JohnnyZeppelin commited on
Commit
d7169af
·
verified ·
1 Parent(s): c78c96b

Upload folder using huggingface_hub

Browse files
Files changed (42) hide show
  1. planB_polish_robust_dec_noise10_cont/ckpt.pt +3 -0
  2. planB_polish_robust_dec_noise10_cont/config.json +26 -0
  3. planB_polish_robust_dec_noise10_cont/diag.jsonl +23 -0
  4. planB_polish_robust_dec_noise10_cont/diag_full.json +19 -0
  5. planB_polish_robust_dec_noise10_cont/logs.jsonl +0 -0
  6. planB_polish_robust_dec_noise10_cont/probe_leakage.json +35 -0
  7. planB_polish_robust_dec_noise10_cont_floor03/ckpt.pt +3 -0
  8. planB_polish_robust_dec_noise10_cont_floor03/config.json +26 -0
  9. planB_polish_robust_dec_noise10_cont_floor03/diag.jsonl +23 -0
  10. planB_polish_robust_dec_noise10_cont_floor03/diag_full.json +19 -0
  11. planB_polish_robust_dec_noise10_cont_floor03/logs.jsonl +0 -0
  12. planB_polish_robust_dec_noise10_cont_floor03/probe_leakage.json +35 -0
  13. planB_polish_robust_dec_noise10_floor03_pow05/ckpt.pt +3 -0
  14. planB_polish_robust_dec_noise10_floor03_pow05/config.json +26 -0
  15. planB_polish_robust_dec_noise10_floor03_pow05/diag.jsonl +22 -0
  16. planB_polish_robust_dec_noise10_floor03_pow05/diag_full.json +19 -0
  17. planB_polish_robust_dec_noise10_floor03_pow05/logs.jsonl +0 -0
  18. planB_polish_robust_dec_noise10_floor03_pow05/probe_leakage.json +35 -0
  19. planB_polish_robust_dec_noise10_r2h_floor02_pow10/ckpt.pt +3 -0
  20. planB_polish_robust_dec_noise10_r2h_floor02_pow10/config.json +26 -0
  21. planB_polish_robust_dec_noise10_r2h_floor02_pow10/diag.jsonl +23 -0
  22. planB_polish_robust_dec_noise10_r2h_floor02_pow10/diag_full.json +19 -0
  23. planB_polish_robust_dec_noise10_r2h_floor02_pow10/logs.jsonl +0 -0
  24. planB_polish_robust_dec_noise10_r2h_floor02_pow10/probe_leakage.json +35 -0
  25. planB_polish_robust_dec_r2i_std08_floor05/ckpt.pt +3 -0
  26. planB_polish_robust_dec_r2i_std08_floor05/config.json +26 -0
  27. planB_polish_robust_dec_r2i_std08_floor05/diag.jsonl +23 -0
  28. planB_polish_robust_dec_r2i_std08_floor05/diag_full.json +19 -0
  29. planB_polish_robust_dec_r2i_std08_floor05/logs.jsonl +0 -0
  30. planB_polish_robust_dec_r2i_std08_floor05/probe_leakage.json +35 -0
  31. planB_polish_robust_dec_r2k_lz02/ckpt.pt +3 -0
  32. planB_polish_robust_dec_r2k_lz02/config.json +30 -0
  33. planB_polish_robust_dec_r2k_lz02/diag.jsonl +11 -0
  34. planB_polish_robust_dec_r2k_lz02/diag_full.json +19 -0
  35. planB_polish_robust_dec_r2k_lz02/logs.jsonl +224 -0
  36. planB_polish_robust_dec_r2k_lz02/probe_leakage.json +35 -0
  37. planB_polish_robust_dec_r2k_zprob01/ckpt.pt +3 -0
  38. planB_polish_robust_dec_r2k_zprob01/config.json +30 -0
  39. planB_polish_robust_dec_r2k_zprob01/diag.jsonl +11 -0
  40. planB_polish_robust_dec_r2k_zprob01/diag_full.json +19 -0
  41. planB_polish_robust_dec_r2k_zprob01/logs.jsonl +224 -0
  42. planB_polish_robust_dec_r2k_zprob01/probe_leakage.json +35 -0
planB_polish_robust_dec_noise10_cont/ckpt.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd04aea55cbb425367253e966148fbdc88d79811af8cde20692a69b629cf5cb2
3
+ size 2439403257
planB_polish_robust_dec_noise10_cont/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train_jsonl": "data/groups_train.jsonl",
3
+ "valid_jsonl": "data/groups_valid.jsonl",
4
+ "run_dir": "runs/planB_polish_robust_dec_noise10_cont",
5
+ "resume": "runs/planB_polish_robust_dec_noise10/ckpt.pt",
6
+ "backbone": "google/mt5-small",
7
+ "num_latents": 16,
8
+ "batch_size": 4,
9
+ "grad_accum": 8,
10
+ "epochs": 2,
11
+ "max_doc_len": 256,
12
+ "max_sum_len": 64,
13
+ "eval_every": 400,
14
+ "max_train_examples": 0,
15
+ "max_valid_examples": 0,
16
+ "lr": 0.0001,
17
+ "weight_decay": 0.0,
18
+ "latent_dropout": 0.3,
19
+ "latent_noise_std": 0.1,
20
+ "noise_warmup_steps": 1600,
21
+ "noise_warmup_start_step": -1,
22
+ "noise_floor": 0.0,
23
+ "noise_power": 1.0,
24
+ "untie_lm_head": true,
25
+ "seed": 42
26
+ }
planB_polish_robust_dec_noise10_cont/diag.jsonl ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"step": 12000, "nll_en": 3.639061490274635, "nll_zh": 3.6403779681086794, "nll": 3.639719729191657}
2
+ {"step": 12400, "nll_en": 3.5795396760452625, "nll_zh": 3.5865511763171463, "nll": 3.5830454261812044}
3
+ {"step": 12800, "nll_en": 3.5674228607726146, "nll_zh": 3.571630957766769, "nll": 3.5695269092696917}
4
+ {"step": 13200, "nll_en": 3.5642870788090324, "nll_zh": 3.5690324331690846, "nll": 3.5666597559890585}
5
+ {"step": 13600, "nll_en": 3.550358092810084, "nll_zh": 3.554855001905252, "nll": 3.552606547357668}
6
+ {"step": 14000, "nll_en": 3.5461637666311123, "nll_zh": 3.548726994189853, "nll": 3.547445380410483}
7
+ {"step": 14400, "nll_en": 3.540014653074817, "nll_zh": 3.546470335623679, "nll": 3.543242494349248}
8
+ {"step": 14800, "nll_en": 3.5325205804933706, "nll_zh": 3.5390495316422768, "nll": 3.5357850560678235}
9
+ {"step": 15200, "nll_en": 3.5269479771749164, "nll_zh": 3.5334743667102515, "nll": 3.5302111719425837}
10
+ {"step": 15600, "nll_en": 3.51795480921707, "nll_zh": 3.5236852123671807, "nll": 3.5208200107921255}
11
+ {"step": 16000, "nll_en": 3.518414297769236, "nll_zh": 3.5234680871378794, "nll": 3.5209411924535576}
12
+ {"step": 16400, "nll_en": 3.5062415937540647, "nll_zh": 3.5118187386188144, "nll": 3.5090301661864394}
13
+ {"step": 16800, "nll_en": 3.6823073280286085, "nll_zh": 3.68175031165996, "nll": 3.6820288198442843}
14
+ {"step": 17200, "nll_en": 3.508091962866753, "nll_zh": 3.5125229877850974, "nll": 3.5103074753259254}
15
+ {"step": 17600, "nll_en": 3.4921103733538574, "nll_zh": 3.4960335929096376, "nll": 3.4940719831317475}
16
+ {"step": 18000, "nll_en": 3.4882996772862893, "nll_zh": 3.4924331850531742, "nll": 3.4903664311697318}
17
+ {"step": 18400, "nll_en": 3.481148925442524, "nll_zh": 3.4867278022443515, "nll": 3.483938363843438}
18
+ {"step": 18800, "nll_en": 3.4727731476878514, "nll_zh": 3.476969939709718, "nll": 3.4748715436987845}
19
+ {"step": 19200, "nll_en": 3.472127875112328, "nll_zh": 3.475698834997907, "nll": 3.4739133550551173}
20
+ {"step": 19600, "nll_en": 3.46572584128027, "nll_zh": 3.4701509989845323, "nll": 3.467938420132401}
21
+ {"step": 20000, "nll_en": 3.4624534852913014, "nll_zh": 3.467584092823698, "nll": 3.4650187890575}
22
+ {"step": 20400, "nll_en": 3.454255983138941, "nll_zh": 3.4573274455161225, "nll": 3.455791714327532}
23
+ {"step": 20800, "nll_en": 3.4493958209080122, "nll_zh": 3.4548085164318123, "nll": 3.4521021686699123}
planB_polish_robust_dec_noise10_cont/diag_full.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ckpt": "runs/planB_polish_robust_dec_noise10_cont/ckpt.pt",
3
+ "nll_en": 3.455909202769241,
4
+ "nll_zh": 3.460962747166575,
5
+ "swap_delta_en": 0.5770713935190729,
6
+ "swap_delta_zh": 0.5570867328764773,
7
+ "ablate_zero_delta_en": 2.3524466347240747,
8
+ "ablate_mean_delta_en": 0.25575099888614067,
9
+ "ablate_noise_delta_en": 3.907243016910351,
10
+ "ablate_zero_delta_zh": 2.3473930903267406,
11
+ "ablate_mean_delta_zh": 0.24409567226063122,
12
+ "ablate_noise_delta_zh": 3.9010576748192687,
13
+ "inv_top1_full": 0.060253698378801346,
14
+ "inv_top5_full": 0.20401690900325775,
15
+ "diag_sim_mean": 0.965337336063385,
16
+ "offdiag_sim_mean": 0.4495629370212555,
17
+ "sim_margin": 0.5157743990421295,
18
+ "n_valid": 946
19
+ }
planB_polish_robust_dec_noise10_cont/logs.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
planB_polish_robust_dec_noise10_cont/probe_leakage.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ckpt": "runs/planB_polish_robust_dec_noise10_cont/ckpt.pt",
3
+ "max_train_groups": 8000,
4
+ "max_valid_groups": 946,
5
+ "probe_hidden": 256,
6
+ "epochs": 20,
7
+ "lr": 0.01,
8
+ "baseline_majority_lang_acc": 0.5,
9
+ "baseline_majority_len_acc": 0.9271249771118164,
10
+ "lang_probe": {
11
+ "train_acc": 0.589,
12
+ "valid_acc": 0.5866807610993657,
13
+ "best_valid_acc": 0.5898520084566596
14
+ },
15
+ "len_probe": {
16
+ "train_acc": 0.9289375,
17
+ "valid_acc": 0.9011627906976745,
18
+ "best_valid_acc": 0.9011627906976745
19
+ },
20
+ "lang_probe_length_matched": {
21
+ "train_acc": 0.588707605344296,
22
+ "valid_acc": 0.5700110253583242,
23
+ "best_valid_acc": 0.5997794928335171
24
+ },
25
+ "lang_bucketwise_valid_acc": {
26
+ "bucket_0_acc": 0.34375,
27
+ "bucket_1_acc": 0.6000000238418579,
28
+ "bucket_2_acc": 0.5565217137336731,
29
+ "bucket_3_acc": 0.5929619073867798
30
+ },
31
+ "length_matched_sizes": {
32
+ "train": 15568,
33
+ "valid": 1814
34
+ }
35
+ }
planB_polish_robust_dec_noise10_cont_floor03/ckpt.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8dc8cecb0cfbcfeb10ceda01de475220067d1b3f3430f9f369a1f135fb247592
3
+ size 2439403257
planB_polish_robust_dec_noise10_cont_floor03/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train_jsonl": "data/groups_train.jsonl",
3
+ "valid_jsonl": "data/groups_valid.jsonl",
4
+ "run_dir": "runs/planB_polish_robust_dec_noise10_cont_floor03",
5
+ "resume": "runs/planB_polish_robust_dec_noise10/ckpt.pt",
6
+ "backbone": "google/mt5-small",
7
+ "num_latents": 16,
8
+ "batch_size": 4,
9
+ "grad_accum": 8,
10
+ "epochs": 2,
11
+ "max_doc_len": 256,
12
+ "max_sum_len": 64,
13
+ "eval_every": 400,
14
+ "max_train_examples": 0,
15
+ "max_valid_examples": 0,
16
+ "lr": 0.0001,
17
+ "weight_decay": 0.0,
18
+ "latent_dropout": 0.3,
19
+ "latent_noise_std": 0.1,
20
+ "noise_warmup_steps": 1600,
21
+ "noise_warmup_start_step": -1,
22
+ "noise_floor": 0.3,
23
+ "noise_power": 1.0,
24
+ "untie_lm_head": true,
25
+ "seed": 42
26
+ }
planB_polish_robust_dec_noise10_cont_floor03/diag.jsonl ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"step": 12000, "nll_en": 3.6392339480603724, "nll_zh": 3.6405573818698747, "nll": 3.6398956649651235}
2
+ {"step": 12400, "nll_en": 3.578603288840038, "nll_zh": 3.585251740622974, "nll": 3.581927514731506}
3
+ {"step": 12800, "nll_en": 3.5680070618998423, "nll_zh": 3.572117379049418, "nll": 3.57006222047463}
4
+ {"step": 13200, "nll_en": 3.562254467171796, "nll_zh": 3.5667225170337122, "nll": 3.564488492102754}
5
+ {"step": 13600, "nll_en": 3.5517645416501713, "nll_zh": 3.557065282261145, "nll": 3.554414911955658}
6
+ {"step": 14000, "nll_en": 3.5448383448239866, "nll_zh": 3.5478629773565378, "nll": 3.546350661090262}
7
+ {"step": 14400, "nll_en": 3.5398652679572398, "nll_zh": 3.5458602008053415, "nll": 3.5428627343812904}
8
+ {"step": 14800, "nll_en": 3.5310517244560775, "nll_zh": 3.537411905494351, "nll": 3.534231814975214}
9
+ {"step": 15200, "nll_en": 3.526407932378273, "nll_zh": 3.532648636977214, "nll": 3.5295282846777436}
10
+ {"step": 15600, "nll_en": 3.520603164809955, "nll_zh": 3.526506723359574, "nll": 3.5235549440847644}
11
+ {"step": 16000, "nll_en": 3.514282572345048, "nll_zh": 3.519213988967476, "nll": 3.516748280656262}
12
+ {"step": 16400, "nll_en": 3.5065276164081083, "nll_zh": 3.5137557449098873, "nll": 3.5101416806589976}
13
+ {"step": 16800, "nll_en": 3.6839124576737965, "nll_zh": 3.684118342450003, "nll": 3.6840154000618996}
14
+ {"step": 17200, "nll_en": 3.508062174153882, "nll_zh": 3.5116925904917164, "nll": 3.509877382322799}
15
+ {"step": 17600, "nll_en": 3.4965766953111452, "nll_zh": 3.500776867533839, "nll": 3.498676781422492}
16
+ {"step": 18000, "nll_en": 3.4861815051346956, "nll_zh": 3.489879660576141, "nll": 3.4880305828554183}
17
+ {"step": 18400, "nll_en": 3.4836242808852087, "nll_zh": 3.4882517344855857, "nll": 3.485938007685397}
18
+ {"step": 18800, "nll_en": 3.4770405156415816, "nll_zh": 3.4816431565718218, "nll": 3.4793418361067014}
19
+ {"step": 19200, "nll_en": 3.4733878069145736, "nll_zh": 3.4769527816369217, "nll": 3.4751702942757476}
20
+ {"step": 19600, "nll_en": 3.4655423083718686, "nll_zh": 3.470435571972966, "nll": 3.4679889401724173}
21
+ {"step": 20000, "nll_en": 3.461458399734336, "nll_zh": 3.4663071380357158, "nll": 3.4638827688850258}
22
+ {"step": 20400, "nll_en": 3.45506925764346, "nll_zh": 3.4580930534427314, "nll": 3.4565811555430956}
23
+ {"step": 20800, "nll_en": 3.4457650204793597, "nll_zh": 3.451642818733199, "nll": 3.448703919606279}
planB_polish_robust_dec_noise10_cont_floor03/diag_full.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ckpt": "runs/planB_polish_robust_dec_noise10_cont_floor03/ckpt.pt",
3
+ "nll_en": 3.4490990709300777,
4
+ "nll_zh": 3.4547658299290855,
5
+ "swap_delta_en": 0.5735126600205016,
6
+ "swap_delta_zh": 0.5527633753689852,
7
+ "ablate_zero_delta_en": 2.4566870532126557,
8
+ "ablate_mean_delta_en": 0.25113583921630084,
9
+ "ablate_noise_delta_en": 3.8464915868343814,
10
+ "ablate_zero_delta_zh": 2.451020294213648,
11
+ "ablate_mean_delta_zh": 0.2398719606137427,
12
+ "ablate_noise_delta_zh": 3.8820612496099813,
13
+ "inv_top1_full": 0.060253698378801346,
14
+ "inv_top5_full": 0.20401690900325775,
15
+ "diag_sim_mean": 0.965337336063385,
16
+ "offdiag_sim_mean": 0.4495629370212555,
17
+ "sim_margin": 0.5157743990421295,
18
+ "n_valid": 946
19
+ }
planB_polish_robust_dec_noise10_cont_floor03/logs.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
planB_polish_robust_dec_noise10_cont_floor03/probe_leakage.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ckpt": "runs/planB_polish_robust_dec_noise10_cont_floor03/ckpt.pt",
3
+ "max_train_groups": 8000,
4
+ "max_valid_groups": 946,
5
+ "probe_hidden": 256,
6
+ "epochs": 20,
7
+ "lr": 0.01,
8
+ "baseline_majority_lang_acc": 0.5,
9
+ "baseline_majority_len_acc": 0.9271249771118164,
10
+ "lang_probe": {
11
+ "train_acc": 0.584,
12
+ "valid_acc": 0.5972515856236786,
13
+ "best_valid_acc": 0.5972515856236786
14
+ },
15
+ "len_probe": {
16
+ "train_acc": 0.9289375,
17
+ "valid_acc": 0.9011627906976745,
18
+ "best_valid_acc": 0.9011627906976745
19
+ },
20
+ "lang_probe_length_matched": {
21
+ "train_acc": 0.5844681397738952,
22
+ "valid_acc": 0.5882028665931642,
23
+ "best_valid_acc": 0.5882028665931642
24
+ },
25
+ "lang_bucketwise_valid_acc": {
26
+ "bucket_0_acc": 0.34375,
27
+ "bucket_1_acc": 0.699999988079071,
28
+ "bucket_2_acc": 0.6173912882804871,
29
+ "bucket_3_acc": 0.5982404947280884
30
+ },
31
+ "length_matched_sizes": {
32
+ "train": 15568,
33
+ "valid": 1814
34
+ }
35
+ }
planB_polish_robust_dec_noise10_floor03_pow05/ckpt.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dfbc03a6394ce2dc095a7251d1adf439d46d3234de6b44857f0dbabadc4a62c0
3
+ size 2439403257
planB_polish_robust_dec_noise10_floor03_pow05/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train_jsonl": "data/groups_train.jsonl",
3
+ "valid_jsonl": "data/groups_valid.jsonl",
4
+ "run_dir": "runs/planB_polish_robust_dec_noise10_floor03_pow05",
5
+ "resume": "runs/planB_polish/ckpt.pt",
6
+ "backbone": "google/mt5-small",
7
+ "num_latents": 16,
8
+ "batch_size": 4,
9
+ "grad_accum": 8,
10
+ "epochs": 2,
11
+ "max_doc_len": 256,
12
+ "max_sum_len": 64,
13
+ "eval_every": 400,
14
+ "max_train_examples": 0,
15
+ "max_valid_examples": 0,
16
+ "lr": 0.0002,
17
+ "weight_decay": 0.0,
18
+ "latent_dropout": 0.3,
19
+ "latent_noise_std": 0.1,
20
+ "noise_warmup_steps": 1600,
21
+ "noise_warmup_start_step": -1,
22
+ "noise_floor": 0.3,
23
+ "noise_power": 0.5,
24
+ "untie_lm_head": true,
25
+ "seed": 42
26
+ }
planB_polish_robust_dec_noise10_floor03_pow05/diag.jsonl ADDED
@@ -0,0 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"step": 3200, "nll_en": 4.815716152715381, "nll_zh": 4.820947251914672, "nll": 4.818331702315026}
2
+ {"step": 3600, "nll_en": 4.564410871985599, "nll_zh": 4.566542262506787, "nll": 4.565476567246193}
3
+ {"step": 4000, "nll_en": 4.412489203519599, "nll_zh": 4.413854398384659, "nll": 4.413171800952129}
4
+ {"step": 4400, "nll_en": 4.290910454683526, "nll_zh": 4.294917765933918, "nll": 4.292914110308722}
5
+ {"step": 4800, "nll_en": 4.196160886050783, "nll_zh": 4.20108481743875, "nll": 4.198622851744767}
6
+ {"step": 5200, "nll_en": 4.123746619920146, "nll_zh": 4.126108768374421, "nll": 4.124927694147283}
7
+ {"step": 5600, "nll_en": 4.067222183904991, "nll_zh": 4.068375767915496, "nll": 4.067798975910243}
8
+ {"step": 6000, "nll_en": 4.009993143867993, "nll_zh": 4.01250443448476, "nll": 4.011248789176376}
9
+ {"step": 6400, "nll_en": 3.958995771710515, "nll_zh": 3.961503709345489, "nll": 3.9602497405280017}
10
+ {"step": 6800, "nll_en": 3.9275849505660396, "nll_zh": 3.9296358291996962, "nll": 3.928610389882868}
11
+ {"step": 7200, "nll_en": 3.8838748014502493, "nll_zh": 3.885269470497115, "nll": 3.8845721359736824}
12
+ {"step": 7600, "nll_en": 4.058586798056991, "nll_zh": 4.057908666058272, "nll": 4.058247732057632}
13
+ {"step": 8000, "nll_en": 3.8380877462552414, "nll_zh": 3.8418902816530514, "nll": 3.8399890139541464}
14
+ {"step": 8400, "nll_en": 3.792902407132042, "nll_zh": 3.7946143956819496, "nll": 3.793758401406996}
15
+ {"step": 8800, "nll_en": 3.7722739417255555, "nll_zh": 3.7769640232997514, "nll": 3.7746189825126537}
16
+ {"step": 9200, "nll_en": 3.743756562408383, "nll_zh": 3.74965866625183, "nll": 3.746707614330106}
17
+ {"step": 9600, "nll_en": 3.7236936571230093, "nll_zh": 3.7264923202563036, "nll": 3.7250929886896564}
18
+ {"step": 10000, "nll_en": 3.693869940574275, "nll_zh": 3.6996631239231745, "nll": 3.696766532248725}
19
+ {"step": 10400, "nll_en": 3.6775053369066426, "nll_zh": 3.681570664016691, "nll": 3.679538000461667}
20
+ {"step": 10800, "nll_en": 3.66168019736292, "nll_zh": 3.6644702290379723, "nll": 3.6630752132004463}
21
+ {"step": 11200, "nll_en": 3.637642835209788, "nll_zh": 3.6413269547024942, "nll": 3.639484894956141}
22
+ {"step": 11600, "nll_en": 3.626983867387187, "nll_zh": 3.629157028036944, "nll": 3.6280704477120658}
planB_polish_robust_dec_noise10_floor03_pow05/diag_full.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ckpt": "runs/planB_polish_robust_dec_noise10_floor03_pow05/ckpt.pt",
3
+ "nll_en": 3.616948983633997,
4
+ "nll_zh": 3.622239927408811,
5
+ "swap_delta_en": 0.5713690824286882,
6
+ "swap_delta_zh": 0.5538681375047874,
7
+ "ablate_zero_delta_en": 1.4302562185372156,
8
+ "ablate_mean_delta_en": 0.26539641656532853,
9
+ "ablate_noise_delta_en": 3.781647785017405,
10
+ "ablate_zero_delta_zh": 1.4249652747624015,
11
+ "ablate_mean_delta_zh": 0.25498392516412394,
12
+ "ablate_noise_delta_zh": 3.7961258767271144,
13
+ "inv_top1_full": 0.060253698378801346,
14
+ "inv_top5_full": 0.20401690900325775,
15
+ "diag_sim_mean": 0.965337336063385,
16
+ "offdiag_sim_mean": 0.4495629370212555,
17
+ "sim_margin": 0.5157743990421295,
18
+ "n_valid": 946
19
+ }
planB_polish_robust_dec_noise10_floor03_pow05/logs.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
planB_polish_robust_dec_noise10_floor03_pow05/probe_leakage.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ckpt": "runs/planB_polish_robust_dec_noise10_floor03_pow05/ckpt.pt",
3
+ "max_train_groups": 8000,
4
+ "max_valid_groups": 946,
5
+ "probe_hidden": 256,
6
+ "epochs": 20,
7
+ "lr": 0.01,
8
+ "baseline_majority_lang_acc": 0.5,
9
+ "baseline_majority_len_acc": 0.9271249771118164,
10
+ "lang_probe": {
11
+ "train_acc": 0.57775,
12
+ "valid_acc": 0.5702959830866807,
13
+ "best_valid_acc": 0.595137420718816
14
+ },
15
+ "len_probe": {
16
+ "train_acc": 0.92875,
17
+ "valid_acc": 0.9011627906976745,
18
+ "best_valid_acc": 0.9011627906976745
19
+ },
20
+ "lang_probe_length_matched": {
21
+ "train_acc": 0.5950668036998972,
22
+ "valid_acc": 0.5749724366041896,
23
+ "best_valid_acc": 0.5915104740904079
24
+ },
25
+ "lang_bucketwise_valid_acc": {
26
+ "bucket_0_acc": 0.40625,
27
+ "bucket_1_acc": 0.6000000238418579,
28
+ "bucket_2_acc": 0.5652173757553101,
29
+ "bucket_3_acc": 0.5730205774307251
30
+ },
31
+ "length_matched_sizes": {
32
+ "train": 15568,
33
+ "valid": 1814
34
+ }
35
+ }
planB_polish_robust_dec_noise10_r2h_floor02_pow10/ckpt.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:804debfa86eb1a9f2ca32a0db4362fc9e3b3162287503f07e9c956d9eda5ba59
3
+ size 2439403321
planB_polish_robust_dec_noise10_r2h_floor02_pow10/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train_jsonl": "data/groups_train.jsonl",
3
+ "valid_jsonl": "data/groups_valid.jsonl",
4
+ "run_dir": "runs/planB_polish_robust_dec_noise10_r2h_floor02_pow10",
5
+ "resume": "runs/planB_polish_robust_dec_noise10/ckpt.pt",
6
+ "backbone": "google/mt5-small",
7
+ "num_latents": 16,
8
+ "batch_size": 4,
9
+ "grad_accum": 8,
10
+ "epochs": 2,
11
+ "max_doc_len": 256,
12
+ "max_sum_len": 64,
13
+ "eval_every": 400,
14
+ "max_train_examples": 0,
15
+ "max_valid_examples": 0,
16
+ "lr": 0.0001,
17
+ "weight_decay": 0.0,
18
+ "latent_dropout": 0.3,
19
+ "latent_noise_std": 0.1,
20
+ "noise_warmup_steps": 1600,
21
+ "noise_warmup_start_step": -1,
22
+ "noise_floor": 0.2,
23
+ "noise_power": 1.0,
24
+ "untie_lm_head": true,
25
+ "seed": 42
26
+ }
planB_polish_robust_dec_noise10_r2h_floor02_pow10/diag.jsonl ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"step": 12000, "nll_en": 3.639122215940383, "nll_zh": 3.640447304062309, "nll": 3.639784760001346}
2
+ {"step": 12400, "nll_en": 3.579483327603491, "nll_zh": 3.5866062464724133, "nll": 3.5830447870379523}
3
+ {"step": 12800, "nll_en": 3.5689073324707548, "nll_zh": 3.5732184101612825, "nll": 3.5710628713160184}
4
+ {"step": 13200, "nll_en": 3.5632989492275247, "nll_zh": 3.5681528420075304, "nll": 3.5657258956175273}
5
+ {"step": 13600, "nll_en": 3.552900061294846, "nll_zh": 3.5579136471415675, "nll": 3.555406854218207}
6
+ {"step": 14000, "nll_en": 3.54472039265058, "nll_zh": 3.547594872892029, "nll": 3.5461576327713047}
7
+ {"step": 14400, "nll_en": 3.5407750581334057, "nll_zh": 3.5465214892623287, "nll": 3.543648273697867}
8
+ {"step": 14800, "nll_en": 3.531735758952812, "nll_zh": 3.5385571332612953, "nll": 3.5351464461070536}
9
+ {"step": 15200, "nll_en": 3.5277884656732734, "nll_zh": 3.5341441152463755, "nll": 3.5309662904598245}
10
+ {"step": 15600, "nll_en": 3.5209051932643383, "nll_zh": 3.526613840341064, "nll": 3.523759516802701}
11
+ {"step": 16000, "nll_en": 3.5111999259690654, "nll_zh": 3.5154007131403144, "nll": 3.51330031955469}
12
+ {"step": 16400, "nll_en": 3.502219701166133, "nll_zh": 3.509295205485241, "nll": 3.505757453325687}
13
+ {"step": 16800, "nll_en": 3.682667057811584, "nll_zh": 3.681769010128481, "nll": 3.6822180339700323}
14
+ {"step": 17200, "nll_en": 3.5064884955988878, "nll_zh": 3.5103246051715744, "nll": 3.508406550385231}
15
+ {"step": 17600, "nll_en": 3.4944138113590424, "nll_zh": 3.4978712749279577, "nll": 3.4961425431435003}
16
+ {"step": 18000, "nll_en": 3.4907478973679007, "nll_zh": 3.4949070878059114, "nll": 3.492827492586906}
17
+ {"step": 18400, "nll_en": 3.480186907988018, "nll_zh": 3.486387956470062, "nll": 3.48328743222904}
18
+ {"step": 18800, "nll_en": 3.4733434852535576, "nll_zh": 3.4780317572660224, "nll": 3.47568762125979}
19
+ {"step": 19200, "nll_en": 3.469461142638773, "nll_zh": 3.4728164622445945, "nll": 3.4711388024416836}
20
+ {"step": 19600, "nll_en": 3.463614428522219, "nll_zh": 3.4687861740967185, "nll": 3.4662003013094687}
21
+ {"step": 20000, "nll_en": 3.4634400371769267, "nll_zh": 3.468289697901139, "nll": 3.465864867539033}
22
+ {"step": 20400, "nll_en": 3.4543008078480373, "nll_zh": 3.4575060255683754, "nll": 3.4559034167082063}
23
+ {"step": 20800, "nll_en": 3.447374842887701, "nll_zh": 3.452513072757842, "nll": 3.4499439578227715}
planB_polish_robust_dec_noise10_r2h_floor02_pow10/diag_full.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ckpt": "runs/planB_polish_robust_dec_noise10_r2h_floor02_pow10/ckpt.pt",
3
+ "nll_en": 3.4513265416183634,
4
+ "nll_zh": 3.456295388445542,
5
+ "swap_delta_en": 0.5774954368397751,
6
+ "swap_delta_zh": 0.5581189512450397,
7
+ "ablate_zero_delta_en": 2.0421293542954686,
8
+ "ablate_mean_delta_en": 0.25529501654885034,
9
+ "ablate_noise_delta_en": 3.7427696163508153,
10
+ "ablate_zero_delta_zh": 2.03716050746829,
11
+ "ablate_mean_delta_zh": 0.24430991428851073,
12
+ "ablate_noise_delta_zh": 3.814268320861881,
13
+ "inv_top1_full": 0.060253698378801346,
14
+ "inv_top5_full": 0.20401690900325775,
15
+ "diag_sim_mean": 0.965337336063385,
16
+ "offdiag_sim_mean": 0.4495629370212555,
17
+ "sim_margin": 0.5157743990421295,
18
+ "n_valid": 946
19
+ }
planB_polish_robust_dec_noise10_r2h_floor02_pow10/logs.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
