electroglyph commited on
Commit
9f3c596
·
0 Parent(s):

squashed: phaseB progress step 29588 total 38.17h (+10.01h launch) 7.70B tok

Browse files
Files changed (7) hide show
  1. .gitattributes +35 -0
  2. body.pt +3 -0
  3. config.json +39 -0
  4. dense_opt.pt +3 -0
  5. metrics.jsonl +153 -0
  6. sparse_opt.pt +3 -0
  7. training_state.json +21 -0
.gitattributes ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ *.7z filter=lfs diff=lfs merge=lfs -text
2
+ *.arrow filter=lfs diff=lfs merge=lfs -text
3
+ *.bin filter=lfs diff=lfs merge=lfs -text
4
+ *.bz2 filter=lfs diff=lfs merge=lfs -text
5
+ *.ckpt filter=lfs diff=lfs merge=lfs -text
6
+ *.ftz filter=lfs diff=lfs merge=lfs -text
7
+ *.gz filter=lfs diff=lfs merge=lfs -text
8
+ *.h5 filter=lfs diff=lfs merge=lfs -text
9
+ *.joblib filter=lfs diff=lfs merge=lfs -text
10
+ *.lfs.* filter=lfs diff=lfs merge=lfs -text
11
+ *.mlmodel filter=lfs diff=lfs merge=lfs -text
12
+ *.model filter=lfs diff=lfs merge=lfs -text
13
+ *.msgpack filter=lfs diff=lfs merge=lfs -text
14
+ *.npy filter=lfs diff=lfs merge=lfs -text
15
+ *.npz filter=lfs diff=lfs merge=lfs -text
16
+ *.onnx filter=lfs diff=lfs merge=lfs -text
17
+ *.ot filter=lfs diff=lfs merge=lfs -text
18
+ *.parquet filter=lfs diff=lfs merge=lfs -text
19
+ *.pb filter=lfs diff=lfs merge=lfs -text
20
+ *.pickle filter=lfs diff=lfs merge=lfs -text
21
+ *.pkl filter=lfs diff=lfs merge=lfs -text
22
+ *.pt filter=lfs diff=lfs merge=lfs -text
23
+ *.pth filter=lfs diff=lfs merge=lfs -text
24
+ *.rar filter=lfs diff=lfs merge=lfs -text
25
+ *.safetensors filter=lfs diff=lfs merge=lfs -text
26
+ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
27
+ *.tar.* filter=lfs diff=lfs merge=lfs -text
28
+ *.tar filter=lfs diff=lfs merge=lfs -text
29
+ *.tflite filter=lfs diff=lfs merge=lfs -text
30
+ *.tgz filter=lfs diff=lfs merge=lfs -text
31
+ *.wasm filter=lfs diff=lfs merge=lfs -text
32
+ *.xz filter=lfs diff=lfs merge=lfs -text
33
+ *.zip filter=lfs diff=lfs merge=lfs -text
34
+ *.zst filter=lfs diff=lfs merge=lfs -text
35
+ *tfevents* filter=lfs diff=lfs merge=lfs -text
body.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7853d0a39023005394e6b14bb5f0dc6d0b3d1210418abf71a5713d771a132d22
3
+ size 8113043905
config.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "layers": 16,
3
+ "hidden": 1024,
4
+ "inter": 1536,
5
+ "heads_q": 8,
6
+ "heads_kv": 4,
7
+ "vocab": 151939,
8
+ "mask_id": 151671,
9
+ "k_heads": 8,
10
+ "n_big": 2097152,
11
+ "n_tri": 12582912,
12
+ "edim": 256,
13
+ "fusion_layers": [
14
+ 1,
15
+ 7
16
+ ],
17
+ "teacher": "jinaai/jina-embeddings-v5-text-small-retrieval",
18
+ "phase": "B",
19
+ "objective": "joint-mlm-0.20-span3-8-mask100+infoNCE",
20
+ "tokenizer": "jinaai/jina-embeddings-v5-text-small-retrieval",
21
+ "schedule_version": 1,
22
+ "books_epoch": 2,
23
+ "seq_len": 1024,
24
+ "mix": {
25
+ "web": 0.35,
26
+ "books": 0.65
27
+ },
28
+ "warmup_steps": 500,
29
+ "finalize": false,
30
+ "final_decay_steps": 20000,
31
+ "base_lr": 0.00015,
32
+ "contrast_lambda": 1.0,
33
+ "contrast_temp": 0.1,
34
+ "false_neg_sim_cap": 0.95,
35
+ "adjacent_frac": 1.0,
36
+ "min_pair_tokens": 128,
37
+ "mrl_lambda": 0.01,
38
+ "phaseA_repo": "electroglyph/berty-phaseA"
39
+ }
dense_opt.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eb6b72c6e3b930203a005ee23805b3adad48e62cedafaafe2b1fe6617022758c
3
+ size 1193754315
metrics.jsonl ADDED
@@ -0,0 +1,153 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"event": "launch", "fresh": true, "step": 0, "time": "2026-10-05T03:54:43Z"}
2
+ {"step": 245, "elapsed_h": 0.298, "seq": 1024, "lr": 7.35e-05, "sched": "warmup", "lr_s": 0.0003675, "lr_e": 3.675e-05, "loss": 5.2062, "mlm": 3.9212, "ctr": 1.285, "ctr_ema": 1.3017, "ret_acc": 0.5069, "loss_ema": 5.3414, "held_ppl": 124.327, "held_ctr": 0.1075, "held_ret_acc": 1.0, "kurt": 7.1, "alpha": {"1": 0.3538, "7": 0.1736}, "step_tok": 262144, "pairs": 144, "fn_masked": 0, "mask_rate": 0.2036, "g_d": 1.628, "g_s": 0.022, "nhev": 8, "held_n": 8, "held_draws": 4036, "fallback": true, "dup_dropped": 374, "samples_seen": {"web": 94905, "books": 692}, "gb": 46.49}
3
+ {"step": 440, "elapsed_h": 0.549, "seq": 1024, "lr": 0.00013199999999999998, "sched": "warmup", "lr_s": 0.0006599999999999999, "lr_e": 6.599999999999999e-05, "loss": 4.9685, "mlm": 4.0035, "ctr": 0.965, "ctr_ema": 0.9794, "ret_acc": 0.5793, "loss_ema": 5.0257, "held_ppl": 134.219, "held_ctr": 0.0898, "held_ret_acc": 1.0, "kurt": 6.49, "alpha": {"1": 0.3594, "7": 0.1742}, "step_tok": 262144, "pairs": 145, "fn_masked": 1, "mask_rate": 0.2042, "g_d": 2.94, "g_s": 0.025, "nhev": 8, "held_n": 23, "held_draws": 11551, "fallback": true, "dup_dropped": 2316, "samples_seen": {"web": 332501, "books": 692}, "gb": 46.49}
4
+ {"step": 649, "elapsed_h": 0.799, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.7454, "mlm": 4.0679, "ctr": 0.6775, "ctr_ema": 0.8534, "ret_acc": 0.7552, "loss_ema": 4.9393, "held_ppl": 119.771, "held_ctr": 0.1119, "held_ret_acc": 1.0, "kurt": 6.37, "alpha": {"1": 0.3702, "7": 0.18}, "step_tok": 262144, "pairs": 143, "fn_masked": 124, "mask_rate": 0.2039, "g_d": 2.611, "g_s": 0.023, "nhev": 8, "held_n": 39, "held_draws": 19749, "fallback": true, "dup_dropped": 5531, "samples_seen": {"web": 591553, "books": 692}, "gb": 46.49}
5
+ {"step": 845, "elapsed_h": 1.05, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.9947, "mlm": 4.2704, "ctr": 0.7244, "ctr_ema": 0.8408, "ret_acc": 0.6972, "loss_ema": 4.9029, "held_ppl": 125.611, "held_ctr": 0.1023, "held_ret_acc": 1.0, "kurt": 6.98, "alpha": {"1": 0.3599, "7": 0.1869}, "step_tok": 196608, "pairs": 109, "fn_masked": 0, "mask_rate": 0.2045, "g_d": 1.665, "g_s": 0.021, "nhev": 8, "held_n": 54, "held_draws": 27247, "fallback": true, "dup_dropped": 10399, "samples_seen": {"web": 831990, "books": 692}, "gb": 46.49}
6
+ {"step": 1041, "elapsed_h": 1.301, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.9562, "mlm": 4.0581, "ctr": 0.8981, "ctr_ema": 0.8043, "ret_acc": 0.6312, "loss_ema": 4.8792, "held_ppl": 109.588, "held_ctr": 0.0814, "held_ret_acc": 1.0, "kurt": 6.93, "alpha": {"1": 0.3697, "7": 0.19}, "step_tok": 262144, "pairs": 141, "fn_masked": 1, "mask_rate": 0.2034, "g_d": 2.118, "g_s": 0.024, "nhev": 8, "held_n": 64, "held_draws": 34796, "fallback": true, "dup_dropped": 15591, "samples_seen": {"web": 1072273, "books": 692}, "gb": 46.49}
7
+ {"step": 1250, "elapsed_h": 1.552, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.7404, "mlm": 4.0552, "ctr": 0.6853, "ctr_ema": 0.7937, "ret_acc": 0.6643, "loss_ema": 4.8762, "held_ppl": 122.411, "held_ctr": 0.0987, "held_ret_acc": 1.0, "kurt": 6.77, "alpha": {"1": 0.3665, "7": 0.1796}, "step_tok": 262144, "pairs": 143, "fn_masked": 3, "mask_rate": 0.2041, "g_d": 1.828, "g_s": 0.024, "nhev": 8, "held_n": 64, "held_draws": 42952, "fallback": true, "dup_dropped": 22016, "samples_seen": {"web": 1334002, "books": 692}, "gb": 46.49}
8
+ {"step": 1445, "elapsed_h": 1.802, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.8791, "mlm": 4.0887, "ctr": 0.7904, "ctr_ema": 0.7879, "ret_acc": 0.7153, "loss_ema": 4.8573, "held_ppl": 109.348, "held_ctr": 0.1094, "held_ret_acc": 1.0, "kurt": 6.83, "alpha": {"1": 0.3644, "7": 0.1884}, "step_tok": 262144, "pairs": 144, "fn_masked": 1, "mask_rate": 0.2033, "g_d": 1.793, "g_s": 0.024, "nhev": 8, "held_n": 64, "held_draws": 50457, "fallback": true, "dup_dropped": 28659, "samples_seen": {"web": 1574501, "books": 692}, "gb": 46.5}
9
+ {"step": 1563, "elapsed_h": 2.053, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6571, "mlm": 3.9786, "ctr": 0.6786, "ctr_ema": 0.7421, "ret_acc": 0.7466, "loss_ema": 4.8045, "held_ppl": 134.2, "held_ctr": 0.1039, "held_ret_acc": 1.0, "kurt": 7.08, "alpha": {"1": 0.3692, "7": 0.196}, "step_tok": 262144, "pairs": 146, "fn_masked": 0, "mask_rate": 0.2036, "g_d": 2.372, "g_s": 0.025, "nhev": 8, "held_n": 64, "held_draws": 55091, "fallback": true, "dup_dropped": 32874, "samples_seen": {"web": 1723085, "books": 692}, "gb": 46.5}
10
+ {"step": 1758, "elapsed_h": 2.304, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.7613, "mlm": 4.0738, "ctr": 0.6874, "ctr_ema": 0.7896, "ret_acc": 0.7211, "loss_ema": 4.8606, "held_ppl": 104.201, "held_ctr": 0.073, "held_ret_acc": 1.0, "kurt": 7.11, "alpha": {"1": 0.3658, "7": 0.1893}, "step_tok": 262144, "pairs": 147, "fn_masked": 2, "mask_rate": 0.2035, "g_d": 1.581, "g_s": 0.022, "nhev": 8, "held_n": 64, "held_draws": 62597, "fallback": true, "dup_dropped": 40293, "samples_seen": {"web": 1965610, "books": 692}, "gb": 46.5}
11
+ {"step": 1953, "elapsed_h": 2.554, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.8145, "mlm": 4.0895, "ctr": 0.725, "ctr_ema": 0.7299, "ret_acc": 0.7241, "loss_ema": 4.8099, "held_ppl": 108.828, "held_ctr": 0.0933, "held_ret_acc": 1.0, "kurt": 7.11, "alpha": {"1": 0.3597, "7": 0.1852}, "step_tok": 262144, "pairs": 145, "fn_masked": 0, "mask_rate": 0.2038, "g_d": 1.862, "g_s": 0.022, "nhev": 8, "held_n": 64, "held_draws": 70152, "fallback": true, "dup_dropped": 48229, "samples_seen": {"web": 2207515, "books": 692}, "gb": 46.5}
12
+ {"step": 2162, "elapsed_h": 2.805, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.7635, "mlm": 4.1081, "ctr": 0.6555, "ctr_ema": 0.7416, "ret_acc": 0.7379, "loss_ema": 4.8077, "held_ppl": 118.388, "held_ctr": 0.113, "held_ret_acc": 1.0, "kurt": 7.08, "alpha": {"1": 0.368, "7": 0.1846}, "step_tok": 262144, "pairs": 145, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.998, "g_s": 0.023, "nhev": 8, "held_n": 64, "held_draws": 78313, "fallback": true, "dup_dropped": 57441, "samples_seen": {"web": 2470103, "books": 692}, "gb": 46.5}
13
+ {"step": 2356, "elapsed_h": 3.056, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 5.1148, "mlm": 4.2355, "ctr": 0.8793, "ctr_ema": 0.7574, "ret_acc": 0.6597, "loss_ema": 4.8235, "held_ppl": 128.16, "held_ctr": 0.0851, "held_ret_acc": 1.0, "kurt": 7.18, "alpha": {"1": 0.3644, "7": 0.1884}, "step_tok": 262144, "pairs": 144, "fn_masked": 0, "mask_rate": 0.2034, "g_d": 1.904, "g_s": 0.024, "nhev": 8, "held_n": 64, "held_draws": 85815, "fallback": true, "dup_dropped": 66752, "samples_seen": {"web": 2715826, "books": 692}, "gb": 46.5}
14
+ {"step": 2565, "elapsed_h": 3.306, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 5.0919, "mlm": 4.1252, "ctr": 0.9667, "ctr_ema": 0.7428, "ret_acc": 0.6667, "loss_ema": 4.8156, "held_ppl": 136.417, "held_ctr": 0.0756, "held_ret_acc": 1.0, "kurt": 7.21, "alpha": {"1": 0.373, "7": 0.1942}, "step_tok": 262144, "pairs": 144, "fn_masked": 3, "mask_rate": 0.2037, "g_d": 1.947, "g_s": 0.023, "nhev": 8, "held_n": 64, "held_draws": 93925, "fallback": true, "dup_dropped": 77646, "samples_seen": {"web": 2982587, "books": 692}, "gb": 46.5}
15
+ {"step": 2760, "elapsed_h": 3.558, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4947, "mlm": 3.8131, "ctr": 0.6816, "ctr_ema": 0.7142, "ret_acc": 0.75, "loss_ema": 4.7317, "held_ppl": 117.878, "held_ctr": 0.1518, "held_ret_acc": 1.0, "kurt": 6.87, "alpha": {"1": 0.3629, "7": 0.1895}, "step_tok": 262144, "pairs": 144, "fn_masked": 0, "mask_rate": 0.2036, "g_d": 1.751, "g_s": 0.023, "nhev": 8, "held_n": 64, "held_draws": 101419, "fallback": true, "dup_dropped": 87758, "samples_seen": {"web": 3225017, "books": 692}, "gb": 46.5}
16
