| { |
| "run_name": "run_large_20251112_150502", |
| "timestamp": "20251112_150502", |
| "phase": "large", |
| "config": { |
| "arch_layout": [ |
| "m4", |
| [ |
| "T22" |
| ], |
| "m4" |
| ], |
| "d_model": [ |
| 1024, |
| 1536 |
| ], |
| "d_intermediate": [ |
| 0, |
| 4096 |
| ], |
| "vocab_size": 256, |
| "ssm_cfg": { |
| "chunk_size": 256, |
| "d_conv": 4, |
| "d_state": 128, |
| "expand": 2 |
| }, |
| "attn_cfg": { |
| "num_heads": [ |
| 16, |
| 16 |
| ], |
| "rotary_emb_dim": [ |
| 32, |
| 48 |
| ], |
| "window_size": [ |
| 1023, |
| -1 |
| ] |
| }, |
| "tie_embeddings": false |
| }, |
| "training_args": { |
| "data": "datasets/PI1M/PI1M_v2.csv", |
| "max_samples": null, |
| "batch_size": 16, |
| "epochs": 22, |
| "lr": 0.0001, |
| "weight_decay": 0.1, |
| "gradient_accumulation": 8, |
| "concatenate": true, |
| "num_concatenate": 10, |
| "concatenate_separator": " ", |
| "checkpoint_bytes": 1000000, |
| "num_test_samples": 5, |
| "num_visualize": 5, |
| "skip_visualization": false |
| }, |
| "dataset_info": { |
| "train_size": 99574, |
| "test_size": 5, |
| "test_smiles_file": "checkpoints/run_large_20251112_150502/test_smiles.txt" |
| }, |
| "model_info": { |
| "num_parameters": 622923776, |
| "device": "cuda", |
| "dtype": "torch.bfloat16", |
| "use_amp": true |
| }, |
| "training_history": [ |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1000000, |
| "cumulative_training_bytes": 1006216, |
| "metrics": { |
| "loss": 2.65240060560631, |
| "ce_loss": 2.6424005681818183, |
| "lb_loss": 0.9999999900658926 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 2000000, |
| "cumulative_training_bytes": 2003501, |
| "metrics": { |
| "loss": 1.91590709904678, |
| "ce_loss": 1.9059070849236641, |
| "lb_loss": 0.9999999920375474 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 3000000, |
| "cumulative_training_bytes": 3003412, |
| "metrics": { |
| "loss": 1.5786327193408218, |
| "ce_loss": 1.5686327131043256, |
| "lb_loss": 0.9999999945400325 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 4000000, |
| "cumulative_training_bytes": 4003745, |
| "metrics": { |
| "loss": 1.373363253724484, |
| "ce_loss": 1.3633632514312977, |
| "lb_loss": 0.9999999956775257 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 5000000, |
| "cumulative_training_bytes": 5000057, |
| "metrics": { |
| "loss": 1.2337313646205703, |
| "ce_loss": 1.223731364678899, |
| "lb_loss": 0.9999999960810402 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 6000000, |
| "cumulative_training_bytes": 6004365, |
| "metrics": { |
| "loss": 1.1314371002707513, |
| "ce_loss": 1.121437101910828, |
| "lb_loss": 0.9999999965831732 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 7000000, |
| "cumulative_training_bytes": 7005346, |
| "metrics": { |
| "loss": 1.0538173757786313, |
| "ce_loss": 1.043817378548035, |
| "lb_loss": 0.999999995575201 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 8000000, |
| "cumulative_training_bytes": 8001141, |
| "metrics": { |
| "loss": 0.9934003011218449, |
| "ce_loss": 0.9834003047323135, |
| "lb_loss": 0.9999999950424435 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 9000000, |
| "cumulative_training_bytes": 9001157, |
| "metrics": { |
| "loss": 0.944607644944041, |
| "ce_loss": 0.9346076492141037, |
| "lb_loss": 0.999999994632037 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 10000000, |
| "cumulative_training_bytes": 10003863, |
| "metrics": { |
| "loss": 0.9043117426221888, |
| "ce_loss": 0.8943117474197247, |
| "lb_loss": 0.9999999956253471 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 11000000, |
| "cumulative_training_bytes": 11001271, |
| "metrics": { |
| "loss": 0.8709831310744411, |
| "ce_loss": 0.8609831363004172, |
| "lb_loss": 0.9999999957306827 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 12000000, |
| "cumulative_training_bytes": 12005542, |
| "metrics": { |
| "loss": 0.8426262691113658, |
| "ce_loss": 0.8326262746972594, |
| "lb_loss": 0.9999999964670288 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 13000000, |
| "cumulative_training_bytes": 13006694, |
| "metrics": { |
| "loss": 0.8181169518302469, |
| "ce_loss": 0.8081169577205882, |
| "lb_loss": 0.9999999966340907 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 14000000, |
| "cumulative_training_bytes": 14004830, |
| "metrics": { |
| "loss": 0.797012246241335, |
| "ce_loss": 0.7870122523907104, |
| "lb_loss": 0.9999999969057698 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 15000000, |
| "cumulative_training_bytes": 15004637, |
| "metrics": { |
| "loss": 0.778617926120515, |
| "ce_loss": 0.7686179324961754, |
| "lb_loss": 0.9999999971732626 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 16000000, |
| "cumulative_training_bytes": 16000188, |
| "metrics": { |
| "loss": 0.7623379522393846, |
| "ce_loss": 0.7523379588115734, |
| "lb_loss": 0.9999999962372964 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 17000000, |
| "cumulative_training_bytes": 17001433, |
| "metrics": { |
| "loss": 0.7477327509037983, |
| "ce_loss": 0.7377327576507651, |
| "lb_loss": 0.9999999957885107 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 18000000, |
| "cumulative_training_bytes": 18004044, |
| "metrics": { |
| "loss": 0.7345919368928714, |
| "ce_loss": 0.7245919437951551, |
| "lb_loss": 0.9999999957190034 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 19000000, |
| "cumulative_training_bytes": 19005595, |
| "metrics": { |
| "loss": 0.7230352938847073, |
| "ce_loss": 0.7130353009259259, |
| "lb_loss": 0.9999999958487908 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 20000000, |
| "cumulative_training_bytes": 20004400, |
| "metrics": { |
| "loss": 0.7124804242391204, |
| "ce_loss": 0.7024804314053538, |
| "lb_loss": 0.9999999958516079 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 21000000, |
| "cumulative_training_bytes": 21006730, |
| "metrics": { |
| "loss": 0.7027192320688224, |
| "ce_loss": 0.6927192393481427, |
| "lb_loss": 0.9999999955936844 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 22000000, |
| "cumulative_training_bytes": 22003387, |
| "metrics": { |
| "loss": 0.6940317696092523, |
| "ce_loss": 0.6840317769906119, |
| "lb_loss": 0.9999999956270584 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 23000000, |
| "cumulative_training_bytes": 23005816, |
| "metrics": { |
| "loss": 0.6859377783162276, |
| "ce_loss": 0.6759377857914866, |
| "lb_loss": 0.9999999957184559 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 24000000, |
| "cumulative_training_bytes": 24002853, |
| "metrics": { |
| "loss": 0.6783933619005801, |
| "ce_loss": 0.6683933694612687, |
| "lb_loss": 0.9999999955348768 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 25000000, |
| "cumulative_training_bytes": 25001613, |
| "metrics": { |
| "loss": 0.6714792498981766, |
| "ce_loss": 0.6614792575374961, |
| "lb_loss": 0.9999999954023966 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 26000000, |
| "cumulative_training_bytes": 26001520, |
| "metrics": { |
| "loss": 0.665006937754656, |
| "ce_loss": 0.6550069454665882, |
| "lb_loss": 0.9999999952274173 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 27000000, |
| "cumulative_training_bytes": 27002890, |
| "metrics": { |
| "loss": 0.6590575319028495, |
| "ce_loss": 0.6490575396825397, |
| "lb_loss": 0.9999999950836305 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 28000000, |
| "cumulative_training_bytes": 28003403, |
| "metrics": { |
| "loss": 0.6535597506296638, |
| "ce_loss": 0.6435597584722602, |
| "lb_loss": 0.9999999951619077 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 29000000, |
| "cumulative_training_bytes": 29006503, |
| "metrics": { |
| "loss": 0.648446149750561, |
| "ce_loss": 0.6384461576517151, |
| "lb_loss": 0.9999999951718666 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 30000000, |
| "cumulative_training_bytes": 30000438, |
| "metrics": { |
| "loss": 0.6435842753673087, |
| "ce_loss": 0.6335842833227041, |
| "lb_loss": 0.9999999954080095 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 31000000, |
| "cumulative_training_bytes": 31000470, |
| "metrics": { |
| "loss": 0.6390364181845314, |
| "ce_loss": 0.6290364261910639, |
| "lb_loss": 0.9999999954240818 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 32000000, |
| "cumulative_training_bytes": 32001821, |
| "metrics": { |
| "loss": 0.6348851022396517, |
| "ce_loss": 0.6248851102941176, |
| "lb_loss": 0.9999999954533999 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 33000000, |
| "cumulative_training_bytes": 33002772, |
| "metrics": { |
| "loss": 0.6308636427589162, |
| "ce_loss": 0.6208636508580705, |
| "lb_loss": 0.9999999956319416 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 34000000, |
| "cumulative_training_bytes": 34007477, |
| "metrics": { |
| "loss": 0.6269874756813264, |
| "ce_loss": 0.6169874838228674, |
| "lb_loss": 0.999999995760732 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 35000000, |
| "cumulative_training_bytes": 35006427, |
| "metrics": { |
| "loss": 0.6233431712141606, |
| "ce_loss": 0.6133431793953641, |
| "lb_loss": 0.9999999954902237 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 36000000, |
| "cumulative_training_bytes": 36003597, |
| "metrics": { |
| "loss": 0.6199277473362917, |
| "ce_loss": 0.6099277555549649, |
| "lb_loss": 0.9999999953360601 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 37000000, |
| "cumulative_training_bytes": 37005705, |
| "metrics": { |
| "loss": 0.6167014650337747, |
| "ce_loss": 0.6067014732881671, |
| "lb_loss": 0.9999999952651668 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 38000000, |
| "cumulative_training_bytes": 38004996, |
| "metrics": { |
| "loss": 0.613597062496674, |
| "ce_loss": 0.6035970707846495, |
| "lb_loss": 0.999999995257084 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 39000000, |
| "cumulative_training_bytes": 39003622, |
| "metrics": { |
| "loss": 0.6106150688542822, |
| "ce_loss": 0.6006150771741264, |
| "lb_loss": 0.9999999950270958 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 40000000, |
| "cumulative_training_bytes": 40003412, |
| "metrics": { |
| "loss": 0.6077792466508302, |
| "ce_loss": 0.5977792550009571, |
| "lb_loss": 0.9999999950253398 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 41000000, |
| "cumulative_training_bytes": 41001660, |
| "metrics": { |
| "loss": 0.6051997902903125, |
| "ce_loss": 0.5951997986694678, |
| "lb_loss": 0.9999999949466837 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 42000000, |
| "cumulative_training_bytes": 42003749, |
| "metrics": { |
| "loss": 0.6027762772521528, |
| "ce_loss": 0.59277628565895, |
| "lb_loss": 0.9999999948935139 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 43000000, |
| "cumulative_training_bytes": 43005510, |
| "metrics": { |
| "loss": 0.600378182215773, |
| "ce_loss": 0.5903781906489229, |
| "lb_loss": 0.9999999950232191 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 44000000, |
| "cumulative_training_bytes": 44002690, |
| "metrics": { |
| "loss": 0.5980829723362137, |
| "ce_loss": 0.5880829807943275, |
| "lb_loss": 0.9999999951150534 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 45000000, |
| "cumulative_training_bytes": 45005962, |
| "metrics": { |
| "loss": 0.5958610677767789, |
| "ce_loss": 0.5858610762589316, |
| "lb_loss": 0.9999999951529397 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 46000000, |
| "cumulative_training_bytes": 46002930, |
| "metrics": { |
| "loss": 0.5937242965882373, |
| "ce_loss": 0.583724305093209, |
| "lb_loss": 0.999999995138769 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 47000000, |
| "cumulative_training_bytes": 47006814, |
| "metrics": { |
| "loss": 0.5916914568989765, |
| "ce_loss": 0.5816914654259652, |
| "lb_loss": 0.9999999951551201 |
| } |
| }, |
| { |
| "epoch": 1, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.590403846496175, |
| "ce_loss": 0.5804038550369537, |
| "lb_loss": 0.999999995077637, |
| "training_bytes": 47653393 |
| }, |
| "cumulative_training_bytes": 47653393, |
| "training_bytes_this_epoch": 47653393 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 48000000, |
| "cumulative_training_bytes": 48005678, |
| "metrics": { |
| "loss": 0.492634161658909, |
| "ce_loss": 0.48263417119565216, |
| "lb_loss": 0.9999999961127406 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 49000000, |
| "cumulative_training_bytes": 49006971, |
| "metrics": { |
| "loss": 0.49733226447455625, |
| "ce_loss": 0.4873322740112994, |
| "lb_loss": 0.9999999989897518 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 50000000, |
| "cumulative_training_bytes": 50003595, |
| "metrics": { |
| "loss": 0.4957809859844295, |
| "ce_loss": 0.48578099552117265, |
| "lb_loss": 0.9999999959228093 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 51000000, |
| "cumulative_training_bytes": 51000980, |
| "metrics": { |
| "loss": 0.49479064406737583, |
| "ce_loss": 0.484790653604119, |
| "lb_loss": 0.9999999971357035 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 52000000, |
| "cumulative_training_bytes": 52001887, |
| "metrics": { |
| "loss": 0.4952002545477639, |
| "ce_loss": 0.48520026408450706, |
| "lb_loss": 0.9999999968518674 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 53000000, |
| "cumulative_training_bytes": 53005646, |
| "metrics": { |
| "loss": 0.49493103034165453, |
| "ce_loss": 0.4849310398783977, |
| "lb_loss": 0.9999999980387599 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 54000000, |
| "cumulative_training_bytes": 54005477, |
| "metrics": { |
| "loss": 0.49478444076446165, |
| "ce_loss": 0.4847844503012048, |
| "lb_loss": 0.9999999974865511 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 55000000, |
| "cumulative_training_bytes": 55002218, |
| "metrics": { |
| "loss": 0.4945906480153402, |
| "ce_loss": 0.48459065755208336, |
| "lb_loss": 0.999999997826914 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 56000000, |
| "cumulative_training_bytes": 56003171, |
| "metrics": { |
| "loss": 0.4947634672047784, |
| "ce_loss": 0.48476347674152154, |
| "lb_loss": 0.9999999967766507 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 57000000, |
| "cumulative_training_bytes": 57006709, |
| "metrics": { |
| "loss": 0.4947378050090834, |
| "ce_loss": 0.48473781454582654, |
| "lb_loss": 0.9999999970734217 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 58000000, |
| "cumulative_training_bytes": 58004016, |
| "metrics": { |
| "loss": 0.49519697872139296, |
| "ce_loss": 0.4851969882581361, |
| "lb_loss": 0.9999999964290117 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 59000000, |
| "cumulative_training_bytes": 59006803, |
| "metrics": { |
| "loss": 0.49550630157249487, |
| "ce_loss": 0.48550631110923803, |
| "lb_loss": 0.9999999969052208 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 60000000, |
| "cumulative_training_bytes": 60002534, |
| "metrics": { |
| "loss": 0.4955277337986567, |
| "ce_loss": 0.4855277433353999, |
| "lb_loss": 0.9999999967851183 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 61000000, |
| "cumulative_training_bytes": 61006527, |
| "metrics": { |
| "loss": 0.49546354407564214, |
| "ce_loss": 0.4854635536123853, |
| "lb_loss": 0.9999999964797716 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 62000000, |
| "cumulative_training_bytes": 62003543, |
| "metrics": { |
| "loss": 0.49532445517777124, |
| "ce_loss": 0.4853244647145144, |
| "lb_loss": 0.999999995547145 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 63000000, |
| "cumulative_training_bytes": 63006878, |
| "metrics": { |
| "loss": 0.495203972695177, |
| "ce_loss": 0.4852039822319202, |
| "lb_loss": 0.9999999965812797 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 64000000, |
| "cumulative_training_bytes": 64003765, |
| "metrics": { |
| "loss": 0.495035485193936, |
| "ce_loss": 0.48503549473067914, |
| "lb_loss": 0.9999999969290347 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 65000000, |
| "cumulative_training_bytes": 65007467, |
| "metrics": { |
| "loss": 0.4949671338326302, |
| "ce_loss": 0.48496714336937335, |
| "lb_loss": 0.9999999965015809 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 66000000, |
| "cumulative_training_bytes": 66003916, |
| "metrics": { |
| "loss": 0.49481925303629526, |
| "ce_loss": 0.4848192625730384, |
| "lb_loss": 0.9999999961689836 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 67000000, |
| "cumulative_training_bytes": 67000894, |
| "metrics": { |
| "loss": 0.4946804078722038, |
| "ce_loss": 0.48468041740894696, |
| "lb_loss": 0.9999999959178133 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 68000000, |
| "cumulative_training_bytes": 68004976, |
| "metrics": { |
| "loss": 0.494755765303302, |
| "ce_loss": 0.4847557748400452, |
| "lb_loss": 0.9999999962985449 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 69000000, |
| "cumulative_training_bytes": 69000639, |
| "metrics": { |
| "loss": 0.4949356291428406, |
| "ce_loss": 0.4849356386795838, |
| "lb_loss": 0.9999999967278398 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 70000000, |
| "cumulative_training_bytes": 70003667, |
| "metrics": { |
| "loss": 0.4949633375760053, |
| "ce_loss": 0.48496334711274847, |
| "lb_loss": 0.9999999967726066 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 71000000, |
| "cumulative_training_bytes": 71005669, |
| "metrics": { |
| "loss": 0.4949598389791637, |
| "ce_loss": 0.48495984851590684, |
| "lb_loss": 0.9999999968330756 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 72000000, |
| "cumulative_training_bytes": 72002076, |
| "metrics": { |
| "loss": 0.4950010472027976, |
| "ce_loss": 0.48500105673954075, |
| "lb_loss": 0.9999999969625818 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 73000000, |
| "cumulative_training_bytes": 73005205, |
| "metrics": { |
| "loss": 0.49494027558410275, |
| "ce_loss": 0.4849402851208459, |
| "lb_loss": 0.9999999970467789 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 74000000, |
| "cumulative_training_bytes": 74001335, |
| "metrics": { |
| "loss": 0.4949183452961057, |
| "ce_loss": 0.48491835483284884, |
| "lb_loss": 0.9999999970370947 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 75000000, |
| "cumulative_training_bytes": 75003123, |
| "metrics": { |
| "loss": 0.49487051867664245, |
| "ce_loss": 0.4848705282133856, |
| "lb_loss": 0.9999999971290958 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 76000000, |
| "cumulative_training_bytes": 76000007, |
| "metrics": { |
| "loss": 0.4947565385245916, |
| "ce_loss": 0.4847565480613348, |
| "lb_loss": 0.999999997036678 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 77000000, |
| "cumulative_training_bytes": 77005449, |
| "metrics": { |
| "loss": 0.4946735897979816, |
| "ce_loss": 0.48467359933472476, |
| "lb_loss": 0.9999999965789143 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 78000000, |
| "cumulative_training_bytes": 78003357, |
| "metrics": { |
| "loss": 0.4945139713489495, |
| "ce_loss": 0.48451398088569264, |
| "lb_loss": 0.9999999965858556 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 79000000, |
| "cumulative_training_bytes": 79005000, |
| "metrics": { |
| "loss": 0.4944432114879271, |
| "ce_loss": 0.48444322102467025, |
| "lb_loss": 0.9999999965495113 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 80000000, |
| "cumulative_training_bytes": 80002503, |
| "metrics": { |
| "loss": 0.49434447288513184, |
| "ce_loss": 0.484344482421875, |
| "lb_loss": 0.9999999965145958 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 81000000, |
| "cumulative_training_bytes": 81006707, |
| "metrics": { |
| "loss": 0.4942619738157253, |
| "ce_loss": 0.48426198335246845, |
| "lb_loss": 0.9999999964552002 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 82000000, |
| "cumulative_training_bytes": 82000170, |
| "metrics": { |
| "loss": 0.4941502329922488, |
| "ce_loss": 0.48415024252899197, |
| "lb_loss": 0.9999999965571804 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 83000000, |
| "cumulative_training_bytes": 83002106, |
| "metrics": { |
| "loss": 0.4940558883505808, |
| "ce_loss": 0.48405589788732395, |
| "lb_loss": 0.9999999969907081 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 84000000, |
| "cumulative_training_bytes": 84002576, |
| "metrics": { |
| "loss": 0.4939967936922918, |
| "ce_loss": 0.483996803229035, |
| "lb_loss": 0.9999999966467678 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 85000000, |
| "cumulative_training_bytes": 85005146, |
| "metrics": { |
| "loss": 0.493969307538303, |
| "ce_loss": 0.48396931707504615, |
| "lb_loss": 0.9999999965168498 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 86000000, |
| "cumulative_training_bytes": 86000950, |
| "metrics": { |
| "loss": 0.49385267483014317, |
| "ce_loss": 0.48385268436688633, |
| "lb_loss": 0.9999999967263277 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 87000000, |
| "cumulative_training_bytes": 87002861, |
| "metrics": { |
| "loss": 0.4937762804836928, |
| "ce_loss": 0.483776290020436, |
| "lb_loss": 0.9999999968097942 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 88000000, |
| "cumulative_training_bytes": 88007699, |
| "metrics": { |
| "loss": 0.49367403024784595, |
| "ce_loss": 0.4836740397845891, |
| "lb_loss": 0.9999999967759872 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 89000000, |
| "cumulative_training_bytes": 89003123, |
| "metrics": { |
| "loss": 0.49364641225432926, |
| "ce_loss": 0.4836464217910724, |
| "lb_loss": 0.9999999966438577 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 90000000, |
| "cumulative_training_bytes": 90001615, |
| "metrics": { |
| "loss": 0.49368486275077944, |
| "ce_loss": 0.4836848722875226, |
| "lb_loss": 0.9999999966694693 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 91000000, |
| "cumulative_training_bytes": 91003122, |
| "metrics": { |
| "loss": 0.49359353345477774, |
| "ce_loss": 0.4835935429915209, |
| "lb_loss": 0.9999999963990834 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 92000000, |
| "cumulative_training_bytes": 92003002, |
| "metrics": { |
| "loss": 0.4935082090351788, |
| "ce_loss": 0.48350821857192194, |
| "lb_loss": 0.9999999964799191 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 93000000, |
| "cumulative_training_bytes": 93004006, |
| "metrics": { |
| "loss": 0.49344559055613085, |
| "ce_loss": 0.483445600092874, |
| "lb_loss": 0.9999999965275916 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 94000000, |
| "cumulative_training_bytes": 94007117, |
| "metrics": { |
| "loss": 0.4933724529818427, |
| "ce_loss": 0.48337246251858584, |
| "lb_loss": 0.9999999964254938 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 95000000, |
| "cumulative_training_bytes": 95006117, |
| "metrics": { |
| "loss": 0.4932983081985414, |
| "ce_loss": 0.4832983177352846, |
| "lb_loss": 0.9999999962602519 |
| } |
| }, |
| { |
| "epoch": 2, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.49327259198558976, |
| "ce_loss": 0.4832726015223329, |
| "lb_loss": 0.9999999961789439, |
| "training_bytes": 47653382 |
| }, |
| "cumulative_training_bytes": 95306775, |
| "training_bytes_this_epoch": 47653382 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 96000000, |
| "cumulative_training_bytes": 96002802, |
| "metrics": { |
| "loss": 0.48830098634237773, |
| "ce_loss": 0.4783009958791209, |
| "lb_loss": 0.9999999927950429 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 97000000, |
| "cumulative_training_bytes": 97003339, |
| "metrics": { |
| "loss": 0.48807572244523884, |
| "ce_loss": 0.478075731981982, |
| "lb_loss": 0.9999999940932334 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 98000000, |
| "cumulative_training_bytes": 98005141, |
| "metrics": { |
| "loss": 0.4878903941459764, |
| "ce_loss": 0.47789040368271957, |
| "lb_loss": 0.9999999947655978 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 99000000, |
| "cumulative_training_bytes": 99004099, |
| "metrics": { |
| "loss": 0.487553206042967, |
| "ce_loss": 0.47755321557971014, |
| "lb_loss": 0.9999999979021139 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 100000000, |
| "cumulative_training_bytes": 100001082, |
| "metrics": { |
| "loss": 0.4875374598759811, |
| "ce_loss": 0.4775374694127243, |
| "lb_loss": 0.9999999975691417 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 101000000, |
| "cumulative_training_bytes": 101002438, |
| "metrics": { |
| "loss": 0.48744717208288046, |
| "ce_loss": 0.47744718161962363, |
| "lb_loss": 0.9999999971960181 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 102000000, |
| "cumulative_training_bytes": 102006394, |
| "metrics": { |
| "loss": 0.48746427617754257, |
| "ce_loss": 0.47746428571428573, |
| "lb_loss": 0.9999999976158143 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 103000000, |
| "cumulative_training_bytes": 103004065, |
| "metrics": { |
| "loss": 0.487491439468232, |
| "ce_loss": 0.47749144900497514, |
| "lb_loss": 0.9999999987545298 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 104000000, |
| "cumulative_training_bytes": 104001074, |
| "metrics": { |
| "loss": 0.4877171573134771, |
| "ce_loss": 0.47771716685022025, |
| "lb_loss": 0.9999999984245468 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 105000000, |
| "cumulative_training_bytes": 105004888, |
| "metrics": { |
| "loss": 0.4877519528447734, |
| "ce_loss": 0.47775196238151657, |
| "lb_loss": 0.999999998210919 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 106000000, |
| "cumulative_training_bytes": 106007608, |
| "metrics": { |
| "loss": 0.4879144388347672, |
| "ce_loss": 0.47791444837151037, |
| "lb_loss": 0.999999997525362 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 107000000, |
| "cumulative_training_bytes": 107001146, |
| "metrics": { |
| "loss": 0.4879950379419733, |
| "ce_loss": 0.47799504747871646, |
| "lb_loss": 0.9999999976970045 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 108000000, |
| "cumulative_training_bytes": 108007350, |
| "metrics": { |
| "loss": 0.4880941454512014, |
| "ce_loss": 0.47809415498794455, |
| "lb_loss": 0.9999999979161728 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 109000000, |
| "cumulative_training_bytes": 109006285, |
| "metrics": { |
| "loss": 0.4880014053598471, |
| "ce_loss": 0.47800141489659026, |
| "lb_loss": 0.9999999974678854 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 110000000, |
