_sec_total": 115.095, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020962, "alpha": 0.03, "t": 533.5} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 543.5} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51154, "bpb": 2.18069, "chunk_loss": 0.030368, "F": 0.1987, "G": 0.21862, "chunks": 203.47, "bytes_per_chunk": 5.033, "boundary_entropy": 0.11898, "router_noise": 0.0, "grad_norm": 0.7412, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.205, "probe_sec": 0.1053, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09893, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.3645, "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": 0.52826, "trunc_abs_frac": 8.54195, "mean_delta_full": 0.60368, "read_hit_rate": {"surface": 0.84658}, "mean_gate": {"surface": 0.323}, "n_reads": 5485, "n_cand": 6479, "read_suppressed_frac": 0.1041, "dropout_reads_logged": 5485, "inventory": {"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": 185.63157653808594, "n_util_obs": 489, "util_count_spearman": 0.2113094505817416}, "probe_overhead_sec_total": 117.823, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.022707, "alpha": 0.03, "t": 543.8} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 553.6} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49454, "bpb": 2.15616, "chunk_loss": 0.03044, "F": 0.20428, "G": 0.21947, "chunks": 209.19, "bytes_per_chunk": 4.895, "boundary_entropy": 0.10633, "router_noise": 0.0, "grad_norm": 0.7701, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.2, "probe_sec": 0.0323, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02433, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09955, "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": -1.10714, "trunc_abs_frac": 2.65131, "mean_delta_full": 0.0012, "read_hit_rate": {"surface": 0.86836}, "mean_gate": {"surface": 0.32182}, "n_reads": 5785, "n_cand": 6662, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 5785, "inventory": {"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": 189.8441619873047, "n_util_obs": 505, "util_count_spearman": 0.21700167739668175}, "probe_overhead_sec_total": 120.711, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.027733, "alpha": 0.03, "t": 553.8} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 564.3} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49081, "bpb": 2.15079, "chunk_loss": 0.030277, "F": 0.19128, "G": 0.2216, "chunks": 195.88, "bytes_per_chunk": 5.228, "boundary_entropy": 0.14199, "router_noise": 0.0, "grad_norm": 0.7635, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.217, "probe_sec": 0.0779, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06973, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24578, "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": 1.05211, "trunc_abs_frac": 1.05211, "mean_delta_full": -0.29749, "read_hit_rate": {"surface": 0.81863}, "mean_gate": {"surface": 0.31751}, "n_reads": 5105, "n_cand": 6236, "read_suppressed_frac": 0.1028, "dropout_reads_logged": 5105, "inventory": {"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": 194.0540008544922, "n_util_obs": 515, "util_count_spearman": 0.21870747319181644}, "probe_overhead_sec_total": 123.866, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.023904, "alpha": 0.03, "t": 564.6} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 575.8} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3476.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199736, "cps": 27.15625, "mean_p": 0.206618, "boundary_entropy": 0.08719, "bpb": 2.035333}, "tr": {"chunks": 3328.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.183837, "cps": 26.0, "mean_p": 0.236466, "boundary_entropy": 0.239562, "bpb": 2.253822}, "fi": {"chunks": 3557.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.189091, "cps": 27.789062, "mean_p": 0.220874, "boundary_entropy": 0.183844, "bpb": 2.435445}}, "bpb": {"en": 2.03533, "tr": 2.25382, "fi": 2.43545}, "gini_cpc": 0.018509, "gini_cps": 0.014735, "bpic": 5.475533, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9204010779132604, "fi": 0.9467086258505432}, "inventory": {"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": 198.19419860839844, "n_util_obs": 528, "util_count_spearman": 0.21799700292279034}, "t": 576.5} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.05179, "tr": 2.26192, "fi": 2.44724}, "bytes_per_chunk": {"en": 4.775, "tr": 5.827, "fi": 5.421}, "bpb_mean": 2.25365, "frozen_inventory": true, "t": 577.4} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 578.4} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44685, "bpb": 2.08736, "chunk_loss": 0.030522, "F": 0.20798, "G": 0.22343, "chunks": 212.97, "bytes_per_chunk": 4.808, "boundary_entropy": 0.10862, "router_noise": 0.0, "grad_norm": 0.7393, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.281, "probe_sec": 0.039, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02411, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08788, "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": -0.31461, "trunc_abs_frac": 0.7118, "mean_delta_full": 0.37642, "read_hit_rate": {"surface": 0.87498}, "mean_gate": {"surface": 0.3155}, "n_reads": 5935, "n_cand": 6783, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 5935, "inventory": {"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": 198.28562927246094, "n_util_obs": 528, "util_count_spearman": 0.21777404066414457}, "probe_overhead_sec_total": 126.649, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.028696, "alpha": 0.03, "t": 578.7} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 588.6} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.63416, "bpb": 2.35759, "chunk_loss": 0.030501, "F": 0.20648, "G": 0.22378, "chunks": 211.44, "bytes_per_chunk": 4.843, "boundary_entropy": 0.10994, "router_noise": 0.0, "grad_norm": 0.7391, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.204, "probe_sec": 0.0888, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.081, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.18051, "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": 0.7962, "trunc_abs_frac": 0.9734, "mean_delta_full": -0.42485, "read_hit_rate": {"surface": 0.81838}, "mean_gate": {"surface": 0.31014}, "n_reads": 5511, "n_cand": 6734, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 5511, "inventory": {"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": 202.53919982910156, "n_util_obs": 540, "util_count_spearman": 0.19641198186108916}, "probe_overhead_sec_total": 129.183, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.021822, "alpha": 0.03, "t": 588.9} |
| {"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 598.7} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48758, "bpb": 2.14613, "chunk_loss": 0.030386, "F": 0.20126, "G": 0.21551, "chunks": 206.09, "bytes_per_chunk": 4.969, "boundary_entropy": 0.10175, "router_noise": 0.0, "grad_norm": 0.7319, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.202, "probe_sec": 0.0821, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06287, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24129, "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": 1.46203, "trunc_abs_frac": 2.75303, "mean_delta_full": 0.4576, "read_hit_rate": {"surface": 0.85571}, "mean_gate": {"surface": 0.31707}, "n_reads": 5616, "n_cand": 6563, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 5616, "inventory": {"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": 206.77285766601562, "n_util_obs": 554, "util_count_spearman": 0.2166141700232834}, "probe_overhead_sec_total": 131.939, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.022192, "alpha": 0.03, "t": 599.0} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 609.5} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47745, "bpb": 2.13151, "chunk_loss": 0.03044, "F": 0.20239, "G": 0.22349, "chunks": 207.25, "bytes_per_chunk": 4.941, "boundary_entropy": 0.12747, "router_noise": 0.0, "grad_norm": 0.7103, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.215, "probe_sec": 0.0596, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05369, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03344, "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": 1.41878, "trunc_abs_frac": 0.81056, "mean_delta_full": -0.09365, "read_hit_rate": {"surface": 0.835}, "mean_gate": {"surface": 0.3262}, "n_reads": 5511, "n_cand": 6600, "read_suppressed_frac": 0.0933, "dropout_reads_logged": 5511, "inventory": {"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": 211.03143310546875, "n_util_obs": 559, "util_count_spearman": 0.22481285066081574}, "probe_overhead_sec_total": 134.651, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.021713, "alpha": 0.03, "t": 609.7} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 619.6} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3449.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198184, "cps": 26.945312, "mean_p": 0.203054, "boundary_entropy": 0.081201, "bpb": 2.019694}, "tr": {"chunks": 3211.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.177374, "cps": 25.085938, "mean_p": 0.228276, "boundary_entropy": 0.231077, "bpb": 2.236445}, "fi": {"chunks": 3352.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.178194, "cps": 26.1875, "mean_p": 0.206505, "boundary_entropy": 0.161453, "bpb": 2.409208}}, "bpb": {"en": 2.01969, "tr": 2.23645, "fi": 2.40921}, "gini_cpc": 0.025054, "gini_cps": 0.015848, "bpic": 5.6664, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8949951460864377, "fi": 0.8991311675699025}, "inventory": {"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": 215.1551971435547, "n_util_obs": 575, "util_count_spearman": 0.23791814697727037}, "t": 620.3} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.03402, "tr": 2.24402, "fi": 2.41902}, "bytes_per_chunk": {"en": 4.801, "tr": 6.039, "fi": 5.683}, "bpb_mean": 2.23235, "frozen_inventory": true, "t": 621.3} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.50719, "bpb": 2.17441, "chunk_loss": 0.030315, "F": 0.19458, "G": 0.21374, "chunks": 199.25, "bytes_per_chunk": 5.139, "boundary_entropy": 0.11404, "router_noise": 0.0, "grad_norm": 0.7708, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 0.261, "probe_sec": 0.0332, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02508, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23048, "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": -1.18517, "trunc_abs_frac": 4.73074, "mean_delta_full": 0.04032, "read_hit_rate": {"surface": 0.82866}, "mean_gate": {"surface": 0.29448}, "n_reads": 5257, "n_cand": 6344, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 5257, "inventory": {"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": 215.232666015625, "n_util_obs": 575, "util_count_spearman": 0.23623355330848436}, "probe_overhead_sec_total": 137.272, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.027622, "alpha": 0.03, "t": 622.8} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 633.5} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46445, "bpb": 2.11276, "chunk_loss": 0.030354, "F": 0.19614, "G": 0.22029, "chunks": 200.84, "bytes_per_chunk": 5.098, "boundary_entropy": 0.12605, "router_noise": 0.0, "grad_norm": 0.7501, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 0.218, "probe_sec": 0.0361, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02977, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09272, "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": 0.14303, "trunc_abs_frac": 2.06097, "mean_delta_full": -0.34062, "read_hit_rate": {"surface": 0.83237}, "mean_gate": {"surface": 0.30958}, "n_reads": 5323, "n_cand": 6395, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 5323, "inventory": {"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": 219.4452362060547, "n_util_obs": 589, "util_count_spearman": 0.2436885558417571}, "probe_overhead_sec_total": 140.281, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.025078, "alpha": 0.03, "t": 633.7} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 643.4} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54167, "bpb": 2.22415, "chunk_loss": 0.030426, "F": 0.20108, "G": 0.22085, "chunks": 205.91, "bytes_per_chunk": 4.973, "boundary_entropy": 0.11502, "router_noise": 0.0, "grad_norm": 0.8655, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.201, "probe_sec": 0.1104, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09765, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14095, "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": 0.42586, "trunc_abs_frac": 0.65046, "mean_delta_full": 0.34407, "read_hit_rate": {"surface": 0.81638}, "mean_gate": {"surface": 0.30751}, "n_reads": 5353, "n_cand": 6557, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 5353, "inventory": {"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": 223.66522216796875, "n_util_obs": 606, "util_count_spearman": 0.2480064831289938}, "probe_overhead_sec_total": 142.985, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.031826, "alpha": 0.03, "t": 643.7} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 653.7} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47286, "bpb": 2.12489, "chunk_loss": 0.030368, "F": 0.1987, "G": 0.21959, "chunks": 203.47, "bytes_per_chunk": 5.033, "boundary_entropy": 0.11431, "router_noise": 0.0, "grad_norm": 0.7073, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.206, "probe_sec": 0.0316, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0222, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06032, "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": 0.84177, "trunc_abs_frac": 1.16249, "mean_delta_full": 0.46802, "read_hit_rate": {"surface": 0.84041}, "mean_gate": {"surface": 0.30566}, "n_reads": 5445, "n_cand": 6479, "read_suppressed_frac": 0.0951, "dropout_reads_logged": 5445, "inventory": {"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": 227.91189575195312, "n_util_obs": 616, "util_count_spearman": 0.23848049616382883}, "probe_overhead_sec_total": 145.758, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.022498, "alpha": 0.03, "t": 654.0} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 663.9} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3489.