+ {"step": 2951, "elapsed_h": 3.808, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.8493, "mlm": 4.0699, "ctr": 0.7794, "ctr_ema": 0.7209, "ret_acc": 0.6507, "loss_ema": 4.7696, "held_ppl": 122.838, "held_ctr": 0.1102, "held_ret_acc": 1.0, "kurt": 7.07, "alpha": {"1": 0.3539, "7": 0.1907}, "step_tok": 262144, "pairs": 146, "fn_masked": 0, "mask_rate": 0.2037, "g_d": 1.76, "g_s": 0.024, "nhev": 8, "held_n": 64, "held_draws": 108732, "fallback": true, "dup_dropped": 98509, "samples_seen": {"web": 3468778, "books": 692}, "gb": 46.5}
17
+ {"step": 3161, "elapsed_h": 4.059, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 5.1399, "mlm": 4.1899, "ctr": 0.95, "ctr_ema": 0.7522, "ret_acc": 0.7055, "loss_ema": 4.8432, "held_ppl": 115.894, "held_ctr": 0.0936, "held_ret_acc": 1.0, "kurt": 7.29, "alpha": {"1": 0.3691, "7": 0.1932}, "step_tok": 262144, "pairs": 146, "fn_masked": 0, "mask_rate": 0.2035, "g_d": 2.21, "g_s": 0.029, "nhev": 8, "held_n": 64, "held_draws": 116749, "fallback": true, "dup_dropped": 110374, "samples_seen": {"web": 3725976, "books": 692}, "gb": 46.5}
18
+ {"event": "launch", "fresh": false, "step": 3263, "time": "2026-10-05T09:13:44Z"}
19
+ {"step": 3320, "elapsed_h": 0.172, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6333, "mlm": 4.1271, "ctr": 0.5062, "ctr_ema": 0.4474, "ret_acc": 0.7935, "loss_ema": 4.591, "held_ppl": 45.78, "held_ctr": 0.0321, "held_ret_acc": 1.0, "kurt": 6.89, "alpha": {"1": 0.3794, "7": 0.1825}, "step_tok": 262144, "pairs": 155, "fn_masked": 0, "mask_rate": 0.2034, "g_d": 1.515, "g_s": 0.019, "nhev": 7, "held_n": 8, "held_draws": 4006, "fallback": true, "dup_dropped": 110, "samples_seen": {"web": 3900567, "books": 692}, "gb": 46.49}
20
+ {"step": 3528, "elapsed_h": 0.422, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.2844, "mlm": 4.0466, "ctr": 0.2378, "ctr_ema": 0.4355, "ret_acc": 0.9193, "loss_ema": 4.5756, "held_ppl": 43.896, "held_ctr": 0.0291, "held_ret_acc": 1.0, "kurt": 7.33, "alpha": {"1": 0.3782, "7": 0.18}, "step_tok": 262144, "pairs": 161, "fn_masked": 0, "mask_rate": 0.2038, "g_d": 1.285, "g_s": 0.019, "nhev": 8, "held_n": 36, "held_draws": 18238, "fallback": true, "dup_dropped": 1155, "samples_seen": {"web": 4057412, "books": 692}, "gb": 46.49}
21
+ {"step": 3718, "elapsed_h": 0.673, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3883, "mlm": 3.9116, "ctr": 0.4767, "ctr_ema": 0.4516, "ret_acc": 0.8039, "loss_ema": 4.5431, "held_ppl": 48.456, "held_ctr": 0.0168, "held_ret_acc": 1.0, "kurt": 6.96, "alpha": {"1": 0.379, "7": 0.1876}, "step_tok": 262144, "pairs": 153, "fn_masked": 0, "mask_rate": 0.2038, "g_d": 1.786, "g_s": 0.022, "nhev": 8, "held_n": 62, "held_draws": 31072, "fallback": true, "dup_dropped": 2628, "samples_seen": {"web": 4200692, "books": 692}, "gb": 46.49}
22
+ {"step": 3930, "elapsed_h": 0.924, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 5.206, "mlm": 3.9607, "ctr": 1.2453, "ctr_ema": 0.5033, "ret_acc": 0.6552, "loss_ema": 4.6431, "held_ppl": 48.228, "held_ctr": 0.0183, "held_ret_acc": 1.0, "kurt": 7.15, "alpha": {"1": 0.3767, "7": 0.1797}, "step_tok": 262144, "pairs": 145, "fn_masked": 0, "mask_rate": 0.2038, "g_d": 2.28, "g_s": 0.023, "nhev": 8, "held_n": 64, "held_draws": 45409, "fallback": true, "dup_dropped": 4645, "samples_seen": {"web": 4361347, "books": 692}, "gb": 46.49}
23
+ {"step": 4125, "elapsed_h": 1.175, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4971, "mlm": 4.1214, "ctr": 0.3756, "ctr_ema": 0.4616, "ret_acc": 0.8831, "loss_ema": 4.602, "held_ppl": 50.714, "held_ctr": 0.0274, "held_ret_acc": 1.0, "kurt": 7.16, "alpha": {"1": 0.3667, "7": 0.1784}, "step_tok": 262144, "pairs": 154, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.218, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 58691, "fallback": true, "dup_dropped": 6821, "samples_seen": {"web": 4513041, "books": 692}, "gb": 46.5}
24
+ {"step": 4317, "elapsed_h": 1.425, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6427, "mlm": 4.1831, "ctr": 0.4596, "ctr_ema": 0.4809, "ret_acc": 0.8312, "loss_ema": 4.5815, "held_ppl": 42.473, "held_ctr": 0.0138, "held_ret_acc": 1.0, "kurt": 7.05, "alpha": {"1": 0.38, "7": 0.1804}, "step_tok": 262144, "pairs": 154, "fn_masked": 0, "mask_rate": 0.2043, "g_d": 1.535, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 71622, "fallback": true, "dup_dropped": 9614, "samples_seen": {"web": 4660236, "books": 692}, "gb": 46.5}
25
+ {"step": 4528, "elapsed_h": 1.676, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 5.0076, "mlm": 4.3531, "ctr": 0.6545, "ctr_ema": 0.4356, "ret_acc": 0.7451, "loss_ema": 4.5501, "held_ppl": 51.921, "held_ctr": 0.021, "held_ret_acc": 1.0, "kurt": 7.2, "alpha": {"1": 0.3791, "7": 0.1778}, "step_tok": 262144, "pairs": 153, "fn_masked": 0, "mask_rate": 0.204, "g_d": 2.987, "g_s": 0.027, "nhev": 8, "held_n": 64, "held_draws": 86127, "fallback": true, "dup_dropped": 13141, "samples_seen": {"web": 4827156, "books": 692}, "gb": 46.5}
26
+ {"step": 4722, "elapsed_h": 1.926, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.614, "mlm": 4.0886, "ctr": 0.5254, "ctr_ema": 0.4653, "ret_acc": 0.8038, "loss_ema": 4.6032, "held_ppl": 42.026, "held_ctr": 0.0305, "held_ret_acc": 1.0, "kurt": 6.87, "alpha": {"1": 0.3787, "7": 0.1815}, "step_tok": 262144, "pairs": 158, "fn_masked": 0, "mask_rate": 0.2041, "g_d": 1.389, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 99295, "fallback": true, "dup_dropped": 16806, "samples_seen": {"web": 4978249, "books": 692}, "gb": 46.5}
27
+ {"step": 4931, "elapsed_h": 2.177, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4922, "mlm": 4.0815, "ctr": 0.4107, "ctr_ema": 0.4571, "ret_acc": 0.891, "loss_ema": 4.5563, "held_ppl": 47.609, "held_ctr": 0.0262, "held_ret_acc": 1.0, "kurt": 6.92, "alpha": {"1": 0.3777, "7": 0.1823}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.501, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 113331, "fallback": true, "dup_dropped": 20689, "samples_seen": {"web": 5133865, "books": 692}, "gb": 46.5}
28
+ {"step": 5125, "elapsed_h": 2.428, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6591, "mlm": 4.161, "ctr": 0.4981, "ctr_ema": 0.444, "ret_acc": 0.8333, "loss_ema": 4.5552, "held_ppl": 47.333, "held_ctr": 0.0259, "held_ret_acc": 1.0, "kurt": 7.28, "alpha": {"1": 0.3828, "7": 0.1827}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.2038, "g_d": 1.602, "g_s": 0.023, "nhev": 8, "held_n": 64, "held_draws": 126449, "fallback": true, "dup_dropped": 24669, "samples_seen": {"web": 5281996, "books": 692}, "gb": 46.5}
29
+ {"step": 5320, "elapsed_h": 2.679, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6333, "mlm": 4.309, "ctr": 0.3243, "ctr_ema": 0.4739, "ret_acc": 0.8839, "loss_ema": 4.6018, "held_ppl": 42.449, "held_ctr": 0.026, "held_ret_acc": 1.0, "kurt": 6.87, "alpha": {"1": 0.3866, "7": 0.1789}, "step_tok": 262144, "pairs": 155, "fn_masked": 0, "mask_rate": 0.2038, "g_d": 2.527, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 139486, "fallback": true, "dup_dropped": 28595, "samples_seen": {"web": 5428927, "books": 692}, "gb": 46.5}
30
+ {"step": 5531, "elapsed_h": 2.929, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4779, "mlm": 4.1099, "ctr": 0.368, "ctr_ema": 0.4182, "ret_acc": 0.9026, "loss_ema": 4.52, "held_ppl": 47.577, "held_ctr": 0.0189, "held_ret_acc": 1.0, "kurt": 6.89, "alpha": {"1": 0.3869, "7": 0.1823}, "step_tok": 262144, "pairs": 154, "fn_masked": 0, "mask_rate": 0.204, "g_d": 1.38, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 153907, "fallback": true, "dup_dropped": 33263, "samples_seen": {"web": 5594637, "books": 692}, "gb": 46.5}
31
+ {"step": 5723, "elapsed_h": 3.18, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.832, "mlm": 4.0547, "ctr": 0.7772, "ctr_ema": 0.4384, "ret_acc": 0.7397, "loss_ema": 4.5479, "held_ppl": 46.697, "held_ctr": 0.0277, "held_ret_acc": 1.0, "kurt": 6.67, "alpha": {"1": 0.3839, "7": 0.1857}, "step_tok": 262144, "pairs": 146, "fn_masked": 5, "mask_rate": 0.2035, "g_d": 1.73, "g_s": 0.023, "nhev": 8, "held_n": 64, "held_draws": 166951, "fallback": true, "dup_dropped": 37737, "samples_seen": {"web": 5743311, "books": 692}, "gb": 46.5}
32
+ {"step": 5934, "elapsed_h": 3.431, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6671, "mlm": 4.2381, "ctr": 0.429, "ctr_ema": 0.4545, "ret_acc": 0.8333, "loss_ema": 4.5844, "held_ppl": 40.921, "held_ctr": 0.0327, "held_ret_acc": 1.0, "kurt": 6.63, "alpha": {"1": 0.3808, "7": 0.1775}, "step_tok": 262144, "pairs": 156, "fn_masked": 1, "mask_rate": 0.2037, "g_d": 1.418, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 181223, "fallback": true, "dup_dropped": 42946, "samples_seen": {"web": 5906571, "books": 692}, "gb": 46.5}
33
+ {"step": 6129, "elapsed_h": 3.681, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4377, "mlm": 4.0245, "ctr": 0.4133, "ctr_ema": 0.4526, "ret_acc": 0.828, "loss_ema": 4.5849, "held_ppl": 51.215, "held_ctr": 0.0306, "held_ret_acc": 1.0, "kurt": 6.9, "alpha": {"1": 0.3806, "7": 0.1834}, "step_tok": 262144, "pairs": 157, "fn_masked": 0, "mask_rate": 0.204, "g_d": 1.294, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 194531, "fallback": true, "dup_dropped": 48030, "samples_seen": {"web": 6058444, "books": 692}, "gb": 46.5}
34
+ {"step": 6321, "elapsed_h": 3.932, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.531, "mlm": 4.1598, "ctr": 0.3712, "ctr_ema": 0.4465, "ret_acc": 0.8688, "loss_ema": 4.5708, "held_ppl": 47.922, "held_ctr": 0.0296, "held_ret_acc": 1.0, "kurt": 7.15, "alpha": {"1": 0.3823, "7": 0.1795}, "step_tok": 262144, "pairs": 160, "fn_masked": 0, "mask_rate": 0.2041, "g_d": 1.236, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 207528, "fallback": true, "dup_dropped": 53114, "samples_seen": {"web": 6206277, "books": 692}, "gb": 46.51}
35
+ {"step": 6529, "elapsed_h": 4.182, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4535, "mlm": 4.1621, "ctr": 0.2914, "ctr_ema": 0.4393, "ret_acc": 0.8774, "loss_ema": 4.5435, "held_ppl": 44.504, "held_ctr": 0.023, "held_ret_acc": 1.0, "kurt": 6.61, "alpha": {"1": 0.3927, "7": 0.1826}, "step_tok": 262144, "pairs": 155, "fn_masked": 0, "mask_rate": 0.2036, "g_d": 1.183, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 221636, "fallback": true, "dup_dropped": 58906, "samples_seen": {"web": 6366510, "books": 692}, "gb": 46.51}
36
+ {"step": 6724, "elapsed_h": 4.433, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6369, "mlm": 4.2769, "ctr": 0.36, "ctr_ema": 0.4389, "ret_acc": 0.8642, "loss_ema": 4.5628, "held_ppl": 51.533, "held_ctr": 0.0218, "held_ret_acc": 1.0, "kurt": 6.83, "alpha": {"1": 0.3883, "7": 0.1839}, "step_tok": 262144, "pairs": 162, "fn_masked": 0, "mask_rate": 0.2041, "g_d": 1.209, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 234908, "fallback": true, "dup_dropped": 64461, "samples_seen": {"web": 6518711, "books": 692}, "gb": 46.51}
37
+ {"step": 6936, "elapsed_h": 4.684, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6465, "mlm": 4.139, "ctr": 0.5075, "ctr_ema": 0.4505, "ret_acc": 0.8387, "loss_ema": 4.5832, "held_ppl": 54.112, "held_ctr": 0.0209, "held_ret_acc": 1.0, "kurt": 6.86, "alpha": {"1": 0.3795, "7": 0.1774}, "step_tok": 262144, "pairs": 155, "fn_masked": 0, "mask_rate": 0.2042, "g_d": 1.699, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 249156, "fallback": true, "dup_dropped": 70811, "samples_seen": {"web": 6682809, "books": 692}, "gb": 46.51}