| "cumulative_training_bytes": 110005030, |
| "metrics": { |
| "loss": 0.4881331284840902, |
| "ce_loss": 0.47813313802083335, |
| "lb_loss": 0.9999999972991646 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 111000000, |
| "cumulative_training_bytes": 111001255, |
| "metrics": { |
| "loss": 0.48820883192667147, |
| "ce_loss": 0.47820884146341464, |
| "lb_loss": 0.9999999975285879 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 112000000, |
| "cumulative_training_bytes": 112000565, |
| "metrics": { |
| "loss": 0.48811692710316507, |
| "ce_loss": 0.47811693663990823, |
| "lb_loss": 0.9999999980860894 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 113000000, |
| "cumulative_training_bytes": 113002042, |
| "metrics": { |
| "loss": 0.4881750229167402, |
| "ce_loss": 0.47817503245348336, |
| "lb_loss": 0.9999999979108714 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 114000000, |
| "cumulative_training_bytes": 114007396, |
| "metrics": { |
| "loss": 0.4890242917531831, |
| "ce_loss": 0.4790243012899263, |
| "lb_loss": 0.999999997876493 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 115000000, |
| "cumulative_training_bytes": 115004141, |
| "metrics": { |
| "loss": 0.48968308777184605, |
| "ce_loss": 0.4796830973085892, |
| "lb_loss": 0.9999999972201488 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 116000000, |
| "cumulative_training_bytes": 116007191, |
| "metrics": { |
| "loss": 0.48999201616591953, |
| "ce_loss": 0.4799920257026627, |
| "lb_loss": 0.9999999972446078 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 117000000, |
| "cumulative_training_bytes": 117006642, |
| "metrics": { |
| "loss": 0.49022899178600815, |
| "ce_loss": 0.4802290013227513, |
| "lb_loss": 0.9999999971827082 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 118000000, |
| "cumulative_training_bytes": 118004076, |
| "metrics": { |
| "loss": 0.49022830517657895, |
| "ce_loss": 0.4802283147133221, |
| "lb_loss": 0.9999999977886979 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 119000000, |
| "cumulative_training_bytes": 119004633, |
| "metrics": { |
| "loss": 0.4901273379996058, |
| "ce_loss": 0.48012734753634895, |
| "lb_loss": 0.9999999980934217 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 120000000, |
| "cumulative_training_bytes": 120007595, |
| "metrics": { |
| "loss": 0.4901242494139698, |
| "ce_loss": 0.480124258950713, |
| "lb_loss": 0.9999999980969378 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 121000000, |
| "cumulative_training_bytes": 121001712, |
| "metrics": { |
| "loss": 0.4901183884817312, |
| "ce_loss": 0.4801183980184744, |
| "lb_loss": 0.9999999977088798 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 122000000, |
| "cumulative_training_bytes": 122006493, |
| "metrics": { |
| "loss": 0.4901035447362709, |
| "ce_loss": 0.48010355427301404, |
| "lb_loss": 0.9999999979146066 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 123000000, |
| "cumulative_training_bytes": 123005264, |
| "metrics": { |
| "loss": 0.4900173131754493, |
| "ce_loss": 0.48001732271219244, |
| "lb_loss": 0.999999997643499 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 124000000, |
| "cumulative_training_bytes": 124002472, |
| "metrics": { |
| "loss": 0.4899558301349942, |
| "ce_loss": 0.4799558396717374, |
| "lb_loss": 0.9999999973275739 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 125000000, |
| "cumulative_training_bytes": 125002266, |
| "metrics": { |
| "loss": 0.4899716454722821, |
| "ce_loss": 0.4799716550090253, |
| "lb_loss": 0.9999999974024794 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 126000000, |
| "cumulative_training_bytes": 126000850, |
| "metrics": { |
| "loss": 0.4900195558162127, |
| "ce_loss": 0.48001956535295587, |
| "lb_loss": 0.9999999972346062 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 127000000, |
| "cumulative_training_bytes": 127007464, |
| "metrics": { |
| "loss": 0.49008660731108294, |
| "ce_loss": 0.4800866168478261, |
| "lb_loss": 0.9999999973796992 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 128000000, |
| "cumulative_training_bytes": 128005894, |
| "metrics": { |
| "loss": 0.49008085990119593, |
| "ce_loss": 0.4800808694379391, |
| "lb_loss": 0.9999999971523776 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 129000000, |
| "cumulative_training_bytes": 129003149, |
| "metrics": { |
| "loss": 0.4900173022530296, |
| "ce_loss": 0.48001731178977275, |
| "lb_loss": 0.9999999971552329 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 130000000, |
| "cumulative_training_bytes": 130006682, |
| "metrics": { |
| "loss": 0.48996655758420143, |
| "ce_loss": 0.4799665671209446, |
| "lb_loss": 0.9999999974084937 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 131000000, |
| "cumulative_training_bytes": 131001772, |
| "metrics": { |
| "loss": 0.4899122611669685, |
| "ce_loss": 0.4799122707037117, |
| "lb_loss": 0.9999999974551975 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 132000000, |
| "cumulative_training_bytes": 132002042, |
| "metrics": { |
| "loss": 0.48989445976104484, |
| "ce_loss": 0.479894469297788, |
| "lb_loss": 0.9999999973630667 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 133000000, |
| "cumulative_training_bytes": 133001074, |
| "metrics": { |
| "loss": 0.48980843977248073, |
| "ce_loss": 0.4798084493092239, |
| "lb_loss": 0.9999999974206035 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 134000000, |
| "cumulative_training_bytes": 134006568, |
| "metrics": { |
| "loss": 0.48973665680008643, |
| "ce_loss": 0.4797366663368296, |
| "lb_loss": 0.9999999974284954 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 135000000, |
| "cumulative_training_bytes": 135007246, |
| "metrics": { |
| "loss": 0.4897220016997538, |
| "ce_loss": 0.47972201123649694, |
| "lb_loss": 0.9999999975164732 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 136000000, |
| "cumulative_training_bytes": 136007771, |
| "metrics": { |
| "loss": 0.4896772011876218, |
| "ce_loss": 0.479677210724365, |
| "lb_loss": 0.9999999976561861 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 137000000, |
| "cumulative_training_bytes": 137003574, |
| "metrics": { |
| "loss": 0.4896326092649097, |
| "ce_loss": 0.47963261880165287, |
| "lb_loss": 0.9999999975807848 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 138000000, |
| "cumulative_training_bytes": 138005969, |
| "metrics": { |
| "loss": 0.48960041247276187, |
| "ce_loss": 0.47960042200950503, |
| "lb_loss": 0.9999999974772783 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 139000000, |
| "cumulative_training_bytes": 139005934, |
| "metrics": { |
| "loss": 0.48958851661535074, |
| "ce_loss": 0.4795885261520939, |
| "lb_loss": 0.999999997409856 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 140000000, |
| "cumulative_training_bytes": 140005424, |
| "metrics": { |
| "loss": 0.48959421686784743, |
| "ce_loss": 0.4795942264045906, |
| "lb_loss": 0.9999999973352497 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 141000000, |
| "cumulative_training_bytes": 141006781, |
| "metrics": { |
| "loss": 0.4895374967457162, |
| "ce_loss": 0.4795375062824594, |
| "lb_loss": 0.9999999973438037 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 142000000, |
| "cumulative_training_bytes": 142002589, |
| "metrics": { |
| "loss": 0.48945935367853804, |
| "ce_loss": 0.4794593632152812, |
| "lb_loss": 0.9999999972538276 |
| } |
| }, |
| { |
| "epoch": 3, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.48934782746518485, |
| "ce_loss": 0.479347837001928, |
| "lb_loss": 0.9999999973377103, |
| "training_bytes": 47653416 |
| }, |
| "cumulative_training_bytes": 142960191, |
| "training_bytes_this_epoch": 47653416 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 143000000, |
| "cumulative_training_bytes": 143005811, |
| "metrics": { |
| "loss": 0.4911197821299235, |
| "ce_loss": 0.4811197916666667, |
| "lb_loss": 0.9999999503294627 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 144000000, |
| "cumulative_training_bytes": 144005584, |
| "metrics": { |
| "loss": 0.4834604684044333, |
| "ce_loss": 0.47346047794117646, |
| "lb_loss": 0.9999999938642278 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 145000000, |
| "cumulative_training_bytes": 145002597, |
| "metrics": { |
| "loss": 0.4844184585083696, |
| "ce_loss": 0.4744184680451128, |
| "lb_loss": 0.9999999959666029 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 146000000, |
| "cumulative_training_bytes": 146003208, |
| "metrics": { |
| "loss": 0.4850127817101082, |
| "ce_loss": 0.47501279124685136, |
| "lb_loss": 0.9999999965468341 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 147000000, |
| "cumulative_training_bytes": 147007482, |
| "metrics": { |
| "loss": 0.48537507924166595, |
| "ce_loss": 0.4753750887784091, |
| "lb_loss": 0.999999994807171 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 148000000, |
| "cumulative_training_bytes": 148004641, |
| "metrics": { |
| "loss": 0.48546422868514133, |
| "ce_loss": 0.4754642382218845, |
| "lb_loss": 0.9999999916662199 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 149000000, |
| "cumulative_training_bytes": 149002815, |
| "metrics": { |
| "loss": 0.48531080502543955, |
| "ce_loss": 0.4753108145621827, |
| "lb_loss": 0.9999999919064759 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 150000000, |
| "cumulative_training_bytes": 150002243, |
| "metrics": { |
| "loss": 0.48544034682887166, |
| "ce_loss": 0.4754403563656148, |
| "lb_loss": 0.9999999926061703 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 151000000, |
| "cumulative_training_bytes": 151002455, |
| "metrics": { |
| "loss": 0.48702938034420923, |
| "ce_loss": 0.4770293898809524, |
| "lb_loss": 0.9999999929609753 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 152000000, |
| "cumulative_training_bytes": 152006519, |
| "metrics": { |
| "loss": 0.487319925899328, |
| "ce_loss": 0.47731993543607115, |
| "lb_loss": 0.9999999927323718 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 153000000, |
| "cumulative_training_bytes": 153001757, |
| "metrics": { |
| "loss": 0.4873133485868266, |
| "ce_loss": 0.4773133581235698, |
| "lb_loss": 0.9999999942259421 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 154000000, |
| "cumulative_training_bytes": 154001996, |
| "metrics": { |
| "loss": 0.4871629306957343, |
| "ce_loss": 0.47716294023247746, |
| "lb_loss": 0.9999999934645845 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 155000000, |
| "cumulative_training_bytes": 155007554, |
| "metrics": { |
| "loss": 0.487115378901552, |
| "ce_loss": 0.47711538843829515, |
| "lb_loss": 0.9999999934404558 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 156000000, |
| "cumulative_training_bytes": 156006842, |
| "metrics": { |
| "loss": 0.4871095489629632, |
| "ce_loss": 0.4771095584997064, |
| "lb_loss": 0.999999993630038 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 157000000, |
| "cumulative_training_bytes": 157000485, |
| "metrics": { |
| "loss": 0.4868759286455712, |
| "ce_loss": 0.4768759381823144, |
| "lb_loss": 0.9999999938833661 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 158000000, |
| "cumulative_training_bytes": 158002706, |
| "metrics": { |
| "loss": 0.4868062579747189, |
| "ce_loss": 0.47680626751146205, |
| "lb_loss": 0.9999999946255619 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 159000000, |
| "cumulative_training_bytes": 159005682, |
| "metrics": { |
| "loss": 0.48680872971827116, |
| "ce_loss": 0.4768087392550143, |
| "lb_loss": 0.9999999952748945 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 160000000, |
| "cumulative_training_bytes": 160001107, |
| "metrics": { |
| "loss": 0.4867460350338504, |
| "ce_loss": 0.47674604457059355, |
| "lb_loss": 0.9999999951758831 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 161000000, |
| "cumulative_training_bytes": 161002648, |
| "metrics": { |
| "loss": 0.4867391425333205, |
| "ce_loss": 0.4767391520700637, |
| "lb_loss": 0.9999999951405131 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 162000000, |
| "cumulative_training_bytes": 162004313, |
| "metrics": { |
| "loss": 0.4866803379089527, |
| "ce_loss": 0.4766803474456959, |
| "lb_loss": 0.9999999951088706 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 163000000, |
| "cumulative_training_bytes": 163005127, |
| "metrics": { |
| "loss": 0.48664981015183156, |
| "ce_loss": 0.4766498196885747, |
| "lb_loss": 0.9999999953081556 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 164000000, |
| "cumulative_training_bytes": 164000917, |
| "metrics": { |
| "loss": 0.4866300357399223, |
| "ce_loss": 0.47663004527666547, |
| "lb_loss": 0.9999999955301941 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 165000000, |
| "cumulative_training_bytes": 165003825, |
| "metrics": { |
| "loss": 0.48657877781558484, |
| "ce_loss": 0.476578787352328, |
| "lb_loss": 0.9999999957129391 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 166000000, |
| "cumulative_training_bytes": 166000316, |
| "metrics": { |
| "loss": 0.4866410308836623, |
| "ce_loss": 0.47664104042040545, |
| "lb_loss": 0.9999999958599632 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 167000000, |
| "cumulative_training_bytes": 167004464, |
| "metrics": { |
| "loss": 0.4865475621193078, |
| "ce_loss": 0.47654757165605094, |
| "lb_loss": 0.9999999961845435 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 168000000, |
| "cumulative_training_bytes": 168000312, |
| "metrics": { |
| "loss": 0.4865374044540825, |
| "ce_loss": 0.47653741399082566, |
| "lb_loss": 0.9999999960081293 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 169000000, |
| "cumulative_training_bytes": 169004041, |
| "metrics": { |
| "loss": 0.48646658576610896, |
| "ce_loss": 0.4764665953028521, |
| "lb_loss": 0.9999999961969398 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 170000000, |
| "cumulative_training_bytes": 170004115, |
| "metrics": { |
| "loss": 0.4864624011017053, |
| "ce_loss": 0.47646241063844846, |
| "lb_loss": 0.9999999961523616 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 171000000, |
| "cumulative_training_bytes": 171002131, |
| "metrics": { |
| "loss": 0.4865150353376767, |
| "ce_loss": 0.47651504487441987, |
| "lb_loss": 0.9999999958831627 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 172000000, |
| "cumulative_training_bytes": 172000593, |
| "metrics": { |
| "loss": 0.48652182180880243, |
| "ce_loss": 0.4765218313455456, |
| "lb_loss": 0.9999999957268151 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 173000000, |
| "cumulative_training_bytes": 173003618, |
| "metrics": { |
| "loss": 0.48656199285179186, |
| "ce_loss": 0.476562002388535, |
| "lb_loss": 0.999999995732763 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 174000000, |
| "cumulative_training_bytes": 174000074, |
| "metrics": { |
| "loss": 0.4865022831574615, |
| "ce_loss": 0.4765022926942047, |
| "lb_loss": 0.9999999959283633 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 175000000, |
| "cumulative_training_bytes": 175004955, |
| "metrics": { |
| "loss": 0.4864411783514556, |
| "ce_loss": 0.4764411878881988, |
| "lb_loss": 0.9999999958849158 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 176000000, |
| "cumulative_training_bytes": 176002879, |
| "metrics": { |
| "loss": 0.4863869083270196, |
| "ce_loss": 0.47638691786376275, |
| "lb_loss": 0.9999999958569524 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 177000000, |
| "cumulative_training_bytes": 177007060, |
| "metrics": { |
| "loss": 0.48635870139197285, |
| "ce_loss": 0.476358710928716, |
| "lb_loss": 0.9999999955903017 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 178000000, |
| "cumulative_training_bytes": 178007425, |
| "metrics": { |
| "loss": 0.4863939705036675, |
| "ce_loss": 0.47639398004041067, |
| "lb_loss": 0.9999999956123273 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 179000000, |
| "cumulative_training_bytes": 179000051, |
| "metrics": { |
| "loss": 0.4864388644543359, |
| "ce_loss": 0.47643887399107904, |
| "lb_loss": 0.9999999957081618 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 180000000, |
| "cumulative_training_bytes": 180003883, |
| "metrics": { |
| "loss": 0.48640507803300265, |
| "ce_loss": 0.4764050875697458, |
| "lb_loss": 0.9999999957381263 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 181000000, |
| "cumulative_training_bytes": 181006010, |
| "metrics": { |
| "loss": 0.48639232886868944, |
| "ce_loss": 0.4763923384054326, |
| "lb_loss": 0.9999999955626321 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 182000000, |
| "cumulative_training_bytes": 182004597, |
| "metrics": { |
| "loss": 0.4863602845809039, |
| "ce_loss": 0.4763602941176471, |
| "lb_loss": 0.9999999954653721 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 183000000, |
| "cumulative_training_bytes": 183005222, |
| "metrics": { |
| "loss": 0.48625701125675586, |
| "ce_loss": 0.476257020793499, |
| "lb_loss": 0.999999995703451 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 184000000, |
| "cumulative_training_bytes": 184003606, |
| "metrics": { |
| "loss": 0.4863026811115777, |
| "ce_loss": 0.47630269064832087, |
| "lb_loss": 0.9999999957186962 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 185000000, |
| "cumulative_training_bytes": 185005739, |
| "metrics": { |
| "loss": 0.48644190962643297, |
| "ce_loss": 0.47644191916317613, |
| "lb_loss": 0.9999999957339964 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 186000000, |
| "cumulative_training_bytes": 186001058, |
| "metrics": { |
| "loss": 0.48650585296548066, |
| "ce_loss": 0.4765058625022238, |
| "lb_loss": 0.999999995779639 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 187000000, |
| "cumulative_training_bytes": 187002105, |
| "metrics": { |
| "loss": 0.4865529829006699, |
| "ce_loss": 0.47655299243741306, |
| "lb_loss": 0.9999999958343068 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 188000000, |
| "cumulative_training_bytes": 188006850, |
| "metrics": { |
| "loss": 0.4865126912536699, |
| "ce_loss": 0.4765127007904131, |
| "lb_loss": 0.9999999957649598 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 189000000, |
| "cumulative_training_bytes": 189001113, |
| "metrics": { |
| "loss": 0.4865072715625417, |
| "ce_loss": 0.47650728109928486, |
| "lb_loss": 0.9999999957573943 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 190000000, |
| "cumulative_training_bytes": 190003438, |
| "metrics": { |
| "loss": 0.4864817460378011, |
| "ce_loss": 0.47648175557454425, |
| "lb_loss": 0.9999999960709829 |
| } |
| }, |
| { |
| "epoch": 4, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.48646113130917584, |
| "ce_loss": 0.476461140845919, |
| "lb_loss": 0.9999999960640249, |
| "training_bytes": 47653407 |
| }, |
| "cumulative_training_bytes": 190613598, |
| "training_bytes_this_epoch": 47653407 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 191000000, |
| "cumulative_training_bytes": 191004593, |
| "metrics": { |
| "loss": 0.4838434340907078, |
| "ce_loss": 0.47384344362745096, |
| "lb_loss": 0.9999999871440962 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 192000000, |
| "cumulative_training_bytes": 192004632, |
| "metrics": { |
| "loss": 0.4837293860676524, |
| "ce_loss": 0.4737293956043956, |
| "lb_loss": 0.9999999944325332 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 193000000, |
| "cumulative_training_bytes": 193001112, |
| "metrics": { |
| "loss": 0.48330102211389786, |
| "ce_loss": 0.473301031650641, |
| "lb_loss": 0.9999999961791894 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 194000000, |
| "cumulative_training_bytes": 194004855, |
| "metrics": { |
| "loss": 0.4841243908582907, |
| "ce_loss": 0.47412440039503384, |
| "lb_loss": 0.9999999960981159 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 195000000, |
| "cumulative_training_bytes": 195000605, |
| "metrics": { |
| "loss": 0.48411852776692177, |
| "ce_loss": 0.47411853730366493, |
| "lb_loss": 0.9999999941747643 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 196000000, |
| "cumulative_training_bytes": 196000298, |
| "metrics": { |
| "loss": 0.4837925535855897, |
| "ce_loss": 0.47379256312233287, |
| "lb_loss": 0.9999999937258268 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 197000000, |
| "cumulative_training_bytes": 197005504, |
| "metrics": { |
| "loss": 0.4837077430112185, |
| "ce_loss": 0.47370775254796166, |
| "lb_loss": 0.9999999942825281 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 198000000, |
| "cumulative_training_bytes": 198005978, |
| "metrics": { |
| "loss": 0.48402241598139156, |
| "ce_loss": 0.4740224255181347, |
| "lb_loss": 0.9999999936380535 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 199000000, |
| "cumulative_training_bytes": 199002040, |
| "metrics": { |
| "loss": 0.4838441685454486, |
| "ce_loss": 0.47384417808219176, |
| "lb_loss": 0.9999999936312846 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 200000000, |
| "cumulative_training_bytes": 200003598, |
| "metrics": { |
| "loss": 0.48375069142165816, |
| "ce_loss": 0.4737507009584013, |
| "lb_loss": 0.999999993242214 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 201000000, |
| "cumulative_training_bytes": 201005092, |
| "metrics": { |
| "loss": 0.48378752730924063, |
| "ce_loss": 0.4737875368459838, |
| "lb_loss": 0.9999999931918055 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 202000000, |
| "cumulative_training_bytes": 202001683, |
| "metrics": { |
| "loss": 0.4836702367645346, |
| "ce_loss": 0.47367024630127774, |
| "lb_loss": 0.9999999932258339 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 203000000, |
| "cumulative_training_bytes": 203001980, |
| "metrics": { |
| "loss": 0.4837692078612791, |
| "ce_loss": 0.47376921739802225, |
| "lb_loss": 0.9999999935164292 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 204000000, |
| "cumulative_training_bytes": 204006015, |
| "metrics": { |
| "loss": 0.4837763001402151, |
| "ce_loss": 0.4737763096769583, |
| "lb_loss": 0.9999999933204629 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 205000000, |
| "cumulative_training_bytes": 205004293, |
| "metrics": { |
| "loss": 0.483796516208334, |
| "ce_loss": 0.4737965257450772, |
| "lb_loss": 0.9999999935288199 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 206000000, |
| "cumulative_training_bytes": 206000909, |
| "metrics": { |
| "loss": 0.48376647876589496, |
| "ce_loss": 0.4737664883026381, |
| "lb_loss": 0.9999999937398805 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 207000000, |
| "cumulative_training_bytes": 207000656, |
| "metrics": { |
| "loss": 0.4837419186423125, |
| "ce_loss": 0.47374192817905564, |
| "lb_loss": 0.9999999936744954 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 208000000, |
| "cumulative_training_bytes": 208006392, |
| "metrics": { |
| "loss": 0.48362419888836694, |
| "ce_loss": 0.4736242084251101, |
| "lb_loss": 0.9999999941183082 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 209000000, |
| "cumulative_training_bytes": 209002552, |
| "metrics": { |
| "loss": 0.4836824448903402, |
| "ce_loss": 0.47368245442708334, |
| "lb_loss": 0.9999999945114056 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 210000000, |
| "cumulative_training_bytes": 210001732, |
| "metrics": { |
| "loss": 0.48379524184610945, |
| "ce_loss": 0.4737952513828526, |
| "lb_loss": 0.9999999946306365 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 211000000, |
| "cumulative_training_bytes": 211005793, |
| "metrics": { |
| "loss": 0.48375900037122077, |
| "ce_loss": 0.47375900990796393, |
| "lb_loss": 0.9999999952083418 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 212000000, |
| "cumulative_training_bytes": 212001830, |
| "metrics": { |
| "loss": 0.48375242487407344, |
| "ce_loss": 0.4737524344108166, |
| "lb_loss": 0.9999999954314491 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 213000000, |
| "cumulative_training_bytes": 213002388, |
| "metrics": { |
| "loss": 0.48380419419230236, |
| "ce_loss": 0.4738042037290455, |
| "lb_loss": 0.9999999950856245 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 214000000, |
| "cumulative_training_bytes": 214006969, |
| "metrics": { |
| "loss": 0.4837882087834926, |
| "ce_loss": 0.47378821832023577, |
| "lb_loss": 0.9999999953159415 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 215000000, |
| "cumulative_training_bytes": 215004172, |
| "metrics": { |
| "loss": 0.483773889254086, |
| "ce_loss": 0.47377389879082915, |
| "lb_loss": 0.9999999951327866 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 216000000, |
| "cumulative_training_bytes": 216001255, |
| "metrics": { |
| "loss": 0.4837848567502212, |
| "ce_loss": 0.47378486628696437, |
| "lb_loss": 0.9999999949999725 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 217000000, |
| "cumulative_training_bytes": 217004188, |
| "metrics": { |
| "loss": 0.4841484393021538, |
| "ce_loss": 0.47414844883889695, |
| "lb_loss": 0.9999999951381988 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 218000000, |
| "cumulative_training_bytes": 218001725, |
| "metrics": { |
| "loss": 0.4844258503625857, |
| "ce_loss": 0.47442585989932884, |
| "lb_loss": 0.9999999952662978 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 219000000, |
| "cumulative_training_bytes": 219003579, |
| "metrics": { |
| "loss": 0.48454192548483765, |
| "ce_loss": 0.4745419350215808, |
| "lb_loss": 0.9999999954175011 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 220000000, |
| "cumulative_training_bytes": 220002944, |
| "metrics": { |
| "loss": 0.4846480412058311, |
| "ce_loss": 0.47464805074257427, |
| "lb_loss": 0.9999999953409605 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 221000000, |
| "cumulative_training_bytes": 221001782, |
| "metrics": { |
| "loss": 0.48471910264516666, |
| "ce_loss": 0.4747191121819098, |
| "lb_loss": 0.9999999951493322 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 222000000, |
| "cumulative_training_bytes": 222000475, |
| "metrics": { |
| "loss": 0.4846779797594266, |
| "ce_loss": 0.4746779892961698, |
| "lb_loss": 0.9999999951286762 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 223000000, |
| "cumulative_training_bytes": 223000890, |
| "metrics": { |
| "loss": 0.484692938897063, |
| "ce_loss": 0.47469294843380616, |