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200483, "cps": 27.257812, "mean_p": 0.205084, "boundary_entropy": 0.084423, "bpb": 1.997271}, "tr": {"chunks": 3284.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.181406, "cps": 25.65625, "mean_p": 0.226146, "boundary_entropy": 0.218983, "bpb": 2.2148}, "fi": {"chunks": 3575.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.190048, "cps": 27.929688, "mean_p": 0.218111, "boundary_entropy": 0.174846, "bpb": 2.370788}}, "bpb": {"en": 1.99727, "tr": 2.2148, "fi": 2.37079}, "gini_cpc": 0.022236, "gini_cps": 0.018748, "bpic": 5.482412, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9048482437056817, "fi": 0.9479541091034851}, "inventory": {"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": 232.0271759033203, "n_util_obs": 626, "util_count_spearman": 0.23422165222010682}, "t": 664.6} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.01446, "tr": 2.22212, "fi": 2.37945}, "bytes_per_chunk": {"en": 4.758, "tr": 5.919, "fi": 5.338}, "bpb_mean": 2.20535, "frozen_inventory": true, "t": 665.7} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45377, "bpb": 2.09735, "chunk_loss": 0.030455, "F": 0.2049, "G": 0.21857, "chunks": 209.81, "bytes_per_chunk": 4.881, "boundary_entropy": 0.10158, "router_noise": 0.0, "grad_norm": 0.6719, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 0.26, "probe_sec": 0.039, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03113, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.25448, "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": 0.78409, "trunc_abs_frac": 3.59591, "mean_delta_full": -0.8602, "read_hit_rate": {"surface": 0.85319}, "mean_gate": {"surface": 0.31164}, "n_reads": 5701, "n_cand": 6682, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 5701, "inventory": {"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": 232.11517333984375, "n_util_obs": 627, "util_count_spearman": 0.2361363226262388}, "probe_overhead_sec_total": 148.526, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.023724, "alpha": 0.03, "t": 667.0} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 676.8} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54208, "bpb": 2.22475, "chunk_loss": 0.030457, "F": 0.20474, "G": 0.21852, "chunks": 209.66, "bytes_per_chunk": 4.884, "boundary_entropy": 0.1022, "router_noise": 0.0, "grad_norm": 0.7395, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.2, "probe_sec": 0.0299, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02078, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.26169, "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": 1.71201, "trunc_abs_frac": 0.83341, "mean_delta_full": 0.14556, "read_hit_rate": {"surface": 0.8366}, "mean_gate": {"surface": 0.3116}, "n_reads": 5586, "n_cand": 6677, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 5586, "inventory": {"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": 236.31312561035156, "n_util_obs": 637, "util_count_spearman": 0.23132260282982361}, "probe_overhead_sec_total": 150.954, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.030035, "alpha": 0.03, "t": 677.0} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 687.1} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41851, "bpb": 2.04647, "chunk_loss": 0.030337, "F": 0.19705, "G": 0.21866, "chunks": 201.78, "bytes_per_chunk": 5.075, "boundary_entropy": 0.11503, "router_noise": 0.0, "grad_norm": 0.69, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.206, "probe_sec": 0.0246, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01827, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.30523, "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": 1.28981, "trunc_abs_frac": 4.08089, "mean_delta_full": 0.55573, "read_hit_rate": {"surface": 0.847}, "mean_gate": {"surface": 0.31605}, "n_reads": 5442, "n_cand": 6425, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 5442, "inventory": {"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": 240.55252075195312, "n_util_obs": 645, "util_count_spearman": 0.23166455302488595}, "probe_overhead_sec_total": 153.57, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.023441, "alpha": 0.03, "t": 687.3} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 697.2} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35728, "bpb": 1.95814, "chunk_loss": 0.030537, "F": 0.21082, "G": 0.22104, "chunks": 215.88, "bytes_per_chunk": 4.743, "boundary_entropy": 0.08746, "router_noise": 0.0, "grad_norm": 0.6487, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.203, "probe_sec": 0.0645, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05625, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.68055, "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": 1.11571, "trunc_abs_frac": 1.11571, "mean_delta_full": 2.42163, "read_hit_rate": {"surface": 0.89252}, "mean_gate": {"surface": 0.31304}, "n_reads": 6137, "n_cand": 6876, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 6137, "inventory": {"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": 244.80479431152344, "n_util_obs": 652, "util_count_spearman": 0.2206298063060312}, "probe_overhead_sec_total": 156.192, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.031189, "alpha": 0.03, "t": 697.4} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 707.5} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3466.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199161, "cps": 27.078125, "mean_p": 0.203277, "boundary_entropy": 0.07807, "bpb": 1.985687}, "tr": {"chunks": 3169.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.175054, "cps": 24.757812, "mean_p": 0.220766, "boundary_entropy": 0.216229, "bpb": 2.189943}, "fi": {"chunks": 3391.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.180267, "cps": 26.492188, "mean_p": 0.208324, "boundary_entropy": 0.160146, "bpb": 2.34221}}, "bpb": {"en": 1.98569, "tr": 2.18994, "fi": 2.34221}, "gini_cpc": 0.028985, "gini_cps": 0.019749, "bpic": 5.658488, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8789562316983418, "fi": 0.9051310599809572}, "inventory": {"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": 248.9450225830078, "n_util_obs": 661, "util_count_spearman": 0.2152180295134802}, "t": 707.9} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.00158, "tr": 2.1987, "fi": 2.35051}, "bytes_per_chunk": {"en": 4.775, "tr": 6.042, "fi": 5.572}, "bpb_mean": 2.1836, "frozen_inventory": true, "t": 708.9} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44756, "bpb": 2.08839, "chunk_loss": 0.030368, "F": 0.19839, "G": 0.21592, "chunks": 203.16, "bytes_per_chunk": 5.04, "boundary_entropy": 0.11001, "router_noise": 0.0, "grad_norm": 0.7965, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.26, "probe_sec": 0.0498, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03843, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16033, "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": 2.00388, "trunc_abs_frac": 0.83296, "mean_delta_full": 0.04132, "read_hit_rate": {"surface": 0.8363}, "mean_gate": {"surface": 0.31211}, "n_reads": 5410, "n_cand": 6469, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 5410, "inventory": {"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": 249.02777099609375, "n_util_obs": 661, "util_count_spearman": 0.21556654750560916}, "probe_overhead_sec_total": 158.792, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.028351, "alpha": 0.03, "t": 710.4} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 720.4} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41312, "bpb": 2.0387, "chunk_loss": 0.0304, "F": 0.20279, "G": 0.21637, "chunks": 207.66, "bytes_per_chunk": 4.931, "boundary_entropy": 0.09623, "router_noise": 0.0, "grad_norm": 0.7252, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.203, "probe_sec": 0.0312, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02522, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14659, "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": 0.8494, "trunc_abs_frac": 3.74034, "mean_delta_full": 0.43488, "read_hit_rate": {"surface": 0.86784}, "mean_gate": {"surface": 0.30654}, "n_reads": 5739, "n_cand": 6613, "read_suppressed_frac": 0.1052, "dropout_reads_logged": 5739, "inventory": {"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": 253.28126525878906, "n_util_obs": 666, "util_count_spearman": 0.22392468596370013}, "probe_overhead_sec_total": 161.092, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.026812, "alpha": 0.03, "t": 720.6} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 730.4} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.50607, "bpb": 2.1728, "chunk_loss": 0.030332, "F": 0.19757, "G": 0.21431, "chunks": 202.31, "bytes_per_chunk": 5.061, "boundary_entropy": 0.10414, "router_noise": 0.0, "grad_norm": 0.8487, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.199, "probe_sec": 0.0597, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05323, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01729, "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": 0.17755, "trunc_abs_frac": 2.80628, "mean_delta_full": -0.1021, "read_hit_rate": {"surface": 0.83608}, "mean_gate": {"surface": 0.30517}, "n_reads": 5386, "n_cand": 6442, "read_suppressed_frac": 0.1038, "dropout_reads_logged": 5386, "inventory": {"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": 257.5057678222656, "n_util_obs": 678, "util_count_spearman": 0.2160104542249287}, "probe_overhead_sec_total": 163.664, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.021254, "alpha": 0.03, "t": 730.5} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 740.8} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41064, "bpb": 2.03513, "chunk_loss": 0.030331, "F": 0.19809, "G": 0.21111, "chunks": 202.84, "bytes_per_chunk": 5.048, "boundary_entropy": 0.09482, "router_noise": 0.0, "grad_norm": 0.7113, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.21, "probe_sec": 0.0555, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04807, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18792, "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": 0.2722, "trunc_abs_frac": 0.69899, "mean_delta_full": 0.13804, "read_hit_rate": {"surface": 0.85153}, "mean_gate": {"surface": 0.30879}, "n_reads": 5500, "n_cand": 6459, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 5500, "inventory": {"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": 261.76043701171875, "n_util_obs": 687, "util_count_spearman": 0.21055616776292596}, "probe_overhead_sec_total": 166.408, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.023992, "alpha": 0.03, "t": 741.1} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 751.6} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3461.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198874, "cps": 27.039062, "mean_p": 0.20294, "boundary_entropy": 0.072576, "bpb": 1.967282}, "tr": {"chunks": 3292.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.181848, "cps": 25.71875, "mean_p": 0.224902, "boundary_entropy": 0.216288, "bpb": 2.156704}, "fi": {"chunks": 3490.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.18553, "cps": 27.265625, "mean_p": 0.210904, "boundary_entropy": 0.158129, "bpb": 2.32307}}, "bpb": {"en": 1.96728, "tr": 2.1567, "fi": 2.32307}, "gini_cpc": 0.020045, "gini_cps": 0.012887, "bpic": 5.538612, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9143906909262981, "fi": 0.9329020865428026}, "inventory": {"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": 265.91754150390625, "n_util_obs": 702, "util_count_spearman": 0.20827764571540744}, "t": 752.3} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 1.98226, "tr": 2.16768, "fi": 2.33429}, "bytes_per_chunk": {"en": 4.786, "tr": 5.903, "fi": 5.477}, "bpb_mean": 2.16141, "frozen_inventory": true, "t": 753.1} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.34143, "bpb": 1.93528, "chunk_loss": 0.030406, "F": 0.20496, "G": 0.21319, "chunks": 209.88, "bytes_per_chunk": 4.879, "boundary_entropy": 0.07978, "router_noise": 0.0, "grad_norm": 0.6911, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.258, "probe_sec": 0.0274, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01903, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09828, "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": 0.8751, "trunc_abs_frac": 1.70097, "mean_delta_full": -0.31318, "read_hit_rate": {"surface": 0.89078}, "mean_gate": {"surface": 0.31683}, "n_reads": 5954, "n_cand": 6684, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 5954, "inventory": {"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": 266.00909423828125, "n_util_obs": 702, "util_count_spearman": 0.208210835018451}, "probe_overhead_sec_total": 169.603, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.024127, "alpha": 0.03, "t": 753.9} |