38
+ {"step": 7125, "elapsed_h": 4.935, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.414, "mlm": 4.1026, "ctr": 0.3114, "ctr_ema": 0.4443, "ret_acc": 0.8625, "loss_ema": 4.5653, "held_ppl": 47.569, "held_ctr": 0.0324, "held_ret_acc": 1.0, "kurt": 6.89, "alpha": {"1": 0.377, "7": 0.1776}, "step_tok": 262144, "pairs": 160, "fn_masked": 1, "mask_rate": 0.2037, "g_d": 1.155, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 261975, "fallback": true, "dup_dropped": 76695, "samples_seen": {"web": 6829955, "books": 692}, "gb": 46.51}
39
+ {"step": 7320, "elapsed_h": 5.185, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4593, "mlm": 4.039, "ctr": 0.4203, "ctr_ema": 0.4463, "ret_acc": 0.8411, "loss_ema": 4.5552, "held_ppl": 46.006, "held_ctr": 0.0247, "held_ret_acc": 1.0, "kurt": 6.86, "alpha": {"1": 0.3828, "7": 0.1811}, "step_tok": 262144, "pairs": 151, "fn_masked": 0, "mask_rate": 0.204, "g_d": 1.526, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 275082, "fallback": true, "dup_dropped": 82809, "samples_seen": {"web": 6979352, "books": 692}, "gb": 46.51}
40
+ {"step": 7530, "elapsed_h": 5.435, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6885, "mlm": 4.1141, "ctr": 0.5745, "ctr_ema": 0.3936, "ret_acc": 0.7818, "loss_ema": 4.5069, "held_ppl": 45.74, "held_ctr": 0.0242, "held_ret_acc": 1.0, "kurt": 6.73, "alpha": {"1": 0.3828, "7": 0.1824}, "step_tok": 196608, "pairs": 110, "fn_masked": 3, "mask_rate": 0.2034, "g_d": 1.635, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 289482, "fallback": true, "dup_dropped": 90178, "samples_seen": {"web": 7147369, "books": 692}, "gb": 46.51}
41
+ {"step": 7724, "elapsed_h": 5.686, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4781, "mlm": 3.9976, "ctr": 0.4805, "ctr_ema": 0.4376, "ret_acc": 0.8456, "loss_ema": 4.5353, "held_ppl": 39.382, "held_ctr": 0.0199, "held_ret_acc": 1.0, "kurt": 7.02, "alpha": {"1": 0.379, "7": 0.1817}, "step_tok": 262144, "pairs": 149, "fn_masked": 3, "mask_rate": 0.2034, "g_d": 1.915, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 302596, "fallback": true, "dup_dropped": 96922, "samples_seen": {"web": 7299796, "books": 692}, "gb": 46.51}
42
+ {"step": 7932, "elapsed_h": 5.936, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3322, "mlm": 3.9899, "ctr": 0.3423, "ctr_ema": 0.3979, "ret_acc": 0.8924, "loss_ema": 4.5327, "held_ppl": 43.222, "held_ctr": 0.032, "held_ret_acc": 1.0, "kurt": 6.74, "alpha": {"1": 0.383, "7": 0.1787}, "step_tok": 262144, "pairs": 158, "fn_masked": 0, "mask_rate": 0.2041, "g_d": 1.272, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 316670, "fallback": true, "dup_dropped": 104336, "samples_seen": {"web": 7461431, "books": 692}, "gb": 46.51}
43
+ {"step": 8127, "elapsed_h": 6.186, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.867, "mlm": 4.3783, "ctr": 0.4886, "ctr_ema": 0.4192, "ret_acc": 0.8182, "loss_ema": 4.544, "held_ppl": 41.732, "held_ctr": 0.0237, "held_ret_acc": 1.0, "kurt": 7.17, "alpha": {"1": 0.3736, "7": 0.1833}, "step_tok": 262144, "pairs": 154, "fn_masked": 2, "mask_rate": 0.2043, "g_d": 1.269, "g_s": 0.017, "nhev": 8, "held_n": 64, "held_draws": 329802, "fallback": true, "dup_dropped": 111575, "samples_seen": {"web": 7614262, "books": 692}, "gb": 46.51}
44
+ {"step": 8321, "elapsed_h": 6.437, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5719, "mlm": 4.1542, "ctr": 0.4177, "ctr_ema": 0.4178, "ret_acc": 0.8645, "loss_ema": 4.5467, "held_ppl": 50.984, "held_ctr": 0.0245, "held_ret_acc": 1.0, "kurt": 7.22, "alpha": {"1": 0.3833, "7": 0.1818}, "step_tok": 262144, "pairs": 155, "fn_masked": 0, "mask_rate": 0.204, "g_d": 1.259, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 342918, "fallback": true, "dup_dropped": 118989, "samples_seen": {"web": 7764810, "books": 692}, "gb": 46.51}
45
+ {"step": 8530, "elapsed_h": 6.688, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5784, "mlm": 4.1945, "ctr": 0.3838, "ctr_ema": 0.43, "ret_acc": 0.859, "loss_ema": 4.5716, "held_ppl": 40.073, "held_ctr": 0.0273, "held_ret_acc": 1.0, "kurt": 7.2, "alpha": {"1": 0.3751, "7": 0.1778}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.2036, "g_d": 1.448, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 357006, "fallback": true, "dup_dropped": 127314, "samples_seen": {"web": 7926165, "books": 692}, "gb": 46.51}
46
+ {"step": 8725, "elapsed_h": 6.939, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.9667, "mlm": 4.2776, "ctr": 0.6892, "ctr_ema": 0.4669, "ret_acc": 0.7632, "loss_ema": 4.6256, "held_ppl": 45.19, "held_ctr": 0.0449, "held_ret_acc": 1.0, "kurt": 7.02, "alpha": {"1": 0.377, "7": 0.1845}, "step_tok": 262144, "pairs": 152, "fn_masked": 4, "mask_rate": 0.2039, "g_d": 1.375, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 370320, "fallback": true, "dup_dropped": 134930, "samples_seen": {"web": 8077189, "books": 692}, "gb": 46.51}
47
+ {"step": 8936, "elapsed_h": 7.19, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5041, "mlm": 4.061, "ctr": 0.4431, "ctr_ema": 0.4339, "ret_acc": 0.8182, "loss_ema": 4.5634, "held_ppl": 49.579, "held_ctr": 0.031, "held_ret_acc": 1.0, "kurt": 6.81, "alpha": {"1": 0.3815, "7": 0.1815}, "step_tok": 262144, "pairs": 154, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.532, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 384502, "fallback": true, "dup_dropped": 143781, "samples_seen": {"web": 8241487, "books": 692}, "gb": 46.51}
48
+ {"step": 9129, "elapsed_h": 7.441, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.8909, "mlm": 4.418, "ctr": 0.473, "ctr_ema": 0.4499, "ret_acc": 0.7834, "loss_ema": 4.607, "held_ppl": 37.61, "held_ctr": 0.0312, "held_ret_acc": 1.0, "kurt": 6.73, "alpha": {"1": 0.3853, "7": 0.1838}, "step_tok": 262144, "pairs": 157, "fn_masked": 0, "mask_rate": 0.2033, "g_d": 1.391, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 397474, "fallback": true, "dup_dropped": 151890, "samples_seen": {"web": 8393318, "books": 692}, "gb": 46.51}
49
+ {"step": 9320, "elapsed_h": 7.691, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3132, "mlm": 3.9715, "ctr": 0.3417, "ctr_ema": 0.4416, "ret_acc": 0.8544, "loss_ema": 4.5557, "held_ppl": 40.419, "held_ctr": 0.0342, "held_ret_acc": 1.0, "kurt": 6.76, "alpha": {"1": 0.3719, "7": 0.1853}, "step_tok": 262144, "pairs": 158, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.326, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 410455, "fallback": true, "dup_dropped": 160444, "samples_seen": {"web": 8546007, "books": 692}, "gb": 46.51}
50
+ {"step": 9530, "elapsed_h": 7.941, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3778, "mlm": 4.0585, "ctr": 0.3193, "ctr_ema": 0.4314, "ret_acc": 0.8917, "loss_ema": 4.5942, "held_ppl": 53.446, "held_ctr": 0.0315, "held_ret_acc": 1.0, "kurt": 6.94, "alpha": {"1": 0.3813, "7": 0.1809}, "step_tok": 262144, "pairs": 157, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.294, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 424674, "fallback": true, "dup_dropped": 169926, "samples_seen": {"web": 8711707, "books": 692}, "gb": 46.51}
51
+ {"step": 9725, "elapsed_h": 8.192, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4238, "mlm": 4.1761, "ctr": 0.2477, "ctr_ema": 0.427, "ret_acc": 0.9032, "loss_ema": 4.5611, "held_ppl": 41.598, "held_ctr": 0.0357, "held_ret_acc": 1.0, "kurt": 6.74, "alpha": {"1": 0.3722, "7": 0.1805}, "step_tok": 196608, "pairs": 124, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 0.924, "g_s": 0.015, "nhev": 8, "held_n": 64, "held_draws": 437618, "fallback": true, "dup_dropped": 178729, "samples_seen": {"web": 8864441, "books": 692}, "gb": 46.51}
52
+ {"step": 9936, "elapsed_h": 8.443, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5082, "mlm": 4.1976, "ctr": 0.3106, "ctr_ema": 0.4437, "ret_acc": 0.9038, "loss_ema": 4.5607, "held_ppl": 43.173, "held_ctr": 0.0258, "held_ret_acc": 1.0, "kurt": 6.62, "alpha": {"1": 0.3853, "7": 0.1793}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.2038, "g_d": 1.351, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 451836, "fallback": true, "dup_dropped": 188570, "samples_seen": {"web": 9030423, "books": 692}, "gb": 46.52}
53
+ {"step": 10128, "elapsed_h": 8.693, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.468, "mlm": 4.2076, "ctr": 0.2605, "ctr_ema": 0.4024, "ret_acc": 0.9125, "loss_ema": 4.5167, "held_ppl": 40.871, "held_ctr": 0.0193, "held_ret_acc": 1.0, "kurt": 6.69, "alpha": {"1": 0.3811, "7": 0.1876}, "step_tok": 262144, "pairs": 160, "fn_masked": 0, "mask_rate": 0.2042, "g_d": 1.34, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 464860, "fallback": true, "dup_dropped": 197813, "samples_seen": {"web": 9182191, "books": 692}, "gb": 46.52}
54
+ {"step": 10323, "elapsed_h": 8.943, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3878, "mlm": 4.0359, "ctr": 0.3519, "ctr_ema": 0.4203, "ret_acc": 0.8571, "loss_ema": 4.524, "held_ppl": 50.456, "held_ctr": 0.0231, "held_ret_acc": 1.0, "kurt": 6.82, "alpha": {"1": 0.3803, "7": 0.1875}, "step_tok": 262144, "pairs": 154, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.389, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 477952, "fallback": true, "dup_dropped": 207521, "samples_seen": {"web": 9336365, "books": 692}, "gb": 46.52}
55
+ {"step": 10534, "elapsed_h": 9.194, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.7472, "mlm": 4.1704, "ctr": 0.5768, "ctr_ema": 0.4269, "ret_acc": 0.7911, "loss_ema": 4.5535, "held_ppl": 40.006, "held_ctr": 0.0315, "held_ret_acc": 1.0, "kurt": 6.83, "alpha": {"1": 0.386, "7": 0.1832}, "step_tok": 262144, "pairs": 158, "fn_masked": 0, "mask_rate": 0.2036, "g_d": 1.475, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 491943, "fallback": true, "dup_dropped": 217964, "samples_seen": {"web": 9501591, "books": 692}, "gb": 46.52}
56
+ {"step": 10726, "elapsed_h": 9.445, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5211, "mlm": 4.2359, "ctr": 0.2852, "ctr_ema": 0.4449, "ret_acc": 0.8981, "loss_ema": 4.5737, "held_ppl": 49.132, "held_ctr": 0.0222, "held_ret_acc": 1.0, "kurt": 7.15, "alpha": {"1": 0.3794, "7": 0.1839}, "step_tok": 262144, "pairs": 157, "fn_masked": 0, "mask_rate": 0.2038, "g_d": 1.26, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 504836, "fallback": true, "dup_dropped": 227789, "samples_seen": {"web": 9653312, "books": 692}, "gb": 46.52}
57
+ {"step": 10937, "elapsed_h": 9.696, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.2956, "mlm": 3.9891, "ctr": 0.3065, "ctr_ema": 0.3918, "ret_acc": 0.8645, "loss_ema": 4.4959, "held_ppl": 45.908, "held_ctr": 0.0194, "held_ret_acc": 1.0, "kurt": 6.89, "alpha": {"1": 0.3839, "7": 0.1799}, "step_tok": 262144, "pairs": 155, "fn_masked": 1, "mask_rate": 0.2039, "g_d": 1.178, "g_s": 0.017, "nhev": 8, "held_n": 64, "held_draws": 519229, "fallback": true, "dup_dropped": 239001, "samples_seen": {"web": 9824248, "books": 692}, "gb": 46.52}
58
+ {"step": 11132, "elapsed_h": 9.946, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.45, "mlm": 4.1169, "ctr": 0.3331, "ctr_ema": 0.4356, "ret_acc": 0.8766, "loss_ema": 4.5364, "held_ppl": 49.869, "held_ctr": 0.0177, "held_ret_acc": 1.0, "kurt": 6.62, "alpha": {"1": 0.3858, "7": 0.1814}, "step_tok": 262144, "pairs": 154, "fn_masked": 0, "mask_rate": 0.2034, "g_d": 1.535, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 532264, "fallback": true, "dup_dropped": 249548, "samples_seen": {"web": 9978550, "books": 692}, "gb": 46.52}