| "lb_loss": 0.9999999951386283 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 224000000, |
| "cumulative_training_bytes": 224000466, |
| "metrics": { |
| "loss": 0.484730315864633, |
| "ce_loss": 0.47473032540137616, |
| "lb_loss": 0.9999999953656021 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 225000000, |
| "cumulative_training_bytes": 225004782, |
| "metrics": { |
| "loss": 0.4847559259453321, |
| "ce_loss": 0.4747559354820753, |
| "lb_loss": 0.999999995487513 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 226000000, |
| "cumulative_training_bytes": 226004131, |
| "metrics": { |
| "loss": 0.48467487820400557, |
| "ce_loss": 0.47467488774074873, |
| "lb_loss": 0.9999999957305481 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 227000000, |
| "cumulative_training_bytes": 227007102, |
| "metrics": { |
| "loss": 0.48469074567159015, |
| "ce_loss": 0.4746907552083333, |
| "lb_loss": 0.9999999959736762 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 228000000, |
| "cumulative_training_bytes": 228004925, |
| "metrics": { |
| "loss": 0.48469536216098363, |
| "ce_loss": 0.4746953716977268, |
| "lb_loss": 0.9999999959108016 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 229000000, |
| "cumulative_training_bytes": 229001388, |
| "metrics": { |
| "loss": 0.4846923528211264, |
| "ce_loss": 0.47469236235786955, |
| "lb_loss": 0.9999999959217248 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 230000000, |
| "cumulative_training_bytes": 230000104, |
| "metrics": { |
| "loss": 0.4846553168778009, |
| "ce_loss": 0.47465532641454405, |
| "lb_loss": 0.9999999958393803 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 231000000, |
| "cumulative_training_bytes": 231004256, |
| "metrics": { |
| "loss": 0.4846512128513873, |
| "ce_loss": 0.47465122238813046, |
| "lb_loss": 0.9999999958523124 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 232000000, |
| "cumulative_training_bytes": 232007293, |
| "metrics": { |
| "loss": 0.484631408132082, |
| "ce_loss": 0.47463141766882516, |
| "lb_loss": 0.9999999958756453 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 233000000, |
| "cumulative_training_bytes": 233002626, |
| "metrics": { |
| "loss": 0.48464676948313035, |
| "ce_loss": 0.4746467790198735, |
| "lb_loss": 0.9999999958325207 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 234000000, |
| "cumulative_training_bytes": 234002992, |
| "metrics": { |
| "loss": 0.4846627954623744, |
| "ce_loss": 0.47466280499911756, |
| "lb_loss": 0.9999999956448424 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 235000000, |
| "cumulative_training_bytes": 235001903, |
| "metrics": { |
| "loss": 0.48463665597990335, |
| "ce_loss": 0.4746366655166465, |
| "lb_loss": 0.999999995661004 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 236000000, |
| "cumulative_training_bytes": 236005937, |
| "metrics": { |
| "loss": 0.4846179318009761, |
| "ce_loss": 0.4746179413377193, |
| "lb_loss": 0.999999995596013 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 237000000, |
| "cumulative_training_bytes": 237000271, |
| "metrics": { |
| "loss": 0.4846171031654688, |
| "ce_loss": 0.47461711270221196, |
| "lb_loss": 0.9999999953855103 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 238000000, |
| "cumulative_training_bytes": 238005016, |
| "metrics": { |
| "loss": 0.48460431618227445, |
| "ce_loss": 0.4746043257190176, |
| "lb_loss": 0.9999999954639218 |
| } |
| }, |
| { |
| "epoch": 5, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.4846178382098828, |
| "ce_loss": 0.47461784774662596, |
| "lb_loss": 0.9999999954128174, |
| "training_bytes": 47653413 |
| }, |
| "cumulative_training_bytes": 238267011, |
| "training_bytes_this_epoch": 47653413 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 239000000, |
| "cumulative_training_bytes": 239004387, |
| "metrics": { |
| "loss": 0.4854953190715043, |
| "ce_loss": 0.47549532860824745, |
| "lb_loss": 0.9999999821800547 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 240000000, |
| "cumulative_training_bytes": 240001325, |
| "metrics": { |
| "loss": 0.4832241094059881, |
| "ce_loss": 0.47322411894273125, |
| "lb_loss": 0.9999999915975831 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 241000000, |
| "cumulative_training_bytes": 241000019, |
| "metrics": { |
| "loss": 0.483070870351525, |
| "ce_loss": 0.4730708798882682, |
| "lb_loss": 0.9999999950051973 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 242000000, |
| "cumulative_training_bytes": 242004419, |
| "metrics": { |
| "loss": 0.48300372896018934, |
| "ce_loss": 0.4730037384969325, |
| "lb_loss": 0.9999999974402913 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 243000000, |
| "cumulative_training_bytes": 243002073, |
| "metrics": { |
| "loss": 0.48269094866196444, |
| "ce_loss": 0.4726909581987076, |
| "lb_loss": 0.9999999974964123 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 244000000, |
| "cumulative_training_bytes": 244003061, |
| "metrics": { |
| "loss": 0.48260155296325685, |
| "ce_loss": 0.4726015625, |
| "lb_loss": 0.9999999978542328 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 245000000, |
| "cumulative_training_bytes": 245004215, |
| "metrics": { |
| "loss": 0.48258308141104345, |
| "ce_loss": 0.4725830909477866, |
| "lb_loss": 0.9999999963465939 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 246000000, |
| "cumulative_training_bytes": 246005003, |
| "metrics": { |
| "loss": 0.48249219340298016, |
| "ce_loss": 0.4724922029397233, |
| "lb_loss": 0.9999999967017192 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 247000000, |
| "cumulative_training_bytes": 247006730, |
| "metrics": { |
| "loss": 0.4825161212594073, |
| "ce_loss": 0.47251613079615046, |
| "lb_loss": 0.9999999975490653 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 248000000, |
| "cumulative_training_bytes": 248005797, |
| "metrics": { |
| "loss": 0.4824859365551657, |
| "ce_loss": 0.47248594609190886, |
| "lb_loss": 0.9999999967692691 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 249000000, |
| "cumulative_training_bytes": 249006395, |
| "metrics": { |
| "loss": 0.48267154272465285, |
| "ce_loss": 0.472671552261396, |
| "lb_loss": 0.9999999966461774 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 250000000, |
| "cumulative_training_bytes": 250006518, |
| "metrics": { |
| "loss": 0.4824703498178853, |
| "ce_loss": 0.47247035935462844, |
| "lb_loss": 0.9999999973578123 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 251000000, |
| "cumulative_training_bytes": 251001307, |
| "metrics": { |
| "loss": 0.482704364336454, |
| "ce_loss": 0.47270437387319714, |
| "lb_loss": 0.9999999973851328 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 252000000, |
| "cumulative_training_bytes": 252005393, |
| "metrics": { |
| "loss": 0.48272043788665514, |
| "ce_loss": 0.4727204474233983, |
| "lb_loss": 0.9999999973767315 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 253000000, |
| "cumulative_training_bytes": 253007637, |
| "metrics": { |
| "loss": 0.4827677896584801, |
| "ce_loss": 0.47276779919522327, |
| "lb_loss": 0.9999999973075784 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 254000000, |
| "cumulative_training_bytes": 254001252, |
| "metrics": { |
| "loss": 0.48291178065051366, |
| "ce_loss": 0.4729117901872568, |
| "lb_loss": 0.9999999969559885 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 255000000, |
| "cumulative_training_bytes": 255003892, |
| "metrics": { |
| "loss": 0.4829045114966359, |
| "ce_loss": 0.47290452103337904, |
| "lb_loss": 0.9999999969475445 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 256000000, |
| "cumulative_training_bytes": 256002122, |
| "metrics": { |
| "loss": 0.48277088224141823, |
| "ce_loss": 0.4727708917781614, |
| "lb_loss": 0.9999999973760579 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 257000000, |
| "cumulative_training_bytes": 257002914, |
| "metrics": { |
| "loss": 0.4828110224281261, |
| "ce_loss": 0.4728110319648693, |
| "lb_loss": 0.9999999978816977 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 258000000, |
| "cumulative_training_bytes": 258007111, |
| "metrics": { |
| "loss": 0.48270698081804553, |
| "ce_loss": 0.4727069903547887, |
| "lb_loss": 0.9999999975501774 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 259000000, |
| "cumulative_training_bytes": 259001323, |
| "metrics": { |
| "loss": 0.4827247332188862, |
| "ce_loss": 0.47272474275562937, |
| "lb_loss": 0.9999999972056884 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 260000000, |
| "cumulative_training_bytes": 260005225, |
| "metrics": { |
| "loss": 0.48267412118508785, |
| "ce_loss": 0.472674130721831, |
| "lb_loss": 0.9999999975654441 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 261000000, |
| "cumulative_training_bytes": 261001447, |
| "metrics": { |
| "loss": 0.48262796289591675, |
| "ce_loss": 0.4726279724326599, |
| "lb_loss": 0.999999997290698 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 262000000, |
| "cumulative_training_bytes": 262002819, |
| "metrics": { |
| "loss": 0.48269655004081247, |
| "ce_loss": 0.47269655957755563, |
| "lb_loss": 0.9999999973474876 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 263000000, |
| "cumulative_training_bytes": 263000855, |
| "metrics": { |
| "loss": 0.4826931146895026, |
| "ce_loss": 0.47269312422624576, |
| "lb_loss": 0.9999999976571371 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 264000000, |
| "cumulative_training_bytes": 264003342, |
| "metrics": { |
| "loss": 0.48269632542012775, |
| "ce_loss": 0.4726963349568709, |
| "lb_loss": 0.9999999976243241 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 265000000, |
| "cumulative_training_bytes": 265002459, |
| "metrics": { |
| "loss": 0.48277645857805784, |
| "ce_loss": 0.472776468114801, |
| "lb_loss": 0.9999999977475484 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 266000000, |
| "cumulative_training_bytes": 266001589, |
| "metrics": { |
| "loss": 0.4827284785946397, |
| "ce_loss": 0.47272848813138285, |
| "lb_loss": 0.9999999978448224 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 267000000, |
| "cumulative_training_bytes": 267007878, |
| "metrics": { |
| "loss": 0.48272075498110445, |
| "ce_loss": 0.4727207645178476, |
| "lb_loss": 0.9999999976659876 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 268000000, |
| "cumulative_training_bytes": 268004701, |
| "metrics": { |
| "loss": 0.48267082352839585, |
| "ce_loss": 0.472670833065139, |
| "lb_loss": 0.9999999977287622 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 269000000, |
| "cumulative_training_bytes": 269000341, |
| "metrics": { |
| "loss": 0.482630938395494, |
| "ce_loss": 0.47263094793223714, |
| "lb_loss": 0.9999999978171692 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 270000000, |
| "cumulative_training_bytes": 270005290, |
| "metrics": { |
| "loss": 0.4826039372364776, |
| "ce_loss": 0.47260394677322076, |
| "lb_loss": 0.9999999979149159 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 271000000, |
| "cumulative_training_bytes": 271001402, |
| "metrics": { |
| "loss": 0.4826388793521457, |
| "ce_loss": 0.4726388888888889, |
| "lb_loss": 0.9999999976436994 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 272000000, |
| "cumulative_training_bytes": 272007043, |
| "metrics": { |
| "loss": 0.4826473747120085, |
| "ce_loss": 0.4726473842487517, |
| "lb_loss": 0.9999999977002293 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 273000000, |
| "cumulative_training_bytes": 273004999, |
| "metrics": { |
| "loss": 0.4826480611873036, |
| "ce_loss": 0.47264807072404674, |
| "lb_loss": 0.9999999978585944 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 274000000, |
| "cumulative_training_bytes": 274003293, |
| "metrics": { |
| "loss": 0.4827403403186553, |
| "ce_loss": 0.4727403498553985, |
| "lb_loss": 0.9999999979697647 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 275000000, |
| "cumulative_training_bytes": 275006178, |
| "metrics": { |
| "loss": 0.4827058927345634, |
| "ce_loss": 0.47270590227130654, |
| "lb_loss": 0.999999997876142 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 276000000, |
| "cumulative_training_bytes": 276000051, |
| "metrics": { |
| "loss": 0.4826930918783512, |
| "ce_loss": 0.47269310141509435, |
| "lb_loss": 0.9999999979442504 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 277000000, |
| "cumulative_training_bytes": 277003305, |
| "metrics": { |
| "loss": 0.48266627628341496, |
| "ce_loss": 0.4726662858201581, |
| "lb_loss": 0.9999999977972196 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 278000000, |
| "cumulative_training_bytes": 278006257, |
| "metrics": { |
| "loss": 0.48266338925443386, |
| "ce_loss": 0.472663398791177, |
| "lb_loss": 0.9999999979102205 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 279000000, |
| "cumulative_training_bytes": 279003019, |
| "metrics": { |
| "loss": 0.48267459345141694, |
| "ce_loss": 0.4726746029881601, |
| "lb_loss": 0.9999999979052681 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 280000000, |
| "cumulative_training_bytes": 280004197, |
| "metrics": { |
| "loss": 0.4827149918388988, |
| "ce_loss": 0.472715001375642, |
| "lb_loss": 0.9999999978790717 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 281000000, |
| "cumulative_training_bytes": 281006273, |
| "metrics": { |
| "loss": 0.4827076661080715, |
| "ce_loss": 0.47270767564481464, |
| "lb_loss": 0.9999999981530354 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 282000000, |
| "cumulative_training_bytes": 282004302, |
| "metrics": { |
| "loss": 0.4826935047070867, |
| "ce_loss": 0.4726935142438299, |
| "lb_loss": 0.999999998111598 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 283000000, |
| "cumulative_training_bytes": 283002020, |
| "metrics": { |
| "loss": 0.4826445411007718, |
| "ce_loss": 0.472644550637515, |
| "lb_loss": 0.9999999981536127 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 284000000, |
| "cumulative_training_bytes": 284001833, |
| "metrics": { |
| "loss": 0.4826856697701378, |
| "ce_loss": 0.472685679306881, |
| "lb_loss": 0.9999999981439036 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 285000000, |
| "cumulative_training_bytes": 285006144, |
| "metrics": { |
| "loss": 0.48266743957449537, |
| "ce_loss": 0.47266744911123854, |
| "lb_loss": 0.9999999983399754 |
| } |
| }, |
| { |
| "epoch": 6, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.4826675374587581, |
| "ce_loss": 0.4726675469955013, |
| "lb_loss": 0.9999999983528279, |
| "training_bytes": 47653391 |
| }, |
| "cumulative_training_bytes": 285920402, |
| "training_bytes_this_epoch": 47653391 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 286000000, |
| "cumulative_training_bytes": 286004101, |
| "metrics": { |
| "loss": 0.4783948768268932, |
| "ce_loss": 0.46839488636363635, |
| "lb_loss": 0.9999999837441877 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 287000000, |
| "cumulative_training_bytes": 287000512, |
| "metrics": { |
| "loss": 0.4791101411724767, |
| "ce_loss": 0.46911015070921985, |
| "lb_loss": 1.0 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 288000000, |
| "cumulative_training_bytes": 288007385, |
| "metrics": { |
| "loss": 0.47894315667204806, |
| "ce_loss": 0.46894316620879123, |
| "lb_loss": 0.9999999986900078 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 289000000, |
| "cumulative_training_bytes": 289001067, |
| "metrics": { |
| "loss": 0.47977744083546525, |
| "ce_loss": 0.4697774503722084, |
| "lb_loss": 1.000000001331121 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 290000000, |
| "cumulative_training_bytes": 290005553, |
| "metrics": { |
| "loss": 0.47966071728909954, |
| "ce_loss": 0.4696607268258427, |
| "lb_loss": 1.0000000011161918 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 291000000, |
| "cumulative_training_bytes": 291003954, |
| "metrics": { |
| "loss": 0.479647133723799, |
| "ce_loss": 0.4696471432605422, |
| "lb_loss": 1.0000000008078942 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 292000000, |
| "cumulative_training_bytes": 292001941, |
| "metrics": { |
| "loss": 0.47993072094184624, |
| "ce_loss": 0.4699307304785894, |
| "lb_loss": 0.9999999990991741 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 293000000, |
| "cumulative_training_bytes": 293006561, |
| "metrics": { |
| "loss": 0.4797550580308244, |
| "ce_loss": 0.4697550675675676, |
| "lb_loss": 0.9999999986468134 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 294000000, |
| "cumulative_training_bytes": 294002191, |
| "metrics": { |
| "loss": 0.4799922149893232, |
| "ce_loss": 0.46999222452606637, |
| "lb_loss": 0.9999999979660974 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 295000000, |
| "cumulative_training_bytes": 295005131, |
| "metrics": { |
| "loss": 0.4803671647887206, |
| "ce_loss": 0.47036717432546377, |
| "lb_loss": 0.9999999961302212 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 296000000, |
| "cumulative_training_bytes": 296006078, |
| "metrics": { |
| "loss": 0.48054443332582325, |
| "ce_loss": 0.4705444428625664, |
| "lb_loss": 0.9999999972392686 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 297000000, |
| "cumulative_training_bytes": 297002743, |
| "metrics": { |
| "loss": 0.4806437253457724, |
| "ce_loss": 0.47064373488251554, |
| "lb_loss": 0.9999999974461037 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 298000000, |
| "cumulative_training_bytes": 298000425, |
| "metrics": { |
| "loss": 0.4805099084245758, |
| "ce_loss": 0.47050991796131897, |
| "lb_loss": 0.9999999969763022 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 299000000, |
| "cumulative_training_bytes": 299006190, |
| "metrics": { |
| "loss": 0.4805110048075191, |
| "ce_loss": 0.4705110143442623, |
| "lb_loss": 0.9999999977665707 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 300000000, |
| "cumulative_training_bytes": 300006090, |
| "metrics": { |
| "loss": 0.4805480549684227, |
| "ce_loss": 0.47054806450516584, |
| "lb_loss": 0.9999999978284333 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 301000000, |
| "cumulative_training_bytes": 301000170, |
| "metrics": { |
| "loss": 0.4807665994652883, |
| "ce_loss": 0.4707666090020315, |
| "lb_loss": 0.9999999979718074 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 302000000, |
| "cumulative_training_bytes": 302000958, |
| "metrics": { |
| "loss": 0.4808426243918283, |
| "ce_loss": 0.47084263392857145, |
| "lb_loss": 0.9999999977009637 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 303000000, |
| "cumulative_training_bytes": 303002933, |
| "metrics": { |
| "loss": 0.4808974649702044, |
| "ce_loss": 0.4708974745069476, |
| "lb_loss": 0.9999999980229746 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 304000000, |
| "cumulative_training_bytes": 304005876, |
| "metrics": { |
| "loss": 0.48091397358137705, |
| "ce_loss": 0.4709139831181202, |
| "lb_loss": 0.9999999976531617 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 305000000, |
| "cumulative_training_bytes": 305007296, |
| "metrics": { |
| "loss": 0.4807987339423901, |
| "ce_loss": 0.47079874347913325, |
| "lb_loss": 0.9999999981582834 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 306000000, |
| "cumulative_training_bytes": 306001572, |
| "metrics": { |
| "loss": 0.4808528401486115, |
| "ce_loss": 0.4708528496853547, |
| "lb_loss": 0.9999999983177942 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 307000000, |
| "cumulative_training_bytes": 307002658, |
| "metrics": { |
| "loss": 0.4808918300473469, |
| "ce_loss": 0.4708918395840901, |
| "lb_loss": 0.999999998073079 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 308000000, |
| "cumulative_training_bytes": 308001046, |
| "metrics": { |
| "loss": 0.4807877983023134, |
| "ce_loss": 0.47078780783905655, |
| "lb_loss": 0.9999999978498498 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 309000000, |
| "cumulative_training_bytes": 309007041, |
| "metrics": { |
| "loss": 0.48085081446823197, |
| "ce_loss": 0.47085082400497513, |
| "lb_loss": 0.9999999974695208 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 310000000, |
| "cumulative_training_bytes": 310000806, |
| "metrics": { |
| "loss": 0.4808527825936225, |
| "ce_loss": 0.4708527921303657, |
| "lb_loss": 0.9999999975172628 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 311000000, |
| "cumulative_training_bytes": 311006686, |
| "metrics": { |
| "loss": 0.4808801844879821, |
| "ce_loss": 0.47088019402472525, |
| "lb_loss": 0.9999999978712627 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 312000000, |
| "cumulative_training_bytes": 312003061, |
| "metrics": { |
| "loss": 0.48086826519901726, |
| "ce_loss": 0.47086827473576043, |
| "lb_loss": 0.9999999976900138 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 313000000, |
| "cumulative_training_bytes": 313000859, |
| "metrics": { |
| "loss": 0.4808953347788677, |
| "ce_loss": 0.47089534431561086, |
| "lb_loss": 0.9999999976232311 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 314000000, |
| "cumulative_training_bytes": 314006832, |
| "metrics": { |
| "loss": 0.4809533795717628, |
| "ce_loss": 0.470953389108506, |
| "lb_loss": 0.9999999977575134 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 315000000, |
| "cumulative_training_bytes": 315007663, |
| "metrics": { |
| "loss": 0.4810331776506746, |
| "ce_loss": 0.47103318718741777, |
| "lb_loss": 0.9999999979289779 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 316000000, |
| "cumulative_training_bytes": 316001449, |
| "metrics": { |
| "loss": 0.4811241632807677, |
| "ce_loss": 0.47112417281751084, |
| "lb_loss": 0.9999999979064798 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 317000000, |
| "cumulative_training_bytes": 317001561, |
| "metrics": { |
| "loss": 0.48121448784626175, |
| "ce_loss": 0.4712144973830049, |
| "lb_loss": 0.9999999978419007 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 318000000, |
| "cumulative_training_bytes": 318006954, |
| "metrics": { |
| "loss": 0.48121900889561187, |
| "ce_loss": 0.47121901843235503, |
| "lb_loss": 0.999999997795581 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 319000000, |
| "cumulative_training_bytes": 319003501, |
| "metrics": { |
| "loss": 0.4812446206125278, |
| "ce_loss": 0.471244630149271, |
| "lb_loss": 0.999999997834314 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 320000000, |
| "cumulative_training_bytes": 320000438, |
| "metrics": { |
| "loss": 0.48125381685058555, |
| "ce_loss": 0.4712538263873287, |
| "lb_loss": 0.999999997991306 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 321000000, |
| "cumulative_training_bytes": 321001768, |
| "metrics": { |
| "loss": 0.48127089427163744, |
| "ce_loss": 0.4712709038083806, |
| "lb_loss": 0.9999999977365325 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 322000000, |
| "cumulative_training_bytes": 322002583, |
| "metrics": { |
| "loss": 0.4812634008441183, |
| "ce_loss": 0.47126341038086145, |
| "lb_loss": 0.9999999978753278 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 323000000, |
| "cumulative_training_bytes": 323002023, |
| "metrics": { |
| "loss": 0.48125308707762515, |
| "ce_loss": 0.4712530966143683, |
| "lb_loss": 0.9999999980927498 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 324000000, |
| "cumulative_training_bytes": 324002741, |
| "metrics": { |
| "loss": 0.4813194785285835, |
| "ce_loss": 0.47131948806532664, |
| "lb_loss": 0.99999999827476 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 325000000, |
| "cumulative_training_bytes": 325006173, |
| "metrics": { |
| "loss": 0.48130787273411463, |
| "ce_loss": 0.4713078822708578, |
| "lb_loss": 0.9999999981555946 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 326000000, |
| "cumulative_training_bytes": 326000306, |
| "metrics": { |
| "loss": 0.4813103896229207, |
| "ce_loss": 0.4713103991596639, |
| "lb_loss": 0.9999999982583059 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 327000000, |
| "cumulative_training_bytes": 327005036, |
| "metrics": { |
| "loss": 0.48133414187465984, |
| "ce_loss": 0.471334151411403, |
| "lb_loss": 0.9999999982786063 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 328000000, |
| "cumulative_training_bytes": 328006990, |
| "metrics": { |
| "loss": 0.4813546307870109, |
| "ce_loss": 0.47135464032375407, |
| "lb_loss": 0.9999999983087805 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 329000000, |
| "cumulative_training_bytes": 329005070, |
| "metrics": { |
| "loss": 0.4813430519415795, |
| "ce_loss": 0.47134306147832267, |
| "lb_loss": 0.9999999982843013 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 330000000, |
| "cumulative_training_bytes": 330001490, |
| "metrics": { |
| "loss": 0.48131808865273235, |
| "ce_loss": 0.4713180981894755, |
| "lb_loss": 0.9999999983954985 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 331000000, |
| "cumulative_training_bytes": 331007691, |
| "metrics": { |
| "loss": 0.48127986607881124, |
| "ce_loss": 0.4712798756155544, |
| "lb_loss": 0.9999999983502196 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 332000000, |
| "cumulative_training_bytes": 332005409, |
| "metrics": { |
| "loss": 0.4814225170457456, |
| "ce_loss": 0.4714225265824888, |
| "lb_loss": 0.9999999983561437 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 333000000, |
| "cumulative_training_bytes": 333001906, |
| "metrics": { |
| "loss": 0.4816718959560199, |
| "ce_loss": 0.47167190549276306, |
| "lb_loss": 0.9999999982551901 |
| } |
| }, |
| { |
| "epoch": 7, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.4817229831126784, |
| "ce_loss": 0.4717229926494216, |
| "lb_loss": 0.9999999982283324, |
| "training_bytes": 47653415 |
| }, |
| "cumulative_training_bytes": 333573817, |
| "training_bytes_this_epoch": 47653415 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 334000000, |
| "cumulative_training_bytes": 334004035, |
| "metrics": { |
| "loss": 0.4830050127846854, |
| "ce_loss": 0.47300502232142855, |
| "lb_loss": 0.9999999872275761 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 335000000, |
| "cumulative_training_bytes": 335002482, |
| "metrics": { |
| "loss": 0.4816376752750848, |
| "ce_loss": 0.47163768481182794, |
| "lb_loss": 0.9999999945522636 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 336000000, |
| "cumulative_training_bytes": 336003192, |
| "metrics": { |