planB_polish_robust_dec_noise10_r2h_floor02_pow10/probe_leakage.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ckpt": "runs/planB_polish_robust_dec_noise10_r2h_floor02_pow10/ckpt.pt",
3
+ "max_train_groups": 8000,
4
+ "max_valid_groups": 946,
5
+ "probe_hidden": 256,
6
+ "epochs": 20,
7
+ "lr": 0.01,
8
+ "baseline_majority_lang_acc": 0.5,
9
+ "baseline_majority_len_acc": 0.9271249771118164,
10
+ "lang_probe": {
11
+ "train_acc": 0.588125,
12
+ "valid_acc": 0.5840380549682875,
13
+ "best_valid_acc": 0.5977801268498943
14
+ },
15
+ "len_probe": {
16
+ "train_acc": 0.9288125,
17
+ "valid_acc": 0.9011627906976745,
18
+ "best_valid_acc": 0.9011627906976745
19
+ },
20
+ "lang_probe_length_matched": {
21
+ "train_acc": 0.5861382322713258,
22
+ "valid_acc": 0.5887541345093715,
23
+ "best_valid_acc": 0.5986769570011026
24
+ },
25
+ "lang_bucketwise_valid_acc": {
26
+ "bucket_0_acc": 0.4375,
27
+ "bucket_1_acc": 0.550000011920929,
28
+ "bucket_2_acc": 0.5652173757553101,
29
+ "bucket_3_acc": 0.5888563394546509
30
+ },
31
+ "length_matched_sizes": {
32
+ "train": 15568,
33
+ "valid": 1814
34
+ }
35
+ }
planB_polish_robust_dec_r2i_std08_floor05/ckpt.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d56b8536108bee017cc5b68c15b248fa2a2a44ea630b5e802402a2850557001
3
+ size 2439403257
planB_polish_robust_dec_r2i_std08_floor05/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train_jsonl": "data/groups_train.jsonl",
3
+ "valid_jsonl": "data/groups_valid.jsonl",
4
+ "run_dir": "runs/planB_polish_robust_dec_r2i_std08_floor05",
5
+ "resume": "runs/planB_polish_robust_dec_noise10/ckpt.pt",
6
+ "backbone": "google/mt5-small",
7
+ "num_latents": 16,
8
+ "batch_size": 4,
9
+ "grad_accum": 8,
10
+ "epochs": 2,
11
+ "max_doc_len": 256,
12
+ "max_sum_len": 64,
13
+ "eval_every": 400,
14
+ "max_train_examples": 0,
15
+ "max_valid_examples": 0,
16
+ "lr": 0.0001,
17
+ "weight_decay": 0.0,
18
+ "latent_dropout": 0.3,
19
+ "latent_noise_std": 0.08,
20
+ "noise_warmup_steps": 1600,
21
+ "noise_warmup_start_step": -1,
22
+ "noise_floor": 0.5,
23
+ "noise_power": 1.0,
24
+ "untie_lm_head": true,
25
+ "seed": 42
26
+ }
planB_polish_robust_dec_r2i_std08_floor05/diag.jsonl ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"step": 12000, "nll_en": 3.639309922434059, "nll_zh": 3.640642463530849, "nll": 3.639976192982454}
2
+ {"step": 12400, "nll_en": 3.579677199712515, "nll_zh": 3.585978259999959, "nll": 3.582827729856237}
3
+ {"step": 12800, "nll_en": 3.565980377963431, "nll_zh": 3.5709019372629567, "nll": 3.5684411576131936}
4
+ {"step": 13200, "nll_en": 3.565598818514866, "nll_zh": 3.570454658968, "nll": 3.568026738741433}
5
+ {"step": 13600, "nll_en": 3.5528014880658203, "nll_zh": 3.558487421362395, "nll": 3.5556444547141077}
6
+ {"step": 14000, "nll_en": 3.5433913497037666, "nll_zh": 3.545671702943443, "nll": 3.544531526323605}
7
+ {"step": 14400, "nll_en": 3.541428400648573, "nll_zh": 3.546172593663157, "nll": 3.543800497155865}
8
+ {"step": 14800, "nll_en": 3.531274393295385, "nll_zh": 3.536736472212487, "nll": 3.534005432753936}
9
+ {"step": 15200, "nll_en": 3.5254173752621414, "nll_zh": 3.5317312990636203, "nll": 3.5285743371628806}
10
+ {"step": 15600, "nll_en": 3.5196065580113998, "nll_zh": 3.526132960652196, "nll": 3.522869759331798}
11
+ {"step": 16000, "nll_en": 3.5124856694808217, "nll_zh": 3.516866459150899, "nll": 3.5146760643158603}
12
+ {"step": 16400, "nll_en": 3.507827251708533, "nll_zh": 3.514900437369155, "nll": 3.511363844538844}
13
+ {"step": 16800, "nll_en": 3.6774377913606093, "nll_zh": 3.676953731077166, "nll": 3.6771957612188877}
14
+ {"step": 17200, "nll_en": 3.5074501017435407, "nll_zh": 3.5112858675499043, "nll": 3.5093679846467225}
15
+ {"step": 17600, "nll_en": 3.4935992918357286, "nll_zh": 3.497349257196987, "nll": 3.495474274516358}
16
+ {"step": 18000, "nll_en": 3.486186614248264, "nll_zh": 3.490686818862818, "nll": 3.4884367165555408}
17
+ {"step": 18400, "nll_en": 3.4840000144522993, "nll_zh": 3.4883164553007164, "nll": 3.4861582348765077}
18
+ {"step": 18800, "nll_en": 3.475878347554116, "nll_zh": 3.478879014230934, "nll": 3.477378680892525}
19
+ {"step": 19200, "nll_en": 3.4722120313261327, "nll_zh": 3.4764316218095903, "nll": 3.4743218265678615}
20
+ {"step": 19600, "nll_en": 3.4655355641000116, "nll_zh": 3.471038711499462, "nll": 3.4682871377997366}
21
+ {"step": 20000, "nll_en": 3.4608596757400867, "nll_zh": 3.4661183024561684, "nll": 3.4634889890981273}
22
+ {"step": 20400, "nll_en": 3.45611714711905, "nll_zh": 3.4590656318825848, "nll": 3.4575913895008177}
23
+ {"step": 20800, "nll_en": 3.4483303822120197, "nll_zh": 3.4539783268095827, "nll": 3.4511543545108014}
planB_polish_robust_dec_r2i_std08_floor05/diag_full.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ckpt": "runs/planB_polish_robust_dec_r2i_std08_floor05/ckpt.pt",
3
+ "nll_en": 3.4535187182920426,
4
+ "nll_zh": 3.45839760318619,
5
+ "swap_delta_en": 0.5759819950160214,
6
+ "swap_delta_zh": 0.5564471712576159,
7
+ "ablate_zero_delta_en": 2.2557294857678403,
8
+ "ablate_mean_delta_en": 0.251795043904988,
9
+ "ablate_noise_delta_en": 3.7902107531122122,
10
+ "ablate_zero_delta_zh": 2.2508506008736933,
11
+ "ablate_mean_delta_zh": 0.24117677055504055,
12
+ "ablate_noise_delta_zh": 3.859085573958544,
13
+ "inv_top1_full": 0.060253698378801346,
14
+ "inv_top5_full": 0.20401690900325775,
15
+ "diag_sim_mean": 0.965337336063385,
16
+ "offdiag_sim_mean": 0.4495629370212555,
17
+ "sim_margin": 0.5157743990421295,
18
+ "n_valid": 946
19
+ }
planB_polish_robust_dec_r2i_std08_floor05/logs.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
planB_polish_robust_dec_r2i_std08_floor05/probe_leakage.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ckpt": "runs/planB_polish_robust_dec_r2i_std08_floor05/ckpt.pt",
3
+ "max_train_groups": 8000,
4
+ "max_valid_groups": 946,
5
+ "probe_hidden": 256,
6
+ "epochs": 20,
7
+ "lr": 0.01,
8
+ "baseline_majority_lang_acc": 0.5,
9
+ "baseline_majority_len_acc": 0.9271249771118164,
10
+ "lang_probe": {
11
+ "train_acc": 0.5855625,
12
+ "valid_acc": 0.5782241014799154,
13
+ "best_valid_acc": 0.5919661733615222
14
+ },
15
+ "len_probe": {
16
+ "train_acc": 0.92875,
17
+ "valid_acc": 0.9011627906976745,
18
+ "best_valid_acc": 0.9011627906976745
19
+ },
20
+ "lang_probe_length_matched": {
21
+ "train_acc": 0.5844681397738952,
22
+ "valid_acc": 0.5920617420066152,
23
+ "best_valid_acc": 0.597574421168688
24
+ },
25
+ "lang_bucketwise_valid_acc": {
26
+ "bucket_0_acc": 0.3125,
27
+ "bucket_1_acc": 0.625,
28
+ "bucket_2_acc": 0.5826086401939392,
29
+ "bucket_3_acc": 0.5818182229995728
30
+ },
31
+ "length_matched_sizes": {
32
+ "train": 15568,
33
+ "valid": 1814
34
+ }
35
+ }
planB_polish_robust_dec_r2k_lz02/ckpt.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:21065d73afe00332add552444c24be28162a74ed07a11a10e5c5447e49b64669
3
+ size 2439403385
planB_polish_robust_dec_r2k_lz02/config.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train_jsonl": "data/groups_train.jsonl",
3
+ "valid_jsonl": "data/groups_valid.jsonl",
4
+ "run_dir": "runs/planB_polish_robust_dec_r2k_lz02",
5
+ "resume": "runs/planB_polish_robust_dec_noise10_cont_floor03/ckpt.pt",
6
+ "backbone": "google/mt5-small",
7
+ "num_latents": 16,
8
+ "batch_size": 4,
9
+ "grad_accum": 8,
10
+ "epochs": 1,
11
+ "max_doc_len": 256,
12
+ "max_sum_len": 64,
13
+ "eval_every": 400,
14
+ "log_every": 20,
15
+ "max_train_examples": 0,
16
+ "max_valid_examples": 0,
17
+ "lr": 0.0003,
18
+ "weight_decay": 0.0,
19
+ "latent_dropout": 0.1,
20
+ "latent_noise_std": 0.1,
21
+ "noise_warmup_steps": 500,
22
+ "noise_warmup_start_step": -1,
23
+ "noise_mult_floor": 0.3,
24
+ "noise_mult_pow": 0.5,
25
+ "z_zero_prob": 0.0,
26
+ "z_zero_loss_weight": 0.2,
27
+ "z_zero_loss_every": 1,
28
+ "untie_lm_head": false,
29
+ "seed": 42
30
+ }
planB_polish_robust_dec_r2k_lz02/diag.jsonl ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"step": 21200, "nll_en": 3.709797738218408, "nll_zh": 3.7108953064642294, "nll": 3.710346522341319}
2
+ {"step": 21600, "nll_en": 3.5124802670065494, "nll_zh": 3.5186123585852207, "nll": 3.515546312795885}
3
+ {"step": 22000, "nll_en": 3.4770791404595083, "nll_zh": 3.4833650246231045, "nll": 3.4802220825413066}
4
+ {"step": 22400, "nll_en": 3.457237751740986, "nll_zh": 3.462468435597974, "nll": 3.45985309366948}
5
+ {"step": 22800, "nll_en": 3.446665392869623, "nll_zh": 3.4524017338016826, "nll": 3.4495335633356525}
6
+ {"step": 23200, "nll_en": 3.420740869534192, "nll_zh": 3.4258902027541436, "nll": 3.423315536144168}
7
+ {"step": 23600, "nll_en": 3.4108592077743176, "nll_zh": 3.412263067781799, "nll": 3.4115611377780586}
8
+ {"step": 24000, "nll_en": 3.3852123014518627, "nll_zh": 3.388282553086069, "nll": 3.3867474272689657}
9
+ {"step": 24400, "nll_en": 3.3782333128044018, "nll_zh": 3.3808601945953694, "nll": 3.3795467536998856}
10
+ {"step": 24800, "nll_en": 3.364263141382068, "nll_zh": 3.36926932667577, "nll": 3.366766234028919}
11
+ {"step": 25200, "nll_en": 3.36157079478903, "nll_zh": 3.364531648083419, "nll": 3.3630512214362245}
planB_polish_robust_dec_r2k_lz02/diag_full.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ckpt": "runs/planB_polish_robust_dec_r2k_lz02/ckpt.pt",
3
+ "nll_en": 3.349451068087814,
4
+ "nll_zh": 3.3522998565851254,
5
+ "swap_delta_en": 0.46363295258729703,
6
+ "swap_delta_zh": 0.4446804024452387,
7
+ "ablate_zero_delta_en": 0.06420324317496624,
8
+ "ablate_mean_delta_en": 0.1719418549890498,
9
+ "ablate_noise_delta_en": 3.9560762530401443,
10
+ "ablate_zero_delta_zh": 0.06135445467765437,
11
+ "ablate_mean_delta_zh": 0.16415762246033103,
12
+ "ablate_noise_delta_zh": 3.986062584165287,
13
+ "inv_top1_full": 0.060253698378801346,
14
+ "inv_top5_full": 0.20401690900325775,
15
+ "diag_sim_mean": 0.965337336063385,
16
+ "offdiag_sim_mean": 0.4495629370212555,
17
+ "sim_margin": 0.5157743990421295,
18
+ "n_valid": 946
19
+ }
planB_polish_robust_dec_r2k_lz02/logs.jsonl ADDED
@@ -0,0 +1,224 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"step": 20940, "epoch": 0, "loss": 6.271872043609619, "loss_main": 4.999202728271484, "loss_zero": 6.363346576690674, "loss_en": 4.989892482757568, "loss_zh": 5.008512496948242, "noise_ramp": 0.024, "noise_mult": 0.6393877863883972, "noise_std_eff": 0.0015345306873321533, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 4.655174016952515}
2
+ {"step": 20960, "epoch": 0, "loss": 5.578813076019287, "loss_main": 4.464287757873535, "loss_zero": 5.572625637054443, "loss_en": 4.45797872543335, "loss_zh": 4.470596790313721, "noise_ramp": 0.064, "noise_mult": 0.7391014099121094, "noise_std_eff": 0.004730249023437501, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 10.735685586929321}
3
+ {"step": 20980, "epoch": 0, "loss": 5.977933883666992, "loss_main": 4.807156562805176, "loss_zero": 5.853886604309082, "loss_en": 4.734373092651367, "loss_zh": 4.879939556121826, "noise_ramp": 0.104, "noise_mult": 0.5647760629653931, "noise_std_eff": 0.005873671054840088, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 16.659618377685547}
4
+ {"step": 21000, "epoch": 0, "loss": 6.143551826477051, "loss_main": 5.010152339935303, "loss_zero": 5.66699743270874, "loss_en": 5.057928562164307, "loss_zh": 4.962376117706299, "noise_ramp": 0.144, "noise_mult": 0.9902229309082031, "noise_std_eff": 0.014259210205078126, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 22.574955463409424}
5
+ {"step": 21020, "epoch": 0, "loss": 6.332795143127441, "loss_main": 5.143570899963379, "loss_zero": 5.946121692657471, "loss_en": 5.233747959136963, "loss_zh": 5.053394317626953, "noise_ramp": 0.184, "noise_mult": 0.8285276889801025, "noise_std_eff": 0.015244909477233887, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 28.40499782562256}
6
+ {"step": 21040, "epoch": 0, "loss": 5.144381046295166, "loss_main": 4.16522216796875, "loss_zero": 4.895794868469238, "loss_en": 4.236445903778076, "loss_zh": 4.093998908996582, "noise_ramp": 0.224, "noise_mult": 0.9277439117431641, "noise_std_eff": 0.020781463623046875, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 34.28642392158508}
7
+ {"step": 21060, "epoch": 0, "loss": 5.05975866317749, "loss_main": 4.081377029418945, "loss_zero": 4.891909122467041, "loss_en": 4.055792331695557, "loss_zh": 4.106961727142334, "noise_ramp": 0.264, "noise_mult": 0.9126155972480774, "noise_std_eff": 0.024093051767349247, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 40.16259217262268}
8
+ {"step": 21080, "epoch": 0, "loss": 5.507012367248535, "loss_main": 4.469840049743652, "loss_zero": 5.185861110687256, "loss_en": 4.6132893562316895, "loss_zh": 4.326390266418457, "noise_ramp": 0.304, "noise_mult": 0.4871520698070526, "noise_std_eff": 0.0148094229221344, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 46.375879526138306}
9
+ {"step": 21100, "epoch": 0, "loss": 5.3095598220825195, "loss_main": 4.376108169555664, "loss_zero": 4.667259216308594, "loss_en": 4.306046009063721, "loss_zh": 4.446170330047607, "noise_ramp": 0.344, "noise_mult": 0.5296347737312317, "noise_std_eff": 0.01821943621635437, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 52.79526615142822}
10
+ {"step": 21120, "epoch": 0, "loss": 5.811549186706543, "loss_main": 4.767222881317139, "loss_zero": 5.221630096435547, "loss_en": 4.732890605926514, "loss_zh": 4.801555156707764, "noise_ramp": 0.384, "noise_mult": 0.9227753281593323, "noise_std_eff": 0.03543457260131836, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 58.57342290878296}
11
+ {"step": 21140, "epoch": 0, "loss": 5.004327297210693, "loss_main": 4.120783805847168, "loss_zero": 4.417717933654785, "loss_en": 3.9832217693328857, "loss_zh": 4.258346080780029, "noise_ramp": 0.424, "noise_mult": 0.5867400765419006, "noise_std_eff": 0.024877779245376586, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 64.33978271484375}
12
+ {"step": 21160, "epoch": 0, "loss": 5.388506889343262, "loss_main": 4.448643207550049, "loss_zero": 4.699317932128906, "loss_en": 4.551909446716309, "loss_zh": 4.345376968383789, "noise_ramp": 0.464, "noise_mult": 0.9461555480957031, "noise_std_eff": 0.04390161743164063, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 70.0869038105011}
13
+ {"step": 21180, "epoch": 0, "loss": 5.211620330810547, "loss_main": 4.345714092254639, "loss_zero": 4.329530715942383, "loss_en": 4.348674774169922, "loss_zh": 4.3427534103393555, "noise_ramp": 0.504, "noise_mult": 0.8105512261390686, "noise_std_eff": 0.04085178179740906, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 75.808345079422}
14
+ {"step": 21200, "epoch": 0, "loss": 6.1579742431640625, "loss_main": 5.134749412536621, "loss_zero": 5.116125106811523, "loss_en": 5.231325626373291, "loss_zh": 5.038173675537109, "noise_ramp": 0.544, "noise_mult": 0.7443344593048096, "noise_std_eff": 0.04049179458618164, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 81.56260752677917}
15
+ {"step": 21220, "epoch": 0, "loss": 3.8347039222717285, "loss_main": 3.1684305667877197, "loss_zero": 3.3313660621643066, "loss_en": 3.1537365913391113, "loss_zh": 3.183124542236328, "noise_ramp": 0.584, "noise_mult": 0.8757113218307495, "noise_std_eff": 0.05114154119491577, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 111.7076506614685}
16
+ {"step": 21240, "epoch": 0, "loss": 5.09527063369751, "loss_main": 4.219029426574707, "loss_zero": 4.381205081939697, "loss_en": 4.2097487449646, "loss_zh": 4.228310585021973, "noise_ramp": 0.624, "noise_mult": 0.6510562896728516, "noise_std_eff": 0.04062591247558594, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 117.20488834381104}
17
+ {"step": 21260, "epoch": 0, "loss": 4.4707465171813965, "loss_main": 3.6781692504882812, "loss_zero": 3.962886095046997, "loss_en": 3.660944938659668, "loss_zh": 3.6953933238983154, "noise_ramp": 0.664, "noise_mult": 0.5684519410133362, "noise_std_eff": 0.03774520888328553, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 122.63625907897949}
18
+ {"step": 21280, "epoch": 0, "loss": 3.4822092056274414, "loss_main": 2.84743595123291, "loss_zero": 3.1738662719726562, "loss_en": 2.8529677391052246, "loss_zh": 2.8419041633605957, "noise_ramp": 0.704, "noise_mult": 0.394264280796051, "noise_std_eff": 0.027756205368041992, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 128.12227821350098}
19
+ {"step": 21300, "epoch": 0, "loss": 4.1023054122924805, "loss_main": 3.3881654739379883, "loss_zero": 3.5706987380981445, "loss_en": 3.3828017711639404, "loss_zh": 3.393528938293457, "noise_ramp": 0.744, "noise_mult": 0.9321831464767456, "noise_std_eff": 0.06935442609786988, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 133.57345271110535}
20
+ {"step": 21320, "epoch": 0, "loss": 3.998100519180298, "loss_main": 3.324305534362793, "loss_zero": 3.368974447250366, "loss_en": 3.3312644958496094, "loss_zh": 3.3173463344573975, "noise_ramp": 0.784, "noise_mult": 0.5544742941856384, "noise_std_eff": 0.04347078466415406, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 139.0582754611969}
21
+ {"step": 21340, "epoch": 0, "loss": 3.7670161724090576, "loss_main": 3.099644184112549, "loss_zero": 3.336860179901123, "loss_en": 3.0858540534973145, "loss_zh": 3.113434314727783, "noise_ramp": 0.824, "noise_mult": 0.9263652563095093, "noise_std_eff": 0.07633249711990356, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 144.52838826179504}
22
+ {"step": 21360, "epoch": 0, "loss": 4.252560615539551, "loss_main": 3.4958319664001465, "loss_zero": 3.783642053604126, "loss_en": 3.536754846572876, "loss_zh": 3.454909324645996, "noise_ramp": 0.864, "noise_mult": 0.682509183883667, "noise_std_eff": 0.05896879348754883, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 150.0637149810791}
23
+ {"step": 21380, "epoch": 0, "loss": 3.947115421295166, "loss_main": 3.2493557929992676, "loss_zero": 3.4887986183166504, "loss_en": 3.247314691543579, "loss_zh": 3.251396656036377, "noise_ramp": 0.904, "noise_mult": 0.8856262564659119, "noise_std_eff": 0.08006061358451844, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 155.49810528755188}
24
+ {"step": 21400, "epoch": 0, "loss": 4.419099807739258, "loss_main": 3.6590147018432617, "loss_zero": 3.8004250526428223, "loss_en": 3.6930346488952637, "loss_zh": 3.6249947547912598, "noise_ramp": 0.944, "noise_mult": 0.7337379455566406, "noise_std_eff": 0.06926486206054687, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 161.0041265487671}
25
+ {"step": 21420, "epoch": 0, "loss": 3.966787338256836, "loss_main": 3.2638163566589355, "loss_zero": 3.5148544311523438, "loss_en": 3.2659504413604736, "loss_zh": 3.2616825103759766, "noise_ramp": 0.984, "noise_mult": 0.9616872072219849, "noise_std_eff": 0.09463002119064332, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 166.5410258769989}
26
+ {"step": 21440, "epoch": 0, "loss": 4.325601577758789, "loss_main": 3.5603604316711426, "loss_zero": 3.826204538345337, "loss_en": 3.5147554874420166, "loss_zh": 3.6059656143188477, "noise_ramp": 1.0, "noise_mult": 0.9851067066192627, "noise_std_eff": 0.09851067066192627, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 172.09621024131775}
27
+ {"step": 21460, "epoch": 0, "loss": 4.690757751464844, "loss_main": 3.8599929809570312, "loss_zero": 4.153824329376221, "loss_en": 3.873178005218506, "loss_zh": 3.8468077182769775, "noise_ramp": 1.0, "noise_mult": 0.9417654275894165, "noise_std_eff": 0.09417654275894166, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 177.64759397506714}
28
+ {"step": 21480, "epoch": 0, "loss": 3.937520742416382, "loss_main": 3.262509822845459, "loss_zero": 3.375054359436035, "loss_en": 3.266911506652832, "loss_zh": 3.258107900619507, "noise_ramp": 1.0, "noise_mult": 0.97783362865448, "noise_std_eff": 0.097783362865448, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 183.30312085151672}
29
+ {"step": 21500, "epoch": 0, "loss": 4.020008563995361, "loss_main": 3.310488224029541, "loss_zero": 3.547602415084839, "loss_en": 3.2823331356048584, "loss_zh": 3.3386430740356445, "noise_ramp": 1.0, "noise_mult": 0.9674955010414124, "noise_std_eff": 0.09674955010414124, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 188.9711492061615}
30
+ {"step": 21520, "epoch": 0, "loss": 3.7418506145477295, "loss_main": 3.0543088912963867, "loss_zero": 3.4377081394195557, "loss_en": 3.1159517765045166, "loss_zh": 2.9926657676696777, "noise_ramp": 1.0, "noise_mult": 0.4809829592704773, "noise_std_eff": 0.04809829592704773, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 194.5644154548645}
31
+ {"step": 21540, "epoch": 0, "loss": 3.641278028488159, "loss_main": 3.002668857574463, "loss_zero": 3.1930456161499023, "loss_en": 3.0031707286834717, "loss_zh": 3.002166986465454, "noise_ramp": 1.0, "noise_mult": 0.6080971956253052, "noise_std_eff": 0.06080971956253052, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 200.08188557624817}
32
+ {"step": 21560, "epoch": 0, "loss": 3.7476255893707275, "loss_main": 3.0886051654815674, "loss_zero": 3.2951016426086426, "loss_en": 3.102020502090454, "loss_zh": 3.0751898288726807, "noise_ramp": 1.0, "noise_mult": 0.997454822063446, "noise_std_eff": 0.0997454822063446, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 205.62211203575134}
33
+ {"step": 21580, "epoch": 0, "loss": 3.8620872497558594, "loss_main": 3.203613758087158, "loss_zero": 3.2923669815063477, "loss_en": 3.199674606323242, "loss_zh": 3.207552909851074, "noise_ramp": 1.0, "noise_mult": 0.8814060688018799, "noise_std_eff": 0.08814060688018799, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 211.19954824447632}
34
+ {"step": 21600, "epoch": 0, "loss": 4.442743301391602, "loss_main": 3.664742946624756, "loss_zero": 3.890002965927124, "loss_en": 3.640730142593384, "loss_zh": 3.688755512237549, "noise_ramp": 1.0, "noise_mult": 0.34293878078460693, "noise_std_eff": 0.034293878078460696, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 217.0545973777771}
35
+ {"step": 21620, "epoch": 0, "loss": 4.384718894958496, "loss_main": 3.6079325675964355, "loss_zero": 3.883932113647461, "loss_en": 3.6205368041992188, "loss_zh": 3.5953283309936523, "noise_ramp": 1.0, "noise_mult": 0.4884576201438904, "noise_std_eff": 0.04884576201438904, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 248.00567865371704}
36
+ {"step": 21640, "epoch": 0, "loss": 4.226445198059082, "loss_main": 3.498170852661133, "loss_zero": 3.641371488571167, "loss_en": 3.4882562160491943, "loss_zh": 3.508085250854492, "noise_ramp": 1.0, "noise_mult": 0.9814921617507935, "noise_std_eff": 0.09814921617507935, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 253.57772850990295}
37
+ {"step": 21660, "epoch": 0, "loss": 4.5649542808532715, "loss_main": 3.7560319900512695, "loss_zero": 4.044610500335693, "loss_en": 3.7675342559814453, "loss_zh": 3.7445294857025146, "noise_ramp": 1.0, "noise_mult": 0.4871969223022461, "noise_std_eff": 0.04871969223022461, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 259.14049768447876}
38
+ {"step": 21680, "epoch": 0, "loss": 4.300538539886475, "loss_main": 3.5533971786499023, "loss_zero": 3.735706329345703, "loss_en": 3.549187421798706, "loss_zh": 3.5576066970825195, "noise_ramp": 1.0, "noise_mult": 0.5862405300140381, "noise_std_eff": 0.05862405300140381, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 264.77500796318054}
39
+ {"step": 21700, "epoch": 0, "loss": 4.209988117218018, "loss_main": 3.492840528488159, "loss_zero": 3.585738182067871, "loss_en": 3.4876983165740967, "loss_zh": 3.4979827404022217, "noise_ramp": 1.0, "noise_mult": 0.8077938556671143, "noise_std_eff": 0.08077938556671144, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 270.40780377388}
40
+ {"step": 21720, "epoch": 0, "loss": 3.75235652923584, "loss_main": 3.080103874206543, "loss_zero": 3.3612635135650635, "loss_en": 3.0877304077148438, "loss_zh": 3.072477102279663, "noise_ramp": 1.0, "noise_mult": 0.7638168334960938, "noise_std_eff": 0.07638168334960938, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 275.9852180480957}
41
+ {"step": 21740, "epoch": 0, "loss": 4.108526706695557, "loss_main": 3.408451557159424, "loss_zero": 3.5003762245178223, "loss_en": 3.4244630336761475, "loss_zh": 3.3924400806427, "noise_ramp": 1.0, "noise_mult": 0.8515471816062927, "noise_std_eff": 0.08515471816062928, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 281.58169984817505}