59
+ {"step": 11326, "elapsed_h": 10.197, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3908, "mlm": 4.1307, "ctr": 0.2601, "ctr_ema": 0.4221, "ret_acc": 0.8861, "loss_ema": 4.5425, "held_ppl": 40.513, "held_ctr": 0.0261, "held_ret_acc": 1.0, "kurt": 7.31, "alpha": {"1": 0.3809, "7": 0.1818}, "step_tok": 262144, "pairs": 158, "fn_masked": 0, "mask_rate": 0.204, "g_d": 1.197, "g_s": 0.017, "nhev": 8, "held_n": 64, "held_draws": 545341, "fallback": true, "dup_dropped": 260413, "samples_seen": {"web": 10135163, "books": 692}, "gb": 46.52}
60
+ {"step": 11535, "elapsed_h": 10.447, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.578, "mlm": 4.1411, "ctr": 0.4369, "ctr_ema": 0.446, "ret_acc": 0.8562, "loss_ema": 4.5646, "held_ppl": 48.026, "held_ctr": 0.0253, "held_ret_acc": 1.0, "kurt": 7.04, "alpha": {"1": 0.3893, "7": 0.1811}, "step_tok": 262144, "pairs": 153, "fn_masked": 0, "mask_rate": 0.204, "g_d": 1.259, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 559430, "fallback": true, "dup_dropped": 271875, "samples_seen": {"web": 10300712, "books": 692}, "gb": 46.52}
61
+ {"step": 11728, "elapsed_h": 10.698, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6376, "mlm": 4.0899, "ctr": 0.5478, "ctr_ema": 0.4666, "ret_acc": 0.7821, "loss_ema": 4.608, "held_ppl": 40.582, "held_ctr": 0.0223, "held_ret_acc": 1.0, "kurt": 6.86, "alpha": {"1": 0.3804, "7": 0.1805}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.204, "g_d": 1.436, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 572334, "fallback": true, "dup_dropped": 282839, "samples_seen": {"web": 10453273, "books": 692}, "gb": 46.52}
62
+ {"step": 11939, "elapsed_h": 10.949, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3302, "mlm": 4.0164, "ctr": 0.3138, "ctr_ema": 0.4017, "ret_acc": 0.8805, "loss_ema": 4.5262, "held_ppl": 61.348, "held_ctr": 0.0228, "held_ret_acc": 1.0, "kurt": 6.86, "alpha": {"1": 0.3744, "7": 0.187}, "step_tok": 262144, "pairs": 159, "fn_masked": 1, "mask_rate": 0.2042, "g_d": 1.274, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 586566, "fallback": true, "dup_dropped": 295460, "samples_seen": {"web": 10622736, "books": 692}, "gb": 46.52}
63
+ {"event": "launch", "fresh": false, "step": 12104, "time": "2026-10-05T21:54:25Z"}
64
+ {"step": 12164, "elapsed_h": 0.263, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4938, "mlm": 4.1003, "ctr": 0.3935, "ctr_ema": 0.4297, "ret_acc": 0.8385, "loss_ema": 4.5728, "held_ppl": 53.864, "held_ctr": 0.0346, "held_ret_acc": 1.0, "kurt": 7.53, "alpha": {"1": 0.3791, "7": 0.226}, "step_tok": 262144, "pairs": 161, "fn_masked": 0, "mask_rate": 0.204, "g_d": 1.327, "g_s": 0.02, "nhev": 8, "held_n": 8, "held_draws": 4059, "fallback": true, "dup_dropped": 92, "pack_dropped": 2, "samples_seen": {"web": 10799565, "books": 692}, "gb": 43.91}
65
+ {"step": 12375, "elapsed_h": 0.513, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4952, "mlm": 4.134, "ctr": 0.3612, "ctr_ema": 0.4156, "ret_acc": 0.879, "loss_ema": 4.5355, "held_ppl": 62.698, "held_ctr": 0.0423, "held_ret_acc": 1.0, "kurt": 7.24, "alpha": {"1": 0.3834, "7": 0.2285}, "step_tok": 262144, "pairs": 157, "fn_masked": 0, "mask_rate": 0.2036, "g_d": 1.307, "g_s": 0.02, "nhev": 8, "held_n": 36, "held_draws": 18232, "fallback": true, "dup_dropped": 1113, "pack_dropped": 8, "samples_seen": {"web": 10956999, "books": 692}, "gb": 46.48}
66
+ {"step": 12568, "elapsed_h": 0.764, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4606, "mlm": 4.1447, "ctr": 0.3159, "ctr_ema": 0.3749, "ret_acc": 0.8938, "loss_ema": 4.5115, "held_ppl": 53.647, "held_ctr": 0.0295, "held_ret_acc": 1.0, "kurt": 7.22, "alpha": {"1": 0.3844, "7": 0.2231}, "step_tok": 262144, "pairs": 160, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.542, "g_s": 0.02, "nhev": 8, "held_n": 62, "held_draws": 31398, "fallback": true, "dup_dropped": 2386, "pack_dropped": 13, "samples_seen": {"web": 11105124, "books": 692}, "gb": 46.5}
67
+ {"step": 12778, "elapsed_h": 1.015, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4989, "mlm": 3.9664, "ctr": 0.5326, "ctr_ema": 0.4438, "ret_acc": 0.7974, "loss_ema": 4.5532, "held_ppl": 53.81, "held_ctr": 0.0271, "held_ret_acc": 1.0, "kurt": 7.15, "alpha": {"1": 0.3815, "7": 0.2249}, "step_tok": 262144, "pairs": 153, "fn_masked": 5, "mask_rate": 0.204, "g_d": 1.379, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 45537, "fallback": true, "dup_dropped": 4145, "pack_dropped": 18, "samples_seen": {"web": 11263794, "books": 692}, "gb": 46.5}
68
+ {"step": 12971, "elapsed_h": 1.265, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5526, "mlm": 4.07, "ctr": 0.4827, "ctr_ema": 0.4275, "ret_acc": 0.8013, "loss_ema": 4.5481, "held_ppl": 50.86, "held_ctr": 0.0236, "held_ret_acc": 1.0, "kurt": 7.65, "alpha": {"1": 0.3826, "7": 0.2243}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.2037, "g_d": 1.631, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 58689, "fallback": true, "dup_dropped": 6605, "pack_dropped": 25, "samples_seen": {"web": 11410492, "books": 692}, "gb": 46.5}
69
+ {"step": 13165, "elapsed_h": 1.515, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5462, "mlm": 4.0141, "ctr": 0.5321, "ctr_ema": 0.4579, "ret_acc": 0.7922, "loss_ema": 4.5738, "held_ppl": 54.721, "held_ctr": 0.0267, "held_ret_acc": 1.0, "kurt": 7.46, "alpha": {"1": 0.3776, "7": 0.2258}, "step_tok": 262144, "pairs": 154, "fn_masked": 0, "mask_rate": 0.204, "g_d": 1.38, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 71527, "fallback": true, "dup_dropped": 9496, "pack_dropped": 31, "samples_seen": {"web": 11555652, "books": 692}, "gb": 46.5}
70
+ {"step": 13374, "elapsed_h": 1.766, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.36, "mlm": 4.0798, "ctr": 0.2802, "ctr_ema": 0.4294, "ret_acc": 0.8846, "loss_ema": 4.5491, "held_ppl": 50.879, "held_ctr": 0.0281, "held_ret_acc": 1.0, "kurt": 7.27, "alpha": {"1": 0.3784, "7": 0.221}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.2036, "g_d": 1.213, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 85475, "fallback": true, "dup_dropped": 12803, "pack_dropped": 34, "samples_seen": {"web": 11712403, "books": 692}, "gb": 46.5}
71
+ {"step": 13568, "elapsed_h": 2.016, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 5.2751, "mlm": 4.4501, "ctr": 0.8251, "ctr_ema": 0.4001, "ret_acc": 0.7171, "loss_ema": 4.5282, "held_ppl": 49.691, "held_ctr": 0.0268, "held_ret_acc": 1.0, "kurt": 7.17, "alpha": {"1": 0.3805, "7": 0.2254}, "step_tok": 262144, "pairs": 152, "fn_masked": 3, "mask_rate": 0.2039, "g_d": 1.196, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 98848, "fallback": true, "dup_dropped": 16201, "pack_dropped": 40, "samples_seen": {"web": 11864994, "books": 692}, "gb": 46.5}
72
+ {"step": 13776, "elapsed_h": 2.267, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.7425, "mlm": 4.0807, "ctr": 0.6618, "ctr_ema": 0.4285, "ret_acc": 0.7467, "loss_ema": 4.5368, "held_ppl": 43.792, "held_ctr": 0.0226, "held_ret_acc": 1.0, "kurt": 7.48, "alpha": {"1": 0.3857, "7": 0.2237}, "step_tok": 262144, "pairs": 150, "fn_masked": 0, "mask_rate": 0.2036, "g_d": 1.379, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 112934, "fallback": true, "dup_dropped": 19997, "pack_dropped": 46, "samples_seen": {"web": 12025389, "books": 692}, "gb": 46.5}
73
+ {"step": 13968, "elapsed_h": 2.517, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3956, "mlm": 4.0438, "ctr": 0.3518, "ctr_ema": 0.3983, "ret_acc": 0.8797, "loss_ema": 4.493, "held_ppl": 56.23, "held_ctr": 0.0202, "held_ret_acc": 1.0, "kurt": 7.41, "alpha": {"1": 0.3898, "7": 0.2305}, "step_tok": 262144, "pairs": 158, "fn_masked": 4, "mask_rate": 0.2038, "g_d": 1.295, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 125979, "fallback": true, "dup_dropped": 23730, "pack_dropped": 50, "samples_seen": {"web": 12174543, "books": 692}, "gb": 46.5}
74
+ {"step": 14162, "elapsed_h": 2.767, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3003, "mlm": 4.015, "ctr": 0.2854, "ctr_ema": 0.4128, "ret_acc": 0.8839, "loss_ema": 4.5081, "held_ppl": 55.434, "held_ctr": 0.0307, "held_ret_acc": 1.0, "kurt": 7.2, "alpha": {"1": 0.3872, "7": 0.2222}, "step_tok": 262144, "pairs": 155, "fn_masked": 1, "mask_rate": 0.2039, "g_d": 1.218, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 139130, "fallback": true, "dup_dropped": 28025, "pack_dropped": 56, "samples_seen": {"web": 12324453, "books": 692}, "gb": 46.5}
75
+ {"step": 14370, "elapsed_h": 3.018, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.7007, "mlm": 4.263, "ctr": 0.4377, "ctr_ema": 0.4371, "ret_acc": 0.8618, "loss_ema": 4.5516, "held_ppl": 48.813, "held_ctr": 0.0201, "held_ret_acc": 1.0, "kurt": 7.21, "alpha": {"1": 0.3847, "7": 0.2224}, "step_tok": 262144, "pairs": 152, "fn_masked": 0, "mask_rate": 0.2035, "g_d": 1.549, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 153061, "fallback": true, "dup_dropped": 32706, "pack_dropped": 64, "samples_seen": {"web": 12485412, "books": 692}, "gb": 46.5}
76
+ {"step": 14564, "elapsed_h": 3.268, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4081, "mlm": 3.9679, "ctr": 0.4401, "ctr_ema": 0.4035, "ret_acc": 0.8553, "loss_ema": 4.5102, "held_ppl": 56.625, "held_ctr": 0.0246, "held_ret_acc": 1.0, "kurt": 7.5, "alpha": {"1": 0.3743, "7": 0.2215}, "step_tok": 262144, "pairs": 152, "fn_masked": 0, "mask_rate": 0.2033, "g_d": 1.359, "g_s": 0.037, "nhev": 8, "held_n": 64, "held_draws": 166351, "fallback": true, "dup_dropped": 37024, "pack_dropped": 71, "samples_seen": {"web": 12636102, "books": 692}, "gb": 46.5}
77
+ {"step": 14774, "elapsed_h": 3.519, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4873, "mlm": 3.9956, "ctr": 0.4917, "ctr_ema": 0.4174, "ret_acc": 0.7974, "loss_ema": 4.5333, "held_ppl": 59.446, "held_ctr": 0.0279, "held_ret_acc": 1.0, "kurt": 7.34, "alpha": {"1": 0.3806, "7": 0.2257}, "step_tok": 262144, "pairs": 153, "fn_masked": 0, "mask_rate": 0.2037, "g_d": 1.353, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 180388, "fallback": true, "dup_dropped": 41958, "pack_dropped": 78, "samples_seen": {"web": 12794620, "books": 692}, "gb": 46.51}
78
+ {"step": 14968, "elapsed_h": 3.769, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5335, "mlm": 4.0772, "ctr": 0.4563, "ctr_ema": 0.4284, "ret_acc": 0.8, "loss_ema": 4.5567, "held_ppl": 54.868, "held_ctr": 0.0371, "held_ret_acc": 1.0, "kurt": 7.12, "alpha": {"1": 0.3856, "7": 0.2224}, "step_tok": 262144, "pairs": 150, "fn_masked": 1, "mask_rate": 0.2038, "g_d": 1.439, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 193474, "fallback": true, "dup_dropped": 47093, "pack_dropped": 85, "samples_seen": {"web": 12944914, "books": 692}, "gb": 46.51}
79
+ {"step": 15159, "elapsed_h": 4.019, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.449, "mlm": 3.9718, "ctr": 0.4773, "ctr_ema": 0.4589, "ret_acc": 0.8269, "loss_ema": 4.602, "held_ppl": 55.829, "held_ctr": 0.018, "held_ret_acc": 1.0, "kurt": 7.34, "alpha": {"1": 0.3744, "7": 0.2186}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.2037, "g_d": 1.34, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 206238, "fallback": true, "dup_dropped": 52151, "pack_dropped": 90, "samples_seen": {"web": 13090625, "books": 692}, "gb": 46.51}