| "loss": 0.48123298533707387, |
| "ce_loss": 0.47123299487381703, |
| "lb_loss": 0.9999999947352364 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 337000000, |
| "cumulative_training_bytes": 337002345, |
| "metrics": { |
| "loss": 0.48121870550799956, |
| "ce_loss": 0.4712187150447427, |
| "lb_loss": 0.9999999915993454 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 338000000, |
| "cumulative_training_bytes": 338006062, |
| "metrics": { |
| "loss": 0.48133054421973354, |
| "ce_loss": 0.4713305537564767, |
| "lb_loss": 0.9999999932056881 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 339000000, |
| "cumulative_training_bytes": 339006251, |
| "metrics": { |
| "loss": 0.48139633729424275, |
| "ce_loss": 0.4713963468309859, |
| "lb_loss": 0.9999999942074359 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 340000000, |
| "cumulative_training_bytes": 340006632, |
| "metrics": { |
| "loss": 0.48136963969036745, |
| "ce_loss": 0.4713696492271106, |
| "lb_loss": 0.9999999957475878 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 341000000, |
| "cumulative_training_bytes": 341002987, |
| "metrics": { |
| "loss": 0.4813286786712898, |
| "ce_loss": 0.471328688208033, |
| "lb_loss": 0.9999999964396814 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 342000000, |
| "cumulative_training_bytes": 342001745, |
| "metrics": { |
| "loss": 0.4810277493188427, |
| "ce_loss": 0.47102775885558584, |
| "lb_loss": 0.9999999965352432 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 343000000, |
| "cumulative_training_bytes": 343001355, |
| "metrics": { |
| "loss": 0.4808982693015996, |
| "ce_loss": 0.4708982788383428, |
| "lb_loss": 0.9999999960780047 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 344000000, |
| "cumulative_training_bytes": 344002997, |
| "metrics": { |
| "loss": 0.4810716566710395, |
| "ce_loss": 0.4710716662077827, |
| "lb_loss": 0.9999999964552304 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 345000000, |
| "cumulative_training_bytes": 345008016, |
| "metrics": { |
| "loss": 0.48147442988762096, |
| "ce_loss": 0.4714744394243641, |
| "lb_loss": 0.9999999960901906 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 346000000, |
| "cumulative_training_bytes": 346005271, |
| "metrics": { |
| "loss": 0.481584667572247, |
| "ce_loss": 0.47158467710899016, |
| "lb_loss": 0.9999999966233822 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 347000000, |
| "cumulative_training_bytes": 347007029, |
| "metrics": { |
| "loss": 0.4815711710188124, |
| "ce_loss": 0.4715711805555556, |
| "lb_loss": 0.9999999961282453 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 348000000, |
| "cumulative_training_bytes": 348004167, |
| "metrics": { |
| "loss": 0.48147504019800486, |
| "ce_loss": 0.471475049734748, |
| "lb_loss": 0.9999999969960524 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 349000000, |
| "cumulative_training_bytes": 349004042, |
| "metrics": { |
| "loss": 0.4814087606245472, |
| "ce_loss": 0.47140877016129035, |
| "lb_loss": 0.9999999970419531 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 350000000, |
| "cumulative_training_bytes": 350007816, |
| "metrics": { |
| "loss": 0.481473079932036, |
| "ce_loss": 0.47147308946877914, |
| "lb_loss": 0.9999999975835955 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 351000000, |
| "cumulative_training_bytes": 351001302, |
| "metrics": { |
| "loss": 0.4814454059399704, |
| "ce_loss": 0.47144541547671354, |
| "lb_loss": 0.9999999974073551 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 352000000, |
| "cumulative_training_bytes": 352004241, |
| "metrics": { |
| "loss": 0.481444767415895, |
| "ce_loss": 0.47144477695263814, |
| "lb_loss": 0.9999999979446674 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 353000000, |
| "cumulative_training_bytes": 353002044, |
| "metrics": { |
| "loss": 0.48147728557752, |
| "ce_loss": 0.4714772951142632, |
| "lb_loss": 0.9999999978863601 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 354000000, |
| "cumulative_training_bytes": 354002090, |
| "metrics": { |
| "loss": 0.48136562600486105, |
| "ce_loss": 0.4713656355416042, |
| "lb_loss": 0.9999999975648777 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 355000000, |
| "cumulative_training_bytes": 355007491, |
| "metrics": { |
| "loss": 0.48128996009526825, |
| "ce_loss": 0.4712899696320114, |
| "lb_loss": 0.9999999974658975 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 356000000, |
| "cumulative_training_bytes": 356005562, |
| "metrics": { |
| "loss": 0.4812372408046703, |
| "ce_loss": 0.47123725034141345, |
| "lb_loss": 0.9999999976190702 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 357000000, |
| "cumulative_training_bytes": 357002412, |
| "metrics": { |
| "loss": 0.48114877785292665, |
| "ce_loss": 0.4711487873896698, |
| "lb_loss": 0.999999997993044 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 358000000, |
| "cumulative_training_bytes": 358000907, |
| "metrics": { |
| "loss": 0.48121007079241046, |
| "ce_loss": 0.4712100803291536, |
| "lb_loss": 0.9999999978512432 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 359000000, |
| "cumulative_training_bytes": 359004191, |
| "metrics": { |
| "loss": 0.4811665489170358, |
| "ce_loss": 0.47116655845377897, |
| "lb_loss": 0.9999999974155167 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 360000000, |
| "cumulative_training_bytes": 360005688, |
| "metrics": { |
| "loss": 0.4811812123998154, |
| "ce_loss": 0.47118122193655854, |
| "lb_loss": 0.9999999971682614 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 361000000, |
| "cumulative_training_bytes": 361001697, |
| "metrics": { |
| "loss": 0.4811540463328827, |
| "ce_loss": 0.4711540558696259, |
| "lb_loss": 0.9999999971379121 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 362000000, |
| "cumulative_training_bytes": 362004914, |
| "metrics": { |
| "loss": 0.4811149918098768, |
| "ce_loss": 0.47111500134662, |
| "lb_loss": 0.9999999970623081 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 363000000, |
| "cumulative_training_bytes": 363004386, |
| "metrics": { |
| "loss": 0.48118021857850135, |
| "ce_loss": 0.4711802281152445, |
| "lb_loss": 0.9999999971469161 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 364000000, |
| "cumulative_training_bytes": 364001806, |
| "metrics": { |
| "loss": 0.4812275201877158, |
| "ce_loss": 0.47122752972445897, |
| "lb_loss": 0.9999999971802533 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 365000000, |
| "cumulative_training_bytes": 365007448, |
| "metrics": { |
| "loss": 0.48119603568250163, |
| "ce_loss": 0.4711960452192448, |
| "lb_loss": 0.9999999972121579 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 366000000, |
| "cumulative_training_bytes": 366007819, |
| "metrics": { |
| "loss": 0.4811812530045694, |
| "ce_loss": 0.47118126254131254, |
| "lb_loss": 0.9999999974109408 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 367000000, |
| "cumulative_training_bytes": 367003761, |
| "metrics": { |
| "loss": 0.4811495695974758, |
| "ce_loss": 0.47114957913421895, |
| "lb_loss": 0.9999999974880315 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 368000000, |
| "cumulative_training_bytes": 368001579, |
| "metrics": { |
| "loss": 0.48113743278591636, |
| "ce_loss": 0.4711374423226595, |
| "lb_loss": 0.9999999974816806 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 369000000, |
| "cumulative_training_bytes": 369002487, |
| "metrics": { |
| "loss": 0.4811002396259621, |
| "ce_loss": 0.47110024916270526, |
| "lb_loss": 0.9999999975400848 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 370000000, |
| "cumulative_training_bytes": 370006595, |
| "metrics": { |
| "loss": 0.4810897184785962, |
| "ce_loss": 0.4710897280153394, |
| "lb_loss": 0.9999999973698308 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 371000000, |
| "cumulative_training_bytes": 371003993, |
| "metrics": { |
| "loss": 0.4811078026436618, |
| "ce_loss": 0.471107812180405, |
| "lb_loss": 0.9999999973422351 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 372000000, |
| "cumulative_training_bytes": 372005909, |
| "metrics": { |
| "loss": 0.4811544351843724, |
| "ce_loss": 0.47115444472111556, |
| "lb_loss": 0.9999999974828317 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 373000000, |
| "cumulative_training_bytes": 373004014, |
| "metrics": { |
| "loss": 0.4811229425263636, |
| "ce_loss": 0.4711229520631068, |
| "lb_loss": 0.9999999972570289 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 374000000, |
| "cumulative_training_bytes": 374004738, |
| "metrics": { |
| "loss": 0.4811517313740692, |
| "ce_loss": 0.47115174091081236, |
| "lb_loss": 0.9999999972347778 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 375000000, |
| "cumulative_training_bytes": 375000531, |
| "metrics": { |
| "loss": 0.4811597229526303, |
| "ce_loss": 0.4711597324893735, |
| "lb_loss": 0.9999999972461354 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 376000000, |
| "cumulative_training_bytes": 376004266, |
| "metrics": { |
| "loss": 0.48114152894214535, |
| "ce_loss": 0.4711415384788885, |
| "lb_loss": 0.9999999971499042 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 377000000, |
| "cumulative_training_bytes": 377004301, |
| "metrics": { |
| "loss": 0.48106536643293246, |
| "ce_loss": 0.4710653759696756, |
| "lb_loss": 0.9999999970260729 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 378000000, |
| "cumulative_training_bytes": 378002475, |
| "metrics": { |
| "loss": 0.48104782418438746, |
| "ce_loss": 0.4710478337211306, |
| "lb_loss": 0.9999999969180638 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 379000000, |
| "cumulative_training_bytes": 379003430, |
| "metrics": { |
| "loss": 0.4810510765495538, |
| "ce_loss": 0.471051086086297, |
| "lb_loss": 0.9999999969760663 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 380000000, |
| "cumulative_training_bytes": 380004459, |
| "metrics": { |
| "loss": 0.4810564429904666, |
| "ce_loss": 0.4710564525272098, |
| "lb_loss": 0.9999999969922458 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 381000000, |
| "cumulative_training_bytes": 381002688, |
| "metrics": { |
| "loss": 0.4810578564424302, |
| "ce_loss": 0.4710578659791734, |
| "lb_loss": 0.9999999970938646 |
| } |
| }, |
| { |
| "epoch": 8, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.4810454771880319, |
| "ce_loss": 0.4710454867247751, |
| "lb_loss": 0.9999999970791426, |
| "training_bytes": 47653397 |
| }, |
| "cumulative_training_bytes": 381227214, |
| "training_bytes_this_epoch": 47653397 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 382000000, |
| "cumulative_training_bytes": 382003673, |
| "metrics": { |
| "loss": 0.47660348675038555, |
| "ce_loss": 0.4666034962871287, |
| "lb_loss": 0.99999999763942 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 383000000, |
| "cumulative_training_bytes": 383005458, |
| "metrics": { |
| "loss": 0.47849743119601545, |
| "ce_loss": 0.4684974407327586, |
| "lb_loss": 0.9999999971739177 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 384000000, |
| "cumulative_training_bytes": 384002363, |
| "metrics": { |
| "loss": 0.4783237560019309, |
| "ce_loss": 0.46832376553867405, |
| "lb_loss": 0.9999999990120777 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 385000000, |
| "cumulative_training_bytes": 385005560, |
| "metrics": { |
| "loss": 0.4780487670859749, |
| "ce_loss": 0.46804877662271804, |
| "lb_loss": 0.999999998549177 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 386000000, |
| "cumulative_training_bytes": 386002393, |
| "metrics": { |
| "loss": 0.47821703500579293, |
| "ce_loss": 0.4682170445425361, |
| "lb_loss": 0.999999998373549 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 387000000, |
| "cumulative_training_bytes": 387003159, |
| "metrics": { |
| "loss": 0.4784002930163072, |
| "ce_loss": 0.4684003025530504, |
| "lb_loss": 0.9999999972332061 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 388000000, |
| "cumulative_training_bytes": 388000670, |
| "metrics": { |
| "loss": 0.47836776124945596, |
| "ce_loss": 0.4683677707861991, |
| "lb_loss": 0.999999998044644 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 389000000, |
| "cumulative_training_bytes": 389001849, |
| "metrics": { |
| "loss": 0.47809766595586767, |
| "ce_loss": 0.46809767549261083, |
| "lb_loss": 0.9999999978272198 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 390000000, |
| "cumulative_training_bytes": 390002629, |
| "metrics": { |
| "loss": 0.47807849722590956, |
| "ce_loss": 0.4680785067626527, |
| "lb_loss": 0.99999999849168 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 391000000, |
| "cumulative_training_bytes": 391006180, |
| "metrics": { |
| "loss": 0.47816267644602894, |
| "ce_loss": 0.4681626859827721, |
| "lb_loss": 0.9999999978062504 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 392000000, |
| "cumulative_training_bytes": 392004730, |
| "metrics": { |
| "loss": 0.478326905857433, |
| "ce_loss": 0.46832691539417615, |
| "lb_loss": 0.9999999977563593 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 393000000, |
| "cumulative_training_bytes": 393003485, |
| "metrics": { |
| "loss": 0.47875887407274353, |
| "ce_loss": 0.4687588836094867, |
| "lb_loss": 0.999999997250208 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 394000000, |
| "cumulative_training_bytes": 394000207, |
| "metrics": { |
| "loss": 0.4788389284500753, |
| "ce_loss": 0.46883893798681847, |
| "lb_loss": 0.9999999963930083 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 395000000, |
| "cumulative_training_bytes": 395003542, |
| "metrics": { |
| "loss": 0.4789453029632568, |
| "ce_loss": 0.4689453125, |
| "lb_loss": 0.9999999956621064 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 396000000, |
| "cumulative_training_bytes": 396000470, |
| "metrics": { |
| "loss": 0.47899286611092523, |
| "ce_loss": 0.4689928756476684, |
| "lb_loss": 0.9999999959542961 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 397000000, |
| "cumulative_training_bytes": 397005545, |
| "metrics": { |
| "loss": 0.4790248115452559, |
| "ce_loss": 0.46902482108199905, |
| "lb_loss": 0.9999999956619618 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 398000000, |
| "cumulative_training_bytes": 398006179, |
| "metrics": { |
| "loss": 0.47905739380495393, |
| "ce_loss": 0.4690574033416971, |
| "lb_loss": 0.9999999957308717 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 399000000, |
| "cumulative_training_bytes": 399000874, |
| "metrics": { |
| "loss": 0.4790856049658408, |
| "ce_loss": 0.46908561450258396, |
| "lb_loss": 0.9999999954051544 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 400000000, |
| "cumulative_training_bytes": 400002954, |
| "metrics": { |
| "loss": 0.4790493687755275, |
| "ce_loss": 0.46904937831227067, |
| "lb_loss": 0.9999999951159667 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 401000000, |
| "cumulative_training_bytes": 401003161, |
| "metrics": { |
| "loss": 0.4790388382178058, |
| "ce_loss": 0.46903884775454896, |
| "lb_loss": 0.9999999944848974 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 402000000, |
| "cumulative_training_bytes": 402007261, |
| "metrics": { |
| "loss": 0.47903641041535705, |
| "ce_loss": 0.4690364199521002, |
| "lb_loss": 0.9999999945973682 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 403000000, |
| "cumulative_training_bytes": 403002393, |
| "metrics": { |
| "loss": 0.47906452325158316, |
| "ce_loss": 0.4690645327883263, |
| "lb_loss": 0.99999999488624 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 404000000, |
| "cumulative_training_bytes": 404005905, |
| "metrics": { |
| "loss": 0.4790848118918283, |
| "ce_loss": 0.46908482142857144, |
| "lb_loss": 0.9999999945704676 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 405000000, |
| "cumulative_training_bytes": 405004571, |
| "metrics": { |
| "loss": 0.4790582130880556, |
| "ce_loss": 0.46905822262479874, |
| "lb_loss": 0.9999999946442203 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 406000000, |
| "cumulative_training_bytes": 406003780, |
| "metrics": { |
| "loss": 0.47909593810622436, |
| "ce_loss": 0.4690959476429675, |
| "lb_loss": 0.9999999946936204 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 407000000, |
| "cumulative_training_bytes": 407001187, |
| "metrics": { |
| "loss": 0.4790645799877739, |
| "ce_loss": 0.4690645895245171, |
| "lb_loss": 0.9999999947037774 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 408000000, |
| "cumulative_training_bytes": 408003413, |
| "metrics": { |
| "loss": 0.47919748849672367, |
| "ce_loss": 0.46919749803346683, |
| "lb_loss": 0.9999999948169874 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 409000000, |
| "cumulative_training_bytes": 409005684, |
| "metrics": { |
| "loss": 0.4793487980177096, |
| "ce_loss": 0.46934880755445274, |
| "lb_loss": 0.9999999948891524 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 410000000, |
| "cumulative_training_bytes": 410004022, |
| "metrics": { |
| "loss": 0.4794526581056198, |
| "ce_loss": 0.469452667642363, |
| "lb_loss": 0.9999999954638296 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 411000000, |
| "cumulative_training_bytes": 411000093, |
| "metrics": { |
| "loss": 0.479534154075654, |
| "ce_loss": 0.46953416361239714, |
| "lb_loss": 0.9999999953702153 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 412000000, |
| "cumulative_training_bytes": 412003081, |
| "metrics": { |
| "loss": 0.479515883720286, |
| "ce_loss": 0.46951589325702914, |
| "lb_loss": 0.9999999951948483 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 413000000, |
| "cumulative_training_bytes": 413002872, |
| "metrics": { |
| "loss": 0.47955900778253396, |
| "ce_loss": 0.4695590173192771, |
| "lb_loss": 0.9999999952316284 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 414000000, |
| "cumulative_training_bytes": 414004353, |
| "metrics": { |
| "loss": 0.47949838415484564, |
| "ce_loss": 0.4694983936915888, |
| "lb_loss": 0.9999999954460937 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 415000000, |
| "cumulative_training_bytes": 415000550, |
| "metrics": { |
| "loss": 0.47950582459950714, |
| "ce_loss": 0.4695058341362503, |
| "lb_loss": 0.9999999952840578 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 416000000, |
| "cumulative_training_bytes": 416004238, |
| "metrics": { |
| "loss": 0.47945263384722864, |
| "ce_loss": 0.4694526433839718, |
| "lb_loss": 0.9999999952625986 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 417000000, |
| "cumulative_training_bytes": 417004582, |
| "metrics": { |
| "loss": 0.47940799961351366, |
| "ce_loss": 0.4694080091502568, |
| "lb_loss": 0.9999999954964812 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 418000000, |
| "cumulative_training_bytes": 418000528, |
| "metrics": { |
| "loss": 0.4794532293480965, |
| "ce_loss": 0.46945323888483964, |
| "lb_loss": 0.9999999955563385 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 419000000, |
| "cumulative_training_bytes": 419005587, |
| "metrics": { |
| "loss": 0.4794745432201999, |
| "ce_loss": 0.46947455275694305, |
| "lb_loss": 0.9999999953964384 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 420000000, |
| "cumulative_training_bytes": 420003878, |
| "metrics": { |
| "loss": 0.47951866702822527, |
| "ce_loss": 0.46951867656496843, |
| "lb_loss": 0.9999999954449057 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 421000000, |
| "cumulative_training_bytes": 421005882, |
| "metrics": { |
| "loss": 0.4795383837500711, |
| "ce_loss": 0.46953839328681424, |
| "lb_loss": 0.999999995399141 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 422000000, |
| "cumulative_training_bytes": 422001156, |
| "metrics": { |
| "loss": 0.4795866248529282, |
| "ce_loss": 0.46958663438967135, |
| "lb_loss": 0.9999999953547554 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 423000000, |
| "cumulative_training_bytes": 423005818, |
| "metrics": { |
| "loss": 0.47955937749129934, |
| "ce_loss": 0.4695593870280425, |
| "lb_loss": 0.9999999953570429 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 424000000, |
| "cumulative_training_bytes": 424005361, |
| "metrics": { |
| "loss": 0.47958130087134704, |
| "ce_loss": 0.4695813104080902, |
| "lb_loss": 0.9999999956472714 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 425000000, |
| "cumulative_training_bytes": 425000747, |
| "metrics": { |
| "loss": 0.4796109096516423, |
| "ce_loss": 0.4696109191883855, |
| "lb_loss": 0.9999999956836937 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 426000000, |
| "cumulative_training_bytes": 426002610, |
| "metrics": { |
| "loss": 0.479592292318592, |
| "ce_loss": 0.4695923018553352, |
| "lb_loss": 0.9999999958517287 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 427000000, |
| "cumulative_training_bytes": 427005445, |
| "metrics": { |
| "loss": 0.4796241437497596, |
| "ce_loss": 0.46962415328650275, |
| "lb_loss": 0.9999999959326484 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 428000000, |
| "cumulative_training_bytes": 428003339, |
| "metrics": { |
| "loss": 0.47960202485923, |
| "ce_loss": 0.46960203439597314, |
| "lb_loss": 0.9999999960387157 |
| } |
| }, |
| { |
| "epoch": 9, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.4795953822932697, |
| "ce_loss": 0.46959539183001286, |
| "lb_loss": 0.9999999961119078, |
| "training_bytes": 47653406 |
| }, |
| "cumulative_training_bytes": 428880620, |
| "training_bytes_this_epoch": 47653406 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 429000000, |
| "cumulative_training_bytes": 429002185, |
| "metrics": { |
| "loss": 0.47997069358825684, |
| "ce_loss": 0.469970703125, |
| "lb_loss": 0.9999999925494194 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 430000000, |
| "cumulative_training_bytes": 430003863, |
| "metrics": { |
| "loss": 0.476517847606114, |
| "ce_loss": 0.46651785714285715, |
| "lb_loss": 0.9999999862138916 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 431000000, |
| "cumulative_training_bytes": 431005314, |
| "metrics": { |
| "loss": 0.47611537940210574, |
| "ce_loss": 0.4661153889388489, |
| "lb_loss": 0.9999999903517661 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 432000000, |
| "cumulative_training_bytes": 432001200, |
| "metrics": { |
| "loss": 0.4762607135024725, |
| "ce_loss": 0.4662607230392157, |
| "lb_loss": 0.9999999903580722 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 433000000, |
| "cumulative_training_bytes": 433004578, |
| "metrics": { |
| "loss": 0.4760105424113088, |
| "ce_loss": 0.46601055194805197, |
| "lb_loss": 0.9999999920379695 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 434000000, |
| "cumulative_training_bytes": 434007002, |
| "metrics": { |
| "loss": 0.4758844354259434, |
| "ce_loss": 0.46588444496268655, |
| "lb_loss": 0.9999999925271789 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 435000000, |
| "cumulative_training_bytes": 435000122, |
| "metrics": { |
| "loss": 0.4758044139613795, |
| "ce_loss": 0.4658044234981227, |
| "lb_loss": 0.9999999929130898 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 436000000, |
| "cumulative_training_bytes": 436000104, |
| "metrics": { |
| "loss": 0.4760492176137945, |
| "ce_loss": 0.46604922715053765, |
| "lb_loss": 0.9999999934627164 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 437000000, |
| "cumulative_training_bytes": 437004132, |
| "metrics": { |
| "loss": 0.4761286426781485, |
| "ce_loss": 0.46612865221489164, |
| "lb_loss": 0.9999999932586641 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 438000000, |
| "cumulative_training_bytes": 438006173, |
| "metrics": { |
| "loss": 0.47627581365956556, |
| "ce_loss": 0.4662758231963087, |
| "lb_loss": 0.999999993699509 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 439000000, |
| "cumulative_training_bytes": 439002702, |
| "metrics": { |
| "loss": 0.4764555856788393, |
| "ce_loss": 0.46645559521558244, |
| "lb_loss": 0.999999995310981 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 440000000, |
| "cumulative_training_bytes": 440005689, |
| "metrics": { |
| "loss": 0.4764527498404076, |
| "ce_loss": 0.46645275937715075, |
| "lb_loss": 0.9999999952414736 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 441000000, |
| "cumulative_training_bytes": 441000190, |
| "metrics": { |
| "loss": 0.47656120729838003, |
| "ce_loss": 0.4665612168351232, |
| "lb_loss": 0.9999999955192971 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 442000000, |
| "cumulative_training_bytes": 442005265, |
| "metrics": { |
| "loss": 0.4766282175636069, |
| "ce_loss": 0.46662822710035007, |
| "lb_loss": 0.9999999957226539 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 443000000, |
| "cumulative_training_bytes": 443003341, |
| "metrics": { |
| "loss": 0.47682650638466606, |
| "ce_loss": 0.4668265159214092, |
| "lb_loss": 0.9999999960586631 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 444000000, |
| "cumulative_training_bytes": 444007542, |
| "metrics": { |
| "loss": 0.47688483516214347, |
| "ce_loss": 0.46688484469888664, |
| "lb_loss": 0.9999999959278203 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 445000000, |
| "cumulative_training_bytes": 445007323, |
| "metrics": { |
| "loss": 0.47694339620874593, |
| "ce_loss": 0.4669434057454891, |
| "lb_loss": 0.999999996235794 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 446000000, |
| "cumulative_training_bytes": 446000020, |
| "metrics": { |
| "loss": 0.4770030092467989, |
| "ce_loss": 0.46700301878354206, |
| "lb_loss": 0.9999999962413887 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 447000000, |
| "cumulative_training_bytes": 447002709, |
| "metrics": { |
| "loss": 0.4770996947302615, |
| "ce_loss": 0.46709970426700465, |
| "lb_loss": 0.9999999963990435 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 448000000, |
| "cumulative_training_bytes": 448004545, |
| "metrics": { |
| "loss": 0.47733323205653144, |
| "ce_loss": 0.4673332415932746, |
| "lb_loss": 0.9999999962538314 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 449000000, |
| "cumulative_training_bytes": 449006756, |
| "metrics": { |
| "loss": 0.4773651958645501, |
| "ce_loss": 0.4673652054012933, |
| "lb_loss": 0.9999999961684348 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 450000000, |
| "cumulative_training_bytes": 450004303, |