| {"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 765.0} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.34137, "bpb": 1.93519, "chunk_loss": 0.03043, "F": 0.20343, "G": 0.21696, "chunks": 208.31, "bytes_per_chunk": 4.916, "boundary_entropy": 0.09652, "router_noise": 0.0, "grad_norm": 0.6399, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.227, "probe_sec": 0.0894, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0764, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23152, "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": 0.99547, "trunc_abs_frac": 1.98815, "mean_delta_full": 0.37964, "read_hit_rate": {"surface": 0.87474}, "mean_gate": {"surface": 0.30805}, "n_reads": 5803, "n_cand": 6634, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 5803, "inventory": {"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": 270.2061767578125, "n_util_obs": 715, "util_count_spearman": 0.19458505266120016}, "probe_overhead_sec_total": 172.799, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.028587, "alpha": 0.03, "t": 765.3} |
| {"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 774.9} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46658, "bpb": 2.11583, "chunk_loss": 0.030427, "F": 0.20374, "G": 0.21673, "chunks": 208.62, "bytes_per_chunk": 4.908, "boundary_entropy": 0.09358, "router_noise": 0.0, "grad_norm": 0.7227, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.199, "probe_sec": 0.1578, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14585, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.33337, "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": 1.67292, "trunc_abs_frac": 0.46743, "mean_delta_full": 0.1507, "read_hit_rate": {"surface": 0.85325}, "mean_gate": {"surface": 0.29917}, "n_reads": 5669, "n_cand": 6644, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 5669, "inventory": {"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": 274.43798828125, "n_util_obs": 727, "util_count_spearman": 0.19198253415551483}, "probe_overhead_sec_total": 175.434, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.02417, "alpha": 0.03, "t": 775.3} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 785.7} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42149, "bpb": 2.05078, "chunk_loss": 0.03034, "F": 0.19888, "G": 0.21289, "chunks": 203.66, "bytes_per_chunk": 5.028, "boundary_entropy": 0.09415, "router_noise": 0.0, "grad_norm": 0.7009, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.214, "probe_sec": 0.0293, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02115, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10739, "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": -3.32256, "trunc_abs_frac": 0.57193, "mean_delta_full": -0.06925, "read_hit_rate": {"surface": 0.8515}, "mean_gate": {"surface": 0.30525}, "n_reads": 5522, "n_cand": 6485, "read_suppressed_frac": 0.0943, "dropout_reads_logged": 5522, "inventory": {"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": 278.66351318359375, "n_util_obs": 744, "util_count_spearman": 0.19389473499873003}, "probe_overhead_sec_total": 178.311, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.022252, "alpha": 0.03, "t": 786.0} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 799.6} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3468.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199276, "cps": 27.09375, "mean_p": 0.203713, "boundary_entropy": 0.071103, "bpb": 1.946798}, "tr": {"chunks": 3213.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.177484, "cps": 25.101562, "mean_p": 0.219959, "boundary_entropy": 0.210961, "bpb": 2.134923}, "fi": {"chunks": 3589.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.190793, "cps": 28.039062, "mean_p": 0.215525, "boundary_entropy": 0.161031, "bpb": 2.299421}}, "bpb": {"en": 1.9468, "tr": 2.13492, "fi": 2.29942}, "gini_cpc": 0.025597, "gini_cps": 0.024408, "bpic": 5.524051, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8906461717427402, "fi": 0.9574290626168632}, "inventory": {"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": 282.8382873535156, "n_util_obs": 753, "util_count_spearman": 0.20015170568913837}, "t": 800.1} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.96517, "tr": 2.14662, "fi": 2.30874}, "bytes_per_chunk": {"en": 4.775, "tr": 5.923, "fi": 5.365}, "bpb_mean": 2.14017, "frozen_inventory": true, "t": 800.9} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38089, "bpb": 1.9922, "chunk_loss": 0.03041, "F": 0.20364, "G": 0.21501, "chunks": 208.53, "bytes_per_chunk": 4.911, "boundary_entropy": 0.0868, "router_noise": 0.0, "grad_norm": 0.7124, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 0.332, "probe_sec": 0.0703, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06221, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.18562, "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": 1.16787, "trunc_abs_frac": 1.16787, "mean_delta_full": -0.20288, "read_hit_rate": {"surface": 0.86342}, "mean_gate": {"surface": 0.30381}, "n_reads": 5734, "n_cand": 6641, "read_suppressed_frac": 0.1039, "dropout_reads_logged": 5734, "inventory": {"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": 282.92645263671875, "n_util_obs": 753, "util_count_spearman": 0.19982871294608295}, "probe_overhead_sec_total": 181.156, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.028339, "alpha": 0.03, "t": 802.6} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 819.6} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.53797, "bpb": 2.21882, "chunk_loss": 0.030436, "F": 0.20163, "G": 0.21664, "chunks": 206.47, "bytes_per_chunk": 4.96, "boundary_entropy": 0.09701, "router_noise": 0.0, "grad_norm": 0.7654, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.346, "probe_sec": 0.0628, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.054, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08347, "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": 0.22483, "trunc_abs_frac": 1.14074, "mean_delta_full": -0.17883, "read_hit_rate": {"surface": 0.8184}, "mean_gate": {"surface": 0.29803}, "n_reads": 5381, "n_cand": 6575, "read_suppressed_frac": 0.1087, "dropout_reads_logged": 5381, "inventory": {"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": 287.18756103515625, "n_util_obs": 765, "util_count_spearman": 0.2065789630980752}, "probe_overhead_sec_total": 183.941, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.025991, "alpha": 0.03, "t": 819.8} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 838.4} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41098, "bpb": 2.03561, "chunk_loss": 0.030383, "F": 0.20114, "G": 0.21381, "chunks": 205.97, "bytes_per_chunk": 4.972, "boundary_entropy": 0.09236, "router_noise": 0.0, "grad_norm": 0.7017, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.376, "probe_sec": 0.0314, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02463, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24837, "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": -0.74556, "trunc_abs_frac": 3.69261, "mean_delta_full": -0.05635, "read_hit_rate": {"surface": 0.85958}, "mean_gate": {"surface": 0.29781}, "n_reads": 5638, "n_cand": 6559, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 5638, "inventory": {"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": 291.473876953125, "n_util_obs": 779, "util_count_spearman": 0.21380206905617266}, "probe_overhead_sec_total": 186.911, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.021667, "alpha": 0.03, "t": 838.6} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 849.8} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40204, "bpb": 2.02272, "chunk_loss": 0.03033, "F": 0.19965, "G": 0.21099, "chunks": 204.44, "bytes_per_chunk": 5.009, "boundary_entropy": 0.08885, "router_noise": 0.0, "grad_norm": 0.7124, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.227, "probe_sec": 0.0279, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02202, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.33197, "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": 0.2611, "trunc_abs_frac": 1.71922, "mean_delta_full": 0.37378, "read_hit_rate": {"surface": 0.84178}, "mean_gate": {"surface": 0.29858}, "n_reads": 5480, "n_cand": 6510, "read_suppressed_frac": 0.0916, "dropout_reads_logged": 5480, "inventory": {"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": 295.73431396484375, "n_util_obs": 789, "util_count_spearman": 0.20289405390313697}, "probe_overhead_sec_total": 189.531, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.025342, "alpha": 0.03, "t": 850.0} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 861.5} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3477.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199793, "cps": 27.164062, "mean_p": 0.203159, "boundary_entropy": 0.069968, "bpb": 1.93308}, "tr": {"chunks": 3230.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.178423, "cps": 25.234375, "mean_p": 0.220092, "boundary_entropy": 0.205567, "bpb": 2.122909}, "fi": {"chunks": 3494.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.185742, "cps": 27.296875, "mean_p": 0.208997, "boundary_entropy": 0.15046, "bpb": 2.27849}}, "bpb": {"en": 1.93308, "tr": 2.12291, "fi": 2.27849}, "gini_cpc": 0.025261, "gini_cps": 0.017253, "bpic": 5.561416, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8930410048873342, "fi": 0.9296734892829994}, "inventory": {"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": 299.8636779785156, "n_util_obs": 793, "util_count_spearman": 0.2016274894673344}, "t": 862.1} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.94911, "tr": 2.13196, "fi": 2.28898}, "bytes_per_chunk": {"en": 4.77, "tr": 5.913, "fi": 5.448}, "bpb_mean": 2.12335, "frozen_inventory": true, "t": 863.2} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42514, "bpb": 2.05604, "chunk_loss": 0.030352, "F": 0.19843, "G": 0.21781, "chunks": 203.19, "bytes_per_chunk": 5.04, "boundary_entropy": 0.10713, "router_noise": 0.0, "grad_norm": 0.7062, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.289, "probe_sec": 0.0463, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03546, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20509, "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": 0.47434, "trunc_abs_frac": 0.49053, "mean_delta_full": 0.16235, "read_hit_rate": {"surface": 0.82782}, "mean_gate": {"surface": 0.30561}, "n_reads": 5356, "n_cand": 6470, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 5356, "inventory": {"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": 299.94403076171875, "n_util_obs": 794, "util_count_spearman": 0.20058008055049079}, "probe_overhead_sec_total": 192.329, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.03407, "alpha": 0.03, "t": 864.4} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 881.3} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35394, "bpb": 1.95332, "chunk_loss": 0.030391, "F": 0.20413, "G": 0.21234, "chunks": 209.03, "bytes_per_chunk": 4.899, "boundary_entropy": 0.07961, "router_noise": 0.0, "grad_norm": 0.669, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.342, "probe_sec": 0.0246, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01834, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.07627, "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": 2.91803, "trunc_abs_frac": 1.06989, "mean_delta_full": -0.04122, "read_hit_rate": {"surface": 0.88583}, "mean_gate": {"surface": 0.29063}, "n_reads": 5897, "n_cand": 6657, "read_suppressed_frac": 0.096, "dropout_reads_logged": 5897, "inventory": {"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": 304.1788330078125, "n_util_obs": 807, "util_count_spearman": 0.20011830189606433}, "probe_overhead_sec_total": 194.881, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.032896, "alpha": 0.03, "t": 881.5} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 895.4} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4606, "bpb": 2.1072, "chunk_loss": 0.030332, "F": 0.19565, "G": 0.21238, "chunks": 200.34, "bytes_per_chunk": 5.111, "boundary_entropy": 0.10205, "router_noise": 0.0, "grad_norm": 0.7748, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.296, "probe_sec": 0.0553, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03177, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02694, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 2.3308, "trunc_abs_frac": 1.46198, "mean_delta_full": 0.10978, "read_hit_rate": {"surface": 0.81533}, "mean_gate": {"surface": 0.29029}, "n_reads": 5201, "n_cand": 6379, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 5201, "inventory": {"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": 308.4486083984375, "n_util_obs": 819, "util_count_spearman": 0.194205193790887}, "probe_overhead_sec_total": 197.391, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.024062, "alpha": 0.03, "t": 896.4} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 909.4} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44039, "bpb": 2.07805, "chunk_loss": 0.030503, "F": 0.20416, "G": 0.21968, "chunks": 209.06, "bytes_per_chunk": 4.898, "boundary_entropy": 0.09757, "router_noise": 0.0, "grad_norm": 0.7485, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.031, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01929, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20575, "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": 0.83058}, "mean_gate": {"surface": 0.30444}, "n_reads": 5530, "n_cand": 6658, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 5530, "inventory": {"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": 312.7161865234375, "n_util_obs": 832, "util_count_spearman": 0.1930777044626675}, "probe_overhead_sec_total": 200.144, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.023964, "alpha": 0.03, "t": 909.5} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 920.2} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3463.