80
+ {"step": 15371, "elapsed_h": 4.27, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.413, "mlm": 3.9999, "ctr": 0.4131, "ctr_ema": 0.4283, "ret_acc": 0.8671, "loss_ema": 4.5404, "held_ppl": 46.033, "held_ctr": 0.0368, "held_ret_acc": 1.0, "kurt": 7.23, "alpha": {"1": 0.4017, "7": 0.2278}, "step_tok": 262144, "pairs": 158, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.326, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 220510, "fallback": true, "dup_dropped": 57984, "pack_dropped": 101, "samples_seen": {"web": 13253201, "books": 692}, "gb": 46.51}
81
+ {"step": 15565, "elapsed_h": 4.52, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.7744, "mlm": 4.2979, "ctr": 0.4765, "ctr_ema": 0.4142, "ret_acc": 0.8235, "loss_ema": 4.5248, "held_ppl": 51.007, "held_ctr": 0.0327, "held_ret_acc": 1.0, "kurt": 7.09, "alpha": {"1": 0.3822, "7": 0.2265}, "step_tok": 262144, "pairs": 153, "fn_masked": 0, "mask_rate": 0.2034, "g_d": 1.403, "g_s": 0.022, "nhev": 8, "held_n": 64, "held_draws": 233628, "fallback": true, "dup_dropped": 63558, "pack_dropped": 105, "samples_seen": {"web": 13402929, "books": 692}, "gb": 46.51}
82
+ {"step": 15777, "elapsed_h": 4.771, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4664, "mlm": 4.2615, "ctr": 0.2049, "ctr_ema": 0.4189, "ret_acc": 0.9383, "loss_ema": 4.5578, "held_ppl": 44.714, "held_ctr": 0.0259, "held_ret_acc": 1.0, "kurt": 7.31, "alpha": {"1": 0.3859, "7": 0.2263}, "step_tok": 262144, "pairs": 162, "fn_masked": 0, "mask_rate": 0.2038, "g_d": 1.074, "g_s": 0.016, "nhev": 8, "held_n": 64, "held_draws": 248052, "fallback": true, "dup_dropped": 70114, "pack_dropped": 117, "samples_seen": {"web": 13568750, "books": 692}, "gb": 46.51}
83
+ {"step": 15971, "elapsed_h": 5.022, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5243, "mlm": 4.247, "ctr": 0.2773, "ctr_ema": 0.3868, "ret_acc": 0.8951, "loss_ema": 4.5273, "held_ppl": 49.975, "held_ctr": 0.0265, "held_ret_acc": 1.0, "kurt": 7.32, "alpha": {"1": 0.3793, "7": 0.2204}, "step_tok": 262144, "pairs": 162, "fn_masked": 4, "mask_rate": 0.2041, "g_d": 1.163, "g_s": 0.019, "nhev": 8, "held_n": 64, "held_draws": 261268, "fallback": true, "dup_dropped": 76100, "pack_dropped": 122, "samples_seen": {"web": 13718787, "books": 692}, "gb": 46.51}
84
+ {"step": 16167, "elapsed_h": 5.272, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4543, "mlm": 4.1746, "ctr": 0.2797, "ctr_ema": 0.3786, "ret_acc": 0.8882, "loss_ema": 4.525, "held_ppl": 53.882, "held_ctr": 0.0237, "held_ret_acc": 1.0, "kurt": 7.4, "alpha": {"1": 0.3872, "7": 0.2225}, "step_tok": 262144, "pairs": 161, "fn_masked": 0, "mask_rate": 0.2038, "g_d": 1.199, "g_s": 0.017, "nhev": 8, "held_n": 64, "held_draws": 274600, "fallback": true, "dup_dropped": 82599, "pack_dropped": 130, "samples_seen": {"web": 13873176, "books": 692}, "gb": 46.51}
85
+ {"step": 16378, "elapsed_h": 5.523, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.8278, "mlm": 4.3307, "ctr": 0.4971, "ctr_ema": 0.4281, "ret_acc": 0.7829, "loss_ema": 4.5469, "held_ppl": 53.966, "held_ctr": 0.0348, "held_ret_acc": 1.0, "kurt": 7.32, "alpha": {"1": 0.3799, "7": 0.2293}, "step_tok": 262144, "pairs": 152, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.512, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 289083, "fallback": true, "dup_dropped": 89633, "pack_dropped": 135, "samples_seen": {"web": 14038612, "books": 692}, "gb": 46.51}
86
+ {"step": 16572, "elapsed_h": 5.774, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5492, "mlm": 4.1028, "ctr": 0.4464, "ctr_ema": 0.4243, "ret_acc": 0.8013, "loss_ema": 4.5711, "held_ppl": 54.396, "held_ctr": 0.0364, "held_ret_acc": 1.0, "kurt": 7.59, "alpha": {"1": 0.3814, "7": 0.2254}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.2036, "g_d": 1.337, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 302199, "fallback": true, "dup_dropped": 96493, "pack_dropped": 143, "samples_seen": {"web": 14192130, "books": 692}, "gb": 46.51}
87
+ {"step": 16783, "elapsed_h": 6.025, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5775, "mlm": 4.0605, "ctr": 0.5171, "ctr_ema": 0.3865, "ret_acc": 0.7962, "loss_ema": 4.4917, "held_ppl": 47.176, "held_ctr": 0.0341, "held_ret_acc": 1.0, "kurt": 7.34, "alpha": {"1": 0.3785, "7": 0.2243}, "step_tok": 262144, "pairs": 157, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.581, "g_s": 0.026, "nhev": 8, "held_n": 64, "held_draws": 316577, "fallback": true, "dup_dropped": 104210, "pack_dropped": 150, "samples_seen": {"web": 14359727, "books": 692}, "gb": 46.51}
88
+ {"step": 16977, "elapsed_h": 6.275, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3659, "mlm": 3.9154, "ctr": 0.4505, "ctr_ema": 0.3896, "ret_acc": 0.8194, "loss_ema": 4.4945, "held_ppl": 58.664, "held_ctr": 0.033, "held_ret_acc": 1.0, "kurt": 7.51, "alpha": {"1": 0.3779, "7": 0.222}, "step_tok": 262144, "pairs": 155, "fn_masked": 0, "mask_rate": 0.2035, "g_d": 1.471, "g_s": 0.022, "nhev": 8, "held_n": 64, "held_draws": 329661, "fallback": true, "dup_dropped": 111448, "pack_dropped": 154, "samples_seen": {"web": 14513148, "books": 692}, "gb": 46.51}
89
+ {"step": 17171, "elapsed_h": 6.525, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.7179, "mlm": 4.2433, "ctr": 0.4746, "ctr_ema": 0.4137, "ret_acc": 0.8442, "loss_ema": 4.529, "held_ppl": 54.358, "held_ctr": 0.041, "held_ret_acc": 1.0, "kurt": 7.6, "alpha": {"1": 0.3816, "7": 0.22}, "step_tok": 262144, "pairs": 154, "fn_masked": 1, "mask_rate": 0.2039, "g_d": 1.529, "g_s": 0.023, "nhev": 8, "held_n": 64, "held_draws": 342491, "fallback": true, "dup_dropped": 118672, "pack_dropped": 161, "samples_seen": {"web": 14661553, "books": 692}, "gb": 46.51}
90
+ {"step": 17377, "elapsed_h": 6.775, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5495, "mlm": 4.1307, "ctr": 0.4188, "ctr_ema": 0.4112, "ret_acc": 0.8471, "loss_ema": 4.4839, "held_ppl": 54.595, "held_ctr": 0.0348, "held_ret_acc": 1.0, "kurt": 7.21, "alpha": {"1": 0.3832, "7": 0.222}, "step_tok": 262144, "pairs": 157, "fn_masked": 1, "mask_rate": 0.2041, "g_d": 1.407, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 356338, "fallback": true, "dup_dropped": 126543, "pack_dropped": 168, "samples_seen": {"web": 14819860, "books": 692}, "gb": 46.51}
91
+ {"step": 17570, "elapsed_h": 7.026, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6587, "mlm": 4.2125, "ctr": 0.4462, "ctr_ema": 0.4291, "ret_acc": 0.8471, "loss_ema": 4.554, "held_ppl": 43.122, "held_ctr": 0.0276, "held_ret_acc": 1.0, "kurt": 7.36, "alpha": {"1": 0.3855, "7": 0.2325}, "step_tok": 262144, "pairs": 157, "fn_masked": 1, "mask_rate": 0.2039, "g_d": 1.476, "g_s": 0.022, "nhev": 8, "held_n": 64, "held_draws": 369502, "fallback": true, "dup_dropped": 134269, "pack_dropped": 173, "samples_seen": {"web": 14972858, "books": 692}, "gb": 46.51}
92
+ {"step": 17780, "elapsed_h": 7.276, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.447, "mlm": 4.1101, "ctr": 0.3369, "ctr_ema": 0.4307, "ret_acc": 0.859, "loss_ema": 4.5477, "held_ppl": 43.633, "held_ctr": 0.0212, "held_ret_acc": 1.0, "kurt": 7.22, "alpha": {"1": 0.381, "7": 0.2272}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.145, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 383715, "fallback": true, "dup_dropped": 142872, "pack_dropped": 176, "samples_seen": {"web": 15136602, "books": 692}, "gb": 46.51}
93
+ {"step": 17973, "elapsed_h": 7.526, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4918, "mlm": 3.9666, "ctr": 0.5252, "ctr_ema": 0.4086, "ret_acc": 0.8345, "loss_ema": 4.5185, "held_ppl": 43.083, "held_ctr": 0.0308, "held_ret_acc": 1.0, "kurt": 7.31, "alpha": {"1": 0.3782, "7": 0.2225}, "step_tok": 262144, "pairs": 145, "fn_masked": 1, "mask_rate": 0.2037, "g_d": 1.649, "g_s": 0.024, "nhev": 8, "held_n": 64, "held_draws": 396669, "fallback": true, "dup_dropped": 151113, "pack_dropped": 187, "samples_seen": {"web": 15286738, "books": 692}, "gb": 46.51}
94
+ {"step": 18161, "elapsed_h": 7.777, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3174, "mlm": 3.9574, "ctr": 0.36, "ctr_ema": 0.383, "ret_acc": 0.8679, "loss_ema": 4.4906, "held_ppl": 47.494, "held_ctr": 0.0337, "held_ret_acc": 1.0, "kurt": 7.33, "alpha": {"1": 0.3885, "7": 0.224}, "step_tok": 262144, "pairs": 159, "fn_masked": 0, "mask_rate": 0.2042, "g_d": 1.358, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 409389, "fallback": true, "dup_dropped": 159388, "pack_dropped": 193, "samples_seen": {"web": 15435594, "books": 692}, "gb": 46.51}
95
+ {"step": 18372, "elapsed_h": 8.028, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3549, "mlm": 4.1039, "ctr": 0.2509, "ctr_ema": 0.394, "ret_acc": 0.9024, "loss_ema": 4.4816, "held_ppl": 41.002, "held_ctr": 0.0546, "held_ret_acc": 1.0, "kurt": 7.0, "alpha": {"1": 0.3798, "7": 0.2236}, "step_tok": 262144, "pairs": 164, "fn_masked": 0, "mask_rate": 0.204, "g_d": 1.17, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 423709, "fallback": true, "dup_dropped": 168761, "pack_dropped": 199, "samples_seen": {"web": 15601909, "books": 692}, "gb": 46.51}
96
+ {"step": 18566, "elapsed_h": 8.278, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3416, "mlm": 4.0233, "ctr": 0.3184, "ctr_ema": 0.4072, "ret_acc": 0.8854, "loss_ema": 4.5315, "held_ppl": 53.55, "held_ctr": 0.0237, "held_ret_acc": 1.0, "kurt": 7.48, "alpha": {"1": 0.3851, "7": 0.2275}, "step_tok": 262144, "pairs": 157, "fn_masked": 0, "mask_rate": 0.2037, "g_d": 1.333, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 436814, "fallback": true, "dup_dropped": 177841, "pack_dropped": 203, "samples_seen": {"web": 15756006, "books": 692}, "gb": 46.51}
97
+ {"step": 18775, "elapsed_h": 8.529, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5219, "mlm": 4.0878, "ctr": 0.4341, "ctr_ema": 0.3554, "ret_acc": 0.8355, "loss_ema": 4.4598, "held_ppl": 47.029, "held_ctr": 0.0279, "held_ret_acc": 1.0, "kurt": 7.39, "alpha": {"1": 0.381, "7": 0.2299}, "step_tok": 262144, "pairs": 152, "fn_masked": 0, "mask_rate": 0.2036, "g_d": 1.441, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 451043, "fallback": true, "dup_dropped": 187786, "pack_dropped": 210, "samples_seen": {"web": 15923387, "books": 692}, "gb": 46.51}
98
+ {"step": 18968, "elapsed_h": 8.779, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6451, "mlm": 4.1689, "ctr": 0.4762, "ctr_ema": 0.3971, "ret_acc": 0.8462, "loss_ema": 4.4923, "held_ppl": 51.87, "held_ctr": 0.0201, "held_ret_acc": 1.0, "kurt": 7.35, "alpha": {"1": 0.3862, "7": 0.2285}, "step_tok": 262144, "pairs": 156, "fn_masked": 1, "mask_rate": 0.2041, "g_d": 1.374, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 464064, "fallback": true, "dup_dropped": 197011, "pack_dropped": 213, "samples_seen": {"web": 16074982, "books": 692}, "gb": 46.51}
99
+ {"step": 19162, "elapsed_h": 9.03, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4922, "mlm": 4.0834, "ctr": 0.4088, "ctr_ema": 0.4111, "ret_acc": 0.8452, "loss_ema": 4.4881, "held_ppl": 47.906, "held_ctr": 0.0212, "held_ret_acc": 1.0, "kurt": 7.31, "alpha": {"1": 0.3822, "7": 0.2232}, "step_tok": 262144, "pairs": 155, "fn_masked": 0, "mask_rate": 0.2037, "g_d": 1.447, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 477101, "fallback": true, "dup_dropped": 206644, "pack_dropped": 215, "samples_seen": {"web": 16226682, "books": 692}, "gb": 46.51}