| "metrics": { |
| "loss": 0.47728603160135036, |
| "ce_loss": 0.4672860411380935, |
| "lb_loss": 0.9999999964785947 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 451000000, |
| "cumulative_training_bytes": 451006142, |
| "metrics": { |
| "loss": 0.47733279215010804, |
| "ce_loss": 0.4673328016868512, |
| "lb_loss": 0.9999999963082244 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 452000000, |
| "cumulative_training_bytes": 452005227, |
| "metrics": { |
| "loss": 0.477412995262992, |
| "ce_loss": 0.4674130047997352, |
| "lb_loss": 0.9999999966853425 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 453000000, |
| "cumulative_training_bytes": 453006680, |
| "metrics": { |
| "loss": 0.47739544374688625, |
| "ce_loss": 0.4673954532836294, |
| "lb_loss": 0.9999999969365633 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 454000000, |
| "cumulative_training_bytes": 454003038, |
| "metrics": { |
| "loss": 0.4773985136579552, |
| "ce_loss": 0.4673985231946984, |
| "lb_loss": 0.999999997039745 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 455000000, |
| "cumulative_training_bytes": 455007108, |
| "metrics": { |
| "loss": 0.4773937331383228, |
| "ce_loss": 0.46739374267506595, |
| "lb_loss": 0.9999999970835114 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 456000000, |
| "cumulative_training_bytes": 456006436, |
| "metrics": { |
| "loss": 0.47737679419596923, |
| "ce_loss": 0.4673768037327124, |
| "lb_loss": 0.9999999970559377 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 457000000, |
| "cumulative_training_bytes": 457003387, |
| "metrics": { |
| "loss": 0.47746527769440306, |
| "ce_loss": 0.4674652872311462, |
| "lb_loss": 0.9999999969816324 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 458000000, |
| "cumulative_training_bytes": 458006870, |
| "metrics": { |
| "loss": 0.4774874419945397, |
| "ce_loss": 0.46748745153128285, |
| "lb_loss": 0.9999999967565296 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 459000000, |
| "cumulative_training_bytes": 459003680, |
| "metrics": { |
| "loss": 0.47746660761501075, |
| "ce_loss": 0.4674666171517539, |
| "lb_loss": 0.9999999966970482 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 460000000, |
| "cumulative_training_bytes": 460005767, |
| "metrics": { |
| "loss": 0.47751757579536014, |
| "ce_loss": 0.4675175853321033, |
| "lb_loss": 0.999999996686187 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 461000000, |
| "cumulative_training_bytes": 461002393, |
| "metrics": { |
| "loss": 0.4775231765478814, |
| "ce_loss": 0.46752318608462456, |
| "lb_loss": 0.9999999966183777 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 462000000, |
| "cumulative_training_bytes": 462006702, |
| "metrics": { |
| "loss": 0.4775458794195675, |
| "ce_loss": 0.46754588895631066, |
| "lb_loss": 0.9999999969136754 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 463000000, |
| "cumulative_training_bytes": 463005861, |
| "metrics": { |
| "loss": 0.47760887816799097, |
| "ce_loss": 0.46760888770473413, |
| "lb_loss": 0.9999999970980014 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 464000000, |
| "cumulative_training_bytes": 464001234, |
| "metrics": { |
| "loss": 0.4776744304567166, |
| "ce_loss": 0.46767443999345976, |
| "lb_loss": 0.9999999970632986 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 465000000, |
| "cumulative_training_bytes": 465004728, |
| "metrics": { |
| "loss": 0.4777479740137972, |
| "ce_loss": 0.46774798355054037, |
| "lb_loss": 0.9999999972338595 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 466000000, |
| "cumulative_training_bytes": 466001776, |
| "metrics": { |
| "loss": 0.477748253636076, |
| "ce_loss": 0.46774826317281915, |
| "lb_loss": 0.9999999973694795 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 467000000, |
| "cumulative_training_bytes": 467002535, |
| "metrics": { |
| "loss": 0.477785585970285, |
| "ce_loss": 0.46778559550702814, |
| "lb_loss": 0.9999999974626135 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 468000000, |
| "cumulative_training_bytes": 468004269, |
| "metrics": { |
| "loss": 0.4778305581114666, |
| "ce_loss": 0.46783056764820974, |
| "lb_loss": 0.9999999974810011 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 469000000, |
| "cumulative_training_bytes": 469006908, |
| "metrics": { |
| "loss": 0.4778490640515819, |
| "ce_loss": 0.46784907358832506, |
| "lb_loss": 0.999999997464358 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 470000000, |
| "cumulative_training_bytes": 470004722, |
| "metrics": { |
| "loss": 0.47785450519240796, |
| "ce_loss": 0.4678545147291511, |
| "lb_loss": 0.9999999974369559 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 471000000, |
| "cumulative_training_bytes": 471002302, |
| "metrics": { |
| "loss": 0.4778841831431914, |
| "ce_loss": 0.4678841926799346, |
| "lb_loss": 0.9999999974975149 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 472000000, |
| "cumulative_training_bytes": 472007065, |
| "metrics": { |
| "loss": 0.4779119462372671, |
| "ce_loss": 0.4679119557740103, |
| "lb_loss": 0.9999999975768749 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 473000000, |
| "cumulative_training_bytes": 473006577, |
| "metrics": { |
| "loss": 0.4779428524080868, |
| "ce_loss": 0.46794286194483, |
| "lb_loss": 0.9999999976112642 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 474000000, |
| "cumulative_training_bytes": 474004871, |
| "metrics": { |
| "loss": 0.4779470685299348, |
| "ce_loss": 0.46794707806667796, |
| "lb_loss": 0.9999999978662063 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 475000000, |
| "cumulative_training_bytes": 475004074, |
| "metrics": { |
| "loss": 0.47796439586249323, |
| "ce_loss": 0.4679644053992364, |
| "lb_loss": 0.9999999978924653 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 476000000, |
| "cumulative_training_bytes": 476004262, |
| "metrics": { |
| "loss": 0.4779159292835363, |
| "ce_loss": 0.4679159388202795, |
| "lb_loss": 0.9999999980532119 |
| } |
| }, |
| { |
| "epoch": 10, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.477952924669555, |
| "ce_loss": 0.4679529342062982, |
| "lb_loss": 0.9999999979314583, |
| "training_bytes": 47653389 |
| }, |
| "cumulative_training_bytes": 476534009, |
| "training_bytes_this_epoch": 47653389 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 477000000, |
| "cumulative_training_bytes": 477001318, |
| "metrics": { |
| "loss": 0.47401126095506013, |
| "ce_loss": 0.4640112704918033, |
| "lb_loss": 1.0000000039085013 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 478000000, |
| "cumulative_training_bytes": 478000704, |
| "metrics": { |
| "loss": 0.47289061546325684, |
| "ce_loss": 0.462890625, |
| "lb_loss": 0.9999999971914042 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 479000000, |
| "cumulative_training_bytes": 479001718, |
| "metrics": { |
| "loss": 0.4746253786620146, |
| "ce_loss": 0.46462538819875776, |
| "lb_loss": 0.999999995927633 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 480000000, |
| "cumulative_training_bytes": 480006675, |
| "metrics": { |
| "loss": 0.47415389498864313, |
| "ce_loss": 0.4641539045253863, |
| "lb_loss": 0.9999999944737415 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 481000000, |
| "cumulative_training_bytes": 481002129, |
| "metrics": { |
| "loss": 0.4743378710378709, |
| "ce_loss": 0.46433788057461406, |
| "lb_loss": 0.9999999941724447 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 482000000, |
| "cumulative_training_bytes": 482003899, |
| "metrics": { |
| "loss": 0.47443615555429325, |
| "ce_loss": 0.4644361650910364, |
| "lb_loss": 0.999999993655528 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 483000000, |
| "cumulative_training_bytes": 483006727, |
| "metrics": { |
| "loss": 0.47455712803722133, |
| "ce_loss": 0.4645571375739645, |
| "lb_loss": 0.9999999937221143 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 484000000, |
| "cumulative_training_bytes": 484005792, |
| "metrics": { |
| "loss": 0.47493579348579784, |
| "ce_loss": 0.464935803022541, |
| "lb_loss": 0.999999994320459 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 485000000, |
| "cumulative_training_bytes": 485004003, |
| "metrics": { |
| "loss": 0.47497618564860944, |
| "ce_loss": 0.4649761951853526, |
| "lb_loss": 0.999999995473065 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 486000000, |
| "cumulative_training_bytes": 486002749, |
| "metrics": { |
| "loss": 0.4750080810781436, |
| "ce_loss": 0.46500809061488674, |
| "lb_loss": 0.9999999949364986 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 487000000, |
| "cumulative_training_bytes": 487001523, |
| "metrics": { |
| "loss": 0.47491094168946474, |
| "ce_loss": 0.4649109512262079, |
| "lb_loss": 0.9999999950256738 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 488000000, |
| "cumulative_training_bytes": 488005452, |
| "metrics": { |
| "loss": 0.4753404965547121, |
| "ce_loss": 0.4653405060914553, |
| "lb_loss": 0.9999999949069462 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 489000000, |
| "cumulative_training_bytes": 489002785, |
| "metrics": { |
| "loss": 0.4756307490627654, |
| "ce_loss": 0.4656307585995086, |
| "lb_loss": 0.9999999955333129 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 490000000, |
| "cumulative_training_bytes": 490003743, |
| "metrics": { |
| "loss": 0.4758719347497833, |
| "ce_loss": 0.46587194428652645, |
| "lb_loss": 0.9999999956287668 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 491000000, |
| "cumulative_training_bytes": 491003982, |
| "metrics": { |
| "loss": 0.4757608510376348, |
| "ce_loss": 0.465760860574378, |
| "lb_loss": 0.9999999958980392 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 492000000, |
| "cumulative_training_bytes": 492005024, |
| "metrics": { |
| "loss": 0.4758957206612766, |
| "ce_loss": 0.4658957301980198, |
| "lb_loss": 0.9999999958984923 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 493000000, |
| "cumulative_training_bytes": 493004726, |
| "metrics": { |
| "loss": 0.4758511895911638, |
| "ce_loss": 0.46585119912790696, |
| "lb_loss": 0.9999999962019366 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 494000000, |
| "cumulative_training_bytes": 494006508, |
| "metrics": { |
| "loss": 0.47593461343563737, |
| "ce_loss": 0.46593462297238053, |
| "lb_loss": 0.9999999958974445 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 495000000, |
| "cumulative_training_bytes": 495003897, |
| "metrics": { |
| "loss": 0.4759487020974335, |
| "ce_loss": 0.46594871163417667, |
| "lb_loss": 0.9999999958219888 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 496000000, |
| "cumulative_training_bytes": 496001530, |
| "metrics": { |
| "loss": 0.4761391648338312, |
| "ce_loss": 0.46613917437057434, |
| "lb_loss": 0.9999999956152366 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 497000000, |
| "cumulative_training_bytes": 497005237, |
| "metrics": { |
| "loss": 0.47611367359082885, |
| "ce_loss": 0.466113683127572, |
| "lb_loss": 0.999999995807829 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 498000000, |
| "cumulative_training_bytes": 498004778, |
| "metrics": { |
| "loss": 0.4762298741116163, |
| "ce_loss": 0.4662298836483595, |
| "lb_loss": 0.9999999961099679 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 499000000, |
| "cumulative_training_bytes": 499000975, |
| "metrics": { |
| "loss": 0.47627962683851244, |
| "ce_loss": 0.4662796363752556, |
| "lb_loss": 0.9999999963635885 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 500000000, |
| "cumulative_training_bytes": 500006091, |
| "metrics": { |
| "loss": 0.47628707753503496, |
| "ce_loss": 0.46628708707177813, |
| "lb_loss": 0.999999996227308 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 501000000, |
| "cumulative_training_bytes": 501000599, |
| "metrics": { |
| "loss": 0.4763518341643523, |
| "ce_loss": 0.46635184370109545, |
| "lb_loss": 0.9999999964181246 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 502000000, |
| "cumulative_training_bytes": 502000922, |
| "metrics": { |
| "loss": 0.4764045962209237, |
| "ce_loss": 0.46640460575766685, |
| "lb_loss": 0.9999999965412819 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 503000000, |
| "cumulative_training_bytes": 503003452, |
| "metrics": { |
| "loss": 0.47639347198625365, |
| "ce_loss": 0.4663934815229968, |
| "lb_loss": 0.999999996551655 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 504000000, |
| "cumulative_training_bytes": 504004414, |
| "metrics": { |
| "loss": 0.4764996507893438, |
| "ce_loss": 0.466499660326087, |
| "lb_loss": 0.9999999965446583 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 505000000, |
| "cumulative_training_bytes": 505006535, |
| "metrics": { |
| "loss": 0.47654373122018073, |
| "ce_loss": 0.4665437407569239, |
| "lb_loss": 0.9999999965541816 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 506000000, |
| "cumulative_training_bytes": 506002066, |
| "metrics": { |
| "loss": 0.476615203018466, |
| "ce_loss": 0.46661521255520916, |
| "lb_loss": 0.9999999968099359 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 507000000, |
| "cumulative_training_bytes": 507007439, |
| "metrics": { |
| "loss": 0.47662020089039253, |
| "ce_loss": 0.4666202104271357, |
| "lb_loss": 0.9999999967951272 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 508000000, |
| "cumulative_training_bytes": 508001710, |
| "metrics": { |
| "loss": 0.47666380585255125, |
| "ce_loss": 0.4666638153892944, |
| "lb_loss": 0.9999999968529908 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 509000000, |
| "cumulative_training_bytes": 509001905, |
| "metrics": { |
| "loss": 0.47671074559176424, |
| "ce_loss": 0.4667107551285074, |
| "lb_loss": 0.9999999967674915 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 510000000, |
| "cumulative_training_bytes": 510006108, |
| "metrics": { |
| "loss": 0.4766892023514087, |
| "ce_loss": 0.4666892118881519, |
| "lb_loss": 0.999999996632583 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 511000000, |
| "cumulative_training_bytes": 511002126, |
| "metrics": { |
| "loss": 0.47665283225579347, |
| "ce_loss": 0.46665284179253663, |
| "lb_loss": 0.9999999965179872 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 512000000, |
| "cumulative_training_bytes": 512004051, |
| "metrics": { |
| "loss": 0.47662187693238467, |
| "ce_loss": 0.46662188646912783, |
| "lb_loss": 0.9999999966157122 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 513000000, |
| "cumulative_training_bytes": 513002597, |
| "metrics": { |
| "loss": 0.47660819736161886, |
| "ce_loss": 0.466608206898362, |
| "lb_loss": 0.9999999966830678 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 514000000, |
| "cumulative_training_bytes": 514004929, |
| "metrics": { |
| "loss": 0.4766356019235062, |
| "ce_loss": 0.46663561146024934, |
| "lb_loss": 0.9999999970276858 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 515000000, |
| "cumulative_training_bytes": 515000236, |
| "metrics": { |
| "loss": 0.47663549837685426, |
| "ce_loss": 0.4666355079135974, |
| "lb_loss": 0.9999999970452804 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 516000000, |
| "cumulative_training_bytes": 516005322, |
| "metrics": { |
| "loss": 0.4766574106708626, |
| "ce_loss": 0.46665742020760576, |
| "lb_loss": 0.9999999971088163 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 517000000, |
| "cumulative_training_bytes": 517005666, |
| "metrics": { |
| "loss": 0.47670004072342714, |
| "ce_loss": 0.4667000502601703, |
| "lb_loss": 0.9999999968872504 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 518000000, |
| "cumulative_training_bytes": 518006360, |
| "metrics": { |
| "loss": 0.47669515147433716, |
| "ce_loss": 0.46669516101108033, |
| "lb_loss": 0.9999999968519061 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 519000000, |
| "cumulative_training_bytes": 519004375, |
| "metrics": { |
| "loss": 0.4766806289889557, |
| "ce_loss": 0.46668063852569885, |
| "lb_loss": 0.9999999968289685 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 520000000, |
| "cumulative_training_bytes": 520002998, |
| "metrics": { |
| "loss": 0.476721162943877, |
| "ce_loss": 0.4667211724806202, |
| "lb_loss": 0.9999999969336582 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 521000000, |
| "cumulative_training_bytes": 521004681, |
| "metrics": { |
| "loss": 0.47674204208199283, |
| "ce_loss": 0.466742051618736, |
| "lb_loss": 0.9999999968796605 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 522000000, |
| "cumulative_training_bytes": 522005267, |
| "metrics": { |
| "loss": 0.4767827219174502, |
| "ce_loss": 0.46678273145419336, |
| "lb_loss": 0.9999999969083478 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 523000000, |
| "cumulative_training_bytes": 523004469, |
| "metrics": { |
| "loss": 0.47677980619743615, |
| "ce_loss": 0.4667798157341793, |
| "lb_loss": 0.9999999967781273 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 524000000, |
| "cumulative_training_bytes": 524001588, |
| "metrics": { |
| "loss": 0.4768444411811146, |
| "ce_loss": 0.46684445071785774, |
| "lb_loss": 0.9999999967404462 |
| } |
| }, |
| { |
| "epoch": 11, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.4768498985810268, |
| "ce_loss": 0.46684990811776994, |
| "lb_loss": 0.9999999967918451, |
| "training_bytes": 47653417 |
| }, |
| "cumulative_training_bytes": 524187426, |
| "training_bytes_this_epoch": 47653417 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 525000000, |
| "cumulative_training_bytes": 525004346, |
| "metrics": { |
| "loss": 0.47418661429503256, |
| "ce_loss": 0.4641866238317757, |
| "lb_loss": 1.0000000077987385 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 526000000, |
| "cumulative_training_bytes": 526005081, |
| "metrics": { |
| "loss": 0.4731121884674585, |
| "ce_loss": 0.46311219800420167, |
| "lb_loss": 1.000000004007035 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 527000000, |
| "cumulative_training_bytes": 527001768, |
| "metrics": { |
| "loss": 0.47275793034097424, |
| "ce_loss": 0.4627579398777174, |
| "lb_loss": 1.0000000037252903 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 528000000, |
| "cumulative_training_bytes": 528002785, |
| "metrics": { |
| "loss": 0.4726709869492006, |
| "ce_loss": 0.4626709964859438, |
| "lb_loss": 0.9999999992818718 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 529000000, |
| "cumulative_training_bytes": 529007587, |
| "metrics": { |
| "loss": 0.47280677702923457, |
| "ce_loss": 0.46280678656597773, |
| "lb_loss": 0.9999999995261952 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 530000000, |
| "cumulative_training_bytes": 530006938, |
| "metrics": { |
| "loss": 0.4731630250027305, |
| "ce_loss": 0.46316303453947366, |
| "lb_loss": 0.9999999993725827 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 531000000, |
| "cumulative_training_bytes": 531005656, |
| "metrics": { |
| "loss": 0.47286867023853774, |
| "ce_loss": 0.4628686797752809, |
| "lb_loss": 0.9999999978569116 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 532000000, |
| "cumulative_training_bytes": 532005754, |
| "metrics": { |
| "loss": 0.47314695116355066, |
| "ce_loss": 0.4631469607002938, |
| "lb_loss": 0.9999999973145801 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 533000000, |
| "cumulative_training_bytes": 533000212, |
| "metrics": { |
| "loss": 0.4732809010844558, |
| "ce_loss": 0.46328091062119897, |
| "lb_loss": 0.999999997307175 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 534000000, |
| "cumulative_training_bytes": 534000062, |
| "metrics": { |
| "loss": 0.47332667615802654, |
| "ce_loss": 0.4633266856947697, |
| "lb_loss": 0.9999999969290347 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 535000000, |
| "cumulative_training_bytes": 535000930, |
| "metrics": { |
| "loss": 0.47340379490730783, |
| "ce_loss": 0.463403804444051, |
| "lb_loss": 0.9999999974250118 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 536000000, |
| "cumulative_training_bytes": 536002691, |
| "metrics": { |
| "loss": 0.47355009577757956, |
| "ce_loss": 0.4635501053143227, |
| "lb_loss": 0.9999999983389503 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 537000000, |
| "cumulative_training_bytes": 537004758, |
| "metrics": { |
| "loss": 0.4735789928527431, |
| "ce_loss": 0.46357900238948624, |
| "lb_loss": 0.9999999973651471 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 538000000, |
| "cumulative_training_bytes": 538000966, |
| "metrics": { |
| "loss": 0.4737231837143655, |
| "ce_loss": 0.46372319325110867, |
| "lb_loss": 0.9999999968611745 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 539000000, |
| "cumulative_training_bytes": 539001127, |
| "metrics": { |
| "loss": 0.4737859239257891, |
| "ce_loss": 0.4637859334625323, |
| "lb_loss": 0.9999999970120669 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 540000000, |
| "cumulative_training_bytes": 540006347, |
| "metrics": { |
| "loss": 0.47382085505667404, |
| "ce_loss": 0.4638208645934172, |
| "lb_loss": 0.9999999970861234 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 541000000, |
| "cumulative_training_bytes": 541005694, |
| "metrics": { |
| "loss": 0.4739520771155099, |
| "ce_loss": 0.46395208665225307, |
| "lb_loss": 0.9999999976668187 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 542000000, |
| "cumulative_training_bytes": 542006701, |
| "metrics": { |
| "loss": 0.4740450398618823, |
| "ce_loss": 0.46404504939862545, |
| "lb_loss": 0.9999999976188866 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 543000000, |
| "cumulative_training_bytes": 543005643, |
| "metrics": { |
| "loss": 0.4742385020533341, |
| "ce_loss": 0.4642385115900773, |
| "lb_loss": 0.9999999976730192 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 544000000, |
| "cumulative_training_bytes": 544004689, |
| "metrics": { |
| "loss": 0.4743254720681236, |
| "ce_loss": 0.46432548160486675, |
| "lb_loss": 0.9999999975366176 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 545000000, |
| "cumulative_training_bytes": 545005233, |
| "metrics": { |
| "loss": 0.474323673342992, |
| "ce_loss": 0.4643236828797352, |
| "lb_loss": 0.9999999969748287 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 546000000, |
| "cumulative_training_bytes": 546004328, |
| "metrics": { |
| "loss": 0.47431586884163524, |
| "ce_loss": 0.4643158783783784, |
| "lb_loss": 0.9999999970291823 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 547000000, |
| "cumulative_training_bytes": 547003640, |
| "metrics": { |
| "loss": 0.4743756198690977, |
| "ce_loss": 0.4643756294058409, |
| "lb_loss": 0.9999999968386929 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 548000000, |
| "cumulative_training_bytes": 548000417, |
| "metrics": { |
| "loss": 0.47438953995666305, |
| "ce_loss": 0.4643895494934062, |
| "lb_loss": 0.999999996817507 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 549000000, |
| "cumulative_training_bytes": 549005076, |
| "metrics": { |
| "loss": 0.4745037520373309, |
| "ce_loss": 0.4645037615740741, |
| "lb_loss": 0.9999999967254238 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 550000000, |
| "cumulative_training_bytes": 550005242, |
| "metrics": { |
| "loss": 0.4746403744368552, |
| "ce_loss": 0.46464038397359836, |
| "lb_loss": 0.9999999965697712 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 551000000, |
| "cumulative_training_bytes": 551001002, |
| "metrics": { |
| "loss": 0.4746668936694836, |
| "ce_loss": 0.4646669032062268, |
| "lb_loss": 0.9999999966120182 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 552000000, |
| "cumulative_training_bytes": 552003787, |
| "metrics": { |
| "loss": 0.47473511401777224, |
| "ce_loss": 0.4647351235545154, |
| "lb_loss": 0.999999996668573 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 553000000, |
| "cumulative_training_bytes": 553002274, |
| "metrics": { |
| "loss": 0.4748042938780854, |
| "ce_loss": 0.4648043034148286, |
| "lb_loss": 0.9999999963410383 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 554000000, |
| "cumulative_training_bytes": 554003365, |
| "metrics": { |
| "loss": 0.47482116968128457, |
| "ce_loss": 0.46482117921802774, |
| "lb_loss": 0.9999999963110633 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 555000000, |
| "cumulative_training_bytes": 555006363, |
| "metrics": { |
| "loss": 0.47514410925207673, |
| "ce_loss": 0.4651441187888199, |
| "lb_loss": 0.9999999963274654 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 556000000, |
| "cumulative_training_bytes": 556003044, |
| "metrics": { |
| "loss": 0.47538666690730014, |
| "ce_loss": 0.4653866764440433, |
| "lb_loss": 0.9999999961698098 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 557000000, |
| "cumulative_training_bytes": 557004116, |
| "metrics": { |
| "loss": 0.47563164366554334, |
| "ce_loss": 0.4656316532022865, |
| "lb_loss": 0.9999999961060895 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 558000000, |
| "cumulative_training_bytes": 558000464, |
| "metrics": { |
| "loss": 0.4758357828941898, |
| "ce_loss": 0.465835792430933, |
| "lb_loss": 0.9999999960857457 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 559000000, |
| "cumulative_training_bytes": 559004482, |
| "metrics": { |
| "loss": 0.47595539375113893, |
| "ce_loss": 0.4659554032878821, |
| "lb_loss": 0.9999999961854076 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 560000000, |
| "cumulative_training_bytes": 560002449, |
| "metrics": { |
| "loss": 0.47605645694818305, |
| "ce_loss": 0.4660564664849262, |
| "lb_loss": 0.9999999963806459 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 561000000, |
| "cumulative_training_bytes": 561007064, |
| "metrics": { |
| "loss": 0.47612497096450473, |
| "ce_loss": 0.4661249805012479, |
| "lb_loss": 0.9999999961693354 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 562000000, |
| "cumulative_training_bytes": 562002742, |
| "metrics": { |
| "loss": 0.4761972426402516, |
| "ce_loss": 0.46619725217699476, |
| "lb_loss": 0.9999999964391717 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 563000000, |
| "cumulative_training_bytes": 563003402, |
| "metrics": { |
| "loss": 0.47623431614386447, |
| "ce_loss": 0.46623432568060763, |
| "lb_loss": 0.9999999965076781 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 564000000, |