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198989, "cps": 27.054688, "mean_p": 0.201621, "boundary_entropy": 0.067317, "bpb": 1.916645}, "tr": {"chunks": 3244.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.179197, "cps": 25.34375, "mean_p": 0.216319, "boundary_entropy": 0.197458, "bpb": 2.112452}, "fi": {"chunks": 3598.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.191271, "cps": 28.109375, "mean_p": 0.211594, "boundary_entropy": 0.149504, "bpb": 2.259434}}, "bpb": {"en": 1.91665, "tr": 2.11245, "fi": 2.25943}, "gini_cpc": 0.02317, "gini_cps": 0.022902, "bpic": 5.505289, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9005377497127205, "fi": 0.961215807322572}, "inventory": {"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": 316.8602600097656, "n_util_obs": 841, "util_count_spearman": 0.1987574344859214}, "t": 920.8} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.93506, "tr": 2.12115, "fi": 2.26891}, "bytes_per_chunk": {"en": 4.781, "tr": 5.928, "fi": 5.359}, "bpb_mean": 2.10837, "frozen_inventory": true, "t": 921.8} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.33661, "bpb": 1.92831, "chunk_loss": 0.030396, "F": 0.20361, "G": 0.2115, "chunks": 208.5, "bytes_per_chunk": 4.911, "boundary_entropy": 0.07649, "router_noise": 0.0, "grad_norm": 0.6865, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.267, "probe_sec": 0.0418, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03474, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02019, "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": 0.83157, "trunc_abs_frac": 0.72782, "mean_delta_full": 0.2607, "read_hit_rate": {"surface": 0.87771}, "mean_gate": {"surface": 0.30901}, "n_reads": 5828, "n_cand": 6640, "read_suppressed_frac": 0.0896, "dropout_reads_logged": 5828, "inventory": {"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": 316.95001220703125, "n_util_obs": 842, "util_count_spearman": 0.19586655593652452}, "probe_overhead_sec_total": 202.768, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.033503, "alpha": 0.03, "t": 922.9} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 933.6} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.3499, "bpb": 1.9475, "chunk_loss": 0.03041, "F": 0.20428, "G": 0.21265, "chunks": 209.19, "bytes_per_chunk": 4.895, "boundary_entropy": 0.08004, "router_noise": 0.0, "grad_norm": 0.7182, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.219, "probe_sec": 0.0337, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02658, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24238, "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": 1.76867, "trunc_abs_frac": 1.94793, "mean_delta_full": 0.27011, "read_hit_rate": {"surface": 0.87316}, "mean_gate": {"surface": 0.30439}, "n_reads": 5817, "n_cand": 6662, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 5817, "inventory": {"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": 321.1507873535156, "n_util_obs": 851, "util_count_spearman": 0.19977171549268968}, "probe_overhead_sec_total": 205.502, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.029243, "alpha": 0.03, "t": 933.9} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 944.1} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42264, "bpb": 2.05243, "chunk_loss": 0.030335, "F": 0.19922, "G": 0.21356, "chunks": 204.0, "bytes_per_chunk": 5.02, "boundary_entropy": 0.09404, "router_noise": 0.0, "grad_norm": 0.6961, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.209, "probe_sec": 0.0607, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.053, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05403, "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": 0.66819, "trunc_abs_frac": 0.52041, "mean_delta_full": 0.50824, "read_hit_rate": {"surface": 0.84467}, "mean_gate": {"surface": 0.30158}, "n_reads": 5487, "n_cand": 6496, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 5487, "inventory": {"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": 325.45257568359375, "n_util_obs": 858, "util_count_spearman": 0.20019894691523718}, "probe_overhead_sec_total": 208.492, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.030078, "alpha": 0.03, "t": 944.3} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 954.5} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41651, "bpb": 2.04358, "chunk_loss": 0.030401, "F": 0.20422, "G": 0.2119, "chunks": 209.12, "bytes_per_chunk": 4.897, "boundary_entropy": 0.07816, "router_noise": 0.0, "grad_norm": 0.6892, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.209, "probe_sec": 0.0819, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07254, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20801, "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": 1.0, "trunc_abs_frac": 1.0, "mean_delta_full": -0.05579, "read_hit_rate": {"surface": 0.84865}, "mean_gate": {"surface": 0.30297}, "n_reads": 5652, "n_cand": 6660, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 5652, "inventory": {"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": 329.7085266113281, "n_util_obs": 867, "util_count_spearman": 0.2059427706038548}, "probe_overhead_sec_total": 211.439, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.031515, "alpha": 0.03, "t": 954.7} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 965.0} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3465.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199104, "cps": 27.070312, "mean_p": 0.202026, "boundary_entropy": 0.068857, "bpb": 1.908597}, "tr": {"chunks": 3203.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.176932, "cps": 25.023438, "mean_p": 0.214561, "boundary_entropy": 0.193883, "bpb": 2.092704}, "fi": {"chunks": 3512.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.186699, "cps": 27.4375, "mean_p": 0.208279, "boundary_entropy": 0.145009, "bpb": 2.253619}}, "bpb": {"en": 1.9086, "tr": 2.0927, "fi": 2.25362}, "gini_cpc": 0.026266, "gini_cps": 0.020236, "bpic": 5.572888, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.888642885958248, "fi": 0.9376991126818356}, "inventory": {"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": 333.865966796875, "n_util_obs": 880, "util_count_spearman": 0.2039782689145709}, "t": 965.4} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.92741, "tr": 2.10148, "fi": 2.26385}, "bytes_per_chunk": {"en": 4.774, "tr": 5.966, "fi": 5.388}, "bpb_mean": 2.09758, "frozen_inventory": true, "t": 966.4} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.3584, "bpb": 1.95975, "chunk_loss": 0.03032, "F": 0.19669, "G": 0.21146, "chunks": 201.41, "bytes_per_chunk": 5.084, "boundary_entropy": 0.09944, "router_noise": 0.0, "grad_norm": 0.6814, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.256, "probe_sec": 0.0325, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0255, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17659, "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": -2.51396, "trunc_abs_frac": 0.94256, "mean_delta_full": -0.02405, "read_hit_rate": {"surface": 0.85015}, "mean_gate": {"surface": 0.30498}, "n_reads": 5452, "n_cand": 6413, "read_suppressed_frac": 0.0921, "dropout_reads_logged": 5452, "inventory": {"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": 333.9502868652344, "n_util_obs": 880, "util_count_spearman": 0.20396206165613096}, "probe_overhead_sec_total": 214.304, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.025403, "alpha": 0.03, "t": 967.5} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 977.7} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49554, "bpb": 2.15761, "chunk_loss": 0.030352, "F": 0.19623, "G": 0.21132, "chunks": 200.94, "bytes_per_chunk": 5.096, "boundary_entropy": 0.09755, "router_noise": 0.0, "grad_norm": 0.76, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.207, "probe_sec": 0.0359, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02317, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14934, "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": 2.37249, "trunc_abs_frac": 0.63829, "mean_delta_full": 0.11536, "read_hit_rate": {"surface": 0.81447}, "mean_gate": {"surface": 0.29106}, "n_reads": 5211, "n_cand": 6398, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 5211, "inventory": {"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": 338.14398193359375, "n_util_obs": 894, "util_count_spearman": 0.18868752232026856}, "probe_overhead_sec_total": 216.82, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.030151, "alpha": 0.03, "t": 977.9} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3469.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199333, "cps": 27.101562, "mean_p": 0.201203, "boundary_entropy": 0.067646, "bpb": 1.902024}, "tr": {"chunks": 3336.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.184279, "cps": 26.0625, "mean_p": 0.217277, "boundary_entropy": 0.193773, "bpb": 2.087184}, "fi": {"chunks": 3460.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.183935, "cps": 27.03125, "mean_p": 0.206204, "boundary_entropy": 0.143358, "bpb": 2.236073}}, "bpb": {"en": 1.90202, "tr": 2.08718, "fi": 2.23607}, "gini_cpc": 0.018088, "gini_cps": 0.008638, "bpic": 5.526741, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9244752971557472, "fi": 0.9227499614246241}, "inventory": {"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": 341.94268798828125, "n_util_obs": 911, "util_count_spearman": 0.181698945501763}, "t": 988.2} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 990.4} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42319, "bpb": 2.05323, "chunk_loss": 0.030364, "F": 0.19962, "G": 0.21567, "chunks": 204.41, "bytes_per_chunk": 5.01, "boundary_entropy": 0.10055, "router_noise": 0.0, "grad_norm": 0.6883, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.0487, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04104, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07801, "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": 0.66628, "trunc_abs_frac": 1.17281, "mean_delta_full": -0.39172, "read_hit_rate": {"surface": 0.82501}, "mean_gate": {"surface": 0.30134}, "n_reads": 5370, "n_cand": 6509, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 5370, "inventory": {"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": 342.37109375, "n_util_obs": 912, "util_count_spearman": 0.18197012323959078}, "probe_overhead_sec_total": 219.725, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.035688, "alpha": 0.03, "t": 990.6} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1000.7} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36251, "bpb": 1.96568, "chunk_loss": 0.030441, "F": 0.20453, "G": 0.21377, "chunks": 209.44, "bytes_per_chunk": 4.889, "boundary_entropy": 0.0797, "router_noise": 0.0, "grad_norm": 0.651, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.206, "probe_sec": 0.0364, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02874, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12918, "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": 0.88366, "trunc_abs_frac": 0.60796, "mean_delta_full": -0.24028, "read_hit_rate": {"surface": 0.87286}, "mean_gate": {"surface": 0.29452}, "n_reads": 5822, "n_cand": 6670, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 5822, "inventory": {"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": 346.61181640625, "n_util_obs": 925, "util_count_spearman": 0.17168629084909545}, "probe_overhead_sec_total": 222.418, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.03218, "alpha": 0.03, "t": 1000.9} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1010.7} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3465.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199104, "cps": 27.070312, "mean_p": 0.201406, "boundary_entropy": 0.066493, "bpb": 1.897419}, "tr": {"chunks": 3268.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.180523, "cps": 25.53125, "mean_p": 0.217724, "boundary_entropy": 0.195304, "bpb": 2.080818}, "fi": {"chunks": 3441.