100
+ {"step": 19373, "elapsed_h": 9.28, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4469, "mlm": 4.0576, "ctr": 0.3893, "ctr_ema": 0.3627, "ret_acc": 0.8599, "loss_ema": 4.4576, "held_ppl": 68.87, "held_ctr": 0.0266, "held_ret_acc": 1.0, "kurt": 7.53, "alpha": {"1": 0.387, "7": 0.2292}, "step_tok": 262144, "pairs": 157, "fn_masked": 0, "mask_rate": 0.2041, "g_d": 1.27, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 491476, "fallback": true, "dup_dropped": 217341, "pack_dropped": 226, "samples_seen": {"web": 16394931, "books": 692}, "gb": 46.51}
101
+ {"step": 19564, "elapsed_h": 9.53, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3204, "mlm": 4.1235, "ctr": 0.1969, "ctr_ema": 0.379, "ret_acc": 0.9427, "loss_ema": 4.486, "held_ppl": 53.436, "held_ctr": 0.0379, "held_ret_acc": 1.0, "kurt": 7.3, "alpha": {"1": 0.3951, "7": 0.2292}, "step_tok": 262144, "pairs": 157, "fn_masked": 0, "mask_rate": 0.2042, "g_d": 1.136, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 504440, "fallback": true, "dup_dropped": 227300, "pack_dropped": 228, "samples_seen": {"web": 16547633, "books": 692}, "gb": 46.52}
102
+ {"step": 19774, "elapsed_h": 9.78, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5984, "mlm": 4.0755, "ctr": 0.5228, "ctr_ema": 0.3847, "ret_acc": 0.8121, "loss_ema": 4.4777, "held_ppl": 53.192, "held_ctr": 0.0317, "held_ret_acc": 1.0, "kurt": 7.09, "alpha": {"1": 0.3798, "7": 0.2311}, "step_tok": 262144, "pairs": 149, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.363, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 518758, "fallback": true, "dup_dropped": 238770, "pack_dropped": 233, "samples_seen": {"web": 16717008, "books": 692}, "gb": 46.52}
103
+ {"step": 19969, "elapsed_h": 10.031, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4762, "mlm": 4.1163, "ctr": 0.3599, "ctr_ema": 0.4078, "ret_acc": 0.8497, "loss_ema": 4.5216, "held_ppl": 52.021, "held_ctr": 0.0277, "held_ret_acc": 1.0, "kurt": 7.41, "alpha": {"1": 0.3897, "7": 0.2311}, "step_tok": 262144, "pairs": 153, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.34, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 532176, "fallback": true, "dup_dropped": 249595, "pack_dropped": 238, "samples_seen": {"web": 16875126, "books": 692}, "gb": 46.52}
104
+ {"step": 20159, "elapsed_h": 10.282, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4735, "mlm": 4.1429, "ctr": 0.3306, "ctr_ema": 0.425, "ret_acc": 0.8839, "loss_ema": 4.5244, "held_ppl": 46.123, "held_ctr": 0.0266, "held_ret_acc": 1.0, "kurt": 7.27, "alpha": {"1": 0.3888, "7": 0.228}, "step_tok": 262144, "pairs": 155, "fn_masked": 9, "mask_rate": 0.2036, "g_d": 1.421, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 545108, "fallback": true, "dup_dropped": 259969, "pack_dropped": 245, "samples_seen": {"web": 17026944, "books": 692}, "gb": 46.52}
105
+ {"step": 20369, "elapsed_h": 10.532, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5742, "mlm": 4.0663, "ctr": 0.5079, "ctr_ema": 0.3955, "ret_acc": 0.8255, "loss_ema": 4.5486, "held_ppl": 50.522, "held_ctr": 0.0409, "held_ret_acc": 1.0, "kurt": 7.01, "alpha": {"1": 0.3814, "7": 0.2273}, "step_tok": 262144, "pairs": 149, "fn_masked": 0, "mask_rate": 0.2037, "g_d": 1.44, "g_s": 0.021, "nhev": 8, "held_n": 64, "held_draws": 559202, "fallback": true, "dup_dropped": 271938, "pack_dropped": 251, "samples_seen": {"web": 17194145, "books": 692}, "gb": 46.52}
106
+ {"step": 20564, "elapsed_h": 10.783, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.2624, "mlm": 3.9352, "ctr": 0.3272, "ctr_ema": 0.3916, "ret_acc": 0.8954, "loss_ema": 4.5119, "held_ppl": 43.912, "held_ctr": 0.0253, "held_ret_acc": 1.0, "kurt": 7.31, "alpha": {"1": 0.3728, "7": 0.2236}, "step_tok": 262144, "pairs": 153, "fn_masked": 1, "mask_rate": 0.2038, "g_d": 1.352, "g_s": 0.02, "nhev": 8, "held_n": 64, "held_draws": 572298, "fallback": true, "dup_dropped": 283210, "pack_dropped": 261, "samples_seen": {"web": 17348868, "books": 692}, "gb": 46.52}
107
+ {"step": 20775, "elapsed_h": 11.034, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.573, "mlm": 4.288, "ctr": 0.285, "ctr_ema": 0.3872, "ret_acc": 0.8882, "loss_ema": 4.5066, "held_ppl": 59.228, "held_ctr": 0.0385, "held_ret_acc": 1.0, "kurt": 7.15, "alpha": {"1": 0.3889, "7": 0.2379}, "step_tok": 262144, "pairs": 161, "fn_masked": 1, "mask_rate": 0.2043, "g_d": 1.22, "g_s": 0.018, "nhev": 8, "held_n": 64, "held_draws": 586729, "fallback": true, "dup_dropped": 295947, "pack_dropped": 267, "samples_seen": {"web": 17519743, "books": 692}, "gb": 46.52}
108
+ {"event": "launch", "fresh": false, "step": 20867, "time": "2026-10-06T09:48:39Z"}
109
+ {"step": 20925, "elapsed_h": 0.463, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3556, "mlm": 4.0448, "ctr": 0.3108, "ctr_ema": 0.3848, "ret_acc": 0.8797, "loss_ema": 4.4935, "held_ppl": 72.428, "held_ctr": 0.0174, "held_ret_acc": 1.0, "kurt": 6.82, "alpha": {"1": 0.377, "7": 0.2206}, "step_tok": 262144, "pairs": 158, "fn_masked": 0, "mask_rate": 0.204, "g_d": 1.39, "g_s": 0.023, "nhev": 7, "held_n": 8, "held_draws": 4002, "fallback": true, "dup_dropped": 111, "pack_dropped": 271, "samples_seen": {"web": 17638104, "books": 692}, "gb": 46.49}
110
+ {"step": 21138, "elapsed_h": 0.713, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5928, "mlm": 4.1881, "ctr": 0.4047, "ctr_ema": 0.3971, "ret_acc": 0.8889, "loss_ema": 4.5034, "held_ppl": 60.154, "held_ctr": 0.0379, "held_ret_acc": 1.0, "kurt": 6.88, "alpha": {"1": 0.3947, "7": 0.2206}, "step_tok": 262144, "pairs": 153, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.589, "g_s": 0.022, "nhev": 31, "held_n": 37, "held_draws": 18638, "fallback": true, "dup_dropped": 1182, "pack_dropped": 281, "samples_seen": {"web": 17799754, "books": 692}, "gb": 46.51}
111
+ {"step": 21334, "elapsed_h": 0.964, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.7311, "mlm": 4.2953, "ctr": 0.4358, "ctr_ema": 0.3972, "ret_acc": 0.8387, "loss_ema": 4.5192, "held_ppl": 64.196, "held_ctr": 0.0451, "held_ret_acc": 1.0, "kurt": 6.88, "alpha": {"1": 0.3854, "7": 0.214}, "step_tok": 262144, "pairs": 155, "fn_masked": 0, "mask_rate": 0.2042, "g_d": 1.513, "g_s": 0.023, "nhev": 31, "held_n": 63, "held_draws": 31871, "fallback": true, "dup_dropped": 2417, "pack_dropped": 288, "samples_seen": {"web": 17947812, "books": 692}, "gb": 46.51}
112
+ {"step": 21546, "elapsed_h": 1.214, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3288, "mlm": 4.0525, "ctr": 0.2763, "ctr_ema": 0.3853, "ret_acc": 0.8974, "loss_ema": 4.5106, "held_ppl": 58.874, "held_ctr": 0.0578, "held_ret_acc": 1.0, "kurt": 6.75, "alpha": {"1": 0.3878, "7": 0.2134}, "step_tok": 262144, "pairs": 156, "fn_masked": 1, "mask_rate": 0.2035, "g_d": 1.363, "g_s": 0.022, "nhev": 31, "held_n": 64, "held_draws": 46195, "fallback": true, "dup_dropped": 4187, "pack_dropped": 294, "samples_seen": {"web": 18107851, "books": 692}, "gb": 46.51}
113
+ {"step": 21739, "elapsed_h": 1.464, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6499, "mlm": 4.335, "ctr": 0.3149, "ctr_ema": 0.3697, "ret_acc": 0.8889, "loss_ema": 4.4669, "held_ppl": 64.556, "held_ctr": 0.0442, "held_ret_acc": 1.0, "kurt": 6.97, "alpha": {"1": 0.3775, "7": 0.2185}, "step_tok": 262144, "pairs": 153, "fn_masked": 0, "mask_rate": 0.2035, "g_d": 1.356, "g_s": 0.024, "nhev": 31, "held_n": 64, "held_draws": 59461, "fallback": true, "dup_dropped": 6844, "pack_dropped": 301, "samples_seen": {"web": 18256646, "books": 692}, "gb": 46.51}
114
+ {"event": "launch", "fresh": false, "step": 21893, "time": "2026-10-06T11:34:51Z"}
115
+ {"held_margin": 0.4865, "step": 21952, "elapsed_h": 0.484, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5183, "mlm": 4.1769, "ctr": 0.3413, "ctr_ema": 0.3933, "ret_acc": 0.8599, "loss_ema": 4.5285, "held_ppl": 80.041, "held_ctr": 0.0339, "held_ret_acc": 1.0, "held_pos": 0.7411, "held_neg": 0.0635, "held_gap": 0.6776, "kurt": 6.8, "alpha": {"1": 0.3871, "7": 0.2255}, "step_tok": 262144, "pairs": 157, "fn_masked": 1, "mask_rate": 0.2038, "g_d": 1.299, "g_s": 0.019, "nhev": 7, "held_n": 8, "held_draws": 4016, "fallback": true, "dup_dropped": 116, "pack_dropped": 306, "samples_seen": {"web": 18420922, "books": 692}, "gb": 46.49}
116
+ {"held_margin": 0.4699, "step": 22165, "elapsed_h": 0.735, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3526, "mlm": 4.0508, "ctr": 0.3018, "ctr_ema": 0.4486, "ret_acc": 0.8782, "loss_ema": 4.578, "held_ppl": 63.039, "held_ctr": 0.0431, "held_ret_acc": 1.0, "held_pos": 0.7857, "held_neg": 0.0165, "held_gap": 0.7692, "kurt": 6.78, "alpha": {"1": 0.3905, "7": 0.2233}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.2038, "g_d": 1.285, "g_s": 0.022, "nhev": 31, "held_n": 36, "held_draws": 18376, "fallback": true, "dup_dropped": 1087, "pack_dropped": 317, "samples_seen": {"web": 18580583, "books": 692}, "gb": 46.51}
117
+ {"held_margin": 0.4748, "step": 22357, "elapsed_h": 0.986, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4172, "mlm": 4.0197, "ctr": 0.3975, "ctr_ema": 0.4317, "ret_acc": 0.8269, "loss_ema": 4.5429, "held_ppl": 66.829, "held_ctr": 0.0477, "held_ret_acc": 1.0, "held_pos": 0.7866, "held_neg": 0.0258, "held_gap": 0.7608, "kurt": 6.64, "alpha": {"1": 0.3906, "7": 0.227}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.2042, "g_d": 1.39, "g_s": 0.021, "nhev": 31, "held_n": 62, "held_draws": 31350, "fallback": true, "dup_dropped": 2466, "pack_dropped": 319, "samples_seen": {"web": 18724511, "books": 692}, "gb": 46.51}
118
+ {"held_margin": 0.4754, "step": 22569, "elapsed_h": 1.237, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.7409, "mlm": 4.3694, "ctr": 0.3715, "ctr_ema": 0.4178, "ret_acc": 0.8733, "loss_ema": 4.5595, "held_ppl": 61.024, "held_ctr": 0.0496, "held_ret_acc": 1.0, "held_pos": 0.7909, "held_neg": 0.0146, "held_gap": 0.7763, "kurt": 6.88, "alpha": {"1": 0.3876, "7": 0.2188}, "step_tok": 262144, "pairs": 150, "fn_masked": 109, "mask_rate": 0.2038, "g_d": 1.371, "g_s": 0.019, "nhev": 31, "held_n": 64, "held_draws": 45684, "fallback": true, "dup_dropped": 4699, "pack_dropped": 324, "samples_seen": {"web": 18885663, "books": 692}, "gb": 46.51}
119
+ {"held_margin": 0.4777, "step": 22764, "elapsed_h": 1.487, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6645, "mlm": 4.1926, "ctr": 0.4719, "ctr_ema": 0.4052, "ret_acc": 0.8067, "loss_ema": 4.5408, "held_ppl": 64.401, "held_ctr": 0.0473, "held_ret_acc": 1.0, "held_pos": 0.7955, "held_neg": 0.0273, "held_gap": 0.7682, "kurt": 6.69, "alpha": {"1": 0.3842, "7": 0.2159}, "step_tok": 262144, "pairs": 150, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.433, "g_s": 0.021, "nhev": 31, "held_n": 64, "held_draws": 58876, "fallback": true, "dup_dropped": 6924, "pack_dropped": 328, "samples_seen": {"web": 19031465, "books": 692}, "gb": 46.51}
120