42
+ {"step": 21760, "epoch": 0, "loss": 4.237359523773193, "loss_main": 3.5161662101745605, "loss_zero": 3.6059675216674805, "loss_en": 3.5189549922943115, "loss_zh": 3.5133771896362305, "noise_ramp": 1.0, "noise_mult": 0.6781389117240906, "noise_std_eff": 0.06781389117240906, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 287.20856261253357}
43
+ {"step": 21780, "epoch": 0, "loss": 3.816683530807495, "loss_main": 3.1595873832702637, "loss_zero": 3.285480499267578, "loss_en": 3.1365997791290283, "loss_zh": 3.182575225830078, "noise_ramp": 1.0, "noise_mult": 0.9141291975975037, "noise_std_eff": 0.09141291975975037, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 292.7787697315216}
44
+ {"step": 21800, "epoch": 0, "loss": 3.19374942779541, "loss_main": 2.6300830841064453, "loss_zero": 2.8183321952819824, "loss_en": 2.643151044845581, "loss_zh": 2.6170148849487305, "noise_ramp": 1.0, "noise_mult": 0.864293098449707, "noise_std_eff": 0.08642930984497071, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 298.3992455005646}
45
+ {"step": 21820, "epoch": 0, "loss": 3.859436511993408, "loss_main": 3.1997761726379395, "loss_zero": 3.298302173614502, "loss_en": 3.204054117202759, "loss_zh": 3.195497989654541, "noise_ramp": 1.0, "noise_mult": 0.8344318270683289, "noise_std_eff": 0.08344318270683289, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 304.03787565231323}
46
+ {"step": 21840, "epoch": 0, "loss": 4.400498390197754, "loss_main": 3.661292552947998, "loss_zero": 3.6960296630859375, "loss_en": 3.6382176876068115, "loss_zh": 3.6843671798706055, "noise_ramp": 1.0, "noise_mult": 0.9956064820289612, "noise_std_eff": 0.09956064820289612, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 310.30675649642944}
47
+ {"step": 21860, "epoch": 0, "loss": 3.5219264030456543, "loss_main": 2.9081637859344482, "loss_zero": 3.0688130855560303, "loss_en": 2.8834166526794434, "loss_zh": 2.932910919189453, "noise_ramp": 1.0, "noise_mult": 0.587268590927124, "noise_std_eff": 0.058726859092712407, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 316.5679347515106}
48
+ {"step": 21880, "epoch": 0, "loss": 4.631665229797363, "loss_main": 3.8592820167541504, "loss_zero": 3.86191725730896, "loss_en": 3.8253397941589355, "loss_zh": 3.8932242393493652, "noise_ramp": 1.0, "noise_mult": 0.8881257176399231, "noise_std_eff": 0.08881257176399232, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 322.38869285583496}
49
+ {"step": 21900, "epoch": 0, "loss": 4.178965091705322, "loss_main": 3.4745872020721436, "loss_zero": 3.5218892097473145, "loss_en": 3.454746961593628, "loss_zh": 3.494427442550659, "noise_ramp": 1.0, "noise_mult": 0.8983040452003479, "noise_std_eff": 0.0898304045200348, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 327.9594693183899}
50
+ {"step": 21920, "epoch": 0, "loss": 3.714785099029541, "loss_main": 3.057072639465332, "loss_zero": 3.288562059402466, "loss_en": 3.057246208190918, "loss_zh": 3.056899070739746, "noise_ramp": 1.0, "noise_mult": 0.886663556098938, "noise_std_eff": 0.0886663556098938, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 333.5695466995239}
51
+ {"step": 21940, "epoch": 0, "loss": 4.336023807525635, "loss_main": 3.608537435531616, "loss_zero": 3.637432813644409, "loss_en": 3.6275975704193115, "loss_zh": 3.589477300643921, "noise_ramp": 1.0, "noise_mult": 0.9921159148216248, "noise_std_eff": 0.09921159148216248, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 339.1360695362091}
52
+ {"step": 21960, "epoch": 0, "loss": 3.883399486541748, "loss_main": 3.217268943786621, "loss_zero": 3.3306527137756348, "loss_en": 3.2382569313049316, "loss_zh": 3.1962811946868896, "noise_ramp": 1.0, "noise_mult": 0.909186601638794, "noise_std_eff": 0.0909186601638794, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 344.7195715904236}
53
+ {"step": 21980, "epoch": 0, "loss": 3.7262701988220215, "loss_main": 3.0723695755004883, "loss_zero": 3.269502878189087, "loss_en": 3.0463356971740723, "loss_zh": 3.098403215408325, "noise_ramp": 1.0, "noise_mult": 0.8542768359184265, "noise_std_eff": 0.08542768359184266, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 350.30107259750366}
54
+ {"step": 22000, "epoch": 0, "loss": 4.433840751647949, "loss_main": 3.6641173362731934, "loss_zero": 3.848616123199463, "loss_en": 3.6910228729248047, "loss_zh": 3.637212038040161, "noise_ramp": 1.0, "noise_mult": 0.903129518032074, "noise_std_eff": 0.0903129518032074, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 355.88102769851685}
55
+ {"step": 22020, "epoch": 0, "loss": 4.269336700439453, "loss_main": 3.5402684211730957, "loss_zero": 3.6453423500061035, "loss_en": 3.524881362915039, "loss_zh": 3.5556557178497314, "noise_ramp": 1.0, "noise_mult": 0.8642053604125977, "noise_std_eff": 0.08642053604125977, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 386.7105529308319}
56
+ {"step": 22040, "epoch": 0, "loss": 3.823829174041748, "loss_main": 3.1721673011779785, "loss_zero": 3.2583093643188477, "loss_en": 3.1773924827575684, "loss_zh": 3.1669423580169678, "noise_ramp": 1.0, "noise_mult": 0.7993564605712891, "noise_std_eff": 0.07993564605712891, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 392.3484106063843}
57
+ {"step": 22060, "epoch": 0, "loss": 4.083178520202637, "loss_main": 3.3631021976470947, "loss_zero": 3.6003825664520264, "loss_en": 3.389423131942749, "loss_zh": 3.3367812633514404, "noise_ramp": 1.0, "noise_mult": 0.9461566209793091, "noise_std_eff": 0.09461566209793092, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 397.95002460479736}
58
+ {"step": 22080, "epoch": 0, "loss": 4.066937446594238, "loss_main": 3.377056121826172, "loss_zero": 3.4494059085845947, "loss_en": 3.3767716884613037, "loss_zh": 3.377340793609619, "noise_ramp": 1.0, "noise_mult": 0.516035258769989, "noise_std_eff": 0.0516035258769989, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 403.5372591018677}
59
+ {"step": 22100, "epoch": 0, "loss": 3.60623836517334, "loss_main": 2.9802584648132324, "loss_zero": 3.129899024963379, "loss_en": 2.983050584793091, "loss_zh": 2.977466344833374, "noise_ramp": 1.0, "noise_mult": 0.7391538023948669, "noise_std_eff": 0.0739153802394867, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 409.1144688129425}
60
+ {"step": 22120, "epoch": 0, "loss": 4.626853942871094, "loss_main": 3.8489584922790527, "loss_zero": 3.8894782066345215, "loss_en": 3.8435490131378174, "loss_zh": 3.854367733001709, "noise_ramp": 1.0, "noise_mult": 0.9748877882957458, "noise_std_eff": 0.09748877882957459, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 414.735657453537}
61
+ {"step": 22140, "epoch": 0, "loss": 4.050447463989258, "loss_main": 3.3717548847198486, "loss_zero": 3.3934621810913086, "loss_en": 3.255591630935669, "loss_zh": 3.4879181385040283, "noise_ramp": 1.0, "noise_mult": 0.8207849264144897, "noise_std_eff": 0.08207849264144898, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 420.28568148612976}
62
+ {"step": 22160, "epoch": 0, "loss": 4.26715612411499, "loss_main": 3.5306124687194824, "loss_zero": 3.682718515396118, "loss_en": 3.528780698776245, "loss_zh": 3.5324440002441406, "noise_ramp": 1.0, "noise_mult": 0.491347074508667, "noise_std_eff": 0.049134707450866705, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 425.86256885528564}
63
+ {"step": 22180, "epoch": 0, "loss": 4.589682102203369, "loss_main": 3.819183826446533, "loss_zero": 3.8524906635284424, "loss_en": 3.8017830848693848, "loss_zh": 3.8365845680236816, "noise_ramp": 1.0, "noise_mult": 0.9930791258811951, "noise_std_eff": 0.09930791258811951, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 431.4286313056946}
64
+ {"step": 22200, "epoch": 0, "loss": 3.935476064682007, "loss_main": 3.275587797164917, "loss_zero": 3.2994415760040283, "loss_en": 3.2575149536132812, "loss_zh": 3.2936606407165527, "noise_ramp": 1.0, "noise_mult": 0.8359575271606445, "noise_std_eff": 0.08359575271606445, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 437.0657982826233}
65
+ {"step": 22220, "epoch": 0, "loss": 3.958986520767212, "loss_main": 3.291731357574463, "loss_zero": 3.336275339126587, "loss_en": 3.300945281982422, "loss_zh": 3.282517433166504, "noise_ramp": 1.0, "noise_mult": 0.8036591410636902, "noise_std_eff": 0.08036591410636902, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 442.62768149375916}
66
+ {"step": 22240, "epoch": 0, "loss": 4.367280006408691, "loss_main": 3.591757297515869, "loss_zero": 3.877614736557007, "loss_en": 3.6112067699432373, "loss_zh": 3.572307825088501, "noise_ramp": 1.0, "noise_mult": 0.8244106769561768, "noise_std_eff": 0.08244106769561768, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 448.26621413230896}
67
+ {"step": 22260, "epoch": 0, "loss": 4.143886566162109, "loss_main": 3.430746555328369, "loss_zero": 3.5657010078430176, "loss_en": 3.45770525932312, "loss_zh": 3.403787851333618, "noise_ramp": 1.0, "noise_mult": 0.9058099985122681, "noise_std_eff": 0.0905809998512268, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 453.8377013206482}
68
+ {"step": 22280, "epoch": 0, "loss": 4.121257781982422, "loss_main": 3.387969970703125, "loss_zero": 3.666438579559326, "loss_en": 3.3796660900115967, "loss_zh": 3.3962740898132324, "noise_ramp": 1.0, "noise_mult": 0.7713720202445984, "noise_std_eff": 0.07713720202445984, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 459.4698076248169}
69
+ {"step": 22300, "epoch": 0, "loss": 4.063778877258301, "loss_main": 3.374488592147827, "loss_zero": 3.4464521408081055, "loss_en": 3.3992574214935303, "loss_zh": 3.349719762802124, "noise_ramp": 1.0, "noise_mult": 0.9565710425376892, "noise_std_eff": 0.09565710425376893, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 465.0435948371887}
70
+ {"step": 22320, "epoch": 0, "loss": 4.304882049560547, "loss_main": 3.585696220397949, "loss_zero": 3.5959300994873047, "loss_en": 3.577251672744751, "loss_zh": 3.5941407680511475, "noise_ramp": 1.0, "noise_mult": 0.5000203251838684, "noise_std_eff": 0.05000203251838684, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 470.63325023651123}
71
+ {"step": 22340, "epoch": 0, "loss": 3.9895431995391846, "loss_main": 3.3188748359680176, "loss_zero": 3.353341579437256, "loss_en": 3.331500768661499, "loss_zh": 3.3062491416931152, "noise_ramp": 1.0, "noise_mult": 0.4292563199996948, "noise_std_eff": 0.04292563199996949, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 476.2089307308197}
72
+ {"step": 22360, "epoch": 0, "loss": 4.804549217224121, "loss_main": 3.982470989227295, "loss_zero": 4.110391616821289, "loss_en": 3.977248191833496, "loss_zh": 3.987694025039673, "noise_ramp": 1.0, "noise_mult": 0.9271335601806641, "noise_std_eff": 0.0927133560180664, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 481.81974172592163}
73
+ {"step": 22380, "epoch": 0, "loss": 3.7246832847595215, "loss_main": 3.091197967529297, "loss_zero": 3.167426109313965, "loss_en": 3.0686933994293213, "loss_zh": 3.1137027740478516, "noise_ramp": 1.0, "noise_mult": 0.88515305519104, "noise_std_eff": 0.08851530551910401, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 487.45599603652954}
74
+ {"step": 22400, "epoch": 0, "loss": 3.766183614730835, "loss_main": 3.1122186183929443, "loss_zero": 3.2698252201080322, "loss_en": 3.1150054931640625, "loss_zh": 3.109431743621826, "noise_ramp": 1.0, "noise_mult": 0.7363093495368958, "noise_std_eff": 0.07363093495368958, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 493.32425260543823}
75
+ {"step": 22420, "epoch": 0, "loss": 4.128057479858398, "loss_main": 3.42683744430542, "loss_zero": 3.506099224090576, "loss_en": 3.4451487064361572, "loss_zh": 3.4085259437561035, "noise_ramp": 1.0, "noise_mult": 0.9996026754379272, "noise_std_eff": 0.09996026754379272, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 524.3457591533661}
76
+ {"step": 22440, "epoch": 0, "loss": 3.5072250366210938, "loss_main": 2.9122886657714844, "loss_zero": 2.9746813774108887, "loss_en": 2.931311845779419, "loss_zh": 2.893265724182129, "noise_ramp": 1.0, "noise_mult": 0.7314436435699463, "noise_std_eff": 0.07314436435699463, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 529.8129503726959}
77
+ {"step": 22460, "epoch": 0, "loss": 4.396843433380127, "loss_main": 3.6398277282714844, "loss_zero": 3.785079002380371, "loss_en": 3.647017002105713, "loss_zh": 3.6326382160186768, "noise_ramp": 1.0, "noise_mult": 0.7915061712265015, "noise_std_eff": 0.07915061712265015, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 535.304041147232}
78
+ {"step": 22480, "epoch": 0, "loss": 3.617982864379883, "loss_main": 2.987143039703369, "loss_zero": 3.15419864654541, "loss_en": 2.9634735584259033, "loss_zh": 3.010812520980835, "noise_ramp": 1.0, "noise_mult": 0.9520891904830933, "noise_std_eff": 0.09520891904830933, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 540.796834230423}
79
+ {"step": 22500, "epoch": 0, "loss": 3.5703091621398926, "loss_main": 2.9484925270080566, "loss_zero": 3.1090831756591797, "loss_en": 2.973165512084961, "loss_zh": 2.9238195419311523, "noise_ramp": 1.0, "noise_mult": 0.7676740884780884, "noise_std_eff": 0.07676740884780885, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 546.250878572464}
80
+ {"step": 22520, "epoch": 0, "loss": 4.230541706085205, "loss_main": 3.4817230701446533, "loss_zero": 3.744093656539917, "loss_en": 3.473177909851074, "loss_zh": 3.4902682304382324, "noise_ramp": 1.0, "noise_mult": 0.9982424378395081, "noise_std_eff": 0.09982424378395081, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 551.7403955459595}
81
+ {"step": 22540, "epoch": 0, "loss": 4.521761417388916, "loss_main": 3.7672576904296875, "loss_zero": 3.7725179195404053, "loss_en": 3.794426202774048, "loss_zh": 3.740089178085327, "noise_ramp": 1.0, "noise_mult": 0.5472862124443054, "noise_std_eff": 0.05472862124443054, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 557.280255317688}
82
+ {"step": 22560, "epoch": 0, "loss": 4.386241436004639, "loss_main": 3.6316096782684326, "loss_zero": 3.7731595039367676, "loss_en": 3.628565549850464, "loss_zh": 3.6346538066864014, "noise_ramp": 1.0, "noise_mult": 0.9875380396842957, "noise_std_eff": 0.09875380396842957, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 562.8268053531647}
83
+ {"step": 22580, "epoch": 0, "loss": 4.867133140563965, "loss_main": 4.034424304962158, "loss_zero": 4.163544178009033, "loss_en": 4.023184299468994, "loss_zh": 4.045664310455322, "noise_ramp": 1.0, "noise_mult": 0.44444310665130615, "noise_std_eff": 0.04444431066513062, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 568.4329149723053}
84
+ {"step": 22600, "epoch": 0, "loss": 4.108560562133789, "loss_main": 3.4134628772735596, "loss_zero": 3.4754886627197266, "loss_en": 3.4177536964416504, "loss_zh": 3.4091720581054688, "noise_ramp": 1.0, "noise_mult": 0.9213575124740601, "noise_std_eff": 0.09213575124740601, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 573.9603877067566}
85
+ {"step": 22620, "epoch": 0, "loss": 4.2754130363464355, "loss_main": 3.5398387908935547, "loss_zero": 3.6778721809387207, "loss_en": 3.5444939136505127, "loss_zh": 3.5351836681365967, "noise_ramp": 1.0, "noise_mult": 0.8259735703468323, "noise_std_eff": 0.08259735703468324, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 579.4632952213287}
86
+ {"step": 22640, "epoch": 0, "loss": 4.090736389160156, "loss_main": 3.404503345489502, "loss_zero": 3.431165933609009, "loss_en": 3.387485980987549, "loss_zh": 3.421520948410034, "noise_ramp": 1.0, "noise_mult": 0.7312448024749756, "noise_std_eff": 0.07312448024749756, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 585.0363347530365}
87
+ {"step": 22660, "epoch": 0, "loss": 4.1763434410095215, "loss_main": 3.4582061767578125, "loss_zero": 3.590686082839966, "loss_en": 3.462019205093384, "loss_zh": 3.4543933868408203, "noise_ramp": 1.0, "noise_mult": 0.899927020072937, "noise_std_eff": 0.0899927020072937, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 590.556479215622}
88
+ {"step": 22680, "epoch": 0, "loss": 3.387646198272705, "loss_main": 2.8006656169891357, "loss_zero": 2.934903144836426, "loss_en": 2.8071322441101074, "loss_zh": 2.794198989868164, "noise_ramp": 1.0, "noise_mult": 0.8097662329673767, "noise_std_eff": 0.08097662329673767, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 596.0707671642303}
89
+ {"step": 22700, "epoch": 0, "loss": 4.3167243003845215, "loss_main": 3.5757665634155273, "loss_zero": 3.7047884464263916, "loss_en": 3.5802738666534424, "loss_zh": 3.5712594985961914, "noise_ramp": 1.0, "noise_mult": 0.9814367294311523, "noise_std_eff": 0.09814367294311524, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 601.5583157539368}
90
+ {"step": 22720, "epoch": 0, "loss": 4.723088264465332, "loss_main": 3.9106247425079346, "loss_zero": 4.062316417694092, "loss_en": 3.8997085094451904, "loss_zh": 3.9215409755706787, "noise_ramp": 1.0, "noise_mult": 0.7763971090316772, "noise_std_eff": 0.07763971090316774, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 607.0941314697266}
91
+ {"step": 22740, "epoch": 0, "loss": 4.212764739990234, "loss_main": 3.4920012950897217, "loss_zero": 3.60381817817688, "loss_en": 3.499056816101074, "loss_zh": 3.484945774078369, "noise_ramp": 1.0, "noise_mult": 0.5956957340240479, "noise_std_eff": 0.05956957340240479, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 612.6412315368652}
92
+ {"step": 22760, "epoch": 0, "loss": 3.916625738143921, "loss_main": 3.255396842956543, "loss_zero": 3.3061439990997314, "loss_en": 3.2607240676879883, "loss_zh": 3.2500698566436768, "noise_ramp": 1.0, "noise_mult": 0.9161072373390198, "noise_std_eff": 0.09161072373390199, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 618.2652742862701}
93
+ {"step": 22780, "epoch": 0, "loss": 4.288612365722656, "loss_main": 3.596343755722046, "loss_zero": 3.46134352684021, "loss_en": 3.5884475708007812, "loss_zh": 3.6042399406433105, "noise_ramp": 1.0, "noise_mult": 0.8041544556617737, "noise_std_eff": 0.08041544556617737, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 623.8436870574951}
94
+ {"step": 22800, "epoch": 0, "loss": 4.394313812255859, "loss_main": 3.640679359436035, "loss_zero": 3.768171548843384, "loss_en": 3.59985089302063, "loss_zh": 3.6815078258514404, "noise_ramp": 1.0, "noise_mult": 0.42410850524902344, "noise_std_eff": 0.042410850524902344, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 629.391973733902}
95
+ {"step": 22820, "epoch": 0, "loss": 4.27274227142334, "loss_main": 3.558807849884033, "loss_zero": 3.569671392440796, "loss_en": 3.55773663520813, "loss_zh": 3.5598788261413574, "noise_ramp": 1.0, "noise_mult": 0.9188005328178406, "noise_std_eff": 0.09188005328178406, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 659.6773760318756}
96
+ {"step": 22840, "epoch": 0, "loss": 4.170602798461914, "loss_main": 3.42691707611084, "loss_zero": 3.718427896499634, "loss_en": 3.4326796531677246, "loss_zh": 3.421154260635376, "noise_ramp": 1.0, "noise_mult": 0.42128482460975647, "noise_std_eff": 0.04212848246097565, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 665.3858525753021}
97
+ {"step": 22860, "epoch": 0, "loss": 4.985819339752197, "loss_main": 4.18576717376709, "loss_zero": 4.0002617835998535, "loss_en": 4.165313720703125, "loss_zh": 4.206220626831055, "noise_ramp": 1.0, "noise_mult": 0.7677230834960938, "noise_std_eff": 0.07677230834960938, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 670.8706324100494}
98
+ {"step": 22880, "epoch": 0, "loss": 3.6251277923583984, "loss_main": 2.990957736968994, "loss_zero": 3.1708502769470215, "loss_en": 2.992722988128662, "loss_zh": 2.989192247390747, "noise_ramp": 1.0, "noise_mult": 0.7845321893692017, "noise_std_eff": 0.07845321893692017, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 676.3769750595093}
99
+ {"step": 22900, "epoch": 0, "loss": 3.8676207065582275, "loss_main": 3.20640230178833, "loss_zero": 3.3060922622680664, "loss_en": 3.225585460662842, "loss_zh": 3.1872193813323975, "noise_ramp": 1.0, "noise_mult": 0.7355393171310425, "noise_std_eff": 0.07355393171310425, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 681.9208271503448}
100
+ {"step": 22920, "epoch": 0, "loss": 4.1099677085876465, "loss_main": 3.420131206512451, "loss_zero": 3.4491817951202393, "loss_en": 3.398676633834839, "loss_zh": 3.4415857791900635, "noise_ramp": 1.0, "noise_mult": 0.8129807114601135, "noise_std_eff": 0.08129807114601136, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 687.4470067024231}
101
+ {"step": 22940, "epoch": 0, "loss": 3.738086223602295, "loss_main": 3.086287498474121, "loss_zero": 3.258993625640869, "loss_en": 3.059798002243042, "loss_zh": 3.1127772331237793, "noise_ramp": 1.0, "noise_mult": 0.737176775932312, "noise_std_eff": 0.07371767759323121, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 693.036238193512}
102
+ {"step": 22960, "epoch": 0, "loss": 4.572447776794434, "loss_main": 3.800600290298462, "loss_zero": 3.859236478805542, "loss_en": 3.7843081951141357, "loss_zh": 3.816892385482788, "noise_ramp": 1.0, "noise_mult": 0.6547698378562927, "noise_std_eff": 0.06547698378562927, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 698.6875736713409}
103
+ {"step": 22980, "epoch": 0, "loss": 3.981555700302124, "loss_main": 3.3172671794891357, "loss_zero": 3.321442127227783, "loss_en": 3.27181339263916, "loss_zh": 3.3627209663391113, "noise_ramp": 1.0, "noise_mult": 0.9904401302337646, "noise_std_eff": 0.09904401302337647, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 704.1949014663696}
104
+ {"step": 23000, "epoch": 0, "loss": 3.8321950435638428, "loss_main": 3.1925411224365234, "loss_zero": 3.1982693672180176, "loss_en": 3.1849920749664307, "loss_zh": 3.200090169906616, "noise_ramp": 1.0, "noise_mult": 0.7176426649093628, "noise_std_eff": 0.07176426649093628, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 709.8429670333862}
105
+ {"step": 23020, "epoch": 0, "loss": 4.6449103355407715, "loss_main": 3.8675119876861572, "loss_zero": 3.8869917392730713, "loss_en": 3.91375732421875, "loss_zh": 3.8212666511535645, "noise_ramp": 1.0, "noise_mult": 0.6732965111732483, "noise_std_eff": 0.06732965111732483, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 715.416906118393}
106
+ {"step": 23040, "epoch": 0, "loss": 3.5195624828338623, "loss_main": 2.940279483795166, "loss_zero": 2.8964149951934814, "loss_en": 2.9146831035614014, "loss_zh": 2.9658756256103516, "noise_ramp": 1.0, "noise_mult": 0.8997371792793274, "noise_std_eff": 0.08997371792793274, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 721.0120406150818}
107
+ {"step": 23060, "epoch": 0, "loss": 3.6491289138793945, "loss_main": 3.0264244079589844, "loss_zero": 3.1135220527648926, "loss_en": 3.029263973236084, "loss_zh": 3.023585081100464, "noise_ramp": 1.0, "noise_mult": 0.6014958620071411, "noise_std_eff": 0.06014958620071412, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 726.5498878955841}
108
+ {"step": 23080, "epoch": 0, "loss": 4.135104179382324, "loss_main": 3.4549965858459473, "loss_zero": 3.400538682937622, "loss_en": 3.507201910018921, "loss_zh": 3.4027910232543945, "noise_ramp": 1.0, "noise_mult": 0.8495233058929443, "noise_std_eff": 0.08495233058929444, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 732.129314661026}
109
+ {"step": 23100, "epoch": 0, "loss": 3.738386631011963, "loss_main": 3.0992066860198975, "loss_zero": 3.19589900970459, "loss_en": 3.106877326965332, "loss_zh": 3.091536045074463, "noise_ramp": 1.0, "noise_mult": 0.8955716490745544, "noise_std_eff": 0.08955716490745545, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 737.6833970546722}
110
+ {"step": 23120, "epoch": 0, "loss": 4.112890243530273, "loss_main": 3.416771411895752, "loss_zero": 3.480592966079712, "loss_en": 3.393714427947998, "loss_zh": 3.439828634262085, "noise_ramp": 1.0, "noise_mult": 0.9344645738601685, "noise_std_eff": 0.09344645738601685, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 743.3410949707031}
111