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.182925, "cps": 26.882812, "mean_p": 0.204766, "boundary_entropy": 0.14103, "bpb": 2.234404}}, "bpb": {"en": 1.89742, "tr": 2.08082, "fi": 2.2344}, "gini_cpc": 0.02202, "gini_cps": 0.012909, "bpic": 5.576175, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9066765380304572, "fi": 0.918742211485819}, "inventory": {"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": 350.7489929199219, "n_util_obs": 936, "util_count_spearman": 0.16419073783553276}, "t": 1011.3} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.91554, "tr": 2.08948, "fi": 2.2412}, "bytes_per_chunk": {"en": 4.776, "tr": 5.906, "fi": 5.528}, "bpb_mean": 2.08207, "frozen_inventory": true, "t": 1012.3} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 1013.3} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.44703, "bpb": 2.08762, "chunk_loss": 0.030424, "F": 0.20441, "G": 0.21546, "chunks": 209.31, "bytes_per_chunk": 4.892, "boundary_entropy": 0.08554, "router_noise": 0.0, "grad_norm": 0.7292, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0311, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02365, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03104, "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": -0.83273, "trunc_abs_frac": 1.03927, "mean_delta_full": 0.03357, "read_hit_rate": {"surface": 0.84698}, "mean_gate": {"surface": 0.29013}, "n_reads": 5646, "n_cand": 6666, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 5646, "inventory": {"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": 350.83642578125, "n_util_obs": 936, "util_count_spearman": 0.16409837180914028}, "probe_overhead_sec_total": 225.169, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.035909, "alpha": 0.03, "t": 1013.6} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1023.7} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33592, "bpb": 1.92733, "chunk_loss": 0.030332, "F": 0.20041, "G": 0.20926, "chunks": 205.22, "bytes_per_chunk": 4.99, "boundary_entropy": 0.08009, "router_noise": 0.0, "grad_norm": 0.6596, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.209, "probe_sec": 0.0992, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09231, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.37254, "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": -0.32808, "trunc_abs_frac": 0.99334, "mean_delta_full": -0.01163, "read_hit_rate": {"surface": 0.86412}, "mean_gate": {"surface": 0.29119}, "n_reads": 5647, "n_cand": 6535, "read_suppressed_frac": 0.102, "dropout_reads_logged": 5647, "inventory": {"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": 355.07550048828125, "n_util_obs": 945, "util_count_spearman": 0.16992012545867005}, "probe_overhead_sec_total": 228.022, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.031497, "alpha": 0.03, "t": 1024.0} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1034.7} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34887, "bpb": 1.946, "chunk_loss": 0.030269, "F": 0.19351, "G": 0.21278, "chunks": 198.16, "bytes_per_chunk": 5.168, "boundary_entropy": 0.10573, "router_noise": 0.0, "grad_norm": 0.7183, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.219, "probe_sec": 0.1037, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09591, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05639, "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": -0.09098, "trunc_abs_frac": 0.95757, "mean_delta_full": -0.23596, "read_hit_rate": {"surface": 0.84007}, "mean_gate": {"surface": 0.29854}, "n_reads": 5300, "n_cand": 6309, "read_suppressed_frac": 0.1, "dropout_reads_logged": 5300, "inventory": {"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": 359.3763122558594, "n_util_obs": 956, "util_count_spearman": 0.1656373606790124}, "probe_overhead_sec_total": 230.718, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.026402, "alpha": 0.03, "t": 1035.0} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1044.9} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32426, "bpb": 1.9105, "chunk_loss": 0.030306, "F": 0.19662, "G": 0.21137, "chunks": 201.34, "bytes_per_chunk": 5.086, "boundary_entropy": 0.09116, "router_noise": 0.0, "grad_norm": 0.7115, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.204, "probe_sec": 0.0308, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02119, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04382, "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": 0.40477, "trunc_abs_frac": 0.83709, "mean_delta_full": 0.2518, "read_hit_rate": {"surface": 0.85322}, "mean_gate": {"surface": 0.28509}, "n_reads": 5470, "n_cand": 6411, "read_suppressed_frac": 0.102, "dropout_reads_logged": 5470, "inventory": {"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": 363.5540466308594, "n_util_obs": 971, "util_count_spearman": 0.1604468723208036}, "probe_overhead_sec_total": 233.299, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.02154, "alpha": 0.03, "t": 1045.1} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1055.0} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3476.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199736, "cps": 27.15625, "mean_p": 0.201818, "boundary_entropy": 0.064516, "bpb": 1.891294}, "tr": {"chunks": 3352.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.185163, "cps": 26.1875, "mean_p": 0.220732, "boundary_entropy": 0.196239, "bpb": 2.067534}, "fi": {"chunks": 3449.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.18335, "cps": 26.945312, "mean_p": 0.204292, "boundary_entropy": 0.137448, "bpb": 2.21547}}, "bpb": {"en": 1.89129, "tr": 2.06753, "fi": 2.21547}, "gini_cpc": 0.019223, "gini_cps": 0.008044, "bpic": 5.520288, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9270385856866732, "fi": 0.9179640288328714}, "inventory": {"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": 367.67230224609375, "n_util_obs": 984, "util_count_spearman": 0.15130347706051456}, "t": 1055.7} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.91018, "tr": 2.0766, "fi": 2.22428}, "bytes_per_chunk": {"en": 4.77, "tr": 5.808, "fi": 5.479}, "bpb_mean": 2.07036, "frozen_inventory": true, "t": 1056.7} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36388, "bpb": 1.96767, "chunk_loss": 0.030455, "F": 0.20801, "G": 0.21392, "chunks": 213.0, "bytes_per_chunk": 4.808, "boundary_entropy": 0.07119, "router_noise": 0.0, "grad_norm": 0.8092, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0358, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02476, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23791, "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": 1.50477, "trunc_abs_frac": 1.08364, "mean_delta_full": -0.08964, "read_hit_rate": {"surface": 0.87529}, "mean_gate": {"surface": 0.29116}, "n_reads": 5938, "n_cand": 6784, "read_suppressed_frac": 0.0941, "dropout_reads_logged": 5938, "inventory": {"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": 367.76409912109375, "n_util_obs": 984, "util_count_spearman": 0.15118498472092548}, "probe_overhead_sec_total": 235.824, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.022756, "alpha": 0.03, "t": 1057.8} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1067.9} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39009, "bpb": 2.00548, "chunk_loss": 0.030283, "F": 0.19562, "G": 0.20839, "chunks": 200.31, "bytes_per_chunk": 5.112, "boundary_entropy": 0.08604, "router_noise": 0.0, "grad_norm": 0.696, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.206, "probe_sec": 0.0239, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01816, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14684, "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": 0.08979, "trunc_abs_frac": 1.17124, "mean_delta_full": -0.14286, "read_hit_rate": {"surface": 0.84823}, "mean_gate": {"surface": 0.28925}, "n_reads": 5410, "n_cand": 6378, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 5410, "inventory": {"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": 372.01678466796875, "n_util_obs": 996, "util_count_spearman": 0.148098771945961}, "probe_overhead_sec_total": 238.467, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.028691, "alpha": 0.03, "t": 1068.1} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1078.2} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40759, "bpb": 2.03073, "chunk_loss": 0.030353, "F": 0.20026, "G": 0.21075, "chunks": 205.06, "bytes_per_chunk": 4.994, "boundary_entropy": 0.08546, "router_noise": 0.0, "grad_norm": 0.702, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.205, "probe_sec": 0.0362, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03001, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0376, "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": 1.03519, "trunc_abs_frac": 5.36556, "mean_delta_full": -0.78931, "read_hit_rate": {"surface": 0.84181}, "mean_gate": {"surface": 0.29071}, "n_reads": 5497, "n_cand": 6530, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 5497, "inventory": {"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": 376.26788330078125, "n_util_obs": 1006, "util_count_spearman": 0.15190117511814186}, "probe_overhead_sec_total": 241.083, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.030622, "alpha": 0.03, "t": 1078.3} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1088.8} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32038, "bpb": 1.9049, "chunk_loss": 0.030296, "F": 0.19485, "G": 0.20842, "chunks": 199.53, "bytes_per_chunk": 5.132, "boundary_entropy": 0.0928, "router_noise": 0.0, "grad_norm": 0.6579, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.212, "probe_sec": 0.0245, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0184, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14763, "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": 2.92173, "trunc_abs_frac": 1.81347, "mean_delta_full": 0.06424, "read_hit_rate": {"surface": 0.84669}, "mean_gate": {"surface": 0.29195}, "n_reads": 5379, "n_cand": 6353, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 5379, "inventory": {"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": 380.52398681640625, "n_util_obs": 1021, "util_count_spearman": 0.1489056826793291}, "probe_overhead_sec_total": 244.175, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.030576, "alpha": 0.03, "t": 1089.0} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1098.6} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3474.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199621, "cps": 27.140625, "mean_p": 0.2011, "boundary_entropy": 0.063203, "bpb": 1.876296}, "tr": {"chunks": 3213.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.177484, "cps": 25.101562, "mean_p": 0.211029, "boundary_entropy": 0.185729, "bpb": 2.057576}, "fi": {"chunks": 3521.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.187178, "cps": 27.507812, "mean_p": 0.205259, "boundary_entropy": 0.134352, "bpb": 2.208786}}, "bpb": {"en": 1.8763, "tr": 2.05758, "fi": 2.20879}, "gini_cpc": 0.026153, "gini_cps": 0.020115, "bpic": 5.557602, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8891079227414573, "fi": 0.9376666024885855}, "inventory": {"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": 384.711669921875, "n_util_obs": 1030, "util_count_spearman": 0.15244641640482104}, "t": 1099.3} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.8959, "tr": 2.06625, "fi": 2.21693}, "bytes_per_chunk": {"en": 4.77, "tr": 5.916, "fi": 5.389}, "bpb_mean": 2.05969, "frozen_inventory": true, "t": 1100.3} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38514, "bpb": 1.99833, "chunk_loss": 0.030329, "F": 0.19934, "G": 0.20801, "chunks": 204.12, "bytes_per_chunk": 5.017, "boundary_entropy": 0.07976, "router_noise": 0.0, "grad_norm": 0.7174, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.25, "probe_sec": 0.0342, "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": 0.24965, "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": 0.1173, "trunc_abs_frac": 1.04836, "mean_delta_full": 0.23422, "read_hit_rate": {"surface": 0.84292}, "mean_gate": {"surface": 0.28985}, "n_reads": 5479, "n_cand": 6500, "read_suppressed_frac": 0.1082, "dropout_reads_logged": 5479, "inventory": {"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": 384.796630859375, "n_util_obs": 1030, "util_count_spearman": 0.15278288507454482}, "probe_overhead_sec_total": 246.715, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.027159, "alpha": 0.03, "t": 1101.5} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1111.0} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3843, "bpb": 1.99712, "chunk_loss": 0.030409, "F": 0.20392, "G": 0.21465, "chunks": 208.81, "bytes_per_chunk": 4.904, "boundary_entropy": 0.08583, "router_noise": 0.0, "grad_norm": 0.7254, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.195, "probe_sec": 0.0261, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01948, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0912, "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": 2.33291, "trunc_abs_frac": 1.37722, "mean_delta_full": -0.10203, "read_hit_rate": {"surface": 0.85323}, "mean_gate": {"surface": 0.28494}, "n_reads": 5674, "n_cand": 6650, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 5674, "inventory": {"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": 389.0602111816406, "n_util_obs": 1041, "util_count_spearman": 0.152620885777713}, "probe_overhead_sec_total": 249.345, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.028273, "alpha": 0.03, "t": 1111.2} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1121.3} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36589, "bpb": 1.97056, "chunk_loss": 0.030318, "F": 0.19702, "G": 0.21209, "chunks": 201.75, "bytes_per_chunk": 