+ {"held_margin": 0.4651, "step": 22960, "elapsed_h": 1.738, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.534, "mlm": 4.0951, "ctr": 0.4389, "ctr_ema": 0.4042, "ret_acc": 0.8054, "loss_ema": 4.5154, "held_ppl": 72.872, "held_ctr": 0.0527, "held_ret_acc": 1.0, "held_pos": 0.8002, "held_neg": 0.0262, "held_gap": 0.774, "kurt": 6.7, "alpha": {"1": 0.3895, "7": 0.2227}, "step_tok": 262144, "pairs": 149, "fn_masked": 0, "mask_rate": 0.2037, "g_d": 1.274, "g_s": 0.021, "nhev": 31, "held_n": 64, "held_draws": 72161, "fallback": true, "dup_dropped": 9391, "pack_dropped": 335, "samples_seen": {"web": 19179687, "books": 692}, "gb": 46.51}
121
+ {"held_margin": 0.4545, "step": 23170, "elapsed_h": 1.988, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4354, "mlm": 4.0351, "ctr": 0.4004, "ctr_ema": 0.4011, "ret_acc": 0.879, "loss_ema": 4.537, "held_ppl": 62.921, "held_ctr": 0.0605, "held_ret_acc": 1.0, "held_pos": 0.7819, "held_neg": 0.0314, "held_gap": 0.7505, "kurt": 6.64, "alpha": {"1": 0.3938, "7": 0.2177}, "step_tok": 262144, "pairs": 157, "fn_masked": 0, "mask_rate": 0.204, "g_d": 1.379, "g_s": 0.02, "nhev": 31, "held_n": 64, "held_draws": 86467, "fallback": true, "dup_dropped": 12764, "pack_dropped": 345, "samples_seen": {"web": 19341543, "books": 692}, "gb": 46.51}
122
+ {"held_margin": 0.4752, "step": 23367, "elapsed_h": 2.239, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6057, "mlm": 4.1249, "ctr": 0.4808, "ctr_ema": 0.4311, "ret_acc": 0.8452, "loss_ema": 4.5397, "held_ppl": 62.249, "held_ctr": 0.0556, "held_ret_acc": 1.0, "held_pos": 0.7986, "held_neg": 0.0267, "held_gap": 0.7719, "kurt": 6.51, "alpha": {"1": 0.3835, "7": 0.2217}, "step_tok": 262144, "pairs": 155, "fn_masked": 0, "mask_rate": 0.2041, "g_d": 1.978, "g_s": 0.023, "nhev": 31, "held_n": 64, "held_draws": 99798, "fallback": true, "dup_dropped": 16130, "pack_dropped": 352, "samples_seen": {"web": 19491469, "books": 692}, "gb": 46.51}
123
+ {"held_margin": 0.4563, "step": 23579, "elapsed_h": 2.49, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4243, "mlm": 4.1314, "ctr": 0.2929, "ctr_ema": 0.3804, "ret_acc": 0.8981, "loss_ema": 4.4957, "held_ppl": 66.316, "held_ctr": 0.0679, "held_ret_acc": 1.0, "held_pos": 0.768, "held_neg": 0.0097, "held_gap": 0.7584, "kurt": 6.6, "alpha": {"1": 0.3845, "7": 0.2217}, "step_tok": 262144, "pairs": 157, "fn_masked": 0, "mask_rate": 0.2037, "g_d": 1.414, "g_s": 0.02, "nhev": 31, "held_n": 64, "held_draws": 114263, "fallback": true, "dup_dropped": 20008, "pack_dropped": 359, "samples_seen": {"web": 19654145, "books": 692}, "gb": 46.51}
124
+ {"held_margin": 0.4838, "step": 23774, "elapsed_h": 2.74, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3668, "mlm": 4.0668, "ctr": 0.3, "ctr_ema": 0.4025, "ret_acc": 0.891, "loss_ema": 4.5134, "held_ppl": 56.818, "held_ctr": 0.0585, "held_ret_acc": 1.0, "held_pos": 0.7638, "held_neg": 0.0027, "held_gap": 0.7611, "kurt": 6.68, "alpha": {"1": 0.3929, "7": 0.2223}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.2043, "g_d": 1.305, "g_s": 0.02, "nhev": 31, "held_n": 64, "held_draws": 127401, "fallback": true, "dup_dropped": 23911, "pack_dropped": 364, "samples_seen": {"web": 19802444, "books": 692}, "gb": 46.51}
125
+ {"held_margin": 0.4798, "step": 23965, "elapsed_h": 2.99, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5456, "mlm": 4.1513, "ctr": 0.3943, "ctr_ema": 0.402, "ret_acc": 0.8634, "loss_ema": 4.5167, "held_ppl": 70.781, "held_ctr": 0.0574, "held_ret_acc": 1.0, "held_pos": 0.7798, "held_neg": 0.0192, "held_gap": 0.7606, "kurt": 6.69, "alpha": {"1": 0.3925, "7": 0.2252}, "step_tok": 262144, "pairs": 161, "fn_masked": 14, "mask_rate": 0.204, "g_d": 1.27, "g_s": 0.018, "nhev": 31, "held_n": 64, "held_draws": 140275, "fallback": true, "dup_dropped": 27769, "pack_dropped": 367, "samples_seen": {"web": 19948798, "books": 692}, "gb": 46.51}
126
+ {"held_margin": 0.4589, "step": 24176, "elapsed_h": 3.241, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3441, "mlm": 4.0228, "ctr": 0.3213, "ctr_ema": 0.4188, "ret_acc": 0.8944, "loss_ema": 4.4931, "held_ppl": 70.863, "held_ctr": 0.0894, "held_ret_acc": 1.0, "held_pos": 0.7626, "held_neg": 0.0282, "held_gap": 0.7344, "kurt": 6.62, "alpha": {"1": 0.3903, "7": 0.2218}, "step_tok": 262144, "pairs": 161, "fn_masked": 3, "mask_rate": 0.2039, "g_d": 1.256, "g_s": 0.02, "nhev": 31, "held_n": 64, "held_draws": 154620, "fallback": true, "dup_dropped": 32442, "pack_dropped": 370, "samples_seen": {"web": 20109403, "books": 692}, "gb": 46.51}
127
+ {"held_margin": 0.479, "step": 24372, "elapsed_h": 3.492, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5071, "mlm": 4.121, "ctr": 0.3861, "ctr_ema": 0.4071, "ret_acc": 0.8544, "loss_ema": 4.5374, "held_ppl": 65.115, "held_ctr": 0.0593, "held_ret_acc": 1.0, "held_pos": 0.8013, "held_neg": 0.0263, "held_gap": 0.775, "kurt": 6.61, "alpha": {"1": 0.3943, "7": 0.2244}, "step_tok": 262144, "pairs": 158, "fn_masked": 0, "mask_rate": 0.2038, "g_d": 1.425, "g_s": 0.025, "nhev": 31, "held_n": 64, "held_draws": 167875, "fallback": true, "dup_dropped": 36927, "pack_dropped": 375, "samples_seen": {"web": 20260633, "books": 692}, "gb": 46.51}
128
+ {"held_margin": 0.4802, "step": 24581, "elapsed_h": 3.742, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5612, "mlm": 4.2118, "ctr": 0.3495, "ctr_ema": 0.3997, "ret_acc": 0.8726, "loss_ema": 4.5268, "held_ppl": 64.51, "held_ctr": 0.061, "held_ret_acc": 1.0, "held_pos": 0.7883, "held_neg": 0.0194, "held_gap": 0.7689, "kurt": 6.58, "alpha": {"1": 0.3838, "7": 0.2209}, "step_tok": 262144, "pairs": 157, "fn_masked": 0, "mask_rate": 0.2033, "g_d": 1.311, "g_s": 0.019, "nhev": 31, "held_n": 64, "held_draws": 182003, "fallback": true, "dup_dropped": 42034, "pack_dropped": 382, "samples_seen": {"web": 20424104, "books": 692}, "gb": 46.52}
129
+ {"held_margin": 0.4602, "step": 24777, "elapsed_h": 3.993, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6513, "mlm": 4.2846, "ctr": 0.3667, "ctr_ema": 0.4024, "ret_acc": 0.8684, "loss_ema": 4.5368, "held_ppl": 71.935, "held_ctr": 0.0693, "held_ret_acc": 1.0, "held_pos": 0.7675, "held_neg": 0.0112, "held_gap": 0.7563, "kurt": 6.66, "alpha": {"1": 0.3926, "7": 0.2234}, "step_tok": 262144, "pairs": 152, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.254, "g_s": 0.02, "nhev": 31, "held_n": 64, "held_draws": 195330, "fallback": true, "dup_dropped": 46788, "pack_dropped": 390, "samples_seen": {"web": 20574297, "books": 692}, "gb": 46.52}
130
+ {"held_margin": 0.4785, "step": 24972, "elapsed_h": 4.244, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6395, "mlm": 4.303, "ctr": 0.3365, "ctr_ema": 0.3804, "ret_acc": 0.8688, "loss_ema": 4.4941, "held_ppl": 62.544, "held_ctr": 0.0559, "held_ret_acc": 1.0, "held_pos": 0.7888, "held_neg": 0.0138, "held_gap": 0.7751, "kurt": 6.6, "alpha": {"1": 0.393, "7": 0.2216}, "step_tok": 262144, "pairs": 160, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.191, "g_s": 0.019, "nhev": 31, "held_n": 64, "held_draws": 208567, "fallback": true, "dup_dropped": 52140, "pack_dropped": 397, "samples_seen": {"web": 20725993, "books": 692}, "gb": 46.52}
131
+ {"held_margin": 0.495, "step": 25184, "elapsed_h": 4.495, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6583, "mlm": 4.1898, "ctr": 0.4685, "ctr_ema": 0.3871, "ret_acc": 0.828, "loss_ema": 4.4873, "held_ppl": 67.404, "held_ctr": 0.0525, "held_ret_acc": 1.0, "held_pos": 0.7672, "held_neg": 0.012, "held_gap": 0.7552, "kurt": 6.42, "alpha": {"1": 0.3918, "7": 0.2231}, "step_tok": 262144, "pairs": 157, "fn_masked": 0, "mask_rate": 0.2042, "g_d": 1.536, "g_s": 0.021, "nhev": 31, "held_n": 64, "held_draws": 222941, "fallback": true, "dup_dropped": 58004, "pack_dropped": 404, "samples_seen": {"web": 20890008, "books": 692}, "gb": 46.52}
132
+ {"held_margin": 0.4771, "step": 25375, "elapsed_h": 4.745, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.9198, "mlm": 4.2625, "ctr": 0.6573, "ctr_ema": 0.4708, "ret_acc": 0.8079, "loss_ema": 4.5848, "held_ppl": 59.793, "held_ctr": 0.055, "held_ret_acc": 1.0, "held_pos": 0.7697, "held_neg": 0.0077, "held_gap": 0.762, "kurt": 6.49, "alpha": {"1": 0.3896, "7": 0.2198}, "step_tok": 262144, "pairs": 151, "fn_masked": 0, "mask_rate": 0.2035, "g_d": 1.618, "g_s": 0.022, "nhev": 31, "held_n": 64, "held_draws": 235693, "fallback": true, "dup_dropped": 63476, "pack_dropped": 408, "samples_seen": {"web": 21034785, "books": 692}, "gb": 46.52}
133
+ {"held_margin": 0.4471, "step": 25587, "elapsed_h": 4.996, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.7522, "mlm": 4.0146, "ctr": 0.7376, "ctr_ema": 0.4378, "ret_acc": 0.76, "loss_ema": 4.542, "held_ppl": 67.834, "held_ctr": 0.0702, "held_ret_acc": 1.0, "held_pos": 0.8005, "held_neg": 0.0285, "held_gap": 0.7721, "kurt": 6.65, "alpha": {"1": 0.3932, "7": 0.2203}, "step_tok": 262144, "pairs": 150, "fn_masked": 0, "mask_rate": 0.204, "g_d": 1.621, "g_s": 0.022, "nhev": 31, "held_n": 64, "held_draws": 250049, "fallback": true, "dup_dropped": 69975, "pack_dropped": 419, "samples_seen": {"web": 21199128, "books": 692}, "gb": 46.52}
134
+ {"held_margin": 0.4771, "step": 25781, "elapsed_h": 5.247, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.2556, "mlm": 3.9423, "ctr": 0.3133, "ctr_ema": 0.3999, "ret_acc": 0.8782, "loss_ema": 4.5201, "held_ppl": 65.351, "held_ctr": 0.0483, "held_ret_acc": 1.0, "held_pos": 0.7856, "held_neg": 0.0193, "held_gap": 0.7663, "kurt": 6.57, "alpha": {"1": 0.3852, "7": 0.221}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.2037, "g_d": 1.43, "g_s": 0.022, "nhev": 31, "held_n": 64, "held_draws": 263159, "fallback": true, "dup_dropped": 75860, "pack_dropped": 423, "samples_seen": {"web": 21349467, "books": 692}, "gb": 46.52}
135
+ {"held_margin": 0.4708, "step": 25976, "elapsed_h": 5.498, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.7945, "mlm": 4.3043, "ctr": 0.4902, "ctr_ema": 0.4087, "ret_acc": 0.8377, "loss_ema": 4.5292, "held_ppl": 58.452, "held_ctr": 0.049, "held_ret_acc": 1.0, "held_pos": 0.8037, "held_neg": 0.0274, "held_gap": 0.7763, "kurt": 6.62, "alpha": {"1": 0.3919, "7": 0.2236}, "step_tok": 262144, "pairs": 154, "fn_masked": 1, "mask_rate": 0.2039, "g_d": 1.396, "g_s": 0.019, "nhev": 31, "held_n": 64, "held_draws": 276268, "fallback": true, "dup_dropped": 82228, "pack_dropped": 425, "samples_seen": {"web": 21502712, "books": 692}, "gb": 46.52}
136
+ {"held_margin": 0.4859, "step": 26184, "elapsed_h": 5.749, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6141, "mlm": 4.0447, "ctr": 0.5695, "ctr_ema": 0.4059, "ret_acc": 0.7908, "loss_ema": 4.4855, "held_ppl": 68.189, "held_ctr": 0.0446, "held_ret_acc": 1.0, "held_pos": 0.7893, "held_neg": 0.0054, "held_gap": 0.7839, "kurt": 6.76, "alpha": {"1": 0.4001, "7": 0.2216}, "step_tok": 262144, "pairs": 153, "fn_masked": 1, "mask_rate": 0.2041, "g_d": 1.276, "g_s": 0.021, "nhev": 31, "held_n": 64, "held_draws": 290304, "fallback": true, "dup_dropped": 89266, "pack_dropped": 431, "samples_seen": {"web": 21666501, "books": 692}, "gb": 46.52}
137
+ {"held_margin": 0.4582, "step": 26379, "elapsed_h": 6.0, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6202, "mlm": 4.2326, "ctr": 0.3876, "ctr_ema": 0.3929, "ret_acc": 0.8452, "loss_ema": 4.5112, "held_ppl": 63.364, "held_ctr": 0.0639, "held_ret_acc": 1.0, "held_pos": 0.7718, "held_neg": 0.0115, "held_gap": 0.7603, "kurt": 6.62, "alpha": {"1": 0.3859, "7": 0.227}, "step_tok": 262144, "pairs": 155, "fn_masked": 2, "mask_rate": 0.2038, "g_d": 1.365, "g_s": 0.02, "nhev": 31, "held_n": 64, "held_draws": 303295, "fallback": true, "dup_dropped": 96215, "pack_dropped": 436, "samples_seen": {"web": 21818690, "books": 692}, "gb": 46.52}