+ {"step": 23140, "epoch": 0, "loss": 4.243632793426514, "loss_main": 3.5339999198913574, "loss_zero": 3.5481643676757812, "loss_en": 3.545022487640381, "loss_zh": 3.522977352142334, "noise_ramp": 1.0, "noise_mult": 0.7159476280212402, "noise_std_eff": 0.07159476280212403, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 748.9217138290405}
112
+ {"step": 23160, "epoch": 0, "loss": 3.9787139892578125, "loss_main": 3.2753067016601562, "loss_zero": 3.517036199569702, "loss_en": 3.286362886428833, "loss_zh": 3.2642502784729004, "noise_ramp": 1.0, "noise_mult": 0.9592054486274719, "noise_std_eff": 0.0959205448627472, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 754.4737238883972}
113
+ {"step": 23180, "epoch": 0, "loss": 4.292906761169434, "loss_main": 3.5698885917663574, "loss_zero": 3.615091323852539, "loss_en": 3.603071928024292, "loss_zh": 3.536705255508423, "noise_ramp": 1.0, "noise_mult": 0.5545864105224609, "noise_std_eff": 0.0554586410522461, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 759.9848651885986}
114
+ {"step": 23200, "epoch": 0, "loss": 4.026883125305176, "loss_main": 3.32263445854187, "loss_zero": 3.5212442874908447, "loss_en": 3.3198046684265137, "loss_zh": 3.3254642486572266, "noise_ramp": 1.0, "noise_mult": 0.7162210941314697, "noise_std_eff": 0.07162210941314698, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 765.5552308559418}
115
+ {"step": 23220, "epoch": 0, "loss": 3.8037710189819336, "loss_main": 3.1561837196350098, "loss_zero": 3.237935781478882, "loss_en": 3.154158115386963, "loss_zh": 3.1582093238830566, "noise_ramp": 1.0, "noise_mult": 0.5116897821426392, "noise_std_eff": 0.051168978214263916, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 796.004478931427}
116
+ {"step": 23240, "epoch": 0, "loss": 4.083685874938965, "loss_main": 3.3886876106262207, "loss_zero": 3.4749910831451416, "loss_en": 3.382032632827759, "loss_zh": 3.3953423500061035, "noise_ramp": 1.0, "noise_mult": 0.9241498708724976, "noise_std_eff": 0.09241498708724977, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 801.5781590938568}
117
+ {"step": 23260, "epoch": 0, "loss": 4.584848403930664, "loss_main": 3.82515287399292, "loss_zero": 3.7984771728515625, "loss_en": 3.818854331970215, "loss_zh": 3.831451177597046, "noise_ramp": 1.0, "noise_mult": 0.8515443801879883, "noise_std_eff": 0.08515443801879884, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 807.1739993095398}
118
+ {"step": 23280, "epoch": 0, "loss": 3.6245672702789307, "loss_main": 3.017584800720215, "loss_zero": 3.034912586212158, "loss_en": 2.9855196475982666, "loss_zh": 3.049650192260742, "noise_ramp": 1.0, "noise_mult": 0.9998478889465332, "noise_std_eff": 0.09998478889465333, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 812.7996118068695}
119
+ {"step": 23300, "epoch": 0, "loss": 3.6708309650421143, "loss_main": 3.040397882461548, "loss_zero": 3.152165412902832, "loss_en": 3.0299339294433594, "loss_zh": 3.0508618354797363, "noise_ramp": 1.0, "noise_mult": 0.8650702834129333, "noise_std_eff": 0.08650702834129334, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 818.3740584850311}
120
+ {"step": 23320, "epoch": 0, "loss": 3.8963069915771484, "loss_main": 3.2483856678009033, "loss_zero": 3.2396063804626465, "loss_en": 3.248464584350586, "loss_zh": 3.2483067512512207, "noise_ramp": 1.0, "noise_mult": 0.539176344871521, "noise_std_eff": 0.0539176344871521, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 823.9524335861206}
121
+ {"step": 23340, "epoch": 0, "loss": 4.156014919281006, "loss_main": 3.4487428665161133, "loss_zero": 3.5363595485687256, "loss_en": 3.429173469543457, "loss_zh": 3.4683122634887695, "noise_ramp": 1.0, "noise_mult": 0.9854084849357605, "noise_std_eff": 0.09854084849357606, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 829.5151264667511}
122
+ {"step": 23360, "epoch": 0, "loss": 3.8787858486175537, "loss_main": 3.2200942039489746, "loss_zero": 3.2934579849243164, "loss_en": 3.205125331878662, "loss_zh": 3.235063076019287, "noise_ramp": 1.0, "noise_mult": 0.7145160436630249, "noise_std_eff": 0.07145160436630249, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 835.0958623886108}
123
+ {"step": 23380, "epoch": 0, "loss": 3.6792287826538086, "loss_main": 3.029315948486328, "loss_zero": 3.2495644092559814, "loss_en": 3.0245864391326904, "loss_zh": 3.034045696258545, "noise_ramp": 1.0, "noise_mult": 0.7330989241600037, "noise_std_eff": 0.07330989241600037, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 840.6516518592834}
124
+ {"step": 23400, "epoch": 0, "loss": 3.87431263923645, "loss_main": 3.193558931350708, "loss_zero": 3.403768539428711, "loss_en": 3.1832566261291504, "loss_zh": 3.2038612365722656, "noise_ramp": 1.0, "noise_mult": 0.7442008256912231, "noise_std_eff": 0.07442008256912232, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 846.2398924827576}
125
+ {"step": 23420, "epoch": 0, "loss": 3.759495735168457, "loss_main": 3.1384525299072266, "loss_zero": 3.1052165031433105, "loss_en": 3.136467933654785, "loss_zh": 3.140436887741089, "noise_ramp": 1.0, "noise_mult": 0.5965988636016846, "noise_std_eff": 0.05965988636016846, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 851.7595701217651}
126
+ {"step": 23440, "epoch": 0, "loss": 3.679863452911377, "loss_main": 3.0276894569396973, "loss_zero": 3.2608695030212402, "loss_en": 3.0071604251861572, "loss_zh": 3.0482184886932373, "noise_ramp": 1.0, "noise_mult": 0.9527906179428101, "noise_std_eff": 0.09527906179428101, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 857.3125085830688}
127
+ {"step": 23460, "epoch": 0, "loss": 3.4049274921417236, "loss_main": 2.8261752128601074, "loss_zero": 2.893761396408081, "loss_en": 2.8172478675842285, "loss_zh": 2.8351023197174072, "noise_ramp": 1.0, "noise_mult": 0.8793360590934753, "noise_std_eff": 0.08793360590934754, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 862.8615922927856}
128
+ {"step": 23480, "epoch": 0, "loss": 4.165307998657227, "loss_main": 3.454756498336792, "loss_zero": 3.5527570247650146, "loss_en": 3.434828519821167, "loss_zh": 3.474684476852417, "noise_ramp": 1.0, "noise_mult": 0.7814388871192932, "noise_std_eff": 0.07814388871192933, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 868.5960423946381}
129
+ {"step": 23500, "epoch": 0, "loss": 4.047581195831299, "loss_main": 3.360588550567627, "loss_zero": 3.434962272644043, "loss_en": 3.3479084968566895, "loss_zh": 3.3732688426971436, "noise_ramp": 1.0, "noise_mult": 0.8311663269996643, "noise_std_eff": 0.08311663269996644, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 874.1675932407379}
130
+ {"step": 23520, "epoch": 0, "loss": 3.4436192512512207, "loss_main": 2.856018543243408, "loss_zero": 2.9380035400390625, "loss_en": 2.8203065395355225, "loss_zh": 2.891730785369873, "noise_ramp": 1.0, "noise_mult": 0.769733190536499, "noise_std_eff": 0.0769733190536499, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 879.7738280296326}
131
+ {"step": 23540, "epoch": 0, "loss": 4.197542190551758, "loss_main": 3.4687561988830566, "loss_zero": 3.643930673599243, "loss_en": 3.4997737407684326, "loss_zh": 3.4377384185791016, "noise_ramp": 1.0, "noise_mult": 0.9852657318115234, "noise_std_eff": 0.09852657318115235, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 885.344114780426}
132
+ {"step": 23560, "epoch": 0, "loss": 4.0753679275512695, "loss_main": 3.3767080307006836, "loss_zero": 3.4932987689971924, "loss_en": 3.348318338394165, "loss_zh": 3.4050979614257812, "noise_ramp": 1.0, "noise_mult": 0.9814148545265198, "noise_std_eff": 0.09814148545265199, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 891.0951459407806}
133
+ {"step": 23580, "epoch": 0, "loss": 4.741475582122803, "loss_main": 3.9508166313171387, "loss_zero": 3.953293800354004, "loss_en": 3.9413723945617676, "loss_zh": 3.960261106491089, "noise_ramp": 1.0, "noise_mult": 0.4066679775714874, "noise_std_eff": 0.040666797757148744, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 896.6784579753876}
134
+ {"step": 23600, "epoch": 0, "loss": 4.330370903015137, "loss_main": 3.6197891235351562, "loss_zero": 3.5529098510742188, "loss_en": 3.6321792602539062, "loss_zh": 3.6073989868164062, "noise_ramp": 1.0, "noise_mult": 0.47746628522872925, "noise_std_eff": 0.047746628522872925, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 902.2448675632477}
135
+ {"step": 23620, "epoch": 0, "loss": 3.830367088317871, "loss_main": 3.169666290283203, "loss_zero": 3.30350399017334, "loss_en": 3.150655508041382, "loss_zh": 3.1886768341064453, "noise_ramp": 1.0, "noise_mult": 0.5108504295349121, "noise_std_eff": 0.051085042953491214, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 934.5106437206268}
136
+ {"step": 23640, "epoch": 0, "loss": 3.578946352005005, "loss_main": 2.957125186920166, "loss_zero": 3.1091060638427734, "loss_en": 2.9469406604766846, "loss_zh": 2.9673099517822266, "noise_ramp": 1.0, "noise_mult": 0.6335480809211731, "noise_std_eff": 0.06335480809211731, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 940.3516838550568}
137
+ {"step": 23660, "epoch": 0, "loss": 4.105408191680908, "loss_main": 3.4237887859344482, "loss_zero": 3.4080963134765625, "loss_en": 3.4167139530181885, "loss_zh": 3.430863618850708, "noise_ramp": 1.0, "noise_mult": 0.985043466091156, "noise_std_eff": 0.09850434660911561, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 946.458349943161}
138
+ {"step": 23680, "epoch": 0, "loss": 4.500675678253174, "loss_main": 3.7298855781555176, "loss_zero": 3.853950023651123, "loss_en": 3.7253496646881104, "loss_zh": 3.734421730041504, "noise_ramp": 1.0, "noise_mult": 0.5602104663848877, "noise_std_eff": 0.056021046638488774, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 952.9681594371796}
139
+ {"step": 23700, "epoch": 0, "loss": 4.467766284942627, "loss_main": 3.7044215202331543, "loss_zero": 3.8167240619659424, "loss_en": 3.7257463932037354, "loss_zh": 3.6830966472625732, "noise_ramp": 1.0, "noise_mult": 0.6183620095252991, "noise_std_eff": 0.06183620095252991, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 958.8115878105164}
140
+ {"step": 23720, "epoch": 0, "loss": 3.4402379989624023, "loss_main": 2.8519339561462402, "loss_zero": 2.9415199756622314, "loss_en": 2.865647554397583, "loss_zh": 2.8382205963134766, "noise_ramp": 1.0, "noise_mult": 0.9199289083480835, "noise_std_eff": 0.09199289083480836, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 964.6595456600189}
141
+ {"step": 23740, "epoch": 0, "loss": 4.180638790130615, "loss_main": 3.4769787788391113, "loss_zero": 3.5182995796203613, "loss_en": 3.4424517154693604, "loss_zh": 3.5115060806274414, "noise_ramp": 1.0, "noise_mult": 0.5165428519248962, "noise_std_eff": 0.05165428519248963, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 970.9361588954926}
142
+ {"step": 23760, "epoch": 0, "loss": 3.5348010063171387, "loss_main": 2.9221549034118652, "loss_zero": 3.063230276107788, "loss_en": 2.9044997692108154, "loss_zh": 2.939809799194336, "noise_ramp": 1.0, "noise_mult": 0.5285784006118774, "noise_std_eff": 0.052857840061187746, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 977.2504413127899}
143
+ {"step": 23780, "epoch": 0, "loss": 4.0819196701049805, "loss_main": 3.384942054748535, "loss_zero": 3.4848883152008057, "loss_en": 3.3805007934570312, "loss_zh": 3.389383554458618, "noise_ramp": 1.0, "noise_mult": 0.9989249110221863, "noise_std_eff": 0.09989249110221864, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 983.5203897953033}
144
+ {"step": 23800, "epoch": 0, "loss": 3.7684972286224365, "loss_main": 3.1450352668762207, "loss_zero": 3.1173095703125, "loss_en": 3.124455451965332, "loss_zh": 3.1656150817871094, "noise_ramp": 1.0, "noise_mult": 0.7263202667236328, "noise_std_eff": 0.07263202667236328, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 989.2871615886688}
145
+ {"step": 23820, "epoch": 0, "loss": 3.7118637561798096, "loss_main": 3.0899648666381836, "loss_zero": 3.109494209289551, "loss_en": 3.057480573654175, "loss_zh": 3.1224489212036133, "noise_ramp": 1.0, "noise_mult": 0.9338898658752441, "noise_std_eff": 0.09338898658752443, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 994.8263993263245}
146
+ {"step": 23840, "epoch": 0, "loss": 3.8563392162323, "loss_main": 3.220154285430908, "loss_zero": 3.1809241771698, "loss_en": 3.2132785320281982, "loss_zh": 3.227030038833618, "noise_ramp": 1.0, "noise_mult": 0.8520193696022034, "noise_std_eff": 0.08520193696022034, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1000.3769311904907}
147
+ {"step": 23860, "epoch": 0, "loss": 3.9809494018554688, "loss_main": 3.308380365371704, "loss_zero": 3.3628451824188232, "loss_en": 3.3086071014404297, "loss_zh": 3.3081536293029785, "noise_ramp": 1.0, "noise_mult": 0.49157798290252686, "noise_std_eff": 0.049157798290252686, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1005.9605793952942}
148
+ {"step": 23880, "epoch": 0, "loss": 4.297205924987793, "loss_main": 3.5784175395965576, "loss_zero": 3.593942880630493, "loss_en": 3.542933225631714, "loss_zh": 3.6139018535614014, "noise_ramp": 1.0, "noise_mult": 0.6004601716995239, "noise_std_eff": 0.06004601716995239, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1011.508350610733}
149
+ {"step": 23900, "epoch": 0, "loss": 4.175784111022949, "loss_main": 3.491522789001465, "loss_zero": 3.4213054180145264, "loss_en": 3.487003803253174, "loss_zh": 3.496042013168335, "noise_ramp": 1.0, "noise_mult": 0.5726016163825989, "noise_std_eff": 0.05726016163825989, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1017.045996427536}
150
+ {"step": 23920, "epoch": 0, "loss": 3.642826557159424, "loss_main": 3.0115957260131836, "loss_zero": 3.1561546325683594, "loss_en": 3.0218863487243652, "loss_zh": 3.001304864883423, "noise_ramp": 1.0, "noise_mult": 0.5807821750640869, "noise_std_eff": 0.058078217506408694, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1022.5654828548431}
151
+ {"step": 23940, "epoch": 0, "loss": 4.031103610992432, "loss_main": 3.368928909301758, "loss_zero": 3.3108744621276855, "loss_en": 3.381895065307617, "loss_zh": 3.3559629917144775, "noise_ramp": 1.0, "noise_mult": 0.8156693577766418, "noise_std_eff": 0.08156693577766419, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1028.1400504112244}
152
+ {"step": 23960, "epoch": 0, "loss": 4.179867267608643, "loss_main": 3.466439723968506, "loss_zero": 3.567138671875, "loss_en": 3.4649457931518555, "loss_zh": 3.4679338932037354, "noise_ramp": 1.0, "noise_mult": 0.5717296600341797, "noise_std_eff": 0.05717296600341797, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1033.9849905967712}
153
+ {"step": 23980, "epoch": 0, "loss": 3.8917837142944336, "loss_main": 3.234902858734131, "loss_zero": 3.2844038009643555, "loss_en": 3.236175775527954, "loss_zh": 3.2336299419403076, "noise_ramp": 1.0, "noise_mult": 0.9729774594306946, "noise_std_eff": 0.09729774594306946, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1039.684490442276}
154
+ {"step": 24000, "epoch": 0, "loss": 4.041860103607178, "loss_main": 3.3489668369293213, "loss_zero": 3.464465379714966, "loss_en": 3.32291841506958, "loss_zh": 3.3750152587890625, "noise_ramp": 1.0, "noise_mult": 0.9124302864074707, "noise_std_eff": 0.09124302864074707, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1045.2195675373077}
155
+ {"step": 24020, "epoch": 0, "loss": 3.7661337852478027, "loss_main": 3.101564407348633, "loss_zero": 3.3228466510772705, "loss_en": 3.0966262817382812, "loss_zh": 3.1065025329589844, "noise_ramp": 1.0, "noise_mult": 0.982207179069519, "noise_std_eff": 0.09822071790695192, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1075.8204703330994}
156
+ {"step": 24040, "epoch": 0, "loss": 3.806785821914673, "loss_main": 3.1500444412231445, "loss_zero": 3.2837069034576416, "loss_en": 3.144474983215332, "loss_zh": 3.155613899230957, "noise_ramp": 1.0, "noise_mult": 0.6426711082458496, "noise_std_eff": 0.06426711082458496, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1081.364192724228}
157
+ {"step": 24060, "epoch": 0, "loss": 4.154338836669922, "loss_main": 3.432291030883789, "loss_zero": 3.6102383136749268, "loss_en": 3.429960012435913, "loss_zh": 3.434621810913086, "noise_ramp": 1.0, "noise_mult": 0.8132129907608032, "noise_std_eff": 0.08132129907608032, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1086.9312183856964}
158
+ {"step": 24080, "epoch": 0, "loss": 3.982252836227417, "loss_main": 3.3044114112854004, "loss_zero": 3.389207363128662, "loss_en": 3.257981777191162, "loss_zh": 3.3508410453796387, "noise_ramp": 1.0, "noise_mult": 0.749858021736145, "noise_std_eff": 0.0749858021736145, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1092.537415266037}
159
+ {"step": 24100, "epoch": 0, "loss": 4.120564937591553, "loss_main": 3.4229846000671387, "loss_zero": 3.4879019260406494, "loss_en": 3.4468119144439697, "loss_zh": 3.3991575241088867, "noise_ramp": 1.0, "noise_mult": 0.9471413493156433, "noise_std_eff": 0.09471413493156433, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1098.2809550762177}
160
+ {"step": 24120, "epoch": 0, "loss": 3.5673232078552246, "loss_main": 2.9402389526367188, "loss_zero": 3.13542103767395, "loss_en": 2.9705007076263428, "loss_zh": 2.9099771976470947, "noise_ramp": 1.0, "noise_mult": 0.9180266857147217, "noise_std_eff": 0.09180266857147218, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1103.8425176143646}
161
+ {"step": 24140, "epoch": 0, "loss": 4.062670707702637, "loss_main": 3.3780250549316406, "loss_zero": 3.4232280254364014, "loss_en": 3.410395622253418, "loss_zh": 3.3456544876098633, "noise_ramp": 1.0, "noise_mult": 0.8556216359138489, "noise_std_eff": 0.0855621635913849, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1109.379933834076}
162
+ {"step": 24160, "epoch": 0, "loss": 3.402831554412842, "loss_main": 2.8053579330444336, "loss_zero": 2.98736834526062, "loss_en": 2.8060920238494873, "loss_zh": 2.80462384223938, "noise_ramp": 1.0, "noise_mult": 0.7174956798553467, "noise_std_eff": 0.07174956798553467, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1114.9175758361816}
163
+ {"step": 24180, "epoch": 0, "loss": 3.3939335346221924, "loss_main": 2.808375120162964, "loss_zero": 2.9277918338775635, "loss_en": 2.8073179721832275, "loss_zh": 2.8094322681427, "noise_ramp": 1.0, "noise_mult": 0.8159123659133911, "noise_std_eff": 0.08159123659133911, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1120.455020904541}
164
+ {"step": 24200, "epoch": 0, "loss": 4.086786270141602, "loss_main": 3.387913703918457, "loss_zero": 3.4943625926971436, "loss_en": 3.379345417022705, "loss_zh": 3.396481990814209, "noise_ramp": 1.0, "noise_mult": 0.9594902992248535, "noise_std_eff": 0.09594902992248536, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1126.0511026382446}
165
+ {"step": 24220, "epoch": 0, "loss": 4.5412373542785645, "loss_main": 3.743797540664673, "loss_zero": 3.987199306488037, "loss_en": 3.741873264312744, "loss_zh": 3.7457218170166016, "noise_ramp": 1.0, "noise_mult": 0.7390864491462708, "noise_std_eff": 0.07390864491462708, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1131.6233723163605}
166
+ {"step": 24240, "epoch": 0, "loss": 4.027472019195557, "loss_main": 3.351304769515991, "loss_zero": 3.380835771560669, "loss_en": 3.34415864944458, "loss_zh": 3.3584508895874023, "noise_ramp": 1.0, "noise_mult": 0.660578727722168, "noise_std_eff": 0.0660578727722168, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1137.2672715187073}
167
+ {"step": 24260, "epoch": 0, "loss": 3.460015296936035, "loss_main": 2.849118947982788, "loss_zero": 3.0544819831848145, "loss_en": 2.864635705947876, "loss_zh": 2.8336021900177, "noise_ramp": 1.0, "noise_mult": 0.7256484627723694, "noise_std_eff": 0.07256484627723694, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1142.825784444809}
168
+ {"step": 24280, "epoch": 0, "loss": 4.11743688583374, "loss_main": 3.429582118988037, "loss_zero": 3.439274549484253, "loss_en": 3.442898750305176, "loss_zh": 3.4162657260894775, "noise_ramp": 1.0, "noise_mult": 0.9543102979660034, "noise_std_eff": 0.09543102979660034, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1148.4019219875336}
169
+ {"step": 24300, "epoch": 0, "loss": 4.188497543334961, "loss_main": 3.4612088203430176, "loss_zero": 3.6364426612854004, "loss_en": 3.472426176071167, "loss_zh": 3.4499917030334473, "noise_ramp": 1.0, "noise_mult": 0.727558434009552, "noise_std_eff": 0.0727558434009552, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1153.9282882213593}
170
+ {"step": 24320, "epoch": 0, "loss": 3.9343888759613037, "loss_main": 3.279693126678467, "loss_zero": 3.2734785079956055, "loss_en": 3.284712553024292, "loss_zh": 3.2746737003326416, "noise_ramp": 1.0, "noise_mult": 0.7017517685890198, "noise_std_eff": 0.07017517685890198, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1159.4646079540253}
171
+ {"step": 24340, "epoch": 0, "loss": 3.765084981918335, "loss_main": 3.125430107116699, "loss_zero": 3.198274612426758, "loss_en": 3.122169017791748, "loss_zh": 3.1286914348602295, "noise_ramp": 1.0, "noise_mult": 0.7187355756759644, "noise_std_eff": 0.07187355756759643, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1164.9863488674164}
172
+ {"step": 24360, "epoch": 0, "loss": 4.0015549659729, "loss_main": 3.329002618789673, "loss_zero": 3.3627617359161377, "loss_en": 3.323275566101074, "loss_zh": 3.3347296714782715, "noise_ramp": 1.0, "noise_mult": 0.6737536787986755, "noise_std_eff": 0.06737536787986756, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1170.5476851463318}
173
+ {"step": 24380, "epoch": 0, "loss": 3.8156087398529053, "loss_main": 3.1706933975219727, "loss_zero": 3.224576234817505, "loss_en": 3.1856746673583984, "loss_zh": 3.1557118892669678, "noise_ramp": 1.0, "noise_mult": 0.8022685647010803, "noise_std_eff": 0.08022685647010804, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1176.0596210956573}
174
+ {"step": 24400, "epoch": 0, "loss": 4.303892135620117, "loss_main": 3.5407347679138184, "loss_zero": 3.8157870769500732, "loss_en": 3.558468818664551, "loss_zh": 3.523000717163086, "noise_ramp": 1.0, "noise_mult": 0.4394839406013489, "noise_std_eff": 0.04394839406013489, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1181.6304018497467}
175
+ {"step": 24420, "epoch": 0, "loss": 3.8031833171844482, "loss_main": 3.1658358573913574, "loss_zero": 3.186737060546875, "loss_en": 3.164726734161377, "loss_zh": 3.166945219039917, "noise_ramp": 1.0, "noise_mult": 0.7647836208343506, "noise_std_eff": 0.07647836208343506, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1212.902856349945}
176
+ {"step": 24440, "epoch": 0, "loss": 4.5047197341918945, "loss_main": 3.7428832054138184, "loss_zero": 3.8091816902160645, "loss_en": 3.753844976425171, "loss_zh": 3.731921434402466, "noise_ramp": 1.0, "noise_mult": 0.8832271695137024, "noise_std_eff": 0.08832271695137024, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1218.4750633239746}
177
+ {"step": 24460, "epoch": 0, "loss": 3.904207944869995, "loss_main": 3.2177605628967285, "loss_zero": 3.432237148284912, "loss_en": 3.20790433883667, "loss_zh": 3.227617025375366, "noise_ramp": 1.0, "noise_mult": 0.8446130156517029, "noise_std_eff": 0.08446130156517029, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1224.027331829071}
178
+ {"step": 24480, "epoch": 0, "loss": 4.258175373077393, "loss_main": 3.5214223861694336, "loss_zero": 3.683764934539795, "loss_en": 3.5413036346435547, "loss_zh": 3.5015413761138916, "noise_ramp": 1.0, "noise_mult": 0.8328529000282288, "noise_std_eff": 0.08328529000282288, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1229.6488692760468}
179
+ {"step": 24500, "epoch": 0, "loss": 3.790485382080078, "loss_main": 3.132014036178589, "loss_zero": 3.292356014251709, "loss_en": 3.1405508518218994, "loss_zh": 3.1234772205352783, "noise_ramp": 1.0, "noise_mult": 0.5995607376098633, "noise_std_eff": 0.05995607376098633, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1235.2065436840057}
180