5.076, "boundary_entropy": 0.09654, "router_noise": 0.0, "grad_norm": 0.7079, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.206, "probe_sec": 0.051, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04358, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16092, "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": 0.96735, "trunc_abs_frac": 0.84788, "mean_delta_full": 0.84061, "read_hit_rate": {"surface": 0.84605}, "mean_gate": {"surface": 0.28925}, "n_reads": 5435, "n_cand": 6424, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 5435, "inventory": {"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": 393.3079833984375, "n_util_obs": 1052, "util_count_spearman": 0.16192747280758832}, "probe_overhead_sec_total": 251.798, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.028666, "alpha": 0.03, "t": 1121.5} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1131.6} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.45446, "bpb": 2.09835, "chunk_loss": 0.030323, "F": 0.19653, "G": 0.20958, "chunks": 201.25, "bytes_per_chunk": 5.088, "boundary_entropy": 0.09159, "router_noise": 0.0, "grad_norm": 0.7754, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.205, "probe_sec": 0.0298, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02309, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.37709, "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": 0.45249, "trunc_abs_frac": 0.64361, "mean_delta_full": 0.31877, "read_hit_rate": {"surface": 0.8196}, "mean_gate": {"surface": 0.27663}, "n_reads": 5252, "n_cand": 6408, "read_suppressed_frac": 0.116, "dropout_reads_logged": 5252, "inventory": {"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": 397.5198974609375, "n_util_obs": 1059, "util_count_spearman": 0.16162442138972558}, "probe_overhead_sec_total": 254.575, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.033331, "alpha": 0.03, "t": 1131.8} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1142.0} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3468.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199276, "cps": 27.09375, "mean_p": 0.200311, "boundary_entropy": 0.064129, "bpb": 1.867752}, "tr": {"chunks": 3266.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.180412, "cps": 25.515625, "mean_p": 0.214268, "boundary_entropy": 0.187969, "bpb": 2.045345}, "fi": {"chunks": 3493.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.185689, "cps": 27.289062, "mean_p": 0.205913, "boundary_entropy": 0.138638, "bpb": 2.197086}}, "bpb": {"en": 1.86775, "tr": 2.04534, "fi": 2.19709}, "gini_cpc": 0.022243, "gini_cps": 0.014797, "bpic": 5.547277, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9053378141648892, "fi": 0.9318193691057964}, "inventory": {"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": 401.669189453125, "n_util_obs": 1071, "util_count_spearman": 0.1638024684084965}, "t": 1142.6} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.88783, "tr": 2.05317, "fi": 2.20406}, "bytes_per_chunk": {"en": 4.781, "tr": 5.887, "fi": 5.457}, "bpb_mean": 2.04835, "frozen_inventory": true, "t": 1143.4} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33638, "bpb": 1.92799, "chunk_loss": 0.030338, "F": 0.20206, "G": 0.20859, "chunks": 206.91, "bytes_per_chunk": 4.949, "boundary_entropy": 0.07514, "router_noise": 0.0, "grad_norm": 0.6911, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.0352, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02616, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12709, "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": -0.29369, "trunc_abs_frac": 0.28733, "mean_delta_full": -0.20221, "read_hit_rate": {"surface": 0.87327}, "mean_gate": {"surface": 0.27816}, "n_reads": 5754, "n_cand": 6589, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 5754, "inventory": {"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": 401.7580871582031, "n_util_obs": 1072, "util_count_spearman": 0.16318908011380112}, "probe_overhead_sec_total": 256.936, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.031614, "alpha": 0.03, "t": 1144.5} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1154.4} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32893, "bpb": 1.91723, "chunk_loss": 0.030452, "F": 0.20712, "G": 0.21525, "chunks": 212.09, "bytes_per_chunk": 4.828, "boundary_entropy": 0.08005, "router_noise": 0.0, "grad_norm": 0.6649, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.202, "probe_sec": 0.0816, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0707, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.30575, "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": 0.40273, "trunc_abs_frac": 0.6098, "mean_delta_full": 0.68744, "read_hit_rate": {"surface": 0.86795}, "mean_gate": {"surface": 0.2837}, "n_reads": 5863, "n_cand": 6755, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 5863, "inventory": {"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": 405.97686767578125, "n_util_obs": 1080, "util_count_spearman": 0.16461316275005053}, "probe_overhead_sec_total": 259.574, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.032351, "alpha": 0.03, "t": 1154.7} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1165.2} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38947, "bpb": 2.00458, "chunk_loss": 0.030329, "F": 0.19897, "G": 0.2105, "chunks": 203.75, "bytes_per_chunk": 5.026, "boundary_entropy": 0.08535, "router_noise": 0.0, "grad_norm": 0.8468, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.215, "probe_sec": 0.0914, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08468, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18128, "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": 1.08737, "trunc_abs_frac": 0.86012, "mean_delta_full": 0.23358, "read_hit_rate": {"surface": 0.85327}, "mean_gate": {"surface": 0.27865}, "n_reads": 5536, "n_cand": 6488, "read_suppressed_frac": 0.101, "dropout_reads_logged": 5536, "inventory": {"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": 410.27398681640625, "n_util_obs": 1095, "util_count_spearman": 0.16574136449353818}, "probe_overhead_sec_total": 262.404, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.025971, "alpha": 0.03, "t": 1165.4} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1175.6} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32421, "bpb": 1.91043, "chunk_loss": 0.030486, "F": 0.20032, "G": 0.21544, "chunks": 205.12, "bytes_per_chunk": 4.992, "boundary_entropy": 0.09609, "router_noise": 0.0, "grad_norm": 0.6825, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.208, "probe_sec": 0.0649, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05734, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24553, "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": 0.35829, "trunc_abs_frac": 0.67965, "mean_delta_full": -0.28483, "read_hit_rate": {"surface": 0.85104}, "mean_gate": {"surface": 0.28059}, "n_reads": 5559, "n_cand": 6532, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 5559, "inventory": {"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": 414.54443359375, "n_util_obs": 1104, "util_count_spearman": 0.1665477926538674}, "probe_overhead_sec_total": 265.254, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.024939, "alpha": 0.03, "t": 1175.8} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1185.9} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3465.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199104, "cps": 27.070312, "mean_p": 0.19997, "boundary_entropy": 0.063075, "bpb": 1.861137}, "tr": {"chunks": 3310.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.182843, "cps": 25.859375, "mean_p": 0.215103, "boundary_entropy": 0.187816, "bpb": 2.034885}, "fi": {"chunks": 3473.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.184626, "cps": 27.132812, "mean_p": 0.20402, "boundary_entropy": 0.134939, "bpb": 2.182919}}, "bpb": {"en": 1.86114, "tr": 2.03489, "fi": 2.18292}, "gini_cpc": 0.019134, "gini_cps": 0.010604, "bpic": 5.535909, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9183290516771155, "fi": 0.9272861669544461}, "inventory": {"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": 418.7005615234375, "n_util_obs": 1114, "util_count_spearman": 0.16342064753177507}, "t": 1186.5} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.88049, "tr": 2.04427, "fi": 2.19197}, "bytes_per_chunk": {"en": 4.783, "tr": 5.819, "fi": 5.483}, "bpb_mean": 2.03891, "frozen_inventory": true, "t": 1187.5} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.43434, "bpb": 2.06932, "chunk_loss": 0.030334, "F": 0.19742, "G": 0.21205, "chunks": 202.16, "bytes_per_chunk": 5.065, "boundary_entropy": 0.09228, "router_noise": 0.0, "grad_norm": 0.7692, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.258, "probe_sec": 0.0544, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04829, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.26141, "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": 0.92825, "trunc_abs_frac": 6.50045, "mean_delta_full": 0.74478, "read_hit_rate": {"surface": 0.80674}, "mean_gate": {"surface": 0.27694}, "n_reads": 5193, "n_cand": 6437, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 5193, "inventory": {"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": 418.7734375, "n_util_obs": 1114, "util_count_spearman": 0.16334646176463707}, "probe_overhead_sec_total": 267.905, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.028845, "alpha": 0.03, "t": 1188.7} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1198.9} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27753, "bpb": 1.84308, "chunk_loss": 0.030356, "F": 0.20255, "G": 0.2116, "chunks": 207.41, "bytes_per_chunk": 4.937, "boundary_entropy": 0.08149, "router_noise": 0.0, "grad_norm": 0.6999, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.207, "probe_sec": 0.0474, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0391, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18737, "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": 0.27863, "trunc_abs_frac": 1.48086, "mean_delta_full": 0.13382, "read_hit_rate": {"surface": 0.87646}, "mean_gate": {"surface": 0.28395}, "n_reads": 5789, "n_cand": 6605, "read_suppressed_frac": 0.108, "dropout_reads_logged": 5789, "inventory": {"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": 423.0754699707031, "n_util_obs": 1126, "util_count_spearman": 0.17251292722028136}, "probe_overhead_sec_total": 270.322, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.029053, "alpha": 0.03, "t": 1199.1} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1209.2} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33606, "bpb": 1.92753, "chunk_loss": 0.030263, "F": 0.19571, "G": 0.20538, "chunks": 200.41, "bytes_per_chunk": 5.11, "boundary_entropy": 0.08108, "router_noise": 0.0, "grad_norm": 0.6941, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.205, "probe_sec": 0.0263, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02046, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23187, "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": 2.10557, "trunc_abs_frac": 2.06399, "mean_delta_full": 0.10406, "read_hit_rate": {"surface": 0.85943}, "mean_gate": {"surface": 0.28047}, "n_reads": 5484, "n_cand": 6381, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 5484, "inventory": {"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": 427.32928466796875, "n_util_obs": 1142, "util_count_spearman": 0.17028902686831351}, "probe_overhead_sec_total": 273.076, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.026591, "alpha": 0.03, "t": 1209.3} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1222.5} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36318, "bpb": 1.96665, "chunk_loss": 0.030431, "F": 0.2027, "G": 0.21276, "chunks": 207.56, "bytes_per_chunk": 4.933, "boundary_entropy": 0.08718, "router_noise": 0.0, "grad_norm": 0.7527, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.272, "probe_sec": 0.0878, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07952, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.26743, "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": 0.96364, "trunc_abs_frac": 1.0733, "mean_delta_full": 0.78117, "read_hit_rate": {"surface": 0.84387}, "mean_gate": {"surface": 0.27672}, "n_reads": 5578, "n_cand": 6610, "read_suppressed_frac": 0.0954, "dropout_reads_logged": 5578, "inventory": {"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": 431.5963439941406, "n_util_obs": 1154, "util_count_spearman": 0.16621648774147385}, "probe_overhead_sec_total": 275.587, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.034785, "alpha": 0.03, "t": 1222.9} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1235.7} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3466.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199161, "cps": 27.078125, "mean_p": 0.199872, "boundary_entropy": 0.062073, "bpb": 1.855393}, "tr": {"chunks": 3207.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.177153, "cps": 25.054688, "mean_p": 0.212238, "boundary_entropy": 0.184499, "bpb": 2.0268}, "fi": {"chunks": 3457.