138
+ {"held_margin": 0.4831, "step": 26590, "elapsed_h": 6.251, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.5168, "mlm": 4.1084, "ctr": 0.4084, "ctr_ema": 0.4235, "ret_acc": 0.8645, "loss_ema": 4.5494, "held_ppl": 70.999, "held_ctr": 0.0464, "held_ret_acc": 1.0, "held_pos": 0.789, "held_neg": 0.0227, "held_gap": 0.7663, "kurt": 6.64, "alpha": {"1": 0.3877, "7": 0.2221}, "step_tok": 262144, "pairs": 155, "fn_masked": 0, "mask_rate": 0.2036, "g_d": 1.297, "g_s": 0.021, "nhev": 31, "held_n": 64, "held_draws": 317550, "fallback": true, "dup_dropped": 103756, "pack_dropped": 442, "samples_seen": {"web": 21983470, "books": 692}, "gb": 46.52}
139
+ {"held_margin": 0.4742, "step": 26783, "elapsed_h": 6.501, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.553, "mlm": 4.0939, "ctr": 0.459, "ctr_ema": 0.4249, "ret_acc": 0.8013, "loss_ema": 4.5274, "held_ppl": 73.652, "held_ctr": 0.0544, "held_ret_acc": 1.0, "held_pos": 0.7848, "held_neg": 0.0205, "held_gap": 0.7643, "kurt": 6.62, "alpha": {"1": 0.3914, "7": 0.2221}, "step_tok": 262144, "pairs": 151, "fn_masked": 0, "mask_rate": 0.2038, "g_d": 1.396, "g_s": 0.021, "nhev": 31, "held_n": 64, "held_draws": 330603, "fallback": true, "dup_dropped": 110794, "pack_dropped": 448, "samples_seen": {"web": 22135171, "books": 692}, "gb": 46.52}
140
+ {"held_margin": 0.4847, "step": 26978, "elapsed_h": 6.752, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.2172, "mlm": 3.9195, "ctr": 0.2977, "ctr_ema": 0.4055, "ret_acc": 0.8435, "loss_ema": 4.5134, "held_ppl": 65.924, "held_ctr": 0.0494, "held_ret_acc": 1.0, "held_pos": 0.7668, "held_neg": 0.0141, "held_gap": 0.7527, "kurt": 6.49, "alpha": {"1": 0.3867, "7": 0.2253}, "step_tok": 196608, "pairs": 115, "fn_masked": 1, "mask_rate": 0.2038, "g_d": 1.092, "g_s": 0.02, "nhev": 31, "held_n": 64, "held_draws": 343717, "fallback": true, "dup_dropped": 118167, "pack_dropped": 453, "samples_seen": {"web": 22289244, "books": 692}, "gb": 46.52}
141
+ {"held_margin": 0.4765, "step": 27189, "elapsed_h": 7.002, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6168, "mlm": 4.206, "ctr": 0.4109, "ctr_ema": 0.4247, "ret_acc": 0.8323, "loss_ema": 4.5388, "held_ppl": 69.513, "held_ctr": 0.062, "held_ret_acc": 1.0, "held_pos": 0.7706, "held_neg": 0.019, "held_gap": 0.7516, "kurt": 6.55, "alpha": {"1": 0.3946, "7": 0.2251}, "step_tok": 262144, "pairs": 155, "fn_masked": 0, "mask_rate": 0.2037, "g_d": 1.304, "g_s": 0.02, "nhev": 31, "held_n": 64, "held_draws": 358190, "fallback": true, "dup_dropped": 126677, "pack_dropped": 461, "samples_seen": {"web": 22455858, "books": 692}, "gb": 46.52}
142
+ {"held_margin": 0.4649, "step": 27383, "elapsed_h": 7.252, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4614, "mlm": 3.9878, "ctr": 0.4736, "ctr_ema": 0.406, "ret_acc": 0.8471, "loss_ema": 4.5073, "held_ppl": 59.369, "held_ctr": 0.0437, "held_ret_acc": 1.0, "held_pos": 0.8113, "held_neg": 0.0173, "held_gap": 0.7941, "kurt": 6.62, "alpha": {"1": 0.3802, "7": 0.2198}, "step_tok": 262144, "pairs": 157, "fn_masked": 0, "mask_rate": 0.2037, "g_d": 1.473, "g_s": 0.025, "nhev": 31, "held_n": 64, "held_draws": 371219, "fallback": true, "dup_dropped": 134442, "pack_dropped": 467, "samples_seen": {"web": 22607198, "books": 692}, "gb": 46.52}
143
+ {"held_margin": 0.4857, "step": 27592, "elapsed_h": 7.503, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4431, "mlm": 4.0698, "ctr": 0.3733, "ctr_ema": 0.4098, "ret_acc": 0.8968, "loss_ema": 4.4922, "held_ppl": 68.125, "held_ctr": 0.0437, "held_ret_acc": 1.0, "held_pos": 0.787, "held_neg": 0.027, "held_gap": 0.7601, "kurt": 6.71, "alpha": {"1": 0.3959, "7": 0.2258}, "step_tok": 262144, "pairs": 155, "fn_masked": 2, "mask_rate": 0.2038, "g_d": 1.4, "g_s": 0.022, "nhev": 31, "held_n": 64, "held_draws": 385295, "fallback": true, "dup_dropped": 143021, "pack_dropped": 475, "samples_seen": {"web": 22768212, "books": 692}, "gb": 46.52}
144
+ {"held_margin": 0.4752, "step": 27785, "elapsed_h": 7.753, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4626, "mlm": 3.983, "ctr": 0.4797, "ctr_ema": 0.3757, "ret_acc": 0.8267, "loss_ema": 4.4994, "held_ppl": 64.942, "held_ctr": 0.0447, "held_ret_acc": 1.0, "held_pos": 0.7983, "held_neg": 0.0088, "held_gap": 0.7895, "kurt": 6.6, "alpha": {"1": 0.396, "7": 0.2219}, "step_tok": 262144, "pairs": 150, "fn_masked": 4, "mask_rate": 0.2034, "g_d": 1.737, "g_s": 0.027, "nhev": 31, "held_n": 64, "held_draws": 398602, "fallback": true, "dup_dropped": 151319, "pack_dropped": 479, "samples_seen": {"web": 22922716, "books": 692}, "gb": 46.52}
145
+ {"held_margin": 0.4687, "step": 27979, "elapsed_h": 8.003, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4779, "mlm": 4.1628, "ctr": 0.3151, "ctr_ema": 0.3756, "ret_acc": 0.8654, "loss_ema": 4.4583, "held_ppl": 64.737, "held_ctr": 0.0557, "held_ret_acc": 1.0, "held_pos": 0.7814, "held_neg": 0.0139, "held_gap": 0.7675, "kurt": 6.76, "alpha": {"1": 0.3964, "7": 0.2236}, "step_tok": 262144, "pairs": 156, "fn_masked": 0, "mask_rate": 0.204, "g_d": 1.232, "g_s": 0.019, "nhev": 31, "held_n": 64, "held_draws": 411729, "fallback": true, "dup_dropped": 159860, "pack_dropped": 486, "samples_seen": {"web": 23076879, "books": 692}, "gb": 46.52}
146
+ {"held_margin": 0.4855, "step": 28187, "elapsed_h": 8.254, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6814, "mlm": 4.2841, "ctr": 0.3972, "ctr_ema": 0.3841, "ret_acc": 0.8387, "loss_ema": 4.4942, "held_ppl": 59.755, "held_ctr": 0.0421, "held_ret_acc": 1.0, "held_pos": 0.7834, "held_neg": 0.0154, "held_gap": 0.7681, "kurt": 6.55, "alpha": {"1": 0.3915, "7": 0.2215}, "step_tok": 262144, "pairs": 155, "fn_masked": 1, "mask_rate": 0.2043, "g_d": 1.274, "g_s": 0.022, "nhev": 31, "held_n": 64, "held_draws": 425828, "fallback": true, "dup_dropped": 169563, "pack_dropped": 492, "samples_seen": {"web": 23242420, "books": 692}, "gb": 46.52}
147
+ {"held_margin": 0.4685, "step": 28380, "elapsed_h": 8.504, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.3461, "mlm": 3.9749, "ctr": 0.3713, "ctr_ema": 0.4176, "ret_acc": 0.8516, "loss_ema": 4.5424, "held_ppl": 65.141, "held_ctr": 0.057, "held_ret_acc": 1.0, "held_pos": 0.78, "held_neg": 0.0173, "held_gap": 0.7627, "kurt": 6.5, "alpha": {"1": 0.3903, "7": 0.2221}, "step_tok": 262144, "pairs": 155, "fn_masked": 0, "mask_rate": 0.2042, "g_d": 1.36, "g_s": 0.023, "nhev": 31, "held_n": 64, "held_draws": 438822, "fallback": true, "dup_dropped": 178607, "pack_dropped": 499, "samples_seen": {"web": 23396930, "books": 692}, "gb": 46.52}
148
+ {"held_margin": 0.475, "step": 28591, "elapsed_h": 8.754, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.1109, "mlm": 3.8766, "ctr": 0.2343, "ctr_ema": 0.3762, "ret_acc": 0.9182, "loss_ema": 4.4413, "held_ppl": 63.97, "held_ctr": 0.0559, "held_ret_acc": 1.0, "held_pos": 0.7852, "held_neg": 0.0229, "held_gap": 0.7623, "kurt": 6.64, "alpha": {"1": 0.3839, "7": 0.2254}, "step_tok": 262144, "pairs": 159, "fn_masked": 0, "mask_rate": 0.2042, "g_d": 1.282, "g_s": 0.019, "nhev": 31, "held_n": 64, "held_draws": 453103, "fallback": true, "dup_dropped": 188672, "pack_dropped": 507, "samples_seen": {"web": 23565489, "books": 692}, "gb": 46.52}
149
+ {"held_margin": 0.4923, "step": 28785, "elapsed_h": 9.005, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.738, "mlm": 4.3773, "ctr": 0.3607, "ctr_ema": 0.4029, "ret_acc": 0.8919, "loss_ema": 4.5151, "held_ppl": 59.16, "held_ctr": 0.0424, "held_ret_acc": 1.0, "held_pos": 0.7973, "held_neg": 0.0126, "held_gap": 0.7846, "kurt": 6.58, "alpha": {"1": 0.392, "7": 0.2266}, "step_tok": 262144, "pairs": 148, "fn_masked": 184, "mask_rate": 0.2037, "g_d": 1.317, "g_s": 0.02, "nhev": 31, "held_n": 64, "held_draws": 466140, "fallback": true, "dup_dropped": 198099, "pack_dropped": 513, "samples_seen": {"web": 23718035, "books": 692}, "gb": 46.52}
150
+ {"held_margin": 0.4603, "step": 28977, "elapsed_h": 9.256, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.4281, "mlm": 4.0575, "ctr": 0.3706, "ctr_ema": 0.3922, "ret_acc": 0.8553, "loss_ema": 4.516, "held_ppl": 57.35, "held_ctr": 0.0807, "held_ret_acc": 1.0, "held_pos": 0.7533, "held_neg": 0.0025, "held_gap": 0.7508, "kurt": 6.71, "alpha": {"1": 0.3863, "7": 0.2212}, "step_tok": 262144, "pairs": 152, "fn_masked": 1, "mask_rate": 0.2042, "g_d": 1.358, "g_s": 0.021, "nhev": 31, "held_n": 64, "held_draws": 479193, "fallback": true, "dup_dropped": 208089, "pack_dropped": 515, "samples_seen": {"web": 23874021, "books": 692}, "gb": 46.52}
151
+ {"held_margin": 0.4769, "step": 29188, "elapsed_h": 9.506, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.6326, "mlm": 4.328, "ctr": 0.3046, "ctr_ema": 0.3856, "ret_acc": 0.8616, "loss_ema": 4.4646, "held_ppl": 68.585, "held_ctr": 0.048, "held_ret_acc": 1.0, "held_pos": 0.7897, "held_neg": 0.0149, "held_gap": 0.7748, "kurt": 6.6, "alpha": {"1": 0.3919, "7": 0.2181}, "step_tok": 262144, "pairs": 159, "fn_masked": 3, "mask_rate": 0.2038, "g_d": 1.173, "g_s": 0.019, "nhev": 31, "held_n": 64, "held_draws": 493407, "fallback": true, "dup_dropped": 218868, "pack_dropped": 524, "samples_seen": {"web": 24043102, "books": 692}, "gb": 46.52}
152
+ {"held_margin": 0.4915, "step": 29378, "elapsed_h": 9.756, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.2974, "mlm": 4.0968, "ctr": 0.2006, "ctr_ema": 0.3897, "ret_acc": 0.9321, "loss_ema": 4.4768, "held_ppl": 70.093, "held_ctr": 0.0465, "held_ret_acc": 1.0, "held_pos": 0.7889, "held_neg": 0.0229, "held_gap": 0.7659, "kurt": 6.6, "alpha": {"1": 0.39, "7": 0.227}, "step_tok": 262144, "pairs": 162, "fn_masked": 0, "mask_rate": 0.2039, "g_d": 1.102, "g_s": 0.017, "nhev": 31, "held_n": 64, "held_draws": 506193, "fallback": true, "dup_dropped": 228626, "pack_dropped": 527, "samples_seen": {"web": 24193958, "books": 692}, "gb": 46.52}
153
+ {"held_margin": 0.4715, "step": 29586, "elapsed_h": 10.007, "seq": 1024, "lr": 0.00015, "sched": "stable", "lr_s": 0.0007499999999999999, "lr_e": 7.5e-05, "loss": 4.2834, "mlm": 3.873, "ctr": 0.4104, "ctr_ema": 0.4066, "ret_acc": 0.8366, "loss_ema": 4.5026, "held_ppl": 68.332, "held_ctr": 0.0498, "held_ret_acc": 1.0, "held_pos": 0.7822, "held_neg": 0.0188, "held_gap": 0.7633, "kurt": 6.79, "alpha": {"1": 0.3954, "7": 0.2243}, "step_tok": 262144, "pairs": 153, "fn_masked": 0, "mask_rate": 0.2038, "g_d": 1.308, "g_s": 0.021, "nhev": 31, "held_n": 64, "held_draws": 520104, "fallback": true, "dup_dropped": 239618, "pack_dropped": 531, "samples_seen": {"web": 24358569, "books": 692}, "gb": 46.52}
sparse_opt.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ae2983004b38f15f90674e1e8cf0ea0780aca1aa1d0859d1c8f4ed6113354d74
3
+ size 15032388235
training_state.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "global_step": 29588,
3
+ "elapsed_offset": 36033.090921878815,
4
+ "elapsed_cum": 137411.46063423157,
5
+ "samples_seen": {
6
+ "web": 24360334,
7
+ "books": 692
8
+ },
9
+ "wrapped": {
10
+ "web": false,
11
+ "books": true
12
+ },
13
+ "loss_ema": 4.498992199493066,
14
+ "ctr_ema": 0.39153454558222694,
15
+ "total_tok": 7698644992,
16
+ "schedule_version": 1,
17
+ "decay_start": null,
18
+ "schedule_phase": "stable",
19
+ "pack_dropped": 531,
20
+ "books_epoch": 2
21
+ }