+ {"step": 24520, "epoch": 0, "loss": 3.923614263534546, "loss_main": 3.249544143676758, "loss_zero": 3.3703501224517822, "loss_en": 3.2419679164886475, "loss_zh": 3.257120370864868, "noise_ramp": 1.0, "noise_mult": 0.5400378108024597, "noise_std_eff": 0.05400378108024598, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1240.780906200409}
181
+ {"step": 24540, "epoch": 0, "loss": 4.287783145904541, "loss_main": 3.567403793334961, "loss_zero": 3.6018974781036377, "loss_en": 3.559112071990967, "loss_zh": 3.575695753097534, "noise_ramp": 1.0, "noise_mult": 0.6988633275032043, "noise_std_eff": 0.06988633275032044, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1246.369514465332}
182
+ {"step": 24560, "epoch": 0, "loss": 4.143819332122803, "loss_main": 3.446690082550049, "loss_zero": 3.4856467247009277, "loss_en": 3.4623355865478516, "loss_zh": 3.431044816970825, "noise_ramp": 1.0, "noise_mult": 0.7092710733413696, "noise_std_eff": 0.07092710733413697, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1251.903310060501}
183
+ {"step": 24580, "epoch": 0, "loss": 4.186366081237793, "loss_main": 3.4800453186035156, "loss_zero": 3.5316038131713867, "loss_en": 3.460685968399048, "loss_zh": 3.4994046688079834, "noise_ramp": 1.0, "noise_mult": 0.8506710529327393, "noise_std_eff": 0.08506710529327394, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1257.3901772499084}
184
+ {"step": 24600, "epoch": 0, "loss": 3.508241891860962, "loss_main": 2.9079933166503906, "loss_zero": 3.0012426376342773, "loss_en": 2.892348051071167, "loss_zh": 2.9236388206481934, "noise_ramp": 1.0, "noise_mult": 0.8901552557945251, "noise_std_eff": 0.08901552557945253, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1262.8872509002686}
185
+ {"step": 24620, "epoch": 0, "loss": 3.5186827182769775, "loss_main": 2.9137845039367676, "loss_zero": 3.02449107170105, "loss_en": 2.8697409629821777, "loss_zh": 2.9578280448913574, "noise_ramp": 1.0, "noise_mult": 0.8995612263679504, "noise_std_eff": 0.08995612263679505, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1268.3866128921509}
186
+ {"step": 24640, "epoch": 0, "loss": 3.893267869949341, "loss_main": 3.232332229614258, "loss_zero": 3.304677963256836, "loss_en": 3.2306344509124756, "loss_zh": 3.23403000831604, "noise_ramp": 1.0, "noise_mult": 0.9092096090316772, "noise_std_eff": 0.09092096090316773, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1273.852080821991}
187
+ {"step": 24660, "epoch": 0, "loss": 4.3309006690979, "loss_main": 3.583850622177124, "loss_zero": 3.735250234603882, "loss_en": 3.5868139266967773, "loss_zh": 3.5808873176574707, "noise_ramp": 1.0, "noise_mult": 0.487322598695755, "noise_std_eff": 0.0487322598695755, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1279.3573446273804}
188
+ {"step": 24680, "epoch": 0, "loss": 3.2609548568725586, "loss_main": 2.701478958129883, "loss_zero": 2.7973787784576416, "loss_en": 2.7148265838623047, "loss_zh": 2.68813157081604, "noise_ramp": 1.0, "noise_mult": 0.7559375166893005, "noise_std_eff": 0.07559375166893006, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1284.8831415176392}
189
+ {"step": 24700, "epoch": 0, "loss": 4.117807388305664, "loss_main": 3.407426595687866, "loss_zero": 3.5519046783447266, "loss_en": 3.402986526489258, "loss_zh": 3.4118666648864746, "noise_ramp": 1.0, "noise_mult": 0.5209697484970093, "noise_std_eff": 0.05209697484970093, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1290.445683479309}
190
+ {"step": 24720, "epoch": 0, "loss": 3.651014566421509, "loss_main": 3.0443859100341797, "loss_zero": 3.0331435203552246, "loss_en": 3.0766828060150146, "loss_zh": 3.0120890140533447, "noise_ramp": 1.0, "noise_mult": 0.6837408542633057, "noise_std_eff": 0.06837408542633057, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1295.9853324890137}
191
+ {"step": 24740, "epoch": 0, "loss": 3.7968311309814453, "loss_main": 3.1592156887054443, "loss_zero": 3.1880767345428467, "loss_en": 3.1703732013702393, "loss_zh": 3.1480581760406494, "noise_ramp": 1.0, "noise_mult": 0.7824752330780029, "noise_std_eff": 0.0782475233078003, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1301.4741582870483}
192
+ {"step": 24760, "epoch": 0, "loss": 3.904627561569214, "loss_main": 3.2551109790802, "loss_zero": 3.2475831508636475, "loss_en": 3.245938301086426, "loss_zh": 3.2642836570739746, "noise_ramp": 1.0, "noise_mult": 0.8251598477363586, "noise_std_eff": 0.08251598477363586, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1306.9403409957886}
193
+ {"step": 24780, "epoch": 0, "loss": 4.10170316696167, "loss_main": 3.4113450050354004, "loss_zero": 3.4517900943756104, "loss_en": 3.4415857791900635, "loss_zh": 3.3811042308807373, "noise_ramp": 1.0, "noise_mult": 0.40364760160446167, "noise_std_eff": 0.04036476016044617, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1312.455843448639}
194
+ {"step": 24800, "epoch": 0, "loss": 4.147869110107422, "loss_main": 3.4518368244171143, "loss_zero": 3.480161190032959, "loss_en": 3.427063226699829, "loss_zh": 3.4766104221343994, "noise_ramp": 1.0, "noise_mult": 0.7373625636100769, "noise_std_eff": 0.0737362563610077, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1317.9750752449036}
195
+ {"step": 24820, "epoch": 0, "loss": 3.872774600982666, "loss_main": 3.217620372772217, "loss_zero": 3.275770902633667, "loss_en": 3.2257604598999023, "loss_zh": 3.209480047225952, "noise_ramp": 1.0, "noise_mult": 0.583233654499054, "noise_std_eff": 0.0583233654499054, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1348.42821931839}
196
+ {"step": 24840, "epoch": 0, "loss": 3.7362122535705566, "loss_main": 3.129997730255127, "loss_zero": 3.0310721397399902, "loss_en": 3.098855495452881, "loss_zh": 3.161139726638794, "noise_ramp": 1.0, "noise_mult": 0.7344589233398438, "noise_std_eff": 0.07344589233398438, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1353.9685044288635}
197
+ {"step": 24860, "epoch": 0, "loss": 4.326812744140625, "loss_main": 3.6008501052856445, "loss_zero": 3.6298141479492188, "loss_en": 3.565999984741211, "loss_zh": 3.635700225830078, "noise_ramp": 1.0, "noise_mult": 0.9807822108268738, "noise_std_eff": 0.09807822108268738, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1359.5365557670593}
198
+ {"step": 24880, "epoch": 0, "loss": 3.9172616004943848, "loss_main": 3.268277168273926, "loss_zero": 3.244921922683716, "loss_en": 3.2693443298339844, "loss_zh": 3.267209768295288, "noise_ramp": 1.0, "noise_mult": 0.7808283567428589, "noise_std_eff": 0.0780828356742859, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1365.0940001010895}
199
+ {"step": 24900, "epoch": 0, "loss": 4.019094944000244, "loss_main": 3.3168768882751465, "loss_zero": 3.511089563369751, "loss_en": 3.2973473072052, "loss_zh": 3.336406707763672, "noise_ramp": 1.0, "noise_mult": 0.845466673374176, "noise_std_eff": 0.08454666733741761, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1370.6297211647034}
200
+ {"step": 24920, "epoch": 0, "loss": 3.4121992588043213, "loss_main": 2.835468053817749, "loss_zero": 2.8836557865142822, "loss_en": 2.8094115257263184, "loss_zh": 2.8615245819091797, "noise_ramp": 1.0, "noise_mult": 0.9619046449661255, "noise_std_eff": 0.09619046449661256, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1376.5106287002563}
201
+ {"step": 24940, "epoch": 0, "loss": 4.548591613769531, "loss_main": 3.7841529846191406, "loss_zero": 3.8221921920776367, "loss_en": 3.7803142070770264, "loss_zh": 3.787992000579834, "noise_ramp": 1.0, "noise_mult": 0.837864339351654, "noise_std_eff": 0.08378643393516541, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1382.19504404068}
202
+ {"step": 24960, "epoch": 0, "loss": 3.762326717376709, "loss_main": 3.133695125579834, "loss_zero": 3.143158197402954, "loss_en": 3.115365505218506, "loss_zh": 3.152024745941162, "noise_ramp": 1.0, "noise_mult": 0.48681649565696716, "noise_std_eff": 0.04868164956569672, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1387.7557718753815}
203
+ {"step": 24980, "epoch": 0, "loss": 4.05138635635376, "loss_main": 3.379668712615967, "loss_zero": 3.3585891723632812, "loss_en": 3.350282669067383, "loss_zh": 3.409054756164551, "noise_ramp": 1.0, "noise_mult": 0.8748701810836792, "noise_std_eff": 0.08748701810836793, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1393.3162095546722}
204
+ {"step": 25000, "epoch": 0, "loss": 3.9687182903289795, "loss_main": 3.2855210304260254, "loss_zero": 3.4159862995147705, "loss_en": 3.263737201690674, "loss_zh": 3.307304620742798, "noise_ramp": 1.0, "noise_mult": 0.6631525754928589, "noise_std_eff": 0.0663152575492859, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1398.8423159122467}
205
+ {"step": 25020, "epoch": 0, "loss": 3.4977080821990967, "loss_main": 2.901082754135132, "loss_zero": 2.983126401901245, "loss_en": 2.9072999954223633, "loss_zh": 2.8948655128479004, "noise_ramp": 1.0, "noise_mult": 0.9737431406974792, "noise_std_eff": 0.09737431406974793, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1404.483494758606}
206
+ {"step": 25040, "epoch": 0, "loss": 3.424121379852295, "loss_main": 2.8415188789367676, "loss_zero": 2.9130125045776367, "loss_en": 2.831798791885376, "loss_zh": 2.8512392044067383, "noise_ramp": 1.0, "noise_mult": 0.6394612789154053, "noise_std_eff": 0.06394612789154053, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1410.098230600357}
207
+ {"step": 25060, "epoch": 0, "loss": 4.194252014160156, "loss_main": 3.49636173248291, "loss_zero": 3.4894516468048096, "loss_en": 3.5320043563842773, "loss_zh": 3.460718870162964, "noise_ramp": 1.0, "noise_mult": 0.9510579705238342, "noise_std_eff": 0.09510579705238342, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1415.6521940231323}
208
+ {"step": 25080, "epoch": 0, "loss": 4.421529769897461, "loss_main": 3.679908037185669, "loss_zero": 3.7081074714660645, "loss_en": 3.6330549716949463, "loss_zh": 3.7267611026763916, "noise_ramp": 1.0, "noise_mult": 0.9369299411773682, "noise_std_eff": 0.09369299411773682, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1421.2613639831543}
209
+ {"step": 25100, "epoch": 0, "loss": 4.174653053283691, "loss_main": 3.4542224407196045, "loss_zero": 3.6021533012390137, "loss_en": 3.4624481201171875, "loss_zh": 3.4459967613220215, "noise_ramp": 1.0, "noise_mult": 0.7639115452766418, "noise_std_eff": 0.07639115452766419, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1426.7920615673065}
210
+ {"step": 25120, "epoch": 0, "loss": 3.332240104675293, "loss_main": 2.7652528285980225, "loss_zero": 2.8349366188049316, "loss_en": 2.7798001766204834, "loss_zh": 2.7507054805755615, "noise_ramp": 1.0, "noise_mult": 0.527439534664154, "noise_std_eff": 0.05274395346641541, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1432.353534936905}
211
+ {"step": 25140, "epoch": 0, "loss": 3.5819168090820312, "loss_main": 2.9842171669006348, "loss_zero": 2.9884979724884033, "loss_en": 2.9791362285614014, "loss_zh": 2.989298105239868, "noise_ramp": 1.0, "noise_mult": 0.6860755085945129, "noise_std_eff": 0.06860755085945129, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1437.9755990505219}
212
+ {"step": 25160, "epoch": 0, "loss": 4.084944725036621, "loss_main": 3.3946003913879395, "loss_zero": 3.4517221450805664, "loss_en": 3.456221103668213, "loss_zh": 3.332979440689087, "noise_ramp": 1.0, "noise_mult": 0.8707074522972107, "noise_std_eff": 0.08707074522972108, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1443.5639255046844}
213
+ {"step": 25180, "epoch": 0, "loss": 4.222889423370361, "loss_main": 3.5194993019104004, "loss_zero": 3.5169501304626465, "loss_en": 3.500173330307007, "loss_zh": 3.538825273513794, "noise_ramp": 1.0, "noise_mult": 0.9043376445770264, "noise_std_eff": 0.09043376445770264, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1449.1410067081451}
214
+ {"step": 25200, "epoch": 0, "loss": 4.156040191650391, "loss_main": 3.4451522827148438, "loss_zero": 3.554439067840576, "loss_en": 3.471616268157959, "loss_zh": 3.4186882972717285, "noise_ramp": 1.0, "noise_mult": 0.9157533049583435, "noise_std_eff": 0.09157533049583436, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1454.7531542778015}
215
+ {"step": 25220, "epoch": 0, "loss": 3.5359134674072266, "loss_main": 2.927280902862549, "loss_zero": 3.0431623458862305, "loss_en": 2.9443531036376953, "loss_zh": 2.9102089405059814, "noise_ramp": 1.0, "noise_mult": 0.9059472680091858, "noise_std_eff": 0.09059472680091858, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1486.1614744663239}
216
+ {"step": 25240, "epoch": 0, "loss": 3.6867687702178955, "loss_main": 3.0671424865722656, "loss_zero": 3.0981311798095703, "loss_en": 3.07118558883667, "loss_zh": 3.0630993843078613, "noise_ramp": 1.0, "noise_mult": 0.9519422054290771, "noise_std_eff": 0.09519422054290771, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1491.772569656372}
217
+ {"step": 25260, "epoch": 0, "loss": 4.43765926361084, "loss_main": 3.6674540042877197, "loss_zero": 3.851025342941284, "loss_en": 3.668987274169922, "loss_zh": 3.6659207344055176, "noise_ramp": 1.0, "noise_mult": 0.5477694272994995, "noise_std_eff": 0.05477694272994996, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1497.3634185791016}
218
+ {"step": 25280, "epoch": 0, "loss": 3.4967939853668213, "loss_main": 2.8983798027038574, "loss_zero": 2.9920711517333984, "loss_en": 2.89105224609375, "loss_zh": 2.9057071208953857, "noise_ramp": 1.0, "noise_mult": 0.7320122718811035, "noise_std_eff": 0.07320122718811035, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1502.9728622436523}
219
+ {"step": 25300, "epoch": 0, "loss": 3.946608066558838, "loss_main": 3.2752370834350586, "loss_zero": 3.3568551540374756, "loss_en": 3.2916479110717773, "loss_zh": 3.2588260173797607, "noise_ramp": 1.0, "noise_mult": 0.8347287178039551, "noise_std_eff": 0.08347287178039552, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1508.5369110107422}
220
+ {"step": 25320, "epoch": 0, "loss": 3.9988110065460205, "loss_main": 3.3172736167907715, "loss_zero": 3.407686710357666, "loss_en": 3.3264005184173584, "loss_zh": 3.3081467151641846, "noise_ramp": 1.0, "noise_mult": 0.9488092064857483, "noise_std_eff": 0.09488092064857484, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1514.1573724746704}
221
+ {"step": 25340, "epoch": 0, "loss": 3.995042324066162, "loss_main": 3.3184995651245117, "loss_zero": 3.382714033126831, "loss_en": 3.309211015701294, "loss_zh": 3.3277878761291504, "noise_ramp": 1.0, "noise_mult": 0.4179893732070923, "noise_std_eff": 0.04179893732070923, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1519.7480919361115}
222
+ {"step": 25360, "epoch": 0, "loss": 3.884035348892212, "loss_main": 3.231996536254883, "loss_zero": 3.2601943016052246, "loss_en": 3.215773820877075, "loss_zh": 3.2482190132141113, "noise_ramp": 1.0, "noise_mult": 0.5935606956481934, "noise_std_eff": 0.05935606956481934, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1525.5334737300873}
223
+ {"step": 25380, "epoch": 0, "loss": 4.714717864990234, "loss_main": 3.937520742416382, "loss_zero": 3.885986804962158, "loss_en": 3.8875205516815186, "loss_zh": 3.987520933151245, "noise_ramp": 1.0, "noise_mult": 0.515251874923706, "noise_std_eff": 0.05152518749237061, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1531.1231591701508}
224
+ {"step": 25400, "epoch": 0, "loss": 3.2208950519561768, "loss_main": 2.672840118408203, "loss_zero": 2.74027419090271, "loss_en": 2.698622941970825, "loss_zh": 2.647057294845581, "noise_ramp": 1.0, "noise_mult": 0.7438763976097107, "noise_std_eff": 0.07438763976097107, "z_zero_prob": 0.0, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.2, "z_zero_loss_every": 1, "elapsed_sec": 1536.656251192093}
planB_polish_robust_dec_r2k_lz02/probe_leakage.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ckpt": "runs/planB_polish_robust_dec_r2k_lz02/ckpt.pt",
3
+ "max_train_groups": 8000,
4
+ "max_valid_groups": 946,
5
+ "probe_hidden": 256,
6
+ "epochs": 20,
7
+ "lr": 0.01,
8
+ "baseline_majority_lang_acc": 0.5,
9
+ "baseline_majority_len_acc": 0.9271249771118164,
10
+ "lang_probe": {
11
+ "train_acc": 0.59225,
12
+ "valid_acc": 0.6035940803382663,
13
+ "best_valid_acc": 0.6035940803382663
14
+ },
15
+ "len_probe": {
16
+ "train_acc": 0.928875,
17
+ "valid_acc": 0.9011627906976745,
18
+ "best_valid_acc": 0.9011627906976745
19
+ },
20
+ "lang_probe_length_matched": {
21
+ "train_acc": 0.5841469681397738,
22
+ "valid_acc": 0.5738699007717751,
23
+ "best_valid_acc": 0.5948180815876516
24
+ },
25
+ "lang_bucketwise_valid_acc": {
26
+ "bucket_0_acc": 0.375,
27
+ "bucket_1_acc": 0.699999988079071,
28
+ "bucket_2_acc": 0.6086956262588501,
29
+ "bucket_3_acc": 0.6052786111831665
30
+ },
31
+ "length_matched_sizes": {
32
+ "train": 15568,
33
+ "valid": 1814
34
+ }
35
+ }
planB_polish_robust_dec_r2k_zprob01/ckpt.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7aaf9471148b6c9c44e75cada22d29689025dd423da3c97e29585bd16ef166fb
3
+ size 2439403449
planB_polish_robust_dec_r2k_zprob01/config.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train_jsonl": "data/groups_train.jsonl",
3
+ "valid_jsonl": "data/groups_valid.jsonl",
4
+ "run_dir": "runs/planB_polish_robust_dec_r2k_zprob01",
5
+ "resume": "runs/planB_polish_robust_dec_noise10_cont_floor03/ckpt.pt",
6
+ "backbone": "google/mt5-small",
7
+ "num_latents": 16,
8
+ "batch_size": 4,
9
+ "grad_accum": 8,
10
+ "epochs": 1,
11
+ "max_doc_len": 256,
12
+ "max_sum_len": 64,
13
+ "eval_every": 400,
14
+ "log_every": 20,
15
+ "max_train_examples": 0,
16
+ "max_valid_examples": 0,
17
+ "lr": 0.0003,
18
+ "weight_decay": 0.0,
19
+ "latent_dropout": 0.1,
20
+ "latent_noise_std": 0.1,
21
+ "noise_warmup_steps": 500,
22
+ "noise_warmup_start_step": -1,
23
+ "noise_mult_floor": 0.3,
24
+ "noise_mult_pow": 0.5,
25
+ "z_zero_prob": 0.1,
26
+ "z_zero_loss_weight": 0.0,
27
+ "z_zero_loss_every": 1,
28
+ "untie_lm_head": false,
29
+ "seed": 42
30
+ }
planB_polish_robust_dec_r2k_zprob01/diag.jsonl ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"step": 21200, "nll_en": 3.720574241864001, "nll_zh": 3.721967575161956, "nll": 3.7212709085129783}
2
+ {"step": 21600, "nll_en": 3.517127914106115, "nll_zh": 3.523969565587114, "nll": 3.520548739846615}
3
+ {"step": 22000, "nll_en": 3.4837384475966084, "nll_zh": 3.490845799194078, "nll": 3.4872921233953433}
4
+ {"step": 22400, "nll_en": 3.4740772247314453, "nll_zh": 3.478010588922158, "nll": 3.4760439068268014}
5
+ {"step": 22800, "nll_en": 3.449692909611708, "nll_zh": 3.4573824380467357, "nll": 3.4535376738292216}
6
+ {"step": 23200, "nll_en": 3.433061654169514, "nll_zh": 3.4387214541687268, "nll": 3.4358915541691206}
7
+ {"step": 23600, "nll_en": 3.4258173363909408, "nll_zh": 3.4284714634272313, "nll": 3.427144399909086}
8
+ {"step": 24000, "nll_en": 3.4078021634204694, "nll_zh": 3.4134877942833284, "nll": 3.410644978851899}
9
+ {"step": 24400, "nll_en": 3.392732555720065, "nll_zh": 3.3975337368237293, "nll": 3.395133146271897}
10
+ {"step": 24800, "nll_en": 3.373830850734267, "nll_zh": 3.378818086539464, "nll": 3.3763244686368656}
11
+ {"step": 25200, "nll_en": 3.3685770024708916, "nll_zh": 3.371672972060157, "nll": 3.3701249872655246}
planB_polish_robust_dec_r2k_zprob01/diag_full.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ckpt": "runs/planB_polish_robust_dec_r2k_zprob01/ckpt.pt",
3
+ "nll_en": 3.362803950118212,
4
+ "nll_zh": 3.3654597969942315,
5
+ "swap_delta_en": 0.47385728182802744,
6
+ "swap_delta_zh": 0.45629846218020415,
7
+ "ablate_zero_delta_en": 0.14164270659077746,
8
+ "ablate_mean_delta_en": 0.18201592205442788,
9
+ "ablate_noise_delta_en": 3.8134631838405357,
10
+ "ablate_zero_delta_zh": 0.13898685971475808,
11
+ "ablate_mean_delta_zh": 0.17577294718639544,
12
+ "ablate_noise_delta_zh": 3.9203611640043037,
13
+ "inv_top1_full": 0.060253698378801346,
14
+ "inv_top5_full": 0.20401690900325775,
15
+ "diag_sim_mean": 0.965337336063385,
16
+ "offdiag_sim_mean": 0.4495629370212555,
17
+ "sim_margin": 0.5157743990421295,
18
+ "n_valid": 946
19
+ }
planB_polish_robust_dec_r2k_zprob01/logs.jsonl ADDED
@@ -0,0 +1,224 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"step": 20940, "epoch": 0, "loss": 5.968393325805664, "loss_main": 5.968393325805664, "loss_zero": 0.0, "loss_en": 6.043532371520996, "loss_zh": 5.893253803253174, "noise_ramp": 0.024, "noise_mult": 0.6759225130081177, "noise_std_eff": 0.0016222140312194826, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 2.6906003952026367}
2
+ {"step": 20960, "epoch": 0, "loss": 4.69274377822876, "loss_main": 4.69274377822876, "loss_zero": 0.0, "loss_en": 4.671468734741211, "loss_zh": 4.714018821716309, "noise_ramp": 0.064, "noise_mult": 0.6551672220230103, "noise_std_eff": 0.004193070220947266, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 6.153272390365601}
3
+ {"step": 20980, "epoch": 0, "loss": 4.693017959594727, "loss_main": 4.693017959594727, "loss_zero": 0.0, "loss_en": 4.578136920928955, "loss_zh": 4.80789852142334, "noise_ramp": 0.104, "noise_mult": 0.9023459553718567, "noise_std_eff": 0.00938439793586731, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 9.610422134399414}
4
+ {"step": 21000, "epoch": 0, "loss": 4.711987495422363, "loss_main": 4.711987495422363, "loss_zero": 0.0, "loss_en": 4.5674920082092285, "loss_zh": 4.85648250579834, "noise_ramp": 0.144, "noise_mult": 0.9806118607521057, "noise_std_eff": 0.014120810794830322, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 13.022950410842896}
5
+ {"step": 21020, "epoch": 0, "loss": 5.499646186828613, "loss_main": 5.499646186828613, "loss_zero": 0.0, "loss_en": 5.5641188621521, "loss_zh": 5.435173511505127, "noise_ramp": 0.184, "noise_mult": 0.7761704921722412, "noise_std_eff": 0.014281537055969238, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 16.38509964942932}
6
+ {"step": 21040, "epoch": 0, "loss": 4.228109359741211, "loss_main": 4.228109359741211, "loss_zero": 0.0, "loss_en": 4.201045513153076, "loss_zh": 4.255173683166504, "noise_ramp": 0.224, "noise_mult": 0.6353203058242798, "noise_std_eff": 0.014231174850463869, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 19.78296136856079}
7
+ {"step": 21060, "epoch": 0, "loss": 4.406501770019531, "loss_main": 4.406501770019531, "loss_zero": 0.0, "loss_en": 4.3309407234191895, "loss_zh": 4.482062339782715, "noise_ramp": 0.264, "noise_mult": 0.6195439100265503, "noise_std_eff": 0.01635595922470093, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 23.183494567871094}
8
+ {"step": 21080, "epoch": 0, "loss": 4.559249401092529, "loss_main": 4.559249401092529, "loss_zero": 0.0, "loss_en": 4.594013214111328, "loss_zh": 4.5244855880737305, "noise_ramp": 0.304, "noise_mult": 0.47016966342926025, "noise_std_eff": 0.014293157768249512, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 26.720442056655884}
9
+ {"step": 21100, "epoch": 0, "loss": 4.309741973876953, "loss_main": 4.309741973876953, "loss_zero": 0.0, "loss_en": 4.18280029296875, "loss_zh": 4.436683177947998, "noise_ramp": 0.344, "noise_mult": 0.9843459725379944, "noise_std_eff": 0.033861501455307, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 30.211848735809326}
10
+ {"step": 21120, "epoch": 0, "loss": 5.145464897155762, "loss_main": 5.145464897155762, "loss_zero": 0.0, "loss_en": 5.12139368057251, "loss_zh": 5.169536113739014, "noise_ramp": 0.384, "noise_mult": 0.7634071707725525, "noise_std_eff": 0.02931483535766602, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 33.66669011116028}
11
+ {"step": 21140, "epoch": 0, "loss": 4.276890754699707, "loss_main": 4.276890754699707, "loss_zero": 0.0, "loss_en": 4.363436698913574, "loss_zh": 4.19034481048584, "noise_ramp": 0.424, "noise_mult": 0.583137035369873, "noise_std_eff": 0.024725010299682618, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 37.06024217605591}
12
+ {"step": 21160, "epoch": 0, "loss": 4.367673873901367, "loss_main": 4.367673873901367, "loss_zero": 0.0, "loss_en": 4.417830944061279, "loss_zh": 4.317516326904297, "noise_ramp": 0.464, "noise_mult": 0.49724942445755005, "noise_std_eff": 0.023072373294830327, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 40.45362973213196}
13