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.183775, "cps": 27.007812, "mean_p": 0.202515, "boundary_entropy": 0.130851, "bpb": 2.176667}}, "bpb": {"en": 1.85539, "tr": 2.0268, "fi": 2.17667}, "gini_cpc": 0.026196, "gini_cps": 0.017045, "bpic": 5.600395, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8894959403775898, "fi": 0.9227478839145294}, "inventory": {"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": 435.72467041015625, "n_util_obs": 1166, "util_count_spearman": 0.15272781815396458}, "t": 1236.3} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.8755, "tr": 2.03568, "fi": 2.18554}, "bytes_per_chunk": {"en": 4.786, "tr": 5.968, "fi": 5.477}, "bpb_mean": 2.03224, "frozen_inventory": true, "t": 1237.1} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31452, "bpb": 1.89645, "chunk_loss": 0.030396, "F": 0.20309, "G": 0.21178, "chunks": 207.97, "bytes_per_chunk": 4.924, "boundary_entropy": 0.07785, "router_noise": 0.0, "grad_norm": 0.6945, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.307, "probe_sec": 0.0708, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0611, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05322, "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": -0.10108, "trunc_abs_frac": 1.95352, "mean_delta_full": -0.17795, "read_hit_rate": {"surface": 0.86411}, "mean_gate": {"surface": 0.27918}, "n_reads": 5723, "n_cand": 6623, "read_suppressed_frac": 0.0945, "dropout_reads_logged": 5723, "inventory": {"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": 435.81298828125, "n_util_obs": 1166, "util_count_spearman": 0.1526873245479694}, "probe_overhead_sec_total": 278.359, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.042099, "alpha": 0.03, "t": 1238.3} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1249.6} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30804, "bpb": 1.8871, "chunk_loss": 0.03034, "F": 0.19934, "G": 0.21125, "chunks": 204.12, "bytes_per_chunk": 5.017, "boundary_entropy": 0.08901, "router_noise": 0.0, "grad_norm": 0.6885, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.229, "probe_sec": 0.0255, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01988, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19556, "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": 0.59969, "trunc_abs_frac": 4.29459, "mean_delta_full": 0.70886, "read_hit_rate": {"surface": 0.85523}, "mean_gate": {"surface": 0.27735}, "n_reads": 5559, "n_cand": 6500, "read_suppressed_frac": 0.1099, "dropout_reads_logged": 5559, "inventory": {"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": 440.0694580078125, "n_util_obs": 1176, "util_count_spearman": 0.15705604371623189}, "probe_overhead_sec_total": 281.156, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.033368, "alpha": 0.03, "t": 1249.8} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1259.9} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30748, "bpb": 1.8863, "chunk_loss": 0.030335, "F": 0.20148, "G": 0.20822, "chunks": 206.31, "bytes_per_chunk": 4.963, "boundary_entropy": 0.0759, "router_noise": 0.0, "grad_norm": 0.6693, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.207, "probe_sec": 0.0282, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02074, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00913, "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": -1.0704, "trunc_abs_frac": 1.0704, "mean_delta_full": -0.33984, "read_hit_rate": {"surface": 0.86941}, "mean_gate": {"surface": 0.28013}, "n_reads": 5712, "n_cand": 6570, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 5712, "inventory": {"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": 444.30535888671875, "n_util_obs": 1191, "util_count_spearman": 0.14867144289173315}, "probe_overhead_sec_total": 284.357, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.025156, "alpha": 0.03, "t": 1260.1} |
| {"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1285.7} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30804, "bpb": 1.8871, "chunk_loss": 0.030337, "F": 0.20065, "G": 0.20846, "chunks": 205.47, "bytes_per_chunk": 4.984, "boundary_entropy": 0.07781, "router_noise": 0.0, "grad_norm": 0.7221, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.525, "probe_sec": 0.0989, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07637, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.2153, "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": 0.79135, "trunc_abs_frac": 1.26604, "mean_delta_full": 1.22797, "read_hit_rate": {"surface": 0.87223}, "mean_gate": {"surface": 0.27797}, "n_reads": 5707, "n_cand": 6543, "read_suppressed_frac": 0.103, "dropout_reads_logged": 5707, "inventory": {"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": 448.5557556152344, "n_util_obs": 1205, "util_count_spearman": 0.148010680651582}, "probe_overhead_sec_total": 287.261, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.027507, "alpha": 0.03, "t": 1286.3} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1295.8} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3462.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198931, "cps": 27.046875, "mean_p": 0.1997, "boundary_entropy": 0.062555, "bpb": 1.850004}, "tr": {"chunks": 3255.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.179804, "cps": 25.429688, "mean_p": 0.215477, "boundary_entropy": 0.188927, "bpb": 2.023166}, "fi": {"chunks": 3511.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.186646, "cps": 27.429688, "mean_p": 0.20431, "boundary_entropy": 0.132282, "bpb": 2.167795}}, "bpb": {"en": 1.85, "tr": 2.02317, "fi": 2.16779}, "gini_cpc": 0.022553, "gini_cps": 0.016686, "bpic": 5.546734, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9038523637751281, "fi": 0.9382444469279241}, "inventory": {"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": 452.7171630859375, "n_util_obs": 1222, "util_count_spearman": 0.15650290571347125}, "t": 1296.4} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.87053, "tr": 2.03179, "fi": 2.17647}, "bytes_per_chunk": {"en": 4.783, "tr": 5.89, "fi": 5.415}, "bpb_mean": 2.02626, "frozen_inventory": true, "t": 1297.5} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37294, "bpb": 1.98073, "chunk_loss": 0.03034, "F": 0.19766, "G": 0.21596, "chunks": 202.41, "bytes_per_chunk": 5.059, "boundary_entropy": 0.10331, "router_noise": 0.0, "grad_norm": 0.913, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.246, "probe_sec": 0.0715, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06086, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13466, "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": 0.41457, "trunc_abs_frac": 0.79126, "mean_delta_full": -0.17236, "read_hit_rate": {"surface": 0.82855}, "mean_gate": {"surface": 0.27933}, "n_reads": 5340, "n_cand": 6445, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 5340, "inventory": {"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": 452.79705810546875, "n_util_obs": 1222, "util_count_spearman": 0.15640796860006304}, "probe_overhead_sec_total": 289.887, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.030631, "alpha": 0.03, "t": 1298.6} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1309.7} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31033, "bpb": 1.89041, "chunk_loss": 0.030351, "F": 0.20108, "G": 0.21042, "chunks": 205.91, "bytes_per_chunk": 4.973, "boundary_entropy": 0.08082, "router_noise": 0.0, "grad_norm": 0.6876, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.226, "probe_sec": 0.0305, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02368, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02453, "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": 1.43896, "trunc_abs_frac": 0.11618, "mean_delta_full": 0.02966, "read_hit_rate": {"surface": 0.86625}, "mean_gate": {"surface": 0.26978}, "n_reads": 5680, "n_cand": 6557, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 5680, "inventory": {"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": 457.03753662109375, "n_util_obs": 1237, "util_count_spearman": 0.16425807040698906}, "probe_overhead_sec_total": 292.708, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.027176, "alpha": 0.03, "t": 1309.9} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1323.7} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31112, "bpb": 1.89155, "chunk_loss": 0.030419, "F": 0.20554, "G": 0.21464, "chunks": 210.47, "bytes_per_chunk": 4.865, "boundary_entropy": 0.07967, "router_noise": 0.0, "grad_norm": 0.6889, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.29, "probe_sec": 0.0495, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02414, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20559, "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": 0.98202, "trunc_abs_frac": 0.98202, "mean_delta_full": 0.38025, "read_hit_rate": {"surface": 0.87394}, "mean_gate": {"surface": 0.27099}, "n_reads": 5858, "n_cand": 6703, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 5858, "inventory": {"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": 461.28924560546875, "n_util_obs": 1246, "util_count_spearman": 0.16096227689807277}, "probe_overhead_sec_total": 295.321, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.034587, "alpha": 0.03, "t": 1324.4} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1338.8} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30677, "bpb": 1.88526, "chunk_loss": 0.030383, "F": 0.20013, "G": 0.21215, "chunks": 204.94, "bytes_per_chunk": 4.997, "boundary_entropy": 0.0857, "router_noise": 0.0, "grad_norm": 0.7048, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.3, "probe_sec": 0.1072, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09858, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24641, "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": -0.1093, "trunc_abs_frac": 0.52991, "mean_delta_full": 0.11724, "read_hit_rate": {"surface": 0.85382}, "mean_gate": {"surface": 0.26616}, "n_reads": 5572, "n_cand": 6526, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 5572, "inventory": {"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": 465.579833984375, "n_util_obs": 1254, "util_count_spearman": 0.1594098111313066}, "probe_overhead_sec_total": 298.13, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.032303, "alpha": 0.03, "t": 1339.4} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1354.8} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3467.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199219, "cps": 27.085938, "mean_p": 0.2, "boundary_entropy": 0.061873, "bpb": 1.844553}, "tr": {"chunks": 3247.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.179363, "cps": 25.367188, "mean_p": 0.214707, "boundary_entropy": 0.187273, "bpb": 2.020095}, "fi": {"chunks": 3545.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.188454, "cps": 27.695312, "mean_p": 0.206089, "boundary_entropy": 0.133839, "bpb": 2.162519}}, "bpb": {"en": 1.84455, "tr": 2.0201, "fi": 2.16252}, "gini_cpc": 0.023345, "gini_cps": 0.019365, "bpic": 5.529974, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9003306098594459, "fi": 0.9459640557598618}, "inventory": {"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": 469.7000732421875, "n_util_obs": 1266, "util_count_spearman": 0.1646862960242852}, "t": 1355.4} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.86542, "tr": 2.02764, "fi": 2.17082}, "bytes_per_chunk": {"en": 4.776, "tr": 5.893, "fi": 5.369}, "bpb_mean": 2.02129, "frozen_inventory": true, "t": 1356.5} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36391, "bpb": 1.96771, "chunk_loss": 0.030281, "F": 0.19531, "G": 0.20921, "chunks": 200.0, "bytes_per_chunk": 5.12, "boundary_entropy": 0.0916, "router_noise": 0.0, "grad_norm": 0.727, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.364, "probe_sec": 0.0597, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05137, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11745, "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": -3.64851, "trunc_abs_frac": 1.22703, "mean_delta_full": -0.04089, "read_hit_rate": {"surface": 0.8381}, "mean_gate": {"surface": 0.26267}, "n_reads": 5337, "n_cand": 6368, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 5337, "inventory": {"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": 469.78289794921875, "n_util_obs": 1266, "util_count_spearman": 0.16460559422859572}, "probe_overhead_sec_total": 300.937, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.034393, "alpha": 0.03, "t": 1357.6} |