| "cumulative_training_bytes": 564003696, |
| "metrics": { |
| "loss": 0.47626276089594916, |
| "ce_loss": 0.4662627704326923, |
| "lb_loss": 0.9999999963664091 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 565000000, |
| "cumulative_training_bytes": 565007276, |
| "metrics": { |
| "loss": 0.47630263068085205, |
| "ce_loss": 0.4663026402175952, |
| "lb_loss": 0.9999999963550714 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 566000000, |
| "cumulative_training_bytes": 566006079, |
| "metrics": { |
| "loss": 0.47630903127537916, |
| "ce_loss": 0.4663090408121223, |
| "lb_loss": 0.9999999962126329 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 567000000, |
| "cumulative_training_bytes": 567007618, |
| "metrics": { |
| "loss": 0.4764510868956603, |
| "ce_loss": 0.46645109643240346, |
| "lb_loss": 0.9999999961627911 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 568000000, |
| "cumulative_training_bytes": 568004347, |
| "metrics": { |
| "loss": 0.47650604713050604, |
| "ce_loss": 0.4665060566672492, |
| "lb_loss": 0.9999999959478841 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 569000000, |
| "cumulative_training_bytes": 569001858, |
| "metrics": { |
| "loss": 0.47657124815293556, |
| "ce_loss": 0.4665712576896787, |
| "lb_loss": 0.9999999959869736 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 570000000, |
| "cumulative_training_bytes": 570002921, |
| "metrics": { |
| "loss": 0.4766179731851355, |
| "ce_loss": 0.46661798272187865, |
| "lb_loss": 0.999999996034991 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 571000000, |
| "cumulative_training_bytes": 571001763, |
| "metrics": { |
| "loss": 0.4765936612263846, |
| "ce_loss": 0.46659367076312774, |
| "lb_loss": 0.9999999959145516 |
| } |
| }, |
| { |
| "epoch": 12, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.47662866575245993, |
| "ce_loss": 0.4666286752892031, |
| "lb_loss": 0.9999999958629167, |
| "training_bytes": 47653410 |
| }, |
| "cumulative_training_bytes": 571840836, |
| "training_bytes_this_epoch": 47653410 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 572000000, |
| "cumulative_training_bytes": 572003805, |
| "metrics": { |
| "loss": 0.46656621070135207, |
| "ce_loss": 0.45656622023809523, |
| "lb_loss": 0.9999999886467343 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 573000000, |
| "cumulative_training_bytes": 573001015, |
| "metrics": { |
| "loss": 0.47032956413875354, |
| "ce_loss": 0.4603295736754967, |
| "lb_loss": 0.999999998421069 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 574000000, |
| "cumulative_training_bytes": 574004305, |
| "metrics": { |
| "loss": 0.4710136761902072, |
| "ce_loss": 0.46101368572695034, |
| "lb_loss": 0.999999999365908 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 575000000, |
| "cumulative_training_bytes": 575005268, |
| "metrics": { |
| "loss": 0.4715617338046612, |
| "ce_loss": 0.4615617433414044, |
| "lb_loss": 0.9999999987011094 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 576000000, |
| "cumulative_training_bytes": 576005624, |
| "metrics": { |
| "loss": 0.47210074873531566, |
| "ce_loss": 0.4621007582720588, |
| "lb_loss": 1.0000000005478369 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 577000000, |
| "cumulative_training_bytes": 577007614, |
| "metrics": { |
| "loss": 0.47244212009288644, |
| "ce_loss": 0.4624421296296296, |
| "lb_loss": 1.0000000013245476 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 578000000, |
| "cumulative_training_bytes": 578000160, |
| "metrics": { |
| "loss": 0.47264556499741833, |
| "ce_loss": 0.4626455745341615, |
| "lb_loss": 1.0000000017029897 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 579000000, |
| "cumulative_training_bytes": 579000407, |
| "metrics": { |
| "loss": 0.47249581265577023, |
| "ce_loss": 0.4624958221925134, |
| "lb_loss": 1.0000000014662105 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 580000000, |
| "cumulative_training_bytes": 580003744, |
| "metrics": { |
| "loss": 0.472489363704345, |
| "ce_loss": 0.4624893732410882, |
| "lb_loss": 1.0000000013419432 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 581000000, |
| "cumulative_training_bytes": 581005566, |
| "metrics": { |
| "loss": 0.4733132214574089, |
| "ce_loss": 0.46331323099415206, |
| "lb_loss": 1.0000000010456955 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 582000000, |
| "cumulative_training_bytes": 582005047, |
| "metrics": { |
| "loss": 0.47375311075338494, |
| "ce_loss": 0.4637531202901281, |
| "lb_loss": 1.0000000007635863 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 583000000, |
| "cumulative_training_bytes": 583001026, |
| "metrics": { |
| "loss": 0.4739308522511772, |
| "ce_loss": 0.46393086178792037, |
| "lb_loss": 1.000000001022729 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 584000000, |
| "cumulative_training_bytes": 584004132, |
| "metrics": { |
| "loss": 0.47396557216980595, |
| "ce_loss": 0.4639655817065491, |
| "lb_loss": 1.000000000638085 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 585000000, |
| "cumulative_training_bytes": 585007650, |
| "metrics": { |
| "loss": 0.4742566674254661, |
| "ce_loss": 0.4642566769622093, |
| "lb_loss": 1.0000000008316927 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 586000000, |
| "cumulative_training_bytes": 586004555, |
| "metrics": { |
| "loss": 0.4743739347200136, |
| "ce_loss": 0.46437394425675677, |
| "lb_loss": 1.000000000483281 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 587000000, |
| "cumulative_training_bytes": 587000297, |
| "metrics": { |
| "loss": 0.474306137874873, |
| "ce_loss": 0.4643061474116162, |
| "lb_loss": 1.000000000752584 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 588000000, |
| "cumulative_training_bytes": 588002607, |
| "metrics": { |
| "loss": 0.4741220741913478, |
| "ce_loss": 0.46412208372809094, |
| "lb_loss": 1.0000000010447048 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 589000000, |
| "cumulative_training_bytes": 589003684, |
| "metrics": { |
| "loss": 0.47414856037851105, |
| "ce_loss": 0.4641485699152542, |
| "lb_loss": 1.0000000009304917 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 590000000, |
| "cumulative_training_bytes": 590003539, |
| "metrics": { |
| "loss": 0.4741836451198097, |
| "ce_loss": 0.46418365465655287, |
| "lb_loss": 1.0000000006279461 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 591000000, |
| "cumulative_training_bytes": 591000699, |
| "metrics": { |
| "loss": 0.4741765723150347, |
| "ce_loss": 0.46417658185177785, |
| "lb_loss": 1.0000000000476266 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 592000000, |
| "cumulative_training_bytes": 592006371, |
| "metrics": { |
| "loss": 0.4741393104139022, |
| "ce_loss": 0.4641393199506454, |
| "lb_loss": 1.0000000001810316 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 593000000, |
| "cumulative_training_bytes": 593001411, |
| "metrics": { |
| "loss": 0.47416184262497896, |
| "ce_loss": 0.4641618521617221, |
| "lb_loss": 0.9999999998490475 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 594000000, |
| "cumulative_training_bytes": 594006305, |
| "metrics": { |
| "loss": 0.47413737712128795, |
| "ce_loss": 0.4641373866580311, |
| "lb_loss": 0.9999999995882235 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 595000000, |
| "cumulative_training_bytes": 595002552, |
| "metrics": { |
| "loss": 0.47414900905829815, |
| "ce_loss": 0.4641490185950413, |
| "lb_loss": 0.9999999991724314 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 596000000, |
| "cumulative_training_bytes": 596006850, |
| "metrics": { |
| "loss": 0.47415680486439754, |
| "ce_loss": 0.4641568144011407, |
| "lb_loss": 0.9999999985079953 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 597000000, |
| "cumulative_training_bytes": 597001688, |
| "metrics": { |
| "loss": 0.4742202382949671, |
| "ce_loss": 0.4642202478317103, |
| "lb_loss": 0.999999998893523 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 598000000, |
| "cumulative_training_bytes": 598005915, |
| "metrics": { |
| "loss": 0.47454308382000254, |
| "ce_loss": 0.4645430933567457, |
| "lb_loss": 0.999999998848725 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 599000000, |
| "cumulative_training_bytes": 599003238, |
| "metrics": { |
| "loss": 0.47462788903951847, |
| "ce_loss": 0.46462789857626163, |
| "lb_loss": 0.9999999990085497 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 600000000, |
| "cumulative_training_bytes": 600001112, |
| "metrics": { |
| "loss": 0.4747784238849534, |
| "ce_loss": 0.46477843342169656, |
| "lb_loss": 0.9999999986711308 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 601000000, |
| "cumulative_training_bytes": 601004350, |
| "metrics": { |
| "loss": 0.47477400431203604, |
| "ce_loss": 0.4647740138487792, |
| "lb_loss": 0.9999999984664597 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 602000000, |
| "cumulative_training_bytes": 602000426, |
| "metrics": { |
| "loss": 0.47477283493520406, |
| "ce_loss": 0.4647728444719472, |
| "lb_loss": 0.9999999983354885 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 603000000, |
| "cumulative_training_bytes": 603006064, |
| "metrics": { |
| "loss": 0.47476264018684405, |
| "ce_loss": 0.4647626497235872, |
| "lb_loss": 0.9999999979204276 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 604000000, |
| "cumulative_training_bytes": 604000130, |
| "metrics": { |
| "loss": 0.4748116534096854, |
| "ce_loss": 0.46481166294642856, |
| "lb_loss": 0.9999999979706038 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 605000000, |
| "cumulative_training_bytes": 605006216, |
| "metrics": { |
| "loss": 0.4747801545419915, |
| "ce_loss": 0.4647801640787347, |
| "lb_loss": 0.9999999979081261 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 606000000, |
| "cumulative_training_bytes": 606000518, |
| "metrics": { |
| "loss": 0.4748008338840145, |
| "ce_loss": 0.46480084342075767, |
| "lb_loss": 0.9999999976484157 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 607000000, |
| "cumulative_training_bytes": 607002557, |
| "metrics": { |
| "loss": 0.47481949653359656, |
| "ce_loss": 0.4648195060703397, |
| "lb_loss": 0.9999999974948396 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 608000000, |
| "cumulative_training_bytes": 608005994, |
| "metrics": { |
| "loss": 0.47476268763666357, |
| "ce_loss": 0.46476269717340674, |
| "lb_loss": 0.9999999975138439 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 609000000, |
| "cumulative_training_bytes": 609006940, |
| "metrics": { |
| "loss": 0.4747737273297587, |
| "ce_loss": 0.46477373686650186, |
| "lb_loss": 0.999999997568661 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 610000000, |
| "cumulative_training_bytes": 610003956, |
| "metrics": { |
| "loss": 0.4748123964311605, |
| "ce_loss": 0.4648124059679037, |
| "lb_loss": 0.9999999975847266 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 611000000, |
| "cumulative_training_bytes": 611004009, |
| "metrics": { |
| "loss": 0.47486817361415595, |
| "ce_loss": 0.4648681831508991, |
| "lb_loss": 0.999999997518415 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 612000000, |
| "cumulative_training_bytes": 612000579, |
| "metrics": { |
| "loss": 0.47488283274785464, |
| "ce_loss": 0.4648828422845978, |
| "lb_loss": 0.9999999974549294 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 613000000, |
| "cumulative_training_bytes": 613001932, |
| "metrics": { |
| "loss": 0.4748709832333889, |
| "ce_loss": 0.46487099277013205, |
| "lb_loss": 0.9999999975058499 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 614000000, |
| "cumulative_training_bytes": 614003917, |
| "metrics": { |
| "loss": 0.47490614968620526, |
| "ce_loss": 0.4649061592229484, |
| "lb_loss": 0.9999999975976341 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 615000000, |
| "cumulative_training_bytes": 615001118, |
| "metrics": { |
| "loss": 0.47489570370376766, |
| "ce_loss": 0.4648957132405108, |
| "lb_loss": 0.9999999976635994 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 616000000, |
| "cumulative_training_bytes": 616000487, |
| "metrics": { |
| "loss": 0.4748908077959545, |
| "ce_loss": 0.46489081733269766, |
| "lb_loss": 0.9999999977575923 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 617000000, |
| "cumulative_training_bytes": 617005273, |
| "metrics": { |
| "loss": 0.474879167548356, |
| "ce_loss": 0.4648791770850992, |
| "lb_loss": 0.9999999978275982 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 618000000, |
| "cumulative_training_bytes": 618005046, |
| "metrics": { |
| "loss": 0.4748712766425569, |
| "ce_loss": 0.46487128617930007, |
| "lb_loss": 0.9999999979436447 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 619000000, |
| "cumulative_training_bytes": 619006994, |
| "metrics": { |
| "loss": 0.4749004952319257, |
| "ce_loss": 0.46490050476866884, |
| "lb_loss": 0.999999997909967 |
| } |
| }, |
| { |
| "epoch": 13, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.47487888667148304, |
| "ce_loss": 0.4648788962082262, |
| "lb_loss": 0.9999999979889178, |
| "training_bytes": 47653395 |
| }, |
| "cumulative_training_bytes": 619494231, |
| "training_bytes_this_epoch": 47653395 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 620000000, |
| "cumulative_training_bytes": 620000323, |
| "metrics": { |
| "loss": 0.46723839008446894, |
| "ce_loss": 0.4572383996212121, |
| "lb_loss": 0.9999999918720939 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 621000000, |
| "cumulative_training_bytes": 621005489, |
| "metrics": { |
| "loss": 0.4674079853871147, |
| "ce_loss": 0.45740799492385786, |
| "lb_loss": 0.9999999990923151 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 622000000, |
| "cumulative_training_bytes": 622007499, |
| "metrics": { |
| "loss": 0.46797802971630564, |
| "ce_loss": 0.4579780392530488, |
| "lb_loss": 1.0000000014537718 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 623000000, |
| "cumulative_training_bytes": 623001614, |
| "metrics": { |
| "loss": 0.4684726296553966, |
| "ce_loss": 0.45847263919213976, |
| "lb_loss": 1.0 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 624000000, |
| "cumulative_training_bytes": 624004859, |
| "metrics": { |
| "loss": 0.46862292127819744, |
| "ce_loss": 0.4586229308149406, |
| "lb_loss": 0.9999999988868402 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 625000000, |
| "cumulative_training_bytes": 625002853, |
| "metrics": { |
| "loss": 0.4686638247817547, |
| "ce_loss": 0.4586638343184979, |
| "lb_loss": 0.9999999982591133 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 626000000, |
| "cumulative_training_bytes": 626004267, |
| "metrics": { |
| "loss": 0.4690142367867862, |
| "ce_loss": 0.4590142463235294, |
| "lb_loss": 0.9999999991585227 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 627000000, |
| "cumulative_training_bytes": 627004393, |
| "metrics": { |
| "loss": 0.4692574043663181, |
| "ce_loss": 0.45925741390306124, |
| "lb_loss": 0.9999999992093261 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 628000000, |
| "cumulative_training_bytes": 628005962, |
| "metrics": { |
| "loss": 0.4694238625379643, |
| "ce_loss": 0.4594238720747075, |
| "lb_loss": 0.9999999987660605 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 629000000, |
| "cumulative_training_bytes": 629001986, |
| "metrics": { |
| "loss": 0.4696217672561658, |
| "ce_loss": 0.45962177679290894, |
| "lb_loss": 0.9999999979827598 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 630000000, |
| "cumulative_training_bytes": 630004200, |
| "metrics": { |
| "loss": 0.4695898199672685, |
| "ce_loss": 0.45958982950401167, |
| "lb_loss": 0.9999999979131853 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 631000000, |
| "cumulative_training_bytes": 631006243, |
| "metrics": { |
| "loss": 0.4696956566583301, |
| "ce_loss": 0.45969566619507324, |
| "lb_loss": 0.9999999982936087 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 632000000, |
| "cumulative_training_bytes": 632002297, |
| "metrics": { |
| "loss": 0.46990467987808526, |
| "ce_loss": 0.4599046894148284, |
| "lb_loss": 0.999999998393012 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 633000000, |
| "cumulative_training_bytes": 633003176, |
| "metrics": { |
| "loss": 0.46981904374078237, |
| "ce_loss": 0.45981905327752554, |
| "lb_loss": 0.9999999986807144 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 634000000, |
| "cumulative_training_bytes": 634003223, |
| "metrics": { |
| "loss": 0.46992636394450354, |
| "ce_loss": 0.4599263734812467, |
| "lb_loss": 0.9999999983311959 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 635000000, |
| "cumulative_training_bytes": 635006003, |
| "metrics": { |
| "loss": 0.4700912006287707, |
| "ce_loss": 0.46009121016551385, |
| "lb_loss": 0.999999997732432 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 636000000, |
| "cumulative_training_bytes": 636004090, |
| "metrics": { |
| "loss": 0.4703039710317023, |
| "ce_loss": 0.46030398056844546, |
| "lb_loss": 0.9999999974000758 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 637000000, |
| "cumulative_training_bytes": 637000138, |
| "metrics": { |
| "loss": 0.47044514910658286, |
| "ce_loss": 0.460445158643326, |
| "lb_loss": 0.999999997443652 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 638000000, |
| "cumulative_training_bytes": 638002271, |
| "metrics": { |
| "loss": 0.4705825973030747, |
| "ce_loss": 0.4605826068398179, |
| "lb_loss": 0.9999999972368707 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 639000000, |
| "cumulative_training_bytes": 639000648, |
| "metrics": { |
| "loss": 0.47067206184974547, |
| "ce_loss": 0.46067207138648864, |
| "lb_loss": 0.9999999969565578 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 640000000, |
| "cumulative_training_bytes": 640003100, |
| "metrics": { |
| "loss": 0.47072955597502375, |
| "ce_loss": 0.4607295655117669, |
| "lb_loss": 0.9999999970164279 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 641000000, |
| "cumulative_training_bytes": 641002993, |
| "metrics": { |
| "loss": 0.47074683985762805, |
| "ce_loss": 0.4607468493943712, |
| "lb_loss": 0.9999999968785598 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 642000000, |
| "cumulative_training_bytes": 642000407, |
| "metrics": { |
| "loss": 0.4708656697278125, |
| "ce_loss": 0.4608656792645557, |
| "lb_loss": 0.9999999967123077 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 643000000, |
| "cumulative_training_bytes": 643000364, |
| "metrics": { |
| "loss": 0.47096550138242244, |
| "ce_loss": 0.4609655109191656, |
| "lb_loss": 0.9999999967166933 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 644000000, |
| "cumulative_training_bytes": 644000345, |
| "metrics": { |
| "loss": 0.47106452298358203, |
| "ce_loss": 0.4610645325203252, |
| "lb_loss": 0.9999999966078658 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 645000000, |
| "cumulative_training_bytes": 645004285, |
| "metrics": { |
| "loss": 0.4711627833440018, |
| "ce_loss": 0.46116279288074497, |
| "lb_loss": 0.9999999969383015 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 646000000, |
| "cumulative_training_bytes": 646004246, |
| "metrics": { |
| "loss": 0.47127110227684066, |
| "ce_loss": 0.4612711118135838, |
| "lb_loss": 0.9999999970886748 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 647000000, |
| "cumulative_training_bytes": 647002840, |
| "metrics": { |
| "loss": 0.4712862239242596, |
| "ce_loss": 0.4612862334610028, |
| "lb_loss": 0.9999999970280693 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 648000000, |
| "cumulative_training_bytes": 648003976, |
| "metrics": { |
| "loss": 0.4713705267774197, |
| "ce_loss": 0.46137053631416286, |
| "lb_loss": 0.9999999969564949 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 649000000, |
| "cumulative_training_bytes": 649000496, |
| "metrics": { |
| "loss": 0.47139901831316594, |
| "ce_loss": 0.4613990278499091, |
| "lb_loss": 0.9999999971211467 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 650000000, |
| "cumulative_training_bytes": 650001399, |
| "metrics": { |
| "loss": 0.4714765372676265, |
| "ce_loss": 0.46147654680436967, |
| "lb_loss": 0.9999999974403831 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 651000000, |
| "cumulative_training_bytes": 651002904, |
| "metrics": { |
| "loss": 0.4715733364090385, |
| "ce_loss": 0.46157334594578164, |
| "lb_loss": 0.9999999976088582 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 652000000, |
| "cumulative_training_bytes": 652005210, |
| "metrics": { |
| "loss": 0.47166369952310605, |
| "ce_loss": 0.4616637090598492, |
| "lb_loss": 0.999999997626488 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 653000000, |
| "cumulative_training_bytes": 653002239, |
| "metrics": { |
| "loss": 0.4716997182581814, |
| "ce_loss": 0.4616997277949246, |
| "lb_loss": 0.9999999975743881 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 654000000, |
| "cumulative_training_bytes": 654006342, |
| "metrics": { |
| "loss": 0.4717738008128684, |
| "ce_loss": 0.46177381034961157, |
| "lb_loss": 0.9999999973670757 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 655000000, |
| "cumulative_training_bytes": 655002681, |
| "metrics": { |
| "loss": 0.47183420027986955, |
| "ce_loss": 0.4618342098166127, |
| "lb_loss": 0.9999999973637643 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 656000000, |
| "cumulative_training_bytes": 656006752, |
| "metrics": { |
| "loss": 0.471994152803303, |
| "ce_loss": 0.46199416234004614, |
| "lb_loss": 0.999999997549295 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 657000000, |
| "cumulative_training_bytes": 657006557, |
| "metrics": { |
| "loss": 0.4720288957562336, |
| "ce_loss": 0.46202890529297674, |
| "lb_loss": 0.9999999973592879 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 658000000, |
| "cumulative_training_bytes": 658003676, |
| "metrics": { |
| "loss": 0.47203408391492746, |
| "ce_loss": 0.4620340934516706, |
| "lb_loss": 0.9999999974157096 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 659000000, |
| "cumulative_training_bytes": 659005764, |
| "metrics": { |
| "loss": 0.47206756957258034, |
| "ce_loss": 0.4620675791093235, |
| "lb_loss": 0.9999999975159918 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 660000000, |
| "cumulative_training_bytes": 660004372, |
| "metrics": { |
| "loss": 0.472095185088895, |
| "ce_loss": 0.46209519462563814, |
| "lb_loss": 0.9999999974530819 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 661000000, |
| "cumulative_training_bytes": 661000728, |
| "metrics": { |
| "loss": 0.47210958171625556, |
| "ce_loss": 0.4621095912529987, |
| "lb_loss": 0.9999999973271952 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 662000000, |
| "cumulative_training_bytes": 662006203, |
| "metrics": { |
| "loss": 0.47222807999667427, |
| "ce_loss": 0.46222808953341743, |
| "lb_loss": 0.9999999972082133 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 663000000, |
| "cumulative_training_bytes": 663003652, |
| "metrics": { |
| "loss": 0.47226043096668935, |
| "ce_loss": 0.4622604405034325, |
| "lb_loss": 0.9999999970307842 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 664000000, |
| "cumulative_training_bytes": 664002530, |
| "metrics": { |
| "loss": 0.47231932976126095, |
| "ce_loss": 0.4623193392980041, |
| "lb_loss": 0.9999999970874537 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 665000000, |
| "cumulative_training_bytes": 665004691, |
| "metrics": { |
| "loss": 0.47233488024956005, |
| "ce_loss": 0.4623348897863032, |
| "lb_loss": 0.9999999971215661 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 666000000, |
| "cumulative_training_bytes": 666005317, |
| "metrics": { |
| "loss": 0.47233496844317124, |
| "ce_loss": 0.4623349779799144, |
| "lb_loss": 0.99999999716402 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 667000000, |
| "cumulative_training_bytes": 667007057, |
| "metrics": { |
| "loss": 0.4723920256163592, |
| "ce_loss": 0.46239203515310234, |
| "lb_loss": 0.9999999972623169 |
| } |
| }, |
| { |
| "epoch": 14, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.4724123759870358, |
| "ce_loss": 0.46241238552377895, |
| "lb_loss": 0.9999999972610977, |
| "training_bytes": 47653399 |
| }, |
| "cumulative_training_bytes": 667147630, |
| "training_bytes_this_epoch": 47653399 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 668000000, |
| "cumulative_training_bytes": 668001125, |
| "metrics": { |
| "loss": 0.46842632974897114, |
| "ce_loss": 0.4584263392857143, |
| "lb_loss": 0.9999999936137881 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 669000000, |
| "cumulative_training_bytes": 669005783, |
| "metrics": { |
| "loss": 0.4678189205044091, |
| "ce_loss": 0.45781893004115226, |
| "lb_loss": 0.9999999923961153 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 670000000, |
| "cumulative_training_bytes": 670007110, |
| "metrics": { |
| "loss": 0.4679608026290322, |
| "ce_loss": 0.4579608121657754, |
| "lb_loss": 0.9999999939439131 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 671000000, |
| "cumulative_training_bytes": 671002720, |
| "metrics": { |
| "loss": 0.46782954155452666, |
| "ce_loss": 0.4578295510912698, |
| "lb_loss": 0.9999999952694726 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 672000000, |
| "cumulative_training_bytes": 672007653, |
| "metrics": { |
| "loss": 0.4680907186472191, |
| "ce_loss": 0.45809072818396224, |
| "lb_loss": 0.9999999966261521 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 673000000, |
| "cumulative_training_bytes": 673000283, |
| "metrics": { |
| "loss": 0.4684463636682177, |
| "ce_loss": 0.45844637320496084, |
| "lb_loss": 0.9999999954090417 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 674000000, |
| "cumulative_training_bytes": 674004321, |
| "metrics": { |
| "loss": 0.46855324088131706, |
| "ce_loss": 0.4585532504180602, |
| "lb_loss": 0.999999994484743 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 675000000, |
| "cumulative_training_bytes": 675006849, |
| "metrics": { |
| "loss": 0.4686119797628677, |
| "ce_loss": 0.4586119892996109, |
| "lb_loss": 0.9999999941439016 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 676000000, |
| "cumulative_training_bytes": 676003287, |
| "metrics": { |
| "loss": 0.4687279967240513, |