+ {"step": 21180, "epoch": 0, "loss": 4.21634578704834, "loss_main": 4.21634578704834, "loss_zero": 0.0, "loss_en": 4.1752471923828125, "loss_zh": 4.257443904876709, "noise_ramp": 0.504, "noise_mult": 0.8637520670890808, "noise_std_eff": 0.043533104181289674, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 43.810283184051514}
14
+ {"step": 21200, "epoch": 0, "loss": 5.094034194946289, "loss_main": 5.094034194946289, "loss_zero": 0.0, "loss_en": 5.185015678405762, "loss_zh": 5.003052711486816, "noise_ramp": 0.544, "noise_mult": 0.876893937587738, "noise_std_eff": 0.047703030204772956, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 47.30783462524414}
15
+ {"step": 21220, "epoch": 0, "loss": 3.139927864074707, "loss_main": 3.139927864074707, "loss_zero": 0.0, "loss_en": 3.122591972351074, "loss_zh": 3.15726375579834, "noise_ramp": 0.584, "noise_mult": 0.87684166431427, "noise_std_eff": 0.051207553195953374, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 75.61918950080872}
16
+ {"step": 21240, "epoch": 0, "loss": 4.2085981369018555, "loss_main": 4.2085981369018555, "loss_zero": 0.0, "loss_en": 4.21269416809082, "loss_zh": 4.204502105712891, "noise_ramp": 0.624, "noise_mult": 0.6665846705436707, "noise_std_eff": 0.04159488344192505, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 79.13465237617493}
17
+ {"step": 21260, "epoch": 0, "loss": 3.8138883113861084, "loss_main": 3.8138883113861084, "loss_zero": 0.0, "loss_en": 3.828432083129883, "loss_zh": 3.799344539642334, "noise_ramp": 0.664, "noise_mult": 0.9746540188789368, "noise_std_eff": 0.06471702685356141, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 82.74431037902832}
18
+ {"step": 21280, "epoch": 0, "loss": 3.008007049560547, "loss_main": 3.008007049560547, "loss_zero": 0.0, "loss_en": 3.001394271850586, "loss_zh": 3.0146195888519287, "noise_ramp": 0.704, "noise_mult": 0.9574816823005676, "noise_std_eff": 0.06740671043395996, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 86.30444264411926}
19
+ {"step": 21300, "epoch": 0, "loss": 3.550321578979492, "loss_main": 3.550321578979492, "loss_zero": 0.0, "loss_en": 3.5527608394622803, "loss_zh": 3.547882318496704, "noise_ramp": 0.744, "noise_mult": 0.6488423347473145, "noise_std_eff": 0.048273869705200194, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 89.78857851028442}
20
+ {"step": 21320, "epoch": 0, "loss": 3.391124963760376, "loss_main": 3.391124963760376, "loss_zero": 0.0, "loss_en": 3.3511054515838623, "loss_zh": 3.4311444759368896, "noise_ramp": 0.784, "noise_mult": 0.9726906418800354, "noise_std_eff": 0.07625894632339478, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 93.23740887641907}
21
+ {"step": 21340, "epoch": 0, "loss": 3.150381088256836, "loss_main": 3.150381088256836, "loss_zero": 0.0, "loss_en": 3.139869451522827, "loss_zh": 3.1608927249908447, "noise_ramp": 0.824, "noise_mult": 0.9978471994400024, "noise_std_eff": 0.0822226092338562, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 96.65329766273499}
22
+ {"step": 21360, "epoch": 0, "loss": 3.645218849182129, "loss_main": 3.645218849182129, "loss_zero": 0.0, "loss_en": 3.6599249839782715, "loss_zh": 3.6305127143859863, "noise_ramp": 0.864, "noise_mult": 0.4106824994087219, "noise_std_eff": 0.03548296794891358, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 100.0449059009552}
23
+ {"step": 21380, "epoch": 0, "loss": 3.3847227096557617, "loss_main": 3.3847227096557617, "loss_zero": 0.0, "loss_en": 3.379271984100342, "loss_zh": 3.3901736736297607, "noise_ramp": 0.904, "noise_mult": 0.6041542887687683, "noise_std_eff": 0.05461554770469666, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 103.40945839881897}
24
+ {"step": 21400, "epoch": 0, "loss": 3.6683616638183594, "loss_main": 3.6683616638183594, "loss_zero": 0.0, "loss_en": 3.6579039096832275, "loss_zh": 3.678819417953491, "noise_ramp": 0.944, "noise_mult": 0.7595483064651489, "noise_std_eff": 0.07170136013031006, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 106.80217695236206}
25
+ {"step": 21420, "epoch": 0, "loss": 3.363452911376953, "loss_main": 3.363452911376953, "loss_zero": 0.0, "loss_en": 3.348710060119629, "loss_zh": 3.3781960010528564, "noise_ramp": 0.984, "noise_mult": 0.9466076493263245, "noise_std_eff": 0.09314619269371033, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 110.22830533981323}
26
+ {"step": 21440, "epoch": 0, "loss": 3.4875717163085938, "loss_main": 3.4875717163085938, "loss_zero": 0.0, "loss_en": 3.466209888458252, "loss_zh": 3.5089335441589355, "noise_ramp": 1.0, "noise_mult": 0.8170216679573059, "noise_std_eff": 0.0817021667957306, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 113.66204714775085}
27
+ {"step": 21460, "epoch": 0, "loss": 3.8374760150909424, "loss_main": 3.8374760150909424, "loss_zero": 0.0, "loss_en": 3.838329553604126, "loss_zh": 3.836622476577759, "noise_ramp": 1.0, "noise_mult": 0.7756085991859436, "noise_std_eff": 0.07756085991859436, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 117.03986978530884}
28
+ {"step": 21480, "epoch": 0, "loss": 3.3151159286499023, "loss_main": 3.3151159286499023, "loss_zero": 0.0, "loss_en": 3.29899525642395, "loss_zh": 3.3312366008758545, "noise_ramp": 1.0, "noise_mult": 0.47319290041923523, "noise_std_eff": 0.04731929004192353, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 120.4391462802887}
29
+ {"step": 21500, "epoch": 0, "loss": 3.371889114379883, "loss_main": 3.371889114379883, "loss_zero": 0.0, "loss_en": 3.3493311405181885, "loss_zh": 3.3944473266601562, "noise_ramp": 1.0, "noise_mult": 0.7031762599945068, "noise_std_eff": 0.07031762599945068, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 123.84822201728821}
30
+ {"step": 21520, "epoch": 0, "loss": 3.0654306411743164, "loss_main": 3.0654306411743164, "loss_zero": 0.0, "loss_en": 3.072023868560791, "loss_zh": 3.058837652206421, "noise_ramp": 1.0, "noise_mult": 0.6508877277374268, "noise_std_eff": 0.06508877277374268, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 127.17781639099121}
31
+ {"step": 21540, "epoch": 0, "loss": 3.220388650894165, "loss_main": 3.220388650894165, "loss_zero": 0.0, "loss_en": 3.232374429702759, "loss_zh": 3.2084028720855713, "noise_ramp": 1.0, "noise_mult": 0.9554324746131897, "noise_std_eff": 0.09554324746131898, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 130.4055688381195}
32
+ {"step": 21560, "epoch": 0, "loss": 3.1542530059814453, "loss_main": 3.1542530059814453, "loss_zero": 0.0, "loss_en": 3.132136821746826, "loss_zh": 3.1763689517974854, "noise_ramp": 1.0, "noise_mult": 0.7686743140220642, "noise_std_eff": 0.07686743140220642, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 133.80457592010498}
33
+ {"step": 21580, "epoch": 0, "loss": 3.1533000469207764, "loss_main": 3.1533000469207764, "loss_zero": 0.0, "loss_en": 3.1457831859588623, "loss_zh": 3.1608169078826904, "noise_ramp": 1.0, "noise_mult": 0.9618533253669739, "noise_std_eff": 0.0961853325366974, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 137.1386752128601}
34
+ {"step": 21600, "epoch": 0, "loss": 3.675626754760742, "loss_main": 3.675626754760742, "loss_zero": 0.0, "loss_en": 3.6814472675323486, "loss_zh": 3.6698060035705566, "noise_ramp": 1.0, "noise_mult": 0.7568153142929077, "noise_std_eff": 0.07568153142929078, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 140.49405312538147}
35
+ {"step": 21620, "epoch": 0, "loss": 3.688657522201538, "loss_main": 3.688657522201538, "loss_zero": 0.0, "loss_en": 3.684624195098877, "loss_zh": 3.692690849304199, "noise_ramp": 1.0, "noise_mult": 0.5477515459060669, "noise_std_eff": 0.054775154590606695, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 169.55257987976074}
36
+ {"step": 21640, "epoch": 0, "loss": 3.5156712532043457, "loss_main": 3.5156712532043457, "loss_zero": 0.0, "loss_en": 3.5120480060577393, "loss_zh": 3.519294261932373, "noise_ramp": 1.0, "noise_mult": 0.9082446694374084, "noise_std_eff": 0.09082446694374086, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 172.90774607658386}
37
+ {"step": 21660, "epoch": 0, "loss": 3.7450833320617676, "loss_main": 3.7450833320617676, "loss_zero": 0.0, "loss_en": 3.7343544960021973, "loss_zh": 3.755812168121338, "noise_ramp": 1.0, "noise_mult": 0.9391080737113953, "noise_std_eff": 0.09391080737113953, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 176.28167629241943}
38
+ {"step": 21680, "epoch": 0, "loss": 3.534811019897461, "loss_main": 3.534811019897461, "loss_zero": 0.0, "loss_en": 3.5434482097625732, "loss_zh": 3.5261738300323486, "noise_ramp": 1.0, "noise_mult": 0.9894654154777527, "noise_std_eff": 0.09894654154777527, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 179.56631135940552}
39
+ {"step": 21700, "epoch": 0, "loss": 3.509016275405884, "loss_main": 3.509016275405884, "loss_zero": 0.0, "loss_en": 3.490868091583252, "loss_zh": 3.5271644592285156, "noise_ramp": 1.0, "noise_mult": 0.4927124083042145, "noise_std_eff": 0.04927124083042145, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 182.86989665031433}
40
+ {"step": 21720, "epoch": 0, "loss": 3.0587005615234375, "loss_main": 3.0587005615234375, "loss_zero": 0.0, "loss_en": 3.0678794384002686, "loss_zh": 3.0495216846466064, "noise_ramp": 1.0, "noise_mult": 0.9210430383682251, "noise_std_eff": 0.09210430383682251, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 186.15391397476196}
41
+ {"step": 21740, "epoch": 0, "loss": 3.4361071586608887, "loss_main": 3.4361071586608887, "loss_zero": 0.0, "loss_en": 3.4413511753082275, "loss_zh": 3.430863380432129, "noise_ramp": 1.0, "noise_mult": 0.7899562120437622, "noise_std_eff": 0.07899562120437623, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 189.46748757362366}
42
+ {"step": 21760, "epoch": 0, "loss": 3.4879794120788574, "loss_main": 3.4879794120788574, "loss_zero": 0.0, "loss_en": 3.4823827743530273, "loss_zh": 3.4935762882232666, "noise_ramp": 1.0, "noise_mult": 0.9215280413627625, "noise_std_eff": 0.09215280413627625, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 192.82418870925903}
43
+ {"step": 21780, "epoch": 0, "loss": 3.196280002593994, "loss_main": 3.196280002593994, "loss_zero": 0.0, "loss_en": 3.220036506652832, "loss_zh": 3.172523260116577, "noise_ramp": 1.0, "noise_mult": 0.5995148420333862, "noise_std_eff": 0.05995148420333862, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 196.13874650001526}
44
+ {"step": 21800, "epoch": 0, "loss": 2.6793127059936523, "loss_main": 2.6793127059936523, "loss_zero": 0.0, "loss_en": 2.7088897228240967, "loss_zh": 2.649735689163208, "noise_ramp": 1.0, "noise_mult": 0.9348920583724976, "noise_std_eff": 0.09348920583724976, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 199.44855761528015}
45
+ {"step": 21820, "epoch": 0, "loss": 3.27639102935791, "loss_main": 3.27639102935791, "loss_zero": 0.0, "loss_en": 3.275148630142212, "loss_zh": 3.2776334285736084, "noise_ramp": 1.0, "noise_mult": 0.45741188526153564, "noise_std_eff": 0.04574118852615357, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 202.72850799560547}
46
+ {"step": 21840, "epoch": 0, "loss": 3.6545708179473877, "loss_main": 3.6545708179473877, "loss_zero": 0.0, "loss_en": 3.6315908432006836, "loss_zh": 3.677550792694092, "noise_ramp": 1.0, "noise_mult": 0.9599159955978394, "noise_std_eff": 0.09599159955978394, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 205.99016046524048}
47
+ {"step": 21860, "epoch": 0, "loss": 2.9207701683044434, "loss_main": 2.9207701683044434, "loss_zero": 0.0, "loss_en": 2.9282584190368652, "loss_zh": 2.9132821559906006, "noise_ramp": 1.0, "noise_mult": 0.6321103572845459, "noise_std_eff": 0.06321103572845459, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 209.2950587272644}
48
+ {"step": 21880, "epoch": 0, "loss": 3.928231954574585, "loss_main": 3.928231954574585, "loss_zero": 0.0, "loss_en": 3.910525321960449, "loss_zh": 3.9459385871887207, "noise_ramp": 1.0, "noise_mult": 0.9126268029212952, "noise_std_eff": 0.09126268029212953, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 212.62294125556946}
49
+ {"step": 21900, "epoch": 0, "loss": 3.4520645141601562, "loss_main": 3.4520645141601562, "loss_zero": 0.0, "loss_en": 3.435272216796875, "loss_zh": 3.4688570499420166, "noise_ramp": 1.0, "noise_mult": 0.8664395809173584, "noise_std_eff": 0.08664395809173585, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 215.96391034126282}
50
+ {"step": 21920, "epoch": 0, "loss": 3.0298352241516113, "loss_main": 3.0298352241516113, "loss_zero": 0.0, "loss_en": 3.0320398807525635, "loss_zh": 3.02763032913208, "noise_ramp": 1.0, "noise_mult": 0.43799713253974915, "noise_std_eff": 0.043799713253974915, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 219.3095850944519}
51
+ {"step": 21940, "epoch": 0, "loss": 3.571084976196289, "loss_main": 3.571084976196289, "loss_zero": 0.0, "loss_en": 3.5826871395111084, "loss_zh": 3.5594828128814697, "noise_ramp": 1.0, "noise_mult": 0.8673273324966431, "noise_std_eff": 0.08673273324966431, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 222.6305112838745}
52
+ {"step": 21960, "epoch": 0, "loss": 3.229288101196289, "loss_main": 3.229288101196289, "loss_zero": 0.0, "loss_en": 3.2368404865264893, "loss_zh": 3.221735954284668, "noise_ramp": 1.0, "noise_mult": 0.8344725370407104, "noise_std_eff": 0.08344725370407105, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 225.97225403785706}
53
+ {"step": 21980, "epoch": 0, "loss": 3.1198477745056152, "loss_main": 3.1198477745056152, "loss_zero": 0.0, "loss_en": 3.0965588092803955, "loss_zh": 3.143136501312256, "noise_ramp": 1.0, "noise_mult": 0.9190050363540649, "noise_std_eff": 0.0919005036354065, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 229.27553462982178}
54
+ {"step": 22000, "epoch": 0, "loss": 3.627474308013916, "loss_main": 3.627474308013916, "loss_zero": 0.0, "loss_en": 3.663422107696533, "loss_zh": 3.591526746749878, "noise_ramp": 1.0, "noise_mult": 0.8233438730239868, "noise_std_eff": 0.08233438730239868, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 232.6033980846405}
55
+ {"step": 22020, "epoch": 0, "loss": 3.57879638671875, "loss_main": 3.57879638671875, "loss_zero": 0.0, "loss_en": 3.5906736850738525, "loss_zh": 3.5669190883636475, "noise_ramp": 1.0, "noise_mult": 0.3688124418258667, "noise_std_eff": 0.03688124418258667, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 261.275705575943}
56
+ {"step": 22040, "epoch": 0, "loss": 3.250476598739624, "loss_main": 3.250476598739624, "loss_zero": 0.0, "loss_en": 3.2534801959991455, "loss_zh": 3.2474730014801025, "noise_ramp": 1.0, "noise_mult": 0.6647882461547852, "noise_std_eff": 0.06647882461547852, "z_zero_prob": 0.1, "z_zero_frac": 0.5, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 264.53069734573364}
57
+ {"step": 22060, "epoch": 0, "loss": 3.3120198249816895, "loss_main": 3.3120198249816895, "loss_zero": 0.0, "loss_en": 3.308744192123413, "loss_zh": 3.3152952194213867, "noise_ramp": 1.0, "noise_mult": 0.9790452122688293, "noise_std_eff": 0.09790452122688294, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 267.7785367965698}
58
+ {"step": 22080, "epoch": 0, "loss": 3.3222298622131348, "loss_main": 3.3222298622131348, "loss_zero": 0.0, "loss_en": 3.352447986602783, "loss_zh": 3.2920117378234863, "noise_ramp": 1.0, "noise_mult": 0.8130279183387756, "noise_std_eff": 0.08130279183387756, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 271.1555132865906}
59
+ {"step": 22100, "epoch": 0, "loss": 3.048081159591675, "loss_main": 3.048081159591675, "loss_zero": 0.0, "loss_en": 3.040804386138916, "loss_zh": 3.0553579330444336, "noise_ramp": 1.0, "noise_mult": 0.5561050176620483, "noise_std_eff": 0.055610501766204835, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 274.4469721317291}
60
+ {"step": 22120, "epoch": 0, "loss": 3.900627613067627, "loss_main": 3.900627613067627, "loss_zero": 0.0, "loss_en": 3.881291389465332, "loss_zh": 3.919964075088501, "noise_ramp": 1.0, "noise_mult": 0.9450612664222717, "noise_std_eff": 0.09450612664222718, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 277.74602913856506}
61
+ {"step": 22140, "epoch": 0, "loss": 3.3879332542419434, "loss_main": 3.3879332542419434, "loss_zero": 0.0, "loss_en": 3.2588539123535156, "loss_zh": 3.517012357711792, "noise_ramp": 1.0, "noise_mult": 0.5169932246208191, "noise_std_eff": 0.05169932246208191, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 281.0312821865082}
62
+ {"step": 22160, "epoch": 0, "loss": 3.554722785949707, "loss_main": 3.554722785949707, "loss_zero": 0.0, "loss_en": 3.582324266433716, "loss_zh": 3.5271215438842773, "noise_ramp": 1.0, "noise_mult": 0.6584380865097046, "noise_std_eff": 0.06584380865097046, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 284.3522152900696}
63
+ {"step": 22180, "epoch": 0, "loss": 3.821605682373047, "loss_main": 3.821605682373047, "loss_zero": 0.0, "loss_en": 3.8323564529418945, "loss_zh": 3.81085467338562, "noise_ramp": 1.0, "noise_mult": 0.8264014720916748, "noise_std_eff": 0.08264014720916749, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 287.6226773262024}
64
+ {"step": 22200, "epoch": 0, "loss": 3.3391079902648926, "loss_main": 3.3391079902648926, "loss_zero": 0.0, "loss_en": 3.343782901763916, "loss_zh": 3.3344333171844482, "noise_ramp": 1.0, "noise_mult": 0.6826056241989136, "noise_std_eff": 0.06826056241989135, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 290.952011346817}
65
+ {"step": 22220, "epoch": 0, "loss": 3.2933406829833984, "loss_main": 3.2933406829833984, "loss_zero": 0.0, "loss_en": 3.281630754470825, "loss_zh": 3.3050506114959717, "noise_ramp": 1.0, "noise_mult": 0.4251408576965332, "noise_std_eff": 0.04251408576965332, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 294.3688807487488}
66
+ {"step": 22240, "epoch": 0, "loss": 3.6499452590942383, "loss_main": 3.6499452590942383, "loss_zero": 0.0, "loss_en": 3.6484320163726807, "loss_zh": 3.651458501815796, "noise_ramp": 1.0, "noise_mult": 0.7445165514945984, "noise_std_eff": 0.07445165514945984, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 297.77895283699036}
67
+ {"step": 22260, "epoch": 0, "loss": 3.4503421783447266, "loss_main": 3.4503421783447266, "loss_zero": 0.0, "loss_en": 3.4647843837738037, "loss_zh": 3.4358997344970703, "noise_ramp": 1.0, "noise_mult": 0.7977756261825562, "noise_std_eff": 0.07977756261825562, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 301.0826609134674}
68
+ {"step": 22280, "epoch": 0, "loss": 3.3982958793640137, "loss_main": 3.3982958793640137, "loss_zero": 0.0, "loss_en": 3.39154052734375, "loss_zh": 3.4050509929656982, "noise_ramp": 1.0, "noise_mult": 0.7721915245056152, "noise_std_eff": 0.07721915245056153, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 304.28115677833557}
69
+ {"step": 22300, "epoch": 0, "loss": 3.455331802368164, "loss_main": 3.455331802368164, "loss_zero": 0.0, "loss_en": 3.469158411026001, "loss_zh": 3.4415054321289062, "noise_ramp": 1.0, "noise_mult": 0.9870381951332092, "noise_std_eff": 0.09870381951332093, "z_zero_prob": 0.1, "z_zero_frac": 0.5, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 307.4645743370056}
70
+ {"step": 22320, "epoch": 0, "loss": 3.4816155433654785, "loss_main": 3.4816155433654785, "loss_zero": 0.0, "loss_en": 3.4581799507141113, "loss_zh": 3.5050508975982666, "noise_ramp": 1.0, "noise_mult": 0.5730099678039551, "noise_std_eff": 0.05730099678039551, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 310.7086625099182}
71
+ {"step": 22340, "epoch": 0, "loss": 3.446791172027588, "loss_main": 3.446791172027588, "loss_zero": 0.0, "loss_en": 3.4970014095306396, "loss_zh": 3.396580696105957, "noise_ramp": 1.0, "noise_mult": 0.8515589833259583, "noise_std_eff": 0.08515589833259583, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 313.9461166858673}
72
+ {"step": 22360, "epoch": 0, "loss": 3.952633857727051, "loss_main": 3.952633857727051, "loss_zero": 0.0, "loss_en": 3.9561941623687744, "loss_zh": 3.949073314666748, "noise_ramp": 1.0, "noise_mult": 0.7239408493041992, "noise_std_eff": 0.07239408493041992, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 317.17104411125183}
73
+ {"step": 22380, "epoch": 0, "loss": 3.1081976890563965, "loss_main": 3.1081976890563965, "loss_zero": 0.0, "loss_en": 3.12026047706604, "loss_zh": 3.096134662628174, "noise_ramp": 1.0, "noise_mult": 0.7171584367752075, "noise_std_eff": 0.07171584367752075, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 320.3518431186676}
74
+ {"step": 22400, "epoch": 0, "loss": 3.1832823753356934, "loss_main": 3.1832823753356934, "loss_zero": 0.0, "loss_en": 3.185795783996582, "loss_zh": 3.1807687282562256, "noise_ramp": 1.0, "noise_mult": 0.9791620969772339, "noise_std_eff": 0.0979162096977234, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 323.5846667289734}
75
+ {"step": 22420, "epoch": 0, "loss": 3.4476799964904785, "loss_main": 3.4476799964904785, "loss_zero": 0.0, "loss_en": 3.43900728225708, "loss_zh": 3.456352949142456, "noise_ramp": 1.0, "noise_mult": 0.6852796077728271, "noise_std_eff": 0.06852796077728272, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 351.6378228664398}
76
+ {"step": 22440, "epoch": 0, "loss": 2.9444832801818848, "loss_main": 2.9444832801818848, "loss_zero": 0.0, "loss_en": 2.9422569274902344, "loss_zh": 2.946709632873535, "noise_ramp": 1.0, "noise_mult": 0.4834885597229004, "noise_std_eff": 0.04834885597229004, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 354.8638184070587}
77
+ {"step": 22460, "epoch": 0, "loss": 3.7421345710754395, "loss_main": 3.7421345710754395, "loss_zero": 0.0, "loss_en": 3.7468509674072266, "loss_zh": 3.7374179363250732, "noise_ramp": 1.0, "noise_mult": 0.6281560659408569, "noise_std_eff": 0.0628156065940857, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 358.1002118587494}
78
+ {"step": 22480, "epoch": 0, "loss": 2.9792089462280273, "loss_main": 2.9792089462280273, "loss_zero": 0.0, "loss_en": 2.9603824615478516, "loss_zh": 2.998035430908203, "noise_ramp": 1.0, "noise_mult": 0.8369330763816833, "noise_std_eff": 0.08369330763816835, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 361.3172025680542}
79
+ {"step": 22500, "epoch": 0, "loss": 3.026412010192871, "loss_main": 3.026412010192871, "loss_zero": 0.0, "loss_en": 3.0241787433624268, "loss_zh": 3.0286452770233154, "noise_ramp": 1.0, "noise_mult": 0.794605016708374, "noise_std_eff": 0.0794605016708374, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 364.5088665485382}
80
+ {"step": 22520, "epoch": 0, "loss": 3.4540646076202393, "loss_main": 3.4540646076202393, "loss_zero": 0.0, "loss_en": 3.437455654144287, "loss_zh": 3.4706735610961914, "noise_ramp": 1.0, "noise_mult": 0.418148398399353, "noise_std_eff": 0.04181483983993531, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 367.6989698410034}
81
+ {"step": 22540, "epoch": 0, "loss": 3.808244228363037, "loss_main": 3.808244228363037, "loss_zero": 0.0, "loss_en": 3.794187068939209, "loss_zh": 3.822301149368286, "noise_ramp": 1.0, "noise_mult": 0.8560588955879211, "noise_std_eff": 0.08560588955879211, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 370.85191822052}
82
+ {"step": 22560, "epoch": 0, "loss": 3.681339740753174, "loss_main": 3.681339740753174, "loss_zero": 0.0, "loss_en": 3.6824209690093994, "loss_zh": 3.6802587509155273, "noise_ramp": 1.0, "noise_mult": 0.8143935203552246, "noise_std_eff": 0.08143935203552247, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 374.10052037239075}
83
+ {"step": 22580, "epoch": 0, "loss": 4.068349838256836, "loss_main": 4.068349838256836, "loss_zero": 0.0, "loss_en": 4.064189434051514, "loss_zh": 4.072509765625, "noise_ramp": 1.0, "noise_mult": 0.5393631458282471, "noise_std_eff": 0.05393631458282471, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 377.30322313308716}
84
+ {"step": 22600, "epoch": 0, "loss": 3.394681930541992, "loss_main": 3.394681930541992, "loss_zero": 0.0, "loss_en": 3.382430076599121, "loss_zh": 3.406933546066284, "noise_ramp": 1.0, "noise_mult": 0.7701174020767212, "noise_std_eff": 0.07701174020767212, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 380.52982664108276}
85