| {"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1365.0} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31695, "bpb": 1.89995, "chunk_loss": 0.030306, "F": 0.19717, "G": 0.20687, "chunks": 201.91, "bytes_per_chunk": 5.072, "boundary_entropy": 0.08314, "router_noise": 0.0, "grad_norm": 0.7517, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.15, "probe_sec": 0.0308, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02495, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04879, "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": -0.72298, "trunc_abs_frac": 0.60065, "mean_delta_full": -0.14535, "read_hit_rate": {"surface": 0.85534}, "mean_gate": {"surface": 0.26854}, "n_reads": 5499, "n_cand": 6429, "read_suppressed_frac": 0.1046, "dropout_reads_logged": 5499, "inventory": {"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": 473.99822998046875, "n_util_obs": 1280, "util_count_spearman": 0.16229721963129484}, "probe_overhead_sec_total": 302.853, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.042692, "alpha": 0.03, "t": 1365.1} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1372.9} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.44239, "bpb": 2.08093, "chunk_loss": 0.030306, "F": 0.19595, "G": 0.21041, "chunks": 200.66, "bytes_per_chunk": 5.103, "boundary_entropy": 0.09494, "router_noise": 0.0, "grad_norm": 0.827, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.159, "probe_sec": 0.0243, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01881, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10085, "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": 0.8549, "trunc_abs_frac": 12.06921, "mean_delta_full": 0.34021, "read_hit_rate": {"surface": 0.81061}, "mean_gate": {"surface": 0.26438}, "n_reads": 5179, "n_cand": 6389, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 5179, "inventory": {"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": 478.24017333984375, "n_util_obs": 1288, "util_count_spearman": 0.15878471083214188}, "probe_overhead_sec_total": 305.18, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.031648, "alpha": 0.03, "t": 1373.0} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1381.8} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.26801, "bpb": 1.82935, "chunk_loss": 0.030437, "F": 0.20486, "G": 0.21485, "chunks": 209.78, "bytes_per_chunk": 4.881, "boundary_entropy": 0.08263, "router_noise": 0.0, "grad_norm": 0.7469, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.178, "probe_sec": 0.027, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02091, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12369, "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": 0.1791, "trunc_abs_frac": 0.50682, "mean_delta_full": 0.50833, "read_hit_rate": {"surface": 0.88145}, "mean_gate": {"surface": 0.2693}, "n_reads": 5889, "n_cand": 6681, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 5889, "inventory": {"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": 482.4591064453125, "n_util_obs": 1300, "util_count_spearman": 0.16667703148980337}, "probe_overhead_sec_total": 307.203, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.02616, "alpha": 0.03, "t": 1382.0} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1393.2} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3469.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199333, "cps": 27.101562, "mean_p": 0.199684, "boundary_entropy": 0.061288, "bpb": 1.840255}, "tr": {"chunks": 3218.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.177761, "cps": 25.140625, "mean_p": 0.213132, "boundary_entropy": 0.183504, "bpb": 2.011919}, "fi": {"chunks": 3480.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.184998, "cps": 27.1875, "mean_p": 0.203378, "boundary_entropy": 0.130553, "bpb": 2.158514}}, "bpb": {"en": 1.84025, "tr": 2.01192, "fi": 2.15851}, "gini_cpc": 0.025586, "gini_cps": 0.01718, "bpic": 5.580014, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8917750318486793, "fi": 0.9280837762305467}, "inventory": {"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": 486.6458740234375, "n_util_obs": 1310, "util_count_spearman": 0.1712425759344214}, "t": 1394.1} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.86092, "tr": 2.02013, "fi": 2.16723}, "bytes_per_chunk": {"en": 4.778, "tr": 5.915, "fi": 5.438}, "bpb_mean": 2.01609, "frozen_inventory": true, "t": 1395.3} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.25651, "bpb": 1.81276, "chunk_loss": 0.030359, "F": 0.20062, "G": 0.21137, "chunks": 205.44, "bytes_per_chunk": 4.984, "boundary_entropy": 0.08332, "router_noise": 0.0, "grad_norm": 0.7361, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.29, "probe_sec": 0.0688, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05697, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.51133, "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": 0.86915}, "mean_gate": {"surface": 0.26934}, "n_reads": 5686, "n_cand": 6542, "read_suppressed_frac": 0.0932, "dropout_reads_logged": 5686, "inventory": {"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": 486.7331848144531, "n_util_obs": 1310, "util_count_spearman": 0.17171715963957712}, "probe_overhead_sec_total": 310.321, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.027197, "alpha": 0.03, "t": 1396.5} |
| {"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1408.3} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37378, "bpb": 1.98195, "chunk_loss": 0.030377, "F": 0.19897, "G": 0.21492, "chunks": 203.75, "bytes_per_chunk": 5.026, "boundary_entropy": 0.09886, "router_noise": 0.0, "grad_norm": 0.7669, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.242, "probe_sec": 0.1401, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13183, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.2115, "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": -0.16518, "trunc_abs_frac": 1.35635, "mean_delta_full": -0.44885, "read_hit_rate": {"surface": 0.81473}, "mean_gate": {"surface": 0.27058}, "n_reads": 5286, "n_cand": 6488, "read_suppressed_frac": 0.1093, "dropout_reads_logged": 5286, "inventory": {"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": 490.9628601074219, "n_util_obs": 1323, "util_count_spearman": 0.17863923141026591}, "probe_overhead_sec_total": 314.174, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.029532, "alpha": 0.03, "t": 1408.6} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1420.0} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35505, "bpb": 1.95493, "chunk_loss": 0.030401, "F": 0.203, "G": 0.2135, "chunks": 207.88, "bytes_per_chunk": 4.926, "boundary_entropy": 0.08521, "router_noise": 0.0, "grad_norm": 0.7547, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.235, "probe_sec": 0.0865, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07516, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13907, "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": 2.38684, "trunc_abs_frac": 0.58645, "mean_delta_full": 0.07756, "read_hit_rate": {"surface": 0.85181}, "mean_gate": {"surface": 0.26551}, "n_reads": 5639, "n_cand": 6620, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 5639, "inventory": {"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": 495.2264404296875, "n_util_obs": 1335, "util_count_spearman": 0.1844876454343104}, "probe_overhead_sec_total": 317.712, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.037513, "alpha": 0.03, "t": 1420.3} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1431.6} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30277, "bpb": 1.87949, "chunk_loss": 0.030284, "F": 0.19464, "G": 0.20683, "chunks": 199.31, "bytes_per_chunk": 5.138, "boundary_entropy": 0.08827, "router_noise": 0.0, "grad_norm": 0.7455, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.232, "probe_sec": 0.0933, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08525, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02533, "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": 10, "trunc_ratio_sum": -1.3796, "trunc_abs_frac": 5.56388, "mean_delta_full": -0.10026, "read_hit_rate": {"surface": 0.83785}, "mean_gate": {"surface": 0.26432}, "n_reads": 5317, "n_cand": 6346, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 5317, "inventory": {"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": 499.4659423828125, "n_util_obs": 1349, "util_count_spearman": 0.1834305185654019}, "probe_overhead_sec_total": 321.122, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.032759, "alpha": 0.03, "t": 1431.9} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1444.2} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35316, "bpb": 1.9522, "chunk_loss": 0.030383, "F": 0.20215, "G": 0.21033, "chunks": 207.0, "bytes_per_chunk": 4.947, "boundary_entropy": 0.08091, "router_noise": 0.0, "grad_norm": 0.7203, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.245, "probe_sec": 0.1387, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12562, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0323, "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": 3.69234, "trunc_abs_frac": 3.69234, "mean_delta_full": 0.07727, "read_hit_rate": {"surface": 0.85983}, "mean_gate": {"surface": 0.2708}, "n_reads": 5668, "n_cand": 6592, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 5668, "inventory": {"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": 503.61553955078125, "n_util_obs": 1356, "util_count_spearman": 0.1854063732308057}, "probe_overhead_sec_total": 324.643, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.058525, "alpha": 0.03, "t": 1444.2} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3460.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198816, "cps": 27.03125, "mean_p": 0.199188, "boundary_entropy": 0.061591, "bpb": 1.839044}, "tr": {"chunks": 3199.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.176711, "cps": 24.992188, "mean_p": 0.212217, "boundary_entropy": 0.183084, "bpb": 2.004495}, "fi": {"chunks": 3450.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.183403, "cps": 26.953125, "mean_p": 0.201623, "boundary_entropy": 0.12877, "bpb": 2.151595}}, "bpb": {"en": 1.83904, "tr": 2.0045, "fi": 2.1516}, "gini_cpc": 0.026366, "gini_cps": 0.017212, "bpic": 5.612029, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8888156850699225, "fi": 0.9224763336419503}, "inventory": {"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": 503.61553955078125, "n_util_obs": 1356, "util_count_spearman": 0.1854063732308057}, "t": 1444.9} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.85868, "tr": 2.01275, "fi": 2.16081}, "bytes_per_chunk": {"en": 4.78, "tr": 5.965, "fi": 5.487}, "bpb_mean": 2.01075, "frozen_inventory": true, "t": 1446.1} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_mem-shuffled_s1/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-shuffled_s1/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-shuffled_s1/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-shuffled_s1/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-shuffled_s1/probe/reps/th0.70"], "t": 1448.2} |
| {"event": "final", "run": "ali_mem-shuffled_s1", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 1, "per_lang": {"en": {"cpc": 0.1988162960409125, "cps": 27.03125, "bpb": 1.8651111317145679, "bytes_per_chunk": 4.769159397087317, "reads_per_chunk": 0.8941173260076593}, "tr": {"cpc": 0.17671104236866816, "cps": 24.9921875, "bpb": 2.0047297433727116, "bytes_per_chunk": 5.996797364688658, "reads_per_chunk": 0.6913345838861692}, "fi": {"cpc": 0.18340332784009356, "cps": 26.953125, "bpb": 2.1788298553818586, "bytes_per_chunk": 5.472849120023382, "reads_per_chunk": 0.6875926428526692}}, "bpb_mean": 2.016223576823046, "gini_cpc": 0.02636612494422219, "gini_cps": 0.017212385003462227, "bpic": 5.612028885151845, "mean_boundary_entropy": 0.08090633898973465, "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": 324.642635345459, "probe_overhead_seq_frac": 0.4999166666666667, "wall_sec": 1450.1279060840607, "inventory": {"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": 503.61553955078125, "n_util_obs": 1356, "util_count_spearman": 0.1854063732308057}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "shuffled", "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": 0.1854063732308057, "paired_vs_dropout": {"n": 3140, "pearson": 0.050820996400106566, "spearman": 0.06833492387311173, "mean_abs_discrepancy": 0.9531561136245728, "mean_paired": 0.11052597314119339, "mean_dropout": 0.10771925747394562}, "predictor_calibration": {"n": 7219, "all": {"n": 7219, "pearson": 0.16164392239999797, "spearman": 0.23251989424959765, "mae": 0.16822393238544464, "reliability": [{"pred": -0.010745412670075893, "obs": 0.009844372048974037, "n": 722}, {"pred": 0.004790930077433586, "obs": 0.022931719198822975, "n": 722}, {"pred": 0.0144640589132905, "obs": 0.04079596698284149, "n": 722}, {"pred": 0.022795040160417557, "obs": 0.062241118401288986, "n": 722}, {"pred": 0.03137506544589996, "obs": 0.07396364957094193, "n": 721}, {"pred": 0.040689099580049515, "obs": 0.08907383680343628, "n": 722}, {"pred": 0.050651539117097855, "obs": 0.07755643129348755, "n": 722}, {"pred": 0.062350910156965256, "obs": 0.13399569690227509, "n": 722}, {"pred": 0.07960300147533417, "obs": 0.13312532007694244, "n": 722}, {"pred": 0.11734567582607269, "obs": 0.15867088735103607, "n": 722}]}, "admitted": {"n": 7219, "pearson": 0.16164392239999797, "spearman": 0.23251989424959765, "mae": 0.16822393238544464, "reliability": [{"pred": -0.010745412670075893, "obs": 0.009844372048974037, "n": 722}, {"pred": 0.004790930077433586, "obs": 0.022931719198822975, "n": 722}, {"pred": 0.0144640589132905, "obs": 0.04079596698284149, "n": 722}, {"pred": 0.022795040160417557, "obs": 0.062241118401288986, "n": 722}, {"pred": 0.03137506544589996, "obs": 0.07396364957094193, "n": 721}, {"pred": 0.040689099580049515, "obs": 0.08907383680343628, "n": 722}, {"pred": 0.050651539117097855, "obs": 0.07755643129348755, "n": 722}, {"pred": 0.062350910156965256, "obs": 0.13399569690227509, "n": 722}, {"pred": 0.07960300147533417, "obs": 0.13312532007694244, "n": 722}, {"pred": 0.11734567582607269, "obs": 0.15867088735103607, "n": 722}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 1450.2} |
| {"event": "hub_done", "repo": "AdaptiveChunking/ali_mem-shuffled_s1", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 1467.1} |
|
|