| "ce_loss": 0.45872800626079446, |
| "lb_loss": 0.9999999935659926 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 677000000, |
| "cumulative_training_bytes": 677005219, |
| "metrics": { |
| "loss": 0.46872374676659273, |
| "ce_loss": 0.4587237563033359, |
| "lb_loss": 0.9999999942198754 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 678000000, |
| "cumulative_training_bytes": 678005845, |
| "metrics": { |
| "loss": 0.4687448702201279, |
| "ce_loss": 0.45874487975687106, |
| "lb_loss": 0.9999999951274569 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 679000000, |
| "cumulative_training_bytes": 679002372, |
| "metrics": { |
| "loss": 0.46877505703846667, |
| "ce_loss": 0.45877506657520983, |
| "lb_loss": 0.9999999955363855 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 680000000, |
| "cumulative_training_bytes": 680006963, |
| "metrics": { |
| "loss": 0.4687262739453997, |
| "ce_loss": 0.45872628348214284, |
| "lb_loss": 0.9999999950329462 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 681000000, |
| "cumulative_training_bytes": 681000974, |
| "metrics": { |
| "loss": 0.468684216900515, |
| "ce_loss": 0.45868422643725815, |
| "lb_loss": 0.9999999951894538 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 682000000, |
| "cumulative_training_bytes": 682006359, |
| "metrics": { |
| "loss": 0.46886864708355563, |
| "ce_loss": 0.4588686566202988, |
| "lb_loss": 0.9999999951481021 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 683000000, |
| "cumulative_training_bytes": 683003602, |
| "metrics": { |
| "loss": 0.4691238743918283, |
| "ce_loss": 0.45912388392857145, |
| "lb_loss": 0.9999999956274584 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 684000000, |
| "cumulative_training_bytes": 684002920, |
| "metrics": { |
| "loss": 0.46929555333207207, |
| "ce_loss": 0.45929556286881523, |
| "lb_loss": 0.9999999955898515 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 685000000, |
| "cumulative_training_bytes": 685007066, |
| "metrics": { |
| "loss": 0.4694569631613296, |
| "ce_loss": 0.4594569726980728, |
| "lb_loss": 0.9999999959412683 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 686000000, |
| "cumulative_training_bytes": 686003334, |
| "metrics": { |
| "loss": 0.46950572702512294, |
| "ce_loss": 0.4595057365618661, |
| "lb_loss": 0.9999999965180248 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 687000000, |
| "cumulative_training_bytes": 687002972, |
| "metrics": { |
| "loss": 0.4696742790322091, |
| "ce_loss": 0.45967428856895226, |
| "lb_loss": 0.9999999965100517 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 688000000, |
| "cumulative_training_bytes": 688007217, |
| "metrics": { |
| "loss": 0.46969413696123996, |
| "ce_loss": 0.4596941464979831, |
| "lb_loss": 0.9999999968744172 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 689000000, |
| "cumulative_training_bytes": 689000339, |
| "metrics": { |
| "loss": 0.46979924995748157, |
| "ce_loss": 0.45979925949422473, |
| "lb_loss": 0.9999999969331876 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 690000000, |
| "cumulative_training_bytes": 690005975, |
| "metrics": { |
| "loss": 0.46974848966841076, |
| "ce_loss": 0.4597484992051539, |
| "lb_loss": 0.999999997187331 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 691000000, |
| "cumulative_training_bytes": 691000262, |
| "metrics": { |
| "loss": 0.4697124794350246, |
| "ce_loss": 0.45971248897176775, |
| "lb_loss": 0.999999997131634 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 692000000, |
| "cumulative_training_bytes": 692001292, |
| "metrics": { |
| "loss": 0.46974805658086766, |
| "ce_loss": 0.4597480661176108, |
| "lb_loss": 0.9999999971555665 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 693000000, |
| "cumulative_training_bytes": 693001879, |
| "metrics": { |
| "loss": 0.46981497825550306, |
| "ce_loss": 0.4598149877922462, |
| "lb_loss": 0.9999999971600037 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 694000000, |
| "cumulative_training_bytes": 694001241, |
| "metrics": { |
| "loss": 0.4697958947451035, |
| "ce_loss": 0.4597959042818467, |
| "lb_loss": 0.9999999970953565 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 695000000, |
| "cumulative_training_bytes": 695005882, |
| "metrics": { |
| "loss": 0.469869174014081, |
| "ce_loss": 0.4598691835508242, |
| "lb_loss": 0.9999999976092643 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 696000000, |
| "cumulative_training_bytes": 696001244, |
| "metrics": { |
| "loss": 0.46987648250569714, |
| "ce_loss": 0.4598764920424403, |
| "lb_loss": 0.9999999974071189 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 697000000, |
| "cumulative_training_bytes": 697002626, |
| "metrics": { |
| "loss": 0.4699041015886093, |
| "ce_loss": 0.45990411112535245, |
| "lb_loss": 0.9999999974483528 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 698000000, |
| "cumulative_training_bytes": 698000141, |
| "metrics": { |
| "loss": 0.46995632496462125, |
| "ce_loss": 0.4599563345013644, |
| "lb_loss": 0.9999999972940585 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 699000000, |
| "cumulative_training_bytes": 699000609, |
| "metrics": { |
| "loss": 0.46998604024095453, |
| "ce_loss": 0.4599860497776977, |
| "lb_loss": 0.9999999971923792 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 700000000, |
| "cumulative_training_bytes": 700001376, |
| "metrics": { |
| "loss": 0.46995432803025755, |
| "ce_loss": 0.4599543375670007, |
| "lb_loss": 0.9999999970968607 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 701000000, |
| "cumulative_training_bytes": 701000514, |
| "metrics": { |
| "loss": 0.46995761458675833, |
| "ce_loss": 0.4599576241235015, |
| "lb_loss": 0.9999999969934775 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 702000000, |
| "cumulative_training_bytes": 702001728, |
| "metrics": { |
| "loss": 0.4700420221251637, |
| "ce_loss": 0.4600420316619069, |
| "lb_loss": 0.9999999970669068 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 703000000, |
| "cumulative_training_bytes": 703001158, |
| "metrics": { |
| "loss": 0.4700812518074806, |
| "ce_loss": 0.4600812613442238, |
| "lb_loss": 0.999999997110772 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 704000000, |
| "cumulative_training_bytes": 704004853, |
| "metrics": { |
| "loss": 0.47010860920546704, |
| "ce_loss": 0.4601086187422102, |
| "lb_loss": 0.9999999968055674 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 705000000, |
| "cumulative_training_bytes": 705003445, |
| "metrics": { |
| "loss": 0.470210351307544, |
| "ce_loss": 0.4602103608442872, |
| "lb_loss": 0.9999999965647475 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 706000000, |
| "cumulative_training_bytes": 706001098, |
| "metrics": { |
| "loss": 0.47026130511842923, |
| "ce_loss": 0.4602613146551724, |
| "lb_loss": 0.9999999964413384 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 707000000, |
| "cumulative_training_bytes": 707007487, |
| "metrics": { |
| "loss": 0.4703792403672138, |
| "ce_loss": 0.46037924990395696, |
| "lb_loss": 0.9999999965079875 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 708000000, |
| "cumulative_training_bytes": 708003669, |
| "metrics": { |
| "loss": 0.47049825695501096, |
| "ce_loss": 0.4604982664917541, |
| "lb_loss": 0.9999999966154034 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 709000000, |
| "cumulative_training_bytes": 709000496, |
| "metrics": { |
| "loss": 0.4705623017969561, |
| "ce_loss": 0.46056231133369924, |
| "lb_loss": 0.9999999967613283 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 710000000, |
| "cumulative_training_bytes": 710001059, |
| "metrics": { |
| "loss": 0.4706237760850185, |
| "ce_loss": 0.46062378562176165, |
| "lb_loss": 0.999999996901027 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 711000000, |
| "cumulative_training_bytes": 711002755, |
| "metrics": { |
| "loss": 0.47068721222477916, |
| "ce_loss": 0.4606872217615223, |
| "lb_loss": 0.9999999968678425 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 712000000, |
| "cumulative_training_bytes": 712004688, |
| "metrics": { |
| "loss": 0.47075681223318344, |
| "ce_loss": 0.4607568217699266, |
| "lb_loss": 0.9999999968870078 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 713000000, |
| "cumulative_training_bytes": 713007034, |
| "metrics": { |
| "loss": 0.4708357584098345, |
| "ce_loss": 0.46083576794657766, |
| "lb_loss": 0.9999999969550882 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 714000000, |
| "cumulative_training_bytes": 714005137, |
| "metrics": { |
| "loss": 0.47085419542649215, |
| "ce_loss": 0.4608542049632353, |
| "lb_loss": 0.9999999971366396 |
| } |
| }, |
| { |
| "epoch": 15, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.4708282861734111, |
| "ce_loss": 0.46082829571015427, |
| "lb_loss": 0.999999997136602, |
| "training_bytes": 47653389 |
| }, |
| "cumulative_training_bytes": 714801019, |
| "training_bytes_this_epoch": 47653389 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 715000000, |
| "cumulative_training_bytes": 715000925, |
| "metrics": { |
| "loss": 0.4613972260401799, |
| "ce_loss": 0.4513972355769231, |
| "lb_loss": 0.9999999954150274 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 716000000, |
| "cumulative_training_bytes": 716000809, |
| "metrics": { |
| "loss": 0.46277442956582093, |
| "ce_loss": 0.4527744391025641, |
| "lb_loss": 0.9999999912121357 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 717000000, |
| "cumulative_training_bytes": 717007272, |
| "metrics": { |
| "loss": 0.4638031588660346, |
| "ce_loss": 0.4538031684027778, |
| "lb_loss": 0.9999999913076559 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 718000000, |
| "cumulative_training_bytes": 718001268, |
| "metrics": { |
| "loss": 0.4635652632450314, |
| "ce_loss": 0.45356527278177455, |
| "lb_loss": 0.9999999925672866 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 719000000, |
| "cumulative_training_bytes": 719000209, |
| "metrics": { |
| "loss": 0.4637926949879369, |
| "ce_loss": 0.45379270452468007, |
| "lb_loss": 0.9999999930261476 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 720000000, |
| "cumulative_training_bytes": 720006860, |
| "metrics": { |
| "loss": 0.46441365025707126, |
| "ce_loss": 0.45441365979381443, |
| "lb_loss": 0.9999999941185402 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 721000000, |
| "cumulative_training_bytes": 721002611, |
| "metrics": { |
| "loss": 0.4646387226944064, |
| "ce_loss": 0.4546387322311496, |
| "lb_loss": 0.9999999942531986 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 722000000, |
| "cumulative_training_bytes": 722004482, |
| "metrics": { |
| "loss": 0.4649596812877249, |
| "ce_loss": 0.4549596908244681, |
| "lb_loss": 0.9999999962588574 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 723000000, |
| "cumulative_training_bytes": 723006477, |
| "metrics": { |
| "loss": 0.46530060029052106, |
| "ce_loss": 0.4553006098272642, |
| "lb_loss": 0.9999999972729902 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 724000000, |
| "cumulative_training_bytes": 724002028, |
| "metrics": { |
| "loss": 0.46535132528840256, |
| "ce_loss": 0.4553513348251457, |
| "lb_loss": 0.9999999976177993 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 725000000, |
| "cumulative_training_bytes": 725004555, |
| "metrics": { |
| "loss": 0.4655385368221157, |
| "ce_loss": 0.4555385463588589, |
| "lb_loss": 0.9999999973598543 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 726000000, |
| "cumulative_training_bytes": 726004006, |
| "metrics": { |
| "loss": 0.4657723225719376, |
| "ce_loss": 0.4557723321086808, |
| "lb_loss": 0.9999999966592065 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 727000000, |
| "cumulative_training_bytes": 727000132, |
| "metrics": { |
| "loss": 0.4658971283320578, |
| "ce_loss": 0.455897137868801, |
| "lb_loss": 0.9999999966699226 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 728000000, |
| "cumulative_training_bytes": 728001078, |
| "metrics": { |
| "loss": 0.4661883605328624, |
| "ce_loss": 0.45618837006960555, |
| "lb_loss": 0.9999999971304029 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 729000000, |
| "cumulative_training_bytes": 729008228, |
| "metrics": { |
| "loss": 0.4664945520203689, |
| "ce_loss": 0.45649456155711204, |
| "lb_loss": 0.9999999964995117 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 730000000, |
| "cumulative_training_bytes": 730005608, |
| "metrics": { |
| "loss": 0.46660442462740587, |
| "ce_loss": 0.45660443416414903, |
| "lb_loss": 0.9999999962184364 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 731000000, |
| "cumulative_training_bytes": 731002789, |
| "metrics": { |
| "loss": 0.46665003033802055, |
| "ce_loss": 0.4566500398747637, |
| "lb_loss": 0.9999999959437292 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 732000000, |
| "cumulative_training_bytes": 732003542, |
| "metrics": { |
| "loss": 0.4667574378931634, |
| "ce_loss": 0.45675744742990654, |
| "lb_loss": 0.9999999955701043 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 733000000, |
| "cumulative_training_bytes": 733005914, |
| "metrics": { |
| "loss": 0.4668850435540895, |
| "ce_loss": 0.45688505309083266, |
| "lb_loss": 0.9999999954130993 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 734000000, |
| "cumulative_training_bytes": 734004082, |
| "metrics": { |
| "loss": 0.4669946388384562, |
| "ce_loss": 0.4569946483751994, |
| "lb_loss": 0.999999995555794 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 735000000, |
| "cumulative_training_bytes": 735003332, |
| "metrics": { |
| "loss": 0.46714373572661416, |
| "ce_loss": 0.4571437452633573, |
| "lb_loss": 0.9999999958667488 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 736000000, |
| "cumulative_training_bytes": 736001573, |
| "metrics": { |
| "loss": 0.467144097090198, |
| "ce_loss": 0.45714410662694116, |
| "lb_loss": 0.9999999956733356 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 737000000, |
| "cumulative_training_bytes": 737001233, |
| "metrics": { |
| "loss": 0.4672433902477396, |
| "ce_loss": 0.45724339978448275, |
| "lb_loss": 0.9999999959098882 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 738000000, |
| "cumulative_training_bytes": 738001247, |
| "metrics": { |
| "loss": 0.4672973706595617, |
| "ce_loss": 0.45729738019630484, |
| "lb_loss": 0.9999999956343678 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 739000000, |
| "cumulative_training_bytes": 739005713, |
| "metrics": { |
| "loss": 0.46737158577476856, |
| "ce_loss": 0.4573715953115117, |
| "lb_loss": 0.9999999955513295 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 740000000, |
| "cumulative_training_bytes": 740004035, |
| "metrics": { |
| "loss": 0.46742578120596645, |
| "ce_loss": 0.4574257907427096, |
| "lb_loss": 0.9999999952924643 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 741000000, |
| "cumulative_training_bytes": 741006421, |
| "metrics": { |
| "loss": 0.4675053996438002, |
| "ce_loss": 0.4575054091805434, |
| "lb_loss": 0.9999999957860578 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 742000000, |
| "cumulative_training_bytes": 742000099, |
| "metrics": { |
| "loss": 0.467600741589066, |
| "ce_loss": 0.45760075112580917, |
| "lb_loss": 0.9999999958563616 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 743000000, |
| "cumulative_training_bytes": 743007078, |
| "metrics": { |
| "loss": 0.4677033062705838, |
| "ce_loss": 0.457703315807327, |
| "lb_loss": 0.9999999960209653 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 744000000, |
| "cumulative_training_bytes": 744007281, |
| "metrics": { |
| "loss": 0.4678138212088019, |
| "ce_loss": 0.45781383074554505, |
| "lb_loss": 0.9999999962512802 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 745000000, |
| "cumulative_training_bytes": 745003239, |
| "metrics": { |
| "loss": 0.46785139457501557, |
| "ce_loss": 0.45785140411175873, |
| "lb_loss": 0.9999999964805164 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 746000000, |
| "cumulative_training_bytes": 746000194, |
| "metrics": { |
| "loss": 0.46787027919608315, |
| "ce_loss": 0.4578702887328263, |
| "lb_loss": 0.9999999962856772 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 747000000, |
| "cumulative_training_bytes": 747000581, |
| "metrics": { |
| "loss": 0.4678921757342982, |
| "ce_loss": 0.45789218527104136, |
| "lb_loss": 0.9999999963154523 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 748000000, |
| "cumulative_training_bytes": 748000684, |
| "metrics": { |
| "loss": 0.46796389480381484, |
| "ce_loss": 0.457963904340558, |
| "lb_loss": 0.9999999963167985 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 749000000, |
| "cumulative_training_bytes": 749004115, |
| "metrics": { |
| "loss": 0.468031408323597, |
| "ce_loss": 0.4580314178603402, |
| "lb_loss": 0.9999999968650198 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 750000000, |
| "cumulative_training_bytes": 750006305, |
| "metrics": { |
| "loss": 0.4680848832388602, |
| "ce_loss": 0.4580848927756034, |
| "lb_loss": 0.9999999969413577 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 751000000, |
| "cumulative_training_bytes": 751001387, |
| "metrics": { |
| "loss": 0.4680606794437983, |
| "ce_loss": 0.4580606889805415, |
| "lb_loss": 0.9999999969491701 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 752000000, |
| "cumulative_training_bytes": 752007732, |
| "metrics": { |
| "loss": 0.4681460488480305, |
| "ce_loss": 0.4581460583847737, |
| "lb_loss": 0.9999999967254238 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 753000000, |
| "cumulative_training_bytes": 753007023, |
| "metrics": { |
| "loss": 0.4681992013851005, |
| "ce_loss": 0.4581992109218437, |
| "lb_loss": 0.9999999967748989 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 754000000, |
| "cumulative_training_bytes": 754007126, |
| "metrics": { |
| "loss": 0.4682459842877052, |
| "ce_loss": 0.45824599382444836, |
| "lb_loss": 0.9999999969039571 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 755000000, |
| "cumulative_training_bytes": 755004818, |
| "metrics": { |
| "loss": 0.46830369154400336, |
| "ce_loss": 0.45830370108074653, |
| "lb_loss": 0.9999999969692553 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 756000000, |
| "cumulative_training_bytes": 756003553, |
| "metrics": { |
| "loss": 0.4682925505125042, |
| "ce_loss": 0.45829256004924734, |
| "lb_loss": 0.9999999967987842 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 757000000, |
| "cumulative_training_bytes": 757001249, |
| "metrics": { |
| "loss": 0.46836309380203384, |
| "ce_loss": 0.458363103338777, |
| "lb_loss": 0.9999999967877736 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 758000000, |
| "cumulative_training_bytes": 758007153, |
| "metrics": { |
| "loss": 0.46841179035691155, |
| "ce_loss": 0.4584117998936547, |
| "lb_loss": 0.9999999969046153 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 759000000, |
| "cumulative_training_bytes": 759008102, |
| "metrics": { |
| "loss": 0.4684474506226194, |
| "ce_loss": 0.45844746015936255, |
| "lb_loss": 0.9999999969232316 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 760000000, |
| "cumulative_training_bytes": 760000290, |
| "metrics": { |
| "loss": 0.46847780365146624, |
| "ce_loss": 0.4584778131882094, |
| "lb_loss": 0.9999999969707957 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 761000000, |
| "cumulative_training_bytes": 761003802, |
| "metrics": { |
| "loss": 0.46849883554943517, |
| "ce_loss": 0.45849884508617833, |
| "lb_loss": 0.9999999969871699 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 762000000, |
| "cumulative_training_bytes": 762007779, |
| "metrics": { |
| "loss": 0.4685566738574174, |
| "ce_loss": 0.45855668339416056, |
| "lb_loss": 0.9999999969931801 |
| } |
| }, |
| { |
| "epoch": 16, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.46860215044879666, |
| "ce_loss": 0.4586021599855398, |
| "lb_loss": 0.9999999970504129, |
| "training_bytes": 47653395 |
| }, |
| "cumulative_training_bytes": 762454414, |
| "training_bytes_this_epoch": 47653395 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 763000000, |
| "cumulative_training_bytes": 763000457, |
| "metrics": { |
| "loss": 0.4623547439843836, |
| "ce_loss": 0.45235475352112675, |
| "lb_loss": 0.9999999974814939 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 764000000, |
| "cumulative_training_bytes": 764004913, |
| "metrics": { |
| "loss": 0.46177133947315785, |
| "ce_loss": 0.451771349009901, |
| "lb_loss": 0.99999999527884 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 765000000, |
| "cumulative_training_bytes": 765007589, |
| "metrics": { |
| "loss": 0.46142407079358716, |
| "ce_loss": 0.4514240803303303, |
| "lb_loss": 0.9999999946302122 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 766000000, |
| "cumulative_training_bytes": 766002665, |
| "metrics": { |
| "loss": 0.46156417782826764, |
| "ce_loss": 0.4515641873650108, |
| "lb_loss": 0.9999999921471202 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 767000000, |
| "cumulative_training_bytes": 767006783, |
| "metrics": { |
| "loss": 0.4617538572561861, |
| "ce_loss": 0.4517538667929293, |
| "lb_loss": 0.9999999925745056 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 768000000, |
| "cumulative_training_bytes": 768006387, |
| "metrics": { |
| "loss": 0.46248921460118786, |
| "ce_loss": 0.452489224137931, |
| "lb_loss": 0.9999999929296559 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 769000000, |
| "cumulative_training_bytes": 769007261, |
| "metrics": { |
| "loss": 0.46278501893872415, |
| "ce_loss": 0.4527850284754673, |
| "lb_loss": 0.9999999917834719 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 770000000, |
| "cumulative_training_bytes": 770004386, |
| "metrics": { |
| "loss": 0.46281795762857125, |
| "ce_loss": 0.4528179671653144, |
| "lb_loss": 0.9999999905696505 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 771000000, |
| "cumulative_training_bytes": 771004582, |
| "metrics": { |
| "loss": 0.46292215921885216, |
| "ce_loss": 0.4529221687555953, |
| "lb_loss": 0.9999999913554588 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 772000000, |
| "cumulative_training_bytes": 772000420, |
| "metrics": { |
| "loss": 0.4632846208719107, |
| "ce_loss": 0.45328463040865385, |
| "lb_loss": 0.9999999921673384 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 773000000, |
| "cumulative_training_bytes": 773003781, |
| "metrics": { |
| "loss": 0.46348049037261146, |
| "ce_loss": 0.45348049990935463, |
| "lb_loss": 0.9999999921333972 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 774000000, |
| "cumulative_training_bytes": 774007887, |
| "metrics": { |
| "loss": 0.4635867557778264, |
| "ce_loss": 0.45358676531456954, |
| "lb_loss": 0.9999999925000778 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 775000000, |
| "cumulative_training_bytes": 775006162, |
| "metrics": { |
| "loss": 0.4637526105089885, |
| "ce_loss": 0.4537526200457317, |
| "lb_loss": 0.999999993857814 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 776000000, |
| "cumulative_training_bytes": 776001976, |
| "metrics": { |
| "loss": 0.46378485981353934, |
| "ce_loss": 0.4537848693502825, |
| "lb_loss": 0.9999999951844835 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 777000000, |
| "cumulative_training_bytes": 777005637, |
| "metrics": { |
| "loss": 0.46390480207556617, |
| "ce_loss": 0.45390481161230933, |
| "lb_loss": 0.9999999949519475 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 778000000, |
| "cumulative_training_bytes": 778003374, |
| "metrics": { |
| "loss": 0.4639972131983627, |
| "ce_loss": 0.45399722273510584, |
| "lb_loss": 0.9999999953337575 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 779000000, |
| "cumulative_training_bytes": 779004889, |
| "metrics": { |
| "loss": 0.4640907123064576, |
| "ce_loss": 0.45409072184320076, |
| "lb_loss": 0.9999999953132332 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 780000000, |
| "cumulative_training_bytes": 780004536, |
| "metrics": { |
| "loss": 0.4642408185269289, |
| "ce_loss": 0.45424082806367205, |
| "lb_loss": 0.9999999955290018 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 781000000, |
| "cumulative_training_bytes": 781001322, |
| "metrics": { |
| "loss": 0.4643268515342432, |
| "ce_loss": 0.45432686107098635, |
| "lb_loss": 0.9999999953752896 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 782000000, |
| "cumulative_training_bytes": 782005205, |
| "metrics": { |
| "loss": 0.4643638559683469, |
| "ce_loss": 0.45436386550509006, |
| "lb_loss": 0.9999999956125007 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 783000000, |
| "cumulative_training_bytes": 783003504, |
| "metrics": { |
| "loss": 0.4644428404539424, |
| "ce_loss": 0.45444284999068557, |
| "lb_loss": 0.9999999955585213 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 784000000, |
| "cumulative_training_bytes": 784007012, |
| "metrics": { |
| "loss": 0.4645556631986032, |
| "ce_loss": 0.45455567273534636, |
| "lb_loss": 0.9999999957228639 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 785000000, |
| "cumulative_training_bytes": 785003493, |
| "metrics": { |
| "loss": 0.4646576464479767, |
| "ce_loss": 0.45465765598471986, |
| "lb_loss": 0.9999999954259254 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 786000000, |
| "cumulative_training_bytes": 786001543, |
| "metrics": { |
| "loss": 0.4646982882274845, |
| "ce_loss": 0.45469829776422765, |
| "lb_loss": 0.9999999956386846 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 787000000, |
| "cumulative_training_bytes": 787006273, |
| "metrics": { |
| "loss": 0.46475888979263924, |
| "ce_loss": 0.4547588993293824, |
| "lb_loss": 0.9999999954264683 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 788000000, |
| "cumulative_training_bytes": 788002324, |
| "metrics": { |
| "loss": 0.46486149188711773, |
| "ce_loss": 0.4548615014238609, |
| "lb_loss": 0.9999999956582948 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 789000000, |
| "cumulative_training_bytes": 789000949, |
| "metrics": { |
| "loss": 0.46496146893267404, |
| "ce_loss": 0.4549614784694172, |
| "lb_loss": 0.9999999959587157 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 790000000, |
| "cumulative_training_bytes": 790004342, |