+ {"step": 22620, "epoch": 0, "loss": 3.599526882171631, "loss_main": 3.599526882171631, "loss_zero": 0.0, "loss_en": 3.6253249645233154, "loss_zh": 3.5737290382385254, "noise_ramp": 1.0, "noise_mult": 0.8626148104667664, "noise_std_eff": 0.08626148104667664, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 383.7280671596527}
86
+ {"step": 22640, "epoch": 0, "loss": 3.4261903762817383, "loss_main": 3.4261903762817383, "loss_zero": 0.0, "loss_en": 3.420870065689087, "loss_zh": 3.4315109252929688, "noise_ramp": 1.0, "noise_mult": 0.7155901193618774, "noise_std_eff": 0.07155901193618774, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 386.94732666015625}
87
+ {"step": 22660, "epoch": 0, "loss": 3.522948741912842, "loss_main": 3.522948741912842, "loss_zero": 0.0, "loss_en": 3.519958734512329, "loss_zh": 3.5259389877319336, "noise_ramp": 1.0, "noise_mult": 0.9095162749290466, "noise_std_eff": 0.09095162749290467, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 390.14401388168335}
88
+ {"step": 22680, "epoch": 0, "loss": 2.8411619663238525, "loss_main": 2.8411619663238525, "loss_zero": 0.0, "loss_en": 2.8659303188323975, "loss_zh": 2.8163936138153076, "noise_ramp": 1.0, "noise_mult": 0.9280558228492737, "noise_std_eff": 0.09280558228492737, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 393.3531491756439}
89
+ {"step": 22700, "epoch": 0, "loss": 3.5915794372558594, "loss_main": 3.5915794372558594, "loss_zero": 0.0, "loss_en": 3.6119916439056396, "loss_zh": 3.571167230606079, "noise_ramp": 1.0, "noise_mult": 0.3601781725883484, "noise_std_eff": 0.03601781725883484, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 396.5375745296478}
90
+ {"step": 22720, "epoch": 0, "loss": 3.9815797805786133, "loss_main": 3.9815797805786133, "loss_zero": 0.0, "loss_en": 3.9636242389678955, "loss_zh": 3.99953556060791, "noise_ramp": 1.0, "noise_mult": 0.9785065054893494, "noise_std_eff": 0.09785065054893494, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 399.7657210826874}
91
+ {"step": 22740, "epoch": 0, "loss": 3.4585065841674805, "loss_main": 3.4585065841674805, "loss_zero": 0.0, "loss_en": 3.445830821990967, "loss_zh": 3.471182107925415, "noise_ramp": 1.0, "noise_mult": 0.6912064552307129, "noise_std_eff": 0.06912064552307129, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 402.93057107925415}
92
+ {"step": 22760, "epoch": 0, "loss": 3.3368513584136963, "loss_main": 3.3368513584136963, "loss_zero": 0.0, "loss_en": 3.35270357131958, "loss_zh": 3.3209991455078125, "noise_ramp": 1.0, "noise_mult": 0.9922390580177307, "noise_std_eff": 0.09922390580177308, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 406.1040368080139}
93
+ {"step": 22780, "epoch": 0, "loss": 3.6119484901428223, "loss_main": 3.6119484901428223, "loss_zero": 0.0, "loss_en": 3.62178635597229, "loss_zh": 3.6021108627319336, "noise_ramp": 1.0, "noise_mult": 0.7261579036712646, "noise_std_eff": 0.07261579036712647, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 409.2622928619385}
94
+ {"step": 22800, "epoch": 0, "loss": 3.6012520790100098, "loss_main": 3.6012520790100098, "loss_zero": 0.0, "loss_en": 3.5783329010009766, "loss_zh": 3.624171018600464, "noise_ramp": 1.0, "noise_mult": 0.8715773224830627, "noise_std_eff": 0.08715773224830628, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 412.4170138835907}
95
+ {"step": 22820, "epoch": 0, "loss": 3.5224344730377197, "loss_main": 3.5224344730377197, "loss_zero": 0.0, "loss_en": 3.5142369270324707, "loss_zh": 3.5306320190429688, "noise_ramp": 1.0, "noise_mult": 0.8924875259399414, "noise_std_eff": 0.08924875259399415, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 440.8631031513214}
96
+ {"step": 22840, "epoch": 0, "loss": 3.41786527633667, "loss_main": 3.41786527633667, "loss_zero": 0.0, "loss_en": 3.455263376235962, "loss_zh": 3.380467414855957, "noise_ramp": 1.0, "noise_mult": 0.48315104842185974, "noise_std_eff": 0.04831510484218598, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 444.0774097442627}
97
+ {"step": 22860, "epoch": 0, "loss": 4.29213285446167, "loss_main": 4.29213285446167, "loss_zero": 0.0, "loss_en": 4.295347213745117, "loss_zh": 4.288918495178223, "noise_ramp": 1.0, "noise_mult": 0.811528205871582, "noise_std_eff": 0.08115282058715821, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 447.2686507701874}
98
+ {"step": 22880, "epoch": 0, "loss": 2.9912872314453125, "loss_main": 2.9912872314453125, "loss_zero": 0.0, "loss_en": 2.9881162643432617, "loss_zh": 2.9944581985473633, "noise_ramp": 1.0, "noise_mult": 0.6395062208175659, "noise_std_eff": 0.0639506220817566, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 450.468847990036}
99
+ {"step": 22900, "epoch": 0, "loss": 3.167079448699951, "loss_main": 3.167079448699951, "loss_zero": 0.0, "loss_en": 3.1797821521759033, "loss_zh": 3.15437650680542, "noise_ramp": 1.0, "noise_mult": 0.5331701636314392, "noise_std_eff": 0.053317016363143926, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 453.6940221786499}
100
+ {"step": 22920, "epoch": 0, "loss": 3.414896011352539, "loss_main": 3.414896011352539, "loss_zero": 0.0, "loss_en": 3.4121975898742676, "loss_zh": 3.4175946712493896, "noise_ramp": 1.0, "noise_mult": 0.6085770726203918, "noise_std_eff": 0.060857707262039186, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 456.9115238189697}
101
+ {"step": 22940, "epoch": 0, "loss": 3.140206813812256, "loss_main": 3.140206813812256, "loss_zero": 0.0, "loss_en": 3.1028895378112793, "loss_zh": 3.1775243282318115, "noise_ramp": 1.0, "noise_mult": 0.9702861905097961, "noise_std_eff": 0.09702861905097962, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 460.1415436267853}
102
+ {"step": 22960, "epoch": 0, "loss": 3.896941661834717, "loss_main": 3.896941661834717, "loss_zero": 0.0, "loss_en": 3.855663299560547, "loss_zh": 3.9382197856903076, "noise_ramp": 1.0, "noise_mult": 0.639333963394165, "noise_std_eff": 0.06393339633941651, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 463.39481687545776}
103
+ {"step": 22980, "epoch": 0, "loss": 3.3219919204711914, "loss_main": 3.3219919204711914, "loss_zero": 0.0, "loss_en": 3.3237948417663574, "loss_zh": 3.3201887607574463, "noise_ramp": 1.0, "noise_mult": 0.5874664187431335, "noise_std_eff": 0.05874664187431336, "z_zero_prob": 0.1, "z_zero_frac": 0.5, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 466.85347294807434}
104
+ {"step": 23000, "epoch": 0, "loss": 3.1096763610839844, "loss_main": 3.1096763610839844, "loss_zero": 0.0, "loss_en": 3.1142051219940186, "loss_zh": 3.1051478385925293, "noise_ramp": 1.0, "noise_mult": 0.9013762474060059, "noise_std_eff": 0.0901376247406006, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 470.09418201446533}
105
+ {"step": 23020, "epoch": 0, "loss": 3.968672752380371, "loss_main": 3.968672752380371, "loss_zero": 0.0, "loss_en": 3.935173749923706, "loss_zh": 4.002171516418457, "noise_ramp": 1.0, "noise_mult": 0.4300421178340912, "noise_std_eff": 0.04300421178340912, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 473.3398311138153}
106
+ {"step": 23040, "epoch": 0, "loss": 2.84181809425354, "loss_main": 2.84181809425354, "loss_zero": 0.0, "loss_en": 2.848538637161255, "loss_zh": 2.835097551345825, "noise_ramp": 1.0, "noise_mult": 0.7302999496459961, "noise_std_eff": 0.07302999496459961, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 476.5456745624542}
107
+ {"step": 23060, "epoch": 0, "loss": 3.0474276542663574, "loss_main": 3.0474276542663574, "loss_zero": 0.0, "loss_en": 3.0707759857177734, "loss_zh": 3.0240790843963623, "noise_ramp": 1.0, "noise_mult": 0.9683941006660461, "noise_std_eff": 0.09683941006660463, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 479.6945822238922}
108
+ {"step": 23080, "epoch": 0, "loss": 3.533545970916748, "loss_main": 3.533545970916748, "loss_zero": 0.0, "loss_en": 3.5653891563415527, "loss_zh": 3.5017025470733643, "noise_ramp": 1.0, "noise_mult": 0.9683284759521484, "noise_std_eff": 0.09683284759521485, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 482.87398290634155}
109
+ {"step": 23100, "epoch": 0, "loss": 3.165747880935669, "loss_main": 3.165747880935669, "loss_zero": 0.0, "loss_en": 3.1776740550994873, "loss_zh": 3.1538217067718506, "noise_ramp": 1.0, "noise_mult": 0.82072514295578, "noise_std_eff": 0.082072514295578, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 486.1027913093567}
110
+ {"step": 23120, "epoch": 0, "loss": 3.437662124633789, "loss_main": 3.437662124633789, "loss_zero": 0.0, "loss_en": 3.399583578109741, "loss_zh": 3.475740671157837, "noise_ramp": 1.0, "noise_mult": 0.790022611618042, "noise_std_eff": 0.0790022611618042, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 489.44156193733215}
111
+ {"step": 23140, "epoch": 0, "loss": 3.4837162494659424, "loss_main": 3.4837162494659424, "loss_zero": 0.0, "loss_en": 3.4752557277679443, "loss_zh": 3.4921767711639404, "noise_ramp": 1.0, "noise_mult": 0.9215710759162903, "noise_std_eff": 0.09215710759162904, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 492.89036655426025}
112
+ {"step": 23160, "epoch": 0, "loss": 3.470722198486328, "loss_main": 3.470722198486328, "loss_zero": 0.0, "loss_en": 3.483288526535034, "loss_zh": 3.458155632019043, "noise_ramp": 1.0, "noise_mult": 0.6545602083206177, "noise_std_eff": 0.06545602083206177, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 496.1330986022949}
113
+ {"step": 23180, "epoch": 0, "loss": 3.5894052982330322, "loss_main": 3.5894052982330322, "loss_zero": 0.0, "loss_en": 3.600494384765625, "loss_zh": 3.5783162117004395, "noise_ramp": 1.0, "noise_mult": 0.6507309675216675, "noise_std_eff": 0.06507309675216676, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 499.32738280296326}
114
+ {"step": 23200, "epoch": 0, "loss": 3.3798813819885254, "loss_main": 3.3798813819885254, "loss_zero": 0.0, "loss_en": 3.375244140625, "loss_zh": 3.38451886177063, "noise_ramp": 1.0, "noise_mult": 0.5825142860412598, "noise_std_eff": 0.05825142860412598, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 502.51959013938904}
115
+ {"step": 23220, "epoch": 0, "loss": 3.1999740600585938, "loss_main": 3.1999740600585938, "loss_zero": 0.0, "loss_en": 3.239055633544922, "loss_zh": 3.1608927249908447, "noise_ramp": 1.0, "noise_mult": 0.9148839116096497, "noise_std_eff": 0.09148839116096497, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 532.6845471858978}
116
+ {"step": 23240, "epoch": 0, "loss": 3.415152072906494, "loss_main": 3.415152072906494, "loss_zero": 0.0, "loss_en": 3.4415688514709473, "loss_zh": 3.388735055923462, "noise_ramp": 1.0, "noise_mult": 0.6043016314506531, "noise_std_eff": 0.06043016314506531, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 536.2663342952728}
117
+ {"step": 23260, "epoch": 0, "loss": 3.794428825378418, "loss_main": 3.794428825378418, "loss_zero": 0.0, "loss_en": 3.8090505599975586, "loss_zh": 3.7798070907592773, "noise_ramp": 1.0, "noise_mult": 0.9164061546325684, "noise_std_eff": 0.09164061546325684, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 539.7669596672058}
118
+ {"step": 23280, "epoch": 0, "loss": 3.0801868438720703, "loss_main": 3.0801868438720703, "loss_zero": 0.0, "loss_en": 3.023973226547241, "loss_zh": 3.1364004611968994, "noise_ramp": 1.0, "noise_mult": 0.8584133386611938, "noise_std_eff": 0.08584133386611939, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 543.1427657604218}
119
+ {"step": 23300, "epoch": 0, "loss": 3.041928291320801, "loss_main": 3.041928291320801, "loss_zero": 0.0, "loss_en": 3.0110676288604736, "loss_zh": 3.072789192199707, "noise_ramp": 1.0, "noise_mult": 0.6397307515144348, "noise_std_eff": 0.06397307515144349, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 546.7315530776978}
120
+ {"step": 23320, "epoch": 0, "loss": 3.29050612449646, "loss_main": 3.29050612449646, "loss_zero": 0.0, "loss_en": 3.3148574829101562, "loss_zh": 3.2661547660827637, "noise_ramp": 1.0, "noise_mult": 0.449756920337677, "noise_std_eff": 0.0449756920337677, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 550.1380186080933}
121
+ {"step": 23340, "epoch": 0, "loss": 3.3957724571228027, "loss_main": 3.3957724571228027, "loss_zero": 0.0, "loss_en": 3.3599042892456055, "loss_zh": 3.431640625, "noise_ramp": 1.0, "noise_mult": 0.9445438981056213, "noise_std_eff": 0.09445438981056214, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 553.6199836730957}
122
+ {"step": 23360, "epoch": 0, "loss": 3.1909708976745605, "loss_main": 3.1909708976745605, "loss_zero": 0.0, "loss_en": 3.174015760421753, "loss_zh": 3.207925796508789, "noise_ramp": 1.0, "noise_mult": 0.850458562374115, "noise_std_eff": 0.08504585623741151, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 556.8494341373444}
123
+ {"step": 23380, "epoch": 0, "loss": 3.0152721405029297, "loss_main": 3.0152721405029297, "loss_zero": 0.0, "loss_en": 3.0288870334625244, "loss_zh": 3.001657247543335, "noise_ramp": 1.0, "noise_mult": 0.7184138298034668, "noise_std_eff": 0.07184138298034669, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 560.0301897525787}
124
+ {"step": 23400, "epoch": 0, "loss": 3.2541933059692383, "loss_main": 3.2541933059692383, "loss_zero": 0.0, "loss_en": 3.2527432441711426, "loss_zh": 3.255643367767334, "noise_ramp": 1.0, "noise_mult": 0.6606534123420715, "noise_std_eff": 0.06606534123420715, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 563.19757771492}
125
+ {"step": 23420, "epoch": 0, "loss": 3.1744394302368164, "loss_main": 3.1744394302368164, "loss_zero": 0.0, "loss_en": 3.173295497894287, "loss_zh": 3.1755833625793457, "noise_ramp": 1.0, "noise_mult": 0.7624515295028687, "noise_std_eff": 0.07624515295028687, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 566.3360040187836}
126
+ {"step": 23440, "epoch": 0, "loss": 3.0862631797790527, "loss_main": 3.0862631797790527, "loss_zero": 0.0, "loss_en": 3.0789737701416016, "loss_zh": 3.093552350997925, "noise_ramp": 1.0, "noise_mult": 0.9578949809074402, "noise_std_eff": 0.09578949809074402, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 569.5112965106964}
127
+ {"step": 23460, "epoch": 0, "loss": 2.8652586936950684, "loss_main": 2.8652586936950684, "loss_zero": 0.0, "loss_en": 2.8659396171569824, "loss_zh": 2.8645777702331543, "noise_ramp": 1.0, "noise_mult": 0.8936758041381836, "noise_std_eff": 0.08936758041381837, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 572.6905772686005}
128
+ {"step": 23480, "epoch": 0, "loss": 3.482895612716675, "loss_main": 3.482895612716675, "loss_zero": 0.0, "loss_en": 3.4677951335906982, "loss_zh": 3.4979960918426514, "noise_ramp": 1.0, "noise_mult": 0.3234814703464508, "noise_std_eff": 0.03234814703464508, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 575.8486065864563}
129
+ {"step": 23500, "epoch": 0, "loss": 3.516509771347046, "loss_main": 3.516509771347046, "loss_zero": 0.0, "loss_en": 3.522313117980957, "loss_zh": 3.5107064247131348, "noise_ramp": 1.0, "noise_mult": 0.9224568605422974, "noise_std_eff": 0.09224568605422974, "z_zero_prob": 0.1, "z_zero_frac": 0.5, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 578.9835343360901}
130
+ {"step": 23520, "epoch": 0, "loss": 2.990009307861328, "loss_main": 2.990009307861328, "loss_zero": 0.0, "loss_en": 2.971485137939453, "loss_zh": 3.0085337162017822, "noise_ramp": 1.0, "noise_mult": 0.8016607165336609, "noise_std_eff": 0.08016607165336609, "z_zero_prob": 0.1, "z_zero_frac": 0.5, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 582.1465728282928}
131
+ {"step": 23540, "epoch": 0, "loss": 3.4653468132019043, "loss_main": 3.4653468132019043, "loss_zero": 0.0, "loss_en": 3.4671995639801025, "loss_zh": 3.463494300842285, "noise_ramp": 1.0, "noise_mult": 0.5342270135879517, "noise_std_eff": 0.05342270135879517, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 585.3217499256134}
132
+ {"step": 23560, "epoch": 0, "loss": 3.3898653984069824, "loss_main": 3.3898653984069824, "loss_zero": 0.0, "loss_en": 3.356994152069092, "loss_zh": 3.422736883163452, "noise_ramp": 1.0, "noise_mult": 0.9466176629066467, "noise_std_eff": 0.09466176629066468, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 588.4770591259003}
133
+ {"step": 23580, "epoch": 0, "loss": 3.985912799835205, "loss_main": 3.985912799835205, "loss_zero": 0.0, "loss_en": 3.9668195247650146, "loss_zh": 4.005006313323975, "noise_ramp": 1.0, "noise_mult": 0.5329269170761108, "noise_std_eff": 0.053292691707611084, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 591.6180241107941}
134
+ {"step": 23600, "epoch": 0, "loss": 3.652456045150757, "loss_main": 3.652456045150757, "loss_zero": 0.0, "loss_en": 3.621485710144043, "loss_zh": 3.6834263801574707, "noise_ramp": 1.0, "noise_mult": 0.8720766305923462, "noise_std_eff": 0.08720766305923462, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 594.7676486968994}
135
+ {"step": 23620, "epoch": 0, "loss": 3.218717098236084, "loss_main": 3.218717098236084, "loss_zero": 0.0, "loss_en": 3.2185919284820557, "loss_zh": 3.218842029571533, "noise_ramp": 1.0, "noise_mult": 0.7849810719490051, "noise_std_eff": 0.07849810719490052, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 623.201990365982}
136
+ {"step": 23640, "epoch": 0, "loss": 3.0710062980651855, "loss_main": 3.0710062980651855, "loss_zero": 0.0, "loss_en": 3.068035840988159, "loss_zh": 3.073976755142212, "noise_ramp": 1.0, "noise_mult": 0.4536603093147278, "noise_std_eff": 0.04536603093147278, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 626.3586850166321}
137
+ {"step": 23660, "epoch": 0, "loss": 3.433985471725464, "loss_main": 3.433985471725464, "loss_zero": 0.0, "loss_en": 3.4480693340301514, "loss_zh": 3.4199016094207764, "noise_ramp": 1.0, "noise_mult": 0.9174969792366028, "noise_std_eff": 0.09174969792366028, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 629.5085210800171}
138
+ {"step": 23680, "epoch": 0, "loss": 3.7526025772094727, "loss_main": 3.7526025772094727, "loss_zero": 0.0, "loss_en": 3.742016077041626, "loss_zh": 3.7631893157958984, "noise_ramp": 1.0, "noise_mult": 0.6277160048484802, "noise_std_eff": 0.06277160048484802, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 632.6773912906647}
139
+ {"step": 23700, "epoch": 0, "loss": 3.635129451751709, "loss_main": 3.635129451751709, "loss_zero": 0.0, "loss_en": 3.6451120376586914, "loss_zh": 3.6251471042633057, "noise_ramp": 1.0, "noise_mult": 0.9574517011642456, "noise_std_eff": 0.09574517011642457, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 635.8820979595184}
140
+ {"step": 23720, "epoch": 0, "loss": 2.8428025245666504, "loss_main": 2.8428025245666504, "loss_zero": 0.0, "loss_en": 2.833507776260376, "loss_zh": 2.8520970344543457, "noise_ramp": 1.0, "noise_mult": 0.9527844190597534, "noise_std_eff": 0.09527844190597534, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 639.0850486755371}
141
+ {"step": 23740, "epoch": 0, "loss": 3.4362597465515137, "loss_main": 3.4362597465515137, "loss_zero": 0.0, "loss_en": 3.43861985206604, "loss_zh": 3.4338996410369873, "noise_ramp": 1.0, "noise_mult": 0.9087368249893188, "noise_std_eff": 0.0908736824989319, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 642.2567234039307}
142
+ {"step": 23760, "epoch": 0, "loss": 2.9452133178710938, "loss_main": 2.9452133178710938, "loss_zero": 0.0, "loss_en": 2.9654109477996826, "loss_zh": 2.925015687942505, "noise_ramp": 1.0, "noise_mult": 0.6588883399963379, "noise_std_eff": 0.06588883399963379, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 645.4757227897644}
143
+ {"step": 23780, "epoch": 0, "loss": 3.4627625942230225, "loss_main": 3.4627625942230225, "loss_zero": 0.0, "loss_en": 3.440406560897827, "loss_zh": 3.4851186275482178, "noise_ramp": 1.0, "noise_mult": 0.8783098459243774, "noise_std_eff": 0.08783098459243775, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 648.6678750514984}
144
+ {"step": 23800, "epoch": 0, "loss": 3.1660614013671875, "loss_main": 3.1660614013671875, "loss_zero": 0.0, "loss_en": 3.189927101135254, "loss_zh": 3.1421959400177, "noise_ramp": 1.0, "noise_mult": 0.5359422564506531, "noise_std_eff": 0.05359422564506531, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 651.9225373268127}
145
+ {"step": 23820, "epoch": 0, "loss": 3.093376636505127, "loss_main": 3.093376636505127, "loss_zero": 0.0, "loss_en": 3.0883665084838867, "loss_zh": 3.098386526107788, "noise_ramp": 1.0, "noise_mult": 0.6405761241912842, "noise_std_eff": 0.06405761241912843, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 655.102260351181}
146
+ {"step": 23840, "epoch": 0, "loss": 3.25927734375, "loss_main": 3.25927734375, "loss_zero": 0.0, "loss_en": 3.247246265411377, "loss_zh": 3.271308422088623, "noise_ramp": 1.0, "noise_mult": 0.8212312459945679, "noise_std_eff": 0.0821231245994568, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 658.3073592185974}
147
+ {"step": 23860, "epoch": 0, "loss": 3.3652708530426025, "loss_main": 3.3652708530426025, "loss_zero": 0.0, "loss_en": 3.358788013458252, "loss_zh": 3.371753692626953, "noise_ramp": 1.0, "noise_mult": 0.8377215266227722, "noise_std_eff": 0.08377215266227722, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 661.5303163528442}
148
+ {"step": 23880, "epoch": 0, "loss": 3.5810394287109375, "loss_main": 3.5810394287109375, "loss_zero": 0.0, "loss_en": 3.580179214477539, "loss_zh": 3.581899642944336, "noise_ramp": 1.0, "noise_mult": 0.534334123134613, "noise_std_eff": 0.053433412313461305, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 664.771411895752}
149
+ {"step": 23900, "epoch": 0, "loss": 3.43782901763916, "loss_main": 3.43782901763916, "loss_zero": 0.0, "loss_en": 3.463139533996582, "loss_zh": 3.412518262863159, "noise_ramp": 1.0, "noise_mult": 0.49483904242515564, "noise_std_eff": 0.049483904242515565, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 667.9771449565887}
150
+ {"step": 23920, "epoch": 0, "loss": 3.0435171127319336, "loss_main": 3.0435171127319336, "loss_zero": 0.0, "loss_en": 3.0628581047058105, "loss_zh": 3.0241761207580566, "noise_ramp": 1.0, "noise_mult": 0.9580323696136475, "noise_std_eff": 0.09580323696136475, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 671.2001805305481}
151
+ {"step": 23940, "epoch": 0, "loss": 3.3355512619018555, "loss_main": 3.3355512619018555, "loss_zero": 0.0, "loss_en": 3.324084997177124, "loss_zh": 3.347017765045166, "noise_ramp": 1.0, "noise_mult": 0.5014905333518982, "noise_std_eff": 0.05014905333518982, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 674.3989458084106}
152
+ {"step": 23960, "epoch": 0, "loss": 3.4915997982025146, "loss_main": 3.4915997982025146, "loss_zero": 0.0, "loss_en": 3.4962165355682373, "loss_zh": 3.486983060836792, "noise_ramp": 1.0, "noise_mult": 0.9905045628547668, "noise_std_eff": 0.09905045628547669, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 677.5985760688782}
153
+ {"step": 23980, "epoch": 0, "loss": 3.2540950775146484, "loss_main": 3.2540950775146484, "loss_zero": 0.0, "loss_en": 3.234405994415283, "loss_zh": 3.2737843990325928, "noise_ramp": 1.0, "noise_mult": 0.7709509134292603, "noise_std_eff": 0.07709509134292603, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 680.7882590293884}
154
+ {"step": 24000, "epoch": 0, "loss": 3.3381974697113037, "loss_main": 3.3381974697113037, "loss_zero": 0.0, "loss_en": 3.332292318344116, "loss_zh": 3.344102621078491, "noise_ramp": 1.0, "noise_mult": 0.6387982368469238, "noise_std_eff": 0.06387982368469239, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 683.9822046756744}
155
+ {"step": 24020, "epoch": 0, "loss": 3.147779941558838, "loss_main": 3.147779941558838, "loss_zero": 0.0, "loss_en": 3.1542680263519287, "loss_zh": 3.141291618347168, "noise_ramp": 1.0, "noise_mult": 0.9558422565460205, "noise_std_eff": 0.09558422565460206, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 712.0714731216431}
156
+ {"step": 24040, "epoch": 0, "loss": 3.1935741901397705, "loss_main": 3.1935741901397705, "loss_zero": 0.0, "loss_en": 3.1726694107055664, "loss_zh": 3.2144789695739746, "noise_ramp": 1.0, "noise_mult": 0.879844605922699, "noise_std_eff": 0.0879844605922699, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 715.2495114803314}
157
+ {"step": 24060, "epoch": 0, "loss": 3.4880833625793457, "loss_main": 3.4880833625793457, "loss_zero": 0.0, "loss_en": 3.489882707595825, "loss_zh": 3.4862842559814453, "noise_ramp": 1.0, "noise_mult": 0.7474002242088318, "noise_std_eff": 0.07474002242088318, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 718.4004056453705}
158
+ {"step": 24080, "epoch": 0, "loss": 3.327589988708496, "loss_main": 3.327589988708496, "loss_zero": 0.0, "loss_en": 3.2670297622680664, "loss_zh": 3.3881499767303467, "noise_ramp": 1.0, "noise_mult": 0.7452157139778137, "noise_std_eff": 0.07452157139778137, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 721.5492136478424}
159
+ {"step": 24100, "epoch": 0, "loss": 3.4973912239074707, "loss_main": 3.4973912239074707, "loss_zero": 0.0, "loss_en": 3.488185405731201, "loss_zh": 3.5065970420837402, "noise_ramp": 1.0, "noise_mult": 0.8264695405960083, "noise_std_eff": 0.08264695405960083, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 724.6895308494568}
160
+ {"step": 24120, "epoch": 0, "loss": 2.955440044403076, "loss_main": 2.955440044403076, "loss_zero": 0.0, "loss_en": 2.9658355712890625, "loss_zh": 2.9450442790985107, "noise_ramp": 1.0, "noise_mult": 0.4690941572189331, "noise_std_eff": 0.04690941572189331, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 727.8637335300446}
161
+ {"step": 24140, "epoch": 0, "loss": 3.4242494106292725, "loss_main": 3.4242494106292725, "loss_zero": 0.0, "loss_en": 3.455263614654541, "loss_zh": 3.393235206604004, "noise_ramp": 1.0, "noise_mult": 0.853327751159668, "noise_std_eff": 0.0853327751159668, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 731.0222110748291}
162
+ {"step": 24160, "epoch": 0, "loss": 2.7727432250976562, "loss_main": 2.7727432250976562, "loss_zero": 0.0, "loss_en": 2.757124423980713, "loss_zh": 2.7883622646331787, "noise_ramp": 1.0, "noise_mult": 0.8953567743301392, "noise_std_eff": 0.08953567743301392, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 734.2486295700073}
163
+ {"step": 24180, "epoch": 0, "loss": 2.924619197845459, "loss_main": 2.924619197845459, "loss_zero": 0.0, "loss_en": 2.914018154144287, "loss_zh": 2.93522047996521, "noise_ramp": 1.0, "noise_mult": 0.7388518452644348, "noise_std_eff": 0.07388518452644348, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 737.3964273929596}
164
+ {"step": 24200, "epoch": 0, "loss": 3.4545092582702637, "loss_main": 3.4545092582702637, "loss_zero": 0.0, "loss_en": 3.429903507232666, "loss_zh": 3.4791150093078613, "noise_ramp": 1.0, "noise_mult": 0.7497230768203735, "noise_std_eff": 0.07497230768203736, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 740.5848820209503}
165
+ {"step": 24220, "epoch": 0, "loss": 3.794590473175049, "loss_main": 3.794590473175049, "loss_zero": 0.0, "loss_en": 3.796346426010132, "loss_zh": 3.792834520339966, "noise_ramp": 1.0, "noise_mult": 0.7213488221168518, "noise_std_eff": 0.07213488221168518, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 743.7302417755127}
166
+ {"step": 24240, "epoch": 0, "loss": 3.3279075622558594, "loss_main": 3.3279075622558594, "loss_zero": 0.0, "loss_en": 3.334239959716797, "loss_zh": 3.3215749263763428, "noise_ramp": 1.0, "noise_mult": 0.9870367050170898, "noise_std_eff": 0.09870367050170899, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 746.9050650596619}
167
+ {"step": 24260, "epoch": 0, "loss": 2.975924015045166, "loss_main": 2.975924015045166, "loss_zero": 0.0, "loss_en": 2.9847302436828613, "loss_zh": 2.9671177864074707, "noise_ramp": 1.0, "noise_mult": 0.7159223556518555, "noise_std_eff": 0.07159223556518556, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 750.0628850460052}
168
+ {"step": 24280, "epoch": 0, "loss": 3.4677140712738037, "loss_main": 3.4677140712738037, "loss_zero": 0.0, "loss_en": 3.463606119155884, "loss_zh": 3.4718220233917236, "noise_ramp": 1.0, "noise_mult": 0.8400489091873169, "noise_std_eff": 0.0840048909187317, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 753.2196667194366}
169
+ {"step": 24300, "epoch": 0, "loss": 3.4649813175201416, "loss_main": 3.4649813175201416, "loss_zero": 0.0, "loss_en": 3.456033229827881, "loss_zh": 3.4739294052124023, "noise_ramp": 1.0, "noise_mult": 0.9431917071342468, "noise_std_eff": 0.09431917071342469, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 756.322515964508}
170
+ {"step": 24320, "epoch": 0, "loss": 3.337216854095459, "loss_main": 3.337216854095459, "loss_zero": 0.0, "loss_en": 3.3493688106536865, "loss_zh": 3.3250646591186523, "noise_ramp": 1.0, "noise_mult": 0.910603404045105, "noise_std_eff": 0.0910603404045105, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 759.4582204818726}
171
+ {"step": 24340, "epoch": 0, "loss": 3.1535606384277344, "loss_main": 3.1535606384277344, "loss_zero": 0.0, "loss_en": 3.1392621994018555, "loss_zh": 3.1678593158721924, "noise_ramp": 1.0, "noise_mult": 0.9942603707313538, "noise_std_eff": 0.09942603707313538, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 762.6335773468018}
172
+ {"step": 24360, "epoch": 0, "loss": 3.2989699840545654, "loss_main": 3.2989699840545654, "loss_zero": 0.0, "loss_en": 3.3039073944091797, "loss_zh": 3.294032573699951, "noise_ramp": 1.0, "noise_mult": 0.7293740510940552, "noise_std_eff": 0.07293740510940552, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 766.0385258197784}
173
+ {"step": 24380, "epoch": 0, "loss": 3.1925575733184814, "loss_main": 3.1925575733184814, "loss_zero": 0.0, "loss_en": 3.171149730682373, "loss_zh": 3.21396541595459, "noise_ramp": 1.0, "noise_mult": 0.6982247829437256, "noise_std_eff": 0.06982247829437256, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 769.3257637023926}
174
+ {"step": 24400, "epoch": 0, "loss": 3.5352072715759277, "loss_main": 3.5352072715759277, "loss_zero": 0.0, "loss_en": 3.553424835205078, "loss_zh": 3.5169897079467773, "noise_ramp": 1.0, "noise_mult": 0.5750742554664612, "noise_std_eff": 0.05750742554664612, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 772.6719534397125}
175
+ {"step": 24420, "epoch": 0, "loss": 3.1582393646240234, "loss_main": 3.1582393646240234, "loss_zero": 0.0, "loss_en": 3.1479504108428955, "loss_zh": 3.1685280799865723, "noise_ramp": 1.0, "noise_mult": 0.6754070520401001, "noise_std_eff": 0.06754070520401001, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 801.0177400112152}
176
+ {"step": 24440, "epoch": 0, "loss": 3.7720651626586914, "loss_main": 3.7720651626586914, "loss_zero": 0.0, "loss_en": 3.774595022201538, "loss_zh": 3.7695353031158447, "noise_ramp": 1.0, "noise_mult": 0.892916202545166, "noise_std_eff": 0.0892916202545166, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 804.2394461631775}
177
+ {"step": 24460, "epoch": 0, "loss": 3.183051586151123, "loss_main": 3.183051586151123, "loss_zero": 0.0, "loss_en": 3.174715518951416, "loss_zh": 3.191387414932251, "noise_ramp": 1.0, "noise_mult": 0.6798707246780396, "noise_std_eff": 0.06798707246780396, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 807.4125242233276}
178
+ {"step": 24480, "epoch": 0, "loss": 3.54618501663208, "loss_main": 3.54618501663208, "loss_zero": 0.0, "loss_en": 3.5687575340270996, "loss_zh": 3.5236122608184814, "noise_ramp": 1.0, "noise_mult": 0.8962011337280273, "noise_std_eff": 0.08962011337280273, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 810.6063241958618}
179
+ {"step": 24500, "epoch": 0, "loss": 3.23215913772583, "loss_main": 3.23215913772583, "loss_zero": 0.0, "loss_en": 3.221224784851074, "loss_zh": 3.243093490600586, "noise_ramp": 1.0, "noise_mult": 0.7503457069396973, "noise_std_eff": 0.07503457069396974, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 813.8309652805328}
180
+ {"step": 24520, "epoch": 0, "loss": 3.284402370452881, "loss_main": 3.284402370452881, "loss_zero": 0.0, "loss_en": 3.2882792949676514, "loss_zh": 3.2805256843566895, "noise_ramp": 1.0, "noise_mult": 0.6605651378631592, "noise_std_eff": 0.06605651378631593, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 817.0578665733337}
181
+ {"step": 24540, "epoch": 0, "loss": 3.5295958518981934, "loss_main": 3.5295958518981934, "loss_zero": 0.0, "loss_en": 3.5113930702209473, "loss_zh": 3.5477986335754395, "noise_ramp": 1.0, "noise_mult": 0.9607135057449341, "noise_std_eff": 0.09607135057449341, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 820.256395816803}
182
+ {"step": 24560, "epoch": 0, "loss": 3.510626792907715, "loss_main": 3.510626792907715, "loss_zero": 0.0, "loss_en": 3.5164289474487305, "loss_zh": 3.5048248767852783, "noise_ramp": 1.0, "noise_mult": 0.6389675736427307, "noise_std_eff": 0.06389675736427307, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 823.4749691486359}
183
+ {"step": 24580, "epoch": 0, "loss": 3.4755449295043945, "loss_main": 3.4755449295043945, "loss_zero": 0.0, "loss_en": 3.4493398666381836, "loss_zh": 3.5017497539520264, "noise_ramp": 1.0, "noise_mult": 0.9782859086990356, "noise_std_eff": 0.09782859086990357, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 826.65407538414}
184
+ {"step": 24600, "epoch": 0, "loss": 2.9173154830932617, "loss_main": 2.9173154830932617, "loss_zero": 0.0, "loss_en": 2.9112799167633057, "loss_zh": 2.923351287841797, "noise_ramp": 1.0, "noise_mult": 0.4624183773994446, "noise_std_eff": 0.04624183773994446, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 829.8720328807831}
185
+ {"step": 24620, "epoch": 0, "loss": 2.893178939819336, "loss_main": 2.893178939819336, "loss_zero": 0.0, "loss_en": 2.864443302154541, "loss_zh": 2.9219143390655518, "noise_ramp": 1.0, "noise_mult": 0.9337574243545532, "noise_std_eff": 0.09337574243545532, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 833.0925414562225}
186
+ {"step": 24640, "epoch": 0, "loss": 3.3755760192871094, "loss_main": 3.3755760192871094, "loss_zero": 0.0, "loss_en": 3.376265048980713, "loss_zh": 3.374886989593506, "noise_ramp": 1.0, "noise_mult": 0.8238125443458557, "noise_std_eff": 0.08238125443458558, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 836.2916479110718}
187
+ {"step": 24660, "epoch": 0, "loss": 3.517399787902832, "loss_main": 3.517399787902832, "loss_zero": 0.0, "loss_en": 3.522636890411377, "loss_zh": 3.512162446975708, "noise_ramp": 1.0, "noise_mult": 0.8709471225738525, "noise_std_eff": 0.08709471225738526, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 839.4994583129883}
188
+ {"step": 24680, "epoch": 0, "loss": 2.775587797164917, "loss_main": 2.775587797164917, "loss_zero": 0.0, "loss_en": 2.773350954055786, "loss_zh": 2.777824640274048, "noise_ramp": 1.0, "noise_mult": 0.5369182229042053, "noise_std_eff": 0.053691822290420535, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 842.7308011054993}
189
+ {"step": 24700, "epoch": 0, "loss": 3.5236167907714844, "loss_main": 3.5236167907714844, "loss_zero": 0.0, "loss_en": 3.524883508682251, "loss_zh": 3.522350311279297, "noise_ramp": 1.0, "noise_mult": 0.9230780005455017, "noise_std_eff": 0.09230780005455018, "z_zero_prob": 0.1, "z_zero_frac": 0.5, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 845.9211158752441}
190
+ {"step": 24720, "epoch": 0, "loss": 3.073603868484497, "loss_main": 3.073603868484497, "loss_zero": 0.0, "loss_en": 3.0679266452789307, "loss_zh": 3.0792810916900635, "noise_ramp": 1.0, "noise_mult": 0.5777028799057007, "noise_std_eff": 0.057770287990570074, "z_zero_prob": 0.1, "z_zero_frac": 0.5, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 849.1322600841522}
191
+ {"step": 24740, "epoch": 0, "loss": 3.260488271713257, "loss_main": 3.260488271713257, "loss_zero": 0.0, "loss_en": 3.2682435512542725, "loss_zh": 3.252732992172241, "noise_ramp": 1.0, "noise_mult": 0.7685015201568604, "noise_std_eff": 0.07685015201568604, "z_zero_prob": 0.1, "z_zero_frac": 0.5, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 852.2957165241241}
192
+ {"step": 24760, "epoch": 0, "loss": 3.243256092071533, "loss_main": 3.243256092071533, "loss_zero": 0.0, "loss_en": 3.2478902339935303, "loss_zh": 3.238621950149536, "noise_ramp": 1.0, "noise_mult": 0.9025610089302063, "noise_std_eff": 0.09025610089302064, "z_zero_prob": 0.1, "z_zero_frac": 0.5, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 855.4940237998962}
193
+ {"step": 24780, "epoch": 0, "loss": 3.4318952560424805, "loss_main": 3.4318952560424805, "loss_zero": 0.0, "loss_en": 3.4167420864105225, "loss_zh": 3.4470484256744385, "noise_ramp": 1.0, "noise_mult": 0.9755181074142456, "noise_std_eff": 0.09755181074142456, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 858.6367032527924}
194
+ {"step": 24800, "epoch": 0, "loss": 3.48720121383667, "loss_main": 3.48720121383667, "loss_zero": 0.0, "loss_en": 3.4923253059387207, "loss_zh": 3.48207688331604, "noise_ramp": 1.0, "noise_mult": 0.9950157999992371, "noise_std_eff": 0.0995015799999237, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 861.8271591663361}
195
+ {"step": 24820, "epoch": 0, "loss": 3.294076442718506, "loss_main": 3.294076442718506, "loss_zero": 0.0, "loss_en": 3.3110337257385254, "loss_zh": 3.2771193981170654, "noise_ramp": 1.0, "noise_mult": 0.9289309978485107, "noise_std_eff": 0.09289309978485108, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 890.0375633239746}
196
+ {"step": 24840, "epoch": 0, "loss": 3.150265693664551, "loss_main": 3.150265693664551, "loss_zero": 0.0, "loss_en": 3.159245014190674, "loss_zh": 3.141286611557007, "noise_ramp": 1.0, "noise_mult": 0.7910135388374329, "noise_std_eff": 0.07910135388374329, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 893.2775406837463}
197
+ {"step": 24860, "epoch": 0, "loss": 3.5842621326446533, "loss_main": 3.5842621326446533, "loss_zero": 0.0, "loss_en": 3.5637271404266357, "loss_zh": 3.604797124862671, "noise_ramp": 1.0, "noise_mult": 0.6699191331863403, "noise_std_eff": 0.06699191331863404, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 896.4913330078125}
198
+ {"step": 24880, "epoch": 0, "loss": 3.2882869243621826, "loss_main": 3.2882869243621826, "loss_zero": 0.0, "loss_en": 3.29468035697937, "loss_zh": 3.281893491744995, "noise_ramp": 1.0, "noise_mult": 0.6186308264732361, "noise_std_eff": 0.06186308264732361, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 899.7295567989349}
199
+ {"step": 24900, "epoch": 0, "loss": 3.366401195526123, "loss_main": 3.366401195526123, "loss_zero": 0.0, "loss_en": 3.3670589923858643, "loss_zh": 3.3657431602478027, "noise_ramp": 1.0, "noise_mult": 0.3458424210548401, "noise_std_eff": 0.03458424210548401, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 902.9915177822113}
200
+ {"step": 24920, "epoch": 0, "loss": 2.8134326934814453, "loss_main": 2.8134326934814453, "loss_zero": 0.0, "loss_en": 2.7983429431915283, "loss_zh": 2.8285226821899414, "noise_ramp": 1.0, "noise_mult": 0.40156078338623047, "noise_std_eff": 0.04015607833862305, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 906.2855637073517}
201
+ {"step": 24940, "epoch": 0, "loss": 3.7265305519104004, "loss_main": 3.7265305519104004, "loss_zero": 0.0, "loss_en": 3.720446825027466, "loss_zh": 3.732614278793335, "noise_ramp": 1.0, "noise_mult": 0.8767857551574707, "noise_std_eff": 0.08767857551574708, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 909.5202181339264}
202
+ {"step": 24960, "epoch": 0, "loss": 3.1482813358306885, "loss_main": 3.1482813358306885, "loss_zero": 0.0, "loss_en": 3.149423122406006, "loss_zh": 3.147139549255371, "noise_ramp": 1.0, "noise_mult": 0.8751060366630554, "noise_std_eff": 0.08751060366630554, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 912.7945420742035}
203
+ {"step": 24980, "epoch": 0, "loss": 3.353475332260132, "loss_main": 3.353475332260132, "loss_zero": 0.0, "loss_en": 3.3355088233947754, "loss_zh": 3.3714418411254883, "noise_ramp": 1.0, "noise_mult": 0.9434627890586853, "noise_std_eff": 0.09434627890586854, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 916.0504076480865}
204
+ {"step": 25000, "epoch": 0, "loss": 3.4673824310302734, "loss_main": 3.4673824310302734, "loss_zero": 0.0, "loss_en": 3.4646918773651123, "loss_zh": 3.4700732231140137, "noise_ramp": 1.0, "noise_mult": 0.830489456653595, "noise_std_eff": 0.0830489456653595, "z_zero_prob": 0.1, "z_zero_frac": 0.5, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 919.2937912940979}
205
+ {"step": 25020, "epoch": 0, "loss": 2.8984994888305664, "loss_main": 2.8984994888305664, "loss_zero": 0.0, "loss_en": 2.922201156616211, "loss_zh": 2.8747975826263428, "noise_ramp": 1.0, "noise_mult": 0.6879863142967224, "noise_std_eff": 0.06879863142967224, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 922.5093760490417}
206
+ {"step": 25040, "epoch": 0, "loss": 2.876084089279175, "loss_main": 2.876084089279175, "loss_zero": 0.0, "loss_en": 2.8885810375213623, "loss_zh": 2.8635871410369873, "noise_ramp": 1.0, "noise_mult": 0.8698585629463196, "noise_std_eff": 0.08698585629463196, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 925.7271449565887}
207
+ {"step": 25060, "epoch": 0, "loss": 3.4724924564361572, "loss_main": 3.4724924564361572, "loss_zero": 0.0, "loss_en": 3.4692258834838867, "loss_zh": 3.4757590293884277, "noise_ramp": 1.0, "noise_mult": 0.877430260181427, "noise_std_eff": 0.08774302601814271, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 928.8787336349487}
208
+ {"step": 25080, "epoch": 0, "loss": 3.6664257049560547, "loss_main": 3.6664257049560547, "loss_zero": 0.0, "loss_en": 3.6895408630371094, "loss_zh": 3.643310308456421, "noise_ramp": 1.0, "noise_mult": 0.6971722841262817, "noise_std_eff": 0.06971722841262817, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 932.087886095047}
209
+ {"step": 25100, "epoch": 0, "loss": 3.4940614700317383, "loss_main": 3.4940614700317383, "loss_zero": 0.0, "loss_en": 3.505350112915039, "loss_zh": 3.4827728271484375, "noise_ramp": 1.0, "noise_mult": 0.37197837233543396, "noise_std_eff": 0.0371978372335434, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 935.3198013305664}
210
+ {"step": 25120, "epoch": 0, "loss": 2.7731027603149414, "loss_main": 2.7731027603149414, "loss_zero": 0.0, "loss_en": 2.7951607704162598, "loss_zh": 2.751044988632202, "noise_ramp": 1.0, "noise_mult": 0.8147488236427307, "noise_std_eff": 0.08147488236427308, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 938.5295042991638}
211
+ {"step": 25140, "epoch": 0, "loss": 3.0608386993408203, "loss_main": 3.0608386993408203, "loss_zero": 0.0, "loss_en": 3.044036626815796, "loss_zh": 3.0776407718658447, "noise_ramp": 1.0, "noise_mult": 0.9365516901016235, "noise_std_eff": 0.09365516901016235, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 941.7268822193146}
212
+ {"step": 25160, "epoch": 0, "loss": 3.411010265350342, "loss_main": 3.411010265350342, "loss_zero": 0.0, "loss_en": 3.432868242263794, "loss_zh": 3.3891525268554688, "noise_ramp": 1.0, "noise_mult": 0.9368941187858582, "noise_std_eff": 0.09368941187858582, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 944.9337575435638}
213
+ {"step": 25180, "epoch": 0, "loss": 3.5255560874938965, "loss_main": 3.5255560874938965, "loss_zero": 0.0, "loss_en": 3.4987874031066895, "loss_zh": 3.5523250102996826, "noise_ramp": 1.0, "noise_mult": 0.5977336764335632, "noise_std_eff": 0.05977336764335633, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 948.112007856369}
214
+ {"step": 25200, "epoch": 0, "loss": 3.4804110527038574, "loss_main": 3.4804110527038574, "loss_zero": 0.0, "loss_en": 3.503249168395996, "loss_zh": 3.4575726985931396, "noise_ramp": 1.0, "noise_mult": 0.5656877756118774, "noise_std_eff": 0.05656877756118775, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 951.3170819282532}
215
+ {"step": 25220, "epoch": 0, "loss": 3.0217173099517822, "loss_main": 3.0217173099517822, "loss_zero": 0.0, "loss_en": 3.0210375785827637, "loss_zh": 3.022397041320801, "noise_ramp": 1.0, "noise_mult": 0.7106748819351196, "noise_std_eff": 0.07106748819351197, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 980.2659814357758}
216
+ {"step": 25240, "epoch": 0, "loss": 3.1885175704956055, "loss_main": 3.1885175704956055, "loss_zero": 0.0, "loss_en": 3.189741611480713, "loss_zh": 3.187293767929077, "noise_ramp": 1.0, "noise_mult": 0.7342385053634644, "noise_std_eff": 0.07342385053634644, "z_zero_prob": 0.1, "z_zero_frac": 0.5, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 983.5931489467621}
217
+ {"step": 25260, "epoch": 0, "loss": 3.740589141845703, "loss_main": 3.740589141845703, "loss_zero": 0.0, "loss_en": 3.7829737663269043, "loss_zh": 3.698204755783081, "noise_ramp": 1.0, "noise_mult": 0.8289331793785095, "noise_std_eff": 0.08289331793785096, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 986.7553017139435}
218
+ {"step": 25280, "epoch": 0, "loss": 2.9544882774353027, "loss_main": 2.9544882774353027, "loss_zero": 0.0, "loss_en": 2.9584670066833496, "loss_zh": 2.950509548187256, "noise_ramp": 1.0, "noise_mult": 0.8880692720413208, "noise_std_eff": 0.08880692720413208, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 989.9015955924988}
219
+ {"step": 25300, "epoch": 0, "loss": 3.3207526206970215, "loss_main": 3.3207526206970215, "loss_zero": 0.0, "loss_en": 3.3330163955688477, "loss_zh": 3.3084888458251953, "noise_ramp": 1.0, "noise_mult": 0.8195693492889404, "noise_std_eff": 0.08195693492889405, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 993.0225133895874}
220
+ {"step": 25320, "epoch": 0, "loss": 3.4083569049835205, "loss_main": 3.4083569049835205, "loss_zero": 0.0, "loss_en": 3.4234957695007324, "loss_zh": 3.3932180404663086, "noise_ramp": 1.0, "noise_mult": 0.7420628070831299, "noise_std_eff": 0.07420628070831299, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 996.1812708377838}
221
+ {"step": 25340, "epoch": 0, "loss": 3.2542824745178223, "loss_main": 3.2542824745178223, "loss_zero": 0.0, "loss_en": 3.229459047317505, "loss_zh": 3.2791061401367188, "noise_ramp": 1.0, "noise_mult": 0.7539697885513306, "noise_std_eff": 0.07539697885513306, "z_zero_prob": 0.1, "z_zero_frac": 0.25, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 999.3025257587433}
222
+ {"step": 25360, "epoch": 0, "loss": 3.2072644233703613, "loss_main": 3.2072644233703613, "loss_zero": 0.0, "loss_en": 3.228450059890747, "loss_zh": 3.1860787868499756, "noise_ramp": 1.0, "noise_mult": 0.5329388976097107, "noise_std_eff": 0.053293889760971075, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 1002.4729702472687}
223
+ {"step": 25380, "epoch": 0, "loss": 3.9240353107452393, "loss_main": 3.9240353107452393, "loss_zero": 0.0, "loss_en": 3.878499746322632, "loss_zh": 3.9695708751678467, "noise_ramp": 1.0, "noise_mult": 0.5806176662445068, "noise_std_eff": 0.058061766624450686, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 1005.6484959125519}
224
+ {"step": 25400, "epoch": 0, "loss": 2.715780735015869, "loss_main": 2.715780735015869, "loss_zero": 0.0, "loss_en": 2.744529962539673, "loss_zh": 2.6870317459106445, "noise_ramp": 1.0, "noise_mult": 0.6807522773742676, "noise_std_eff": 0.06807522773742676, "z_zero_prob": 0.1, "z_zero_frac": 0.0, "z_zero_loss_weight": 0.0, "z_zero_loss_every": 1, "elapsed_sec": 1008.8126285076141}
planB_polish_robust_dec_r2k_zprob01/probe_leakage.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ckpt": "runs/planB_polish_robust_dec_r2k_zprob01/ckpt.pt",
3
+ "max_train_groups": 8000,
4
+ "max_valid_groups": 946,
5
+ "probe_hidden": 256,
6
+ "epochs": 20,
7
+ "lr": 0.01,
8
+ "baseline_majority_lang_acc": 0.5,
9
+ "baseline_majority_len_acc": 0.9271249771118164,
10
+ "lang_probe": {
11
+ "train_acc": 0.5845,
12
+ "valid_acc": 0.5835095137420718,
13
+ "best_valid_acc": 0.6025369978858351
14
+ },
15
+ "len_probe": {
16
+ "train_acc": 0.928875,
17
+ "valid_acc": 0.9011627906976745,
18
+ "best_valid_acc": 0.9011627906976745
19
+ },
20
+ "lang_probe_length_matched": {
21
+ "train_acc": 0.591983556012333,
22
+ "valid_acc": 0.5948180815876516,
23
+ "best_valid_acc": 0.5948180815876516
24
+ },
25
+ "lang_bucketwise_valid_acc": {
26
+ "bucket_0_acc": 0.375,
27
+ "bucket_1_acc": 0.675000011920929,
28
+ "bucket_2_acc": 0.5826086401939392,
29
+ "bucket_3_acc": 0.5853372812271118
30
+ },
31
+ "length_matched_sizes": {
32
+ "train": 15568,
33
+ "valid": 1814
34
+ }
35
+ }