| "metrics": { |
| "loss": 0.46504716519219763, |
| "ce_loss": 0.4550471747289408, |
| "lb_loss": 0.9999999959567603 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 791000000, |
| "cumulative_training_bytes": 791006479, |
| "metrics": { |
| "loss": 0.46513784085220533, |
| "ce_loss": 0.4551378503889485, |
| "lb_loss": 0.9999999957790702 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 792000000, |
| "cumulative_training_bytes": 792007573, |
| "metrics": { |
| "loss": 0.4652086949527619, |
| "ce_loss": 0.45520870448950507, |
| "lb_loss": 0.9999999958760196 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 793000000, |
| "cumulative_training_bytes": 793000192, |
| "metrics": { |
| "loss": 0.46531999156002296, |
| "ce_loss": 0.4553200010967661, |
| "lb_loss": 0.9999999958012271 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 794000000, |
| "cumulative_training_bytes": 794002455, |
| "metrics": { |
| "loss": 0.4653909947108297, |
| "ce_loss": 0.45539100424757284, |
| "lb_loss": 0.9999999960070675 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 795000000, |
| "cumulative_training_bytes": 795002572, |
| "metrics": { |
| "loss": 0.46549575616544453, |
| "ce_loss": 0.4554957657021877, |
| "lb_loss": 0.9999999956814325 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 796000000, |
| "cumulative_training_bytes": 796000398, |
| "metrics": { |
| "loss": 0.4655890219343822, |
| "ce_loss": 0.45558903147112534, |
| "lb_loss": 0.9999999957279483 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 797000000, |
| "cumulative_training_bytes": 797004240, |
| "metrics": { |
| "loss": 0.4656447468074501, |
| "ce_loss": 0.45564475634419327, |
| "lb_loss": 0.999999996076556 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 798000000, |
| "cumulative_training_bytes": 798001241, |
| "metrics": { |
| "loss": 0.4657365903932439, |
| "ce_loss": 0.45573659992998705, |
| "lb_loss": 0.9999999960323491 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 799000000, |
| "cumulative_training_bytes": 799006684, |
| "metrics": { |
| "loss": 0.46579053755104227, |
| "ce_loss": 0.45579054708778544, |
| "lb_loss": 0.9999999962786121 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 800000000, |
| "cumulative_training_bytes": 800005908, |
| "metrics": { |
| "loss": 0.46589497096386856, |
| "ce_loss": 0.4558949805006117, |
| "lb_loss": 0.9999999962564783 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 801000000, |
| "cumulative_training_bytes": 801004990, |
| "metrics": { |
| "loss": 0.4659691438892747, |
| "ce_loss": 0.45596915342601785, |
| "lb_loss": 0.9999999963183626 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 802000000, |
| "cumulative_training_bytes": 802006894, |
| "metrics": { |
| "loss": 0.4660316155116502, |
| "ce_loss": 0.45603162504839334, |
| "lb_loss": 0.9999999962848053 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 803000000, |
| "cumulative_training_bytes": 803004532, |
| "metrics": { |
| "loss": 0.4660554155482266, |
| "ce_loss": 0.45605542508496977, |
| "lb_loss": 0.9999999960833806 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 804000000, |
| "cumulative_training_bytes": 804005297, |
| "metrics": { |
| "loss": 0.4661466299510033, |
| "ce_loss": 0.4561466394877465, |
| "lb_loss": 0.9999999963426669 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 805000000, |
| "cumulative_training_bytes": 805002675, |
| "metrics": { |
| "loss": 0.4661778685899439, |
| "ce_loss": 0.4561778781266871, |
| "lb_loss": 0.9999999964389523 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 806000000, |
| "cumulative_training_bytes": 806004255, |
| "metrics": { |
| "loss": 0.46626139055995164, |
| "ce_loss": 0.4562614000966948, |
| "lb_loss": 0.9999999964371344 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 807000000, |
| "cumulative_training_bytes": 807001576, |
| "metrics": { |
| "loss": 0.4663353263819574, |
| "ce_loss": 0.45633533591870057, |
| "lb_loss": 0.9999999966601729 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 808000000, |
| "cumulative_training_bytes": 808007129, |
| "metrics": { |
| "loss": 0.4663545905010783, |
| "ce_loss": 0.45635460003782147, |
| "lb_loss": 0.9999999965934477 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 809000000, |
| "cumulative_training_bytes": 809006953, |
| "metrics": { |
| "loss": 0.4663725949212269, |
| "ce_loss": 0.45637260445797007, |
| "lb_loss": 0.9999999967741523 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 810000000, |
| "cumulative_training_bytes": 810005884, |
| "metrics": { |
| "loss": 0.46644152880866746, |
| "ce_loss": 0.4564415383454106, |
| "lb_loss": 0.9999999967942107 |
| } |
| }, |
| { |
| "epoch": 17, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.4664472485569265, |
| "ce_loss": 0.45644725809366965, |
| "lb_loss": 0.9999999967535388, |
| "training_bytes": 47653413 |
| }, |
| "cumulative_training_bytes": 810107827, |
| "training_bytes_this_epoch": 47653413 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 811000000, |
| "cumulative_training_bytes": 811000418, |
| "metrics": { |
| "loss": 0.4582590152477396, |
| "ce_loss": 0.44825902478448276, |
| "lb_loss": 0.9999999866403383 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 812000000, |
| "cumulative_training_bytes": 812004574, |
| "metrics": { |
| "loss": 0.458675330685031, |
| "ce_loss": 0.4486753402217742, |
| "lb_loss": 0.9999999891846411 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 813000000, |
| "cumulative_training_bytes": 813005709, |
| "metrics": { |
| "loss": 0.45849726948700353, |
| "ce_loss": 0.4484972790237467, |
| "lb_loss": 0.9999999976409771 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 814000000, |
| "cumulative_training_bytes": 814001225, |
| "metrics": { |
| "loss": 0.45856258255559473, |
| "ce_loss": 0.4485625920923379, |
| "lb_loss": 0.9999999977750722 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 815000000, |
| "cumulative_training_bytes": 815003952, |
| "metrics": { |
| "loss": 0.4582559323646653, |
| "ce_loss": 0.44825594190140844, |
| "lb_loss": 0.9999999988806639 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 816000000, |
| "cumulative_training_bytes": 816003923, |
| "metrics": { |
| "loss": 0.4589549417619581, |
| "ce_loss": 0.4489549512987013, |
| "lb_loss": 0.9999999982196015 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 817000000, |
| "cumulative_training_bytes": 817007266, |
| "metrics": { |
| "loss": 0.4592534241480515, |
| "ce_loss": 0.44925343368479465, |
| "lb_loss": 0.9999999976846142 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 818000000, |
| "cumulative_training_bytes": 818004841, |
| "metrics": { |
| "loss": 0.4594157579220347, |
| "ce_loss": 0.44941576745877787, |
| "lb_loss": 0.999999997109377 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 819000000, |
| "cumulative_training_bytes": 819004493, |
| "metrics": { |
| "loss": 0.45983224358287816, |
| "ce_loss": 0.44983225311962133, |
| "lb_loss": 0.9999999966145382 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 820000000, |
| "cumulative_training_bytes": 820003619, |
| "metrics": { |
| "loss": 0.46002989910594827, |
| "ce_loss": 0.45002990864269143, |
| "lb_loss": 0.9999999966809479 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 821000000, |
| "cumulative_training_bytes": 821007131, |
| "metrics": { |
| "loss": 0.460265253367049, |
| "ce_loss": 0.45026526290379215, |
| "lb_loss": 0.9999999963165669 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 822000000, |
| "cumulative_training_bytes": 822001154, |
| "metrics": { |
| "loss": 0.4605283648197562, |
| "ce_loss": 0.45052837435649934, |
| "lb_loss": 0.9999999952439023 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 823000000, |
| "cumulative_training_bytes": 823004363, |
| "metrics": { |
| "loss": 0.46066069150183253, |
| "ce_loss": 0.4506607010385757, |
| "lb_loss": 0.9999999947646959 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 824000000, |
| "cumulative_training_bytes": 824005319, |
| "metrics": { |
| "loss": 0.46092772483825684, |
| "ce_loss": 0.450927734375, |
| "lb_loss": 0.999999995011065 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 825000000, |
| "cumulative_training_bytes": 825005826, |
| "metrics": { |
| "loss": 0.46111766774999885, |
| "ce_loss": 0.451117677286742, |
| "lb_loss": 0.9999999947011287 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 826000000, |
| "cumulative_training_bytes": 826000335, |
| "metrics": { |
| "loss": 0.46198096348830503, |
| "ce_loss": 0.4519809730250482, |
| "lb_loss": 0.9999999948606785 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 827000000, |
| "cumulative_training_bytes": 827007343, |
| "metrics": { |
| "loss": 0.46243060498997784, |
| "ce_loss": 0.452430614526721, |
| "lb_loss": 0.9999999950869359 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 828000000, |
| "cumulative_training_bytes": 828002571, |
| "metrics": { |
| "loss": 0.4624639199134066, |
| "ce_loss": 0.45246392945014974, |
| "lb_loss": 0.9999999954601512 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 829000000, |
| "cumulative_training_bytes": 829005468, |
| "metrics": { |
| "loss": 0.4625939747116361, |
| "ce_loss": 0.45259398424837927, |
| "lb_loss": 0.9999999954113118 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 830000000, |
| "cumulative_training_bytes": 830003520, |
| "metrics": { |
| "loss": 0.4626671563119866, |
| "ce_loss": 0.4526671658487298, |
| "lb_loss": 0.9999999956409228 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 831000000, |
| "cumulative_training_bytes": 831006123, |
| "metrics": { |
| "loss": 0.46278288941891826, |
| "ce_loss": 0.45278289895566143, |
| "lb_loss": 0.9999999959156949 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 832000000, |
| "cumulative_training_bytes": 832004095, |
| "metrics": { |
| "loss": 0.4628742745223684, |
| "ce_loss": 0.45287428405911156, |
| "lb_loss": 0.9999999959763216 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 833000000, |
| "cumulative_training_bytes": 833003109, |
| "metrics": { |
| "loss": 0.46298912089803945, |
| "ce_loss": 0.4529891304347826, |
| "lb_loss": 0.9999999960330019 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 834000000, |
| "cumulative_training_bytes": 834000963, |
| "metrics": { |
| "loss": 0.46311497764685183, |
| "ce_loss": 0.453114987183595, |
| "lb_loss": 0.9999999959130427 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 835000000, |
| "cumulative_training_bytes": 835000614, |
| "metrics": { |
| "loss": 0.4632120762950527, |
| "ce_loss": 0.45321208583179584, |
| "lb_loss": 0.9999999960226913 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 836000000, |
| "cumulative_training_bytes": 836003119, |
| "metrics": { |
| "loss": 0.46328433497995797, |
| "ce_loss": 0.45328434451670113, |
| "lb_loss": 0.9999999961943236 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 837000000, |
| "cumulative_training_bytes": 837001175, |
| "metrics": { |
| "loss": 0.46331735631865106, |
| "ce_loss": 0.4533173658553942, |
| "lb_loss": 0.9999999961315517 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 838000000, |
| "cumulative_training_bytes": 838002124, |
| "metrics": { |
| "loss": 0.4633801187563414, |
| "ce_loss": 0.45338012829308455, |
| "lb_loss": 0.999999996156122 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 839000000, |
| "cumulative_training_bytes": 839006609, |
| "metrics": { |
| "loss": 0.46342972970166746, |
| "ce_loss": 0.4534297392384106, |
| "lb_loss": 0.9999999963684587 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 840000000, |
| "cumulative_training_bytes": 840000100, |
| "metrics": { |
| "loss": 0.46349160778079845, |
| "ce_loss": 0.4534916173175416, |
| "lb_loss": 0.9999999965046188 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 841000000, |
| "cumulative_training_bytes": 841000215, |
| "metrics": { |
| "loss": 0.46359052719516963, |
| "ce_loss": 0.4535905367319128, |
| "lb_loss": 0.9999999966623762 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 842000000, |
| "cumulative_training_bytes": 842004728, |
| "metrics": { |
| "loss": 0.46364198035406556, |
| "ce_loss": 0.45364198989080873, |
| "lb_loss": 0.9999999965670471 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 843000000, |
| "cumulative_training_bytes": 843006127, |
| "metrics": { |
| "loss": 0.4637348307959697, |
| "ce_loss": 0.4537348403327129, |
| "lb_loss": 0.9999999966439451 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 844000000, |
| "cumulative_training_bytes": 844003435, |
| "metrics": { |
| "loss": 0.46380382049374463, |
| "ce_loss": 0.4538038300304878, |
| "lb_loss": 0.999999996850161 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 845000000, |
| "cumulative_training_bytes": 845006695, |
| "metrics": { |
| "loss": 0.4638635707440202, |
| "ce_loss": 0.45386358028076335, |
| "lb_loss": 0.999999997019114 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 846000000, |
| "cumulative_training_bytes": 846002322, |
| "metrics": { |
| "loss": 0.4639164038776309, |
| "ce_loss": 0.45391641341437405, |
| "lb_loss": 0.9999999970381995 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 847000000, |
| "cumulative_training_bytes": 847004506, |
| "metrics": { |
| "loss": 0.46396459108566346, |
| "ce_loss": 0.4539646006224066, |
| "lb_loss": 0.9999999970939644 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 848000000, |
| "cumulative_training_bytes": 848000971, |
| "metrics": { |
| "loss": 0.4640376325087114, |
| "ce_loss": 0.45403764204545455, |
| "lb_loss": 0.9999999971100778 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 849000000, |
| "cumulative_training_bytes": 849004995, |
| "metrics": { |
| "loss": 0.4640902132294446, |
| "ce_loss": 0.45409022276618777, |
| "lb_loss": 0.9999999971142014 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 850000000, |
| "cumulative_training_bytes": 850005350, |
| "metrics": { |
| "loss": 0.4640859984271793, |
| "ce_loss": 0.45408600796392246, |
| "lb_loss": 0.9999999969002382 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 851000000, |
| "cumulative_training_bytes": 851000453, |
| "metrics": { |
| "loss": 0.464119288698606, |
| "ce_loss": 0.4541192982353492, |
| "lb_loss": 0.9999999969198873 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 852000000, |
| "cumulative_training_bytes": 852004187, |
| "metrics": { |
| "loss": 0.4641575910891706, |
| "ce_loss": 0.45415760062591376, |
| "lb_loss": 0.9999999968846988 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 853000000, |
| "cumulative_training_bytes": 853003588, |
| "metrics": { |
| "loss": 0.4642327954338083, |
| "ce_loss": 0.45423280497055146, |
| "lb_loss": 0.9999999971702953 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 854000000, |
| "cumulative_training_bytes": 854000038, |
| "metrics": { |
| "loss": 0.464275811368542, |
| "ce_loss": 0.4542758209052852, |
| "lb_loss": 0.9999999970785095 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 855000000, |
| "cumulative_training_bytes": 855002152, |
| "metrics": { |
| "loss": 0.464321046376456, |
| "ce_loss": 0.45432105591319916, |
| "lb_loss": 0.9999999970218006 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 856000000, |
| "cumulative_training_bytes": 856002523, |
| "metrics": { |
| "loss": 0.4643201960855458, |
| "ce_loss": 0.454320205622289, |
| "lb_loss": 0.9999999968079595 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 857000000, |
| "cumulative_training_bytes": 857006768, |
| "metrics": { |
| "loss": 0.4643526690346854, |
| "ce_loss": 0.4543526785714286, |
| "lb_loss": 0.9999999967789164 |
| } |
| }, |
| { |
| "epoch": 18, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.4643767603258854, |
| "ce_loss": 0.45437676986262854, |
| "lb_loss": 0.9999999966960793, |
| "training_bytes": 47653406 |
| }, |
| "cumulative_training_bytes": 857761233, |
| "training_bytes_this_epoch": 47653406 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 858000000, |
| "cumulative_training_bytes": 858005351, |
| "metrics": { |
| "loss": 0.45714354515075684, |
| "ce_loss": 0.4471435546875, |
| "lb_loss": 1.0000000167638063 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 859000000, |
| "cumulative_training_bytes": 859002822, |
| "metrics": { |
| "loss": 0.4560117574385655, |
| "ce_loss": 0.44601176697530864, |
| "lb_loss": 1.0000000051510187 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 860000000, |
| "cumulative_training_bytes": 860005316, |
| "metrics": { |
| "loss": 0.45533248833015105, |
| "ce_loss": 0.4453324978668942, |
| "lb_loss": 0.9999999993897135 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 861000000, |
| "cumulative_training_bytes": 861001866, |
| "metrics": { |
| "loss": 0.4553494289975358, |
| "ce_loss": 0.44534943853427894, |
| "lb_loss": 1.0000000015500026 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 862000000, |
| "cumulative_training_bytes": 862004590, |
| "metrics": { |
| "loss": 0.4559224013387081, |
| "ce_loss": 0.44592241087545126, |
| "lb_loss": 0.9999999984937454 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 863000000, |
| "cumulative_training_bytes": 863001335, |
| "metrics": { |
| "loss": 0.45603777372349075, |
| "ce_loss": 0.4460377832602339, |
| "lb_loss": 1.0000000004357064 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 864000000, |
| "cumulative_training_bytes": 864007195, |
| "metrics": { |
| "loss": 0.4564063362046784, |
| "ce_loss": 0.4464063457414216, |
| "lb_loss": 0.9999999999269551 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 865000000, |
| "cumulative_training_bytes": 865006139, |
| "metrics": { |
| "loss": 0.45668132965458874, |
| "ce_loss": 0.4466813391913319, |
| "lb_loss": 0.9999999995589508 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 866000000, |
| "cumulative_training_bytes": 866001721, |
| "metrics": { |
| "loss": 0.45680093144838696, |
| "ce_loss": 0.4468009409851301, |
| "lb_loss": 0.9999999989475016 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 867000000, |
| "cumulative_training_bytes": 867002974, |
| "metrics": { |
| "loss": 0.4570681993489238, |
| "ce_loss": 0.44706820888566695, |
| "lb_loss": 0.999999999061733 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 868000000, |
| "cumulative_training_bytes": 868004509, |
| "metrics": { |
| "loss": 0.45706402140940494, |
| "ce_loss": 0.4470640309461481, |
| "lb_loss": 0.9999999983950881 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 869000000, |
| "cumulative_training_bytes": 869007120, |
| "metrics": { |
| "loss": 0.45723368426434674, |
| "ce_loss": 0.4472336938010899, |
| "lb_loss": 0.9999999979292664 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 870000000, |
| "cumulative_training_bytes": 870003043, |
| "metrics": { |
| "loss": 0.4574648391021805, |
| "ce_loss": 0.44746484863892366, |
| "lb_loss": 0.9999999979858255 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 871000000, |
| "cumulative_training_bytes": 871006700, |
| "metrics": { |
| "loss": 0.4577389291561429, |
| "ce_loss": 0.44773893869288606, |
| "lb_loss": 0.9999999981039587 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 872000000, |
| "cumulative_training_bytes": 872007461, |
| "metrics": { |
| "loss": 0.4579849114982031, |
| "ce_loss": 0.44798492103494625, |
| "lb_loss": 0.9999999977247689 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 873000000, |
| "cumulative_training_bytes": 873006742, |
| "metrics": { |
| "loss": 0.45822795379826436, |
| "ce_loss": 0.44822796333500753, |
| "lb_loss": 0.999999997754722 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 874000000, |
| "cumulative_training_bytes": 874001226, |
| "metrics": { |
| "loss": 0.45845995922799043, |
| "ce_loss": 0.4484599687647336, |
| "lb_loss": 0.9999999979766457 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 875000000, |
| "cumulative_training_bytes": 875004724, |
| "metrics": { |
| "loss": 0.45866020915775063, |
| "ce_loss": 0.4486602186944938, |
| "lb_loss": 0.999999998014943 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 876000000, |
| "cumulative_training_bytes": 876004731, |
| "metrics": { |
| "loss": 0.45884581898822535, |
| "ce_loss": 0.4488458285249685, |
| "lb_loss": 0.9999999977488804 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 877000000, |
| "cumulative_training_bytes": 877006989, |
| "metrics": { |
| "loss": 0.45900198554082244, |
| "ce_loss": 0.4490019950775656, |
| "lb_loss": 0.999999997818764 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 878000000, |
| "cumulative_training_bytes": 878005540, |
| "metrics": { |
| "loss": 0.45909980962929314, |
| "ce_loss": 0.4490998191660363, |
| "lb_loss": 0.9999999976329471 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 879000000, |
| "cumulative_training_bytes": 879001520, |
| "metrics": { |
| "loss": 0.45923634252706363, |
| "ce_loss": 0.4492363520638068, |
| "lb_loss": 0.9999999974000858 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 880000000, |
| "cumulative_training_bytes": 880003773, |
| "metrics": { |
| "loss": 0.45938951283847196, |
| "ce_loss": 0.4493895223752151, |
| "lb_loss": 0.9999999969017895 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 881000000, |
| "cumulative_training_bytes": 881006052, |
| "metrics": { |
| "loss": 0.4595082352755098, |
| "ce_loss": 0.449508244812253, |
| "lb_loss": 0.999999996976576 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 882000000, |
| "cumulative_training_bytes": 882002195, |
| "metrics": { |
| "loss": 0.4596030730201109, |
| "ce_loss": 0.44960308255685405, |
| "lb_loss": 0.9999999969124568 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 883000000, |
| "cumulative_training_bytes": 883005745, |
| "metrics": { |
| "loss": 0.4596205058607083, |
| "ce_loss": 0.44962051539745146, |
| "lb_loss": 0.9999999968895634 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 884000000, |
| "cumulative_training_bytes": 884007789, |
| "metrics": { |
| "loss": 0.4597436392449901, |
| "ce_loss": 0.4497436487817333, |
| "lb_loss": 0.9999999968693213 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 885000000, |
| "cumulative_training_bytes": 885004439, |
| "metrics": { |
| "loss": 0.45980132849390065, |
| "ce_loss": 0.4498013380306438, |
| "lb_loss": 0.9999999972518522 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 886000000, |
| "cumulative_training_bytes": 886001168, |
| "metrics": { |
| "loss": 0.4598883227795031, |
| "ce_loss": 0.44988833231624625, |
| "lb_loss": 0.9999999972194199 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 887000000, |
| "cumulative_training_bytes": 887003211, |
| "metrics": { |
| "loss": 0.45996510226917614, |
| "ce_loss": 0.4499651118059193, |
| "lb_loss": 0.9999999972523789 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 888000000, |
| "cumulative_training_bytes": 888000146, |
| "metrics": { |
| "loss": 0.46011987840957236, |
| "ce_loss": 0.4501198879463155, |
| "lb_loss": 0.9999999971624023 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 889000000, |
| "cumulative_training_bytes": 889004224, |
| "metrics": { |
| "loss": 0.4601905158921784, |
| "ce_loss": 0.4501905254289216, |
| "lb_loss": 0.9999999969613318 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 890000000, |
| "cumulative_training_bytes": 890005885, |
| "metrics": { |
| "loss": 0.46025722226686644, |
| "ce_loss": 0.4502572318036096, |
| "lb_loss": 0.9999999972964885 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 891000000, |
| "cumulative_training_bytes": 891002684, |
| "metrics": { |
| "loss": 0.46033590456714996, |
| "ce_loss": 0.4503359141038931, |
| "lb_loss": 0.9999999971577606 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 892000000, |
| "cumulative_training_bytes": 892004908, |
| "metrics": { |
| "loss": 0.4604197900495717, |
| "ce_loss": 0.45041979958631484, |
| "lb_loss": 0.999999997360975 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 893000000, |
| "cumulative_training_bytes": 893001735, |
| "metrics": { |
| "loss": 0.4605195505729917, |
| "ce_loss": 0.4505195601097349, |
| "lb_loss": 0.9999999973060048 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 894000000, |
| "cumulative_training_bytes": 894003726, |
| "metrics": { |
| "loss": 0.4606263274852302, |
| "ce_loss": 0.45062633702197336, |
| "lb_loss": 0.9999999972042614 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 895000000, |
| "cumulative_training_bytes": 895004512, |
| "metrics": { |
| "loss": 0.46070968477349533, |
| "ce_loss": 0.4507096943102385, |
| "lb_loss": 0.9999999971202526 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 896000000, |
| "cumulative_training_bytes": 896007365, |
| "metrics": { |
| "loss": 0.46081095768047403, |
| "ce_loss": 0.4508109672172172, |
| "lb_loss": 0.9999999972196432 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 897000000, |
| "cumulative_training_bytes": 897002215, |
| "metrics": { |
| "loss": 0.4608837557071593, |
| "ce_loss": 0.45088376524390245, |
| "lb_loss": 0.999999997197128 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 898000000, |
| "cumulative_training_bytes": 898006553, |
| "metrics": { |
| "loss": 0.4609418456049996, |
| "ce_loss": 0.4509418551417428, |
| "lb_loss": 0.9999999970855415 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 899000000, |
| "cumulative_training_bytes": 899002574, |
| "metrics": { |
| "loss": 0.4610199235072598, |
| "ce_loss": 0.45101993304400295, |
| "lb_loss": 0.9999999969566882 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 900000000, |
| "cumulative_training_bytes": 900001892, |
| "metrics": { |
| "loss": 0.4611180545152779, |
| "ce_loss": 0.45111806405202104, |
| "lb_loss": 0.9999999968668938 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 901000000, |
| "cumulative_training_bytes": 901002457, |
| "metrics": { |
| "loss": 0.4611618370239863, |
| "ce_loss": 0.45116184656072944, |
| "lb_loss": 0.999999996834031 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 902000000, |
| "cumulative_training_bytes": 902005917, |
| "metrics": { |
| "loss": 0.4612114078797649, |
| "ce_loss": 0.45121141741650805, |
| "lb_loss": 0.999999996926426 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 903000000, |
| "cumulative_training_bytes": 903002754, |
| "metrics": { |
| "loss": 0.4612944935517659, |
| "ce_loss": 0.45129450308850905, |
| "lb_loss": 0.999999996913349 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 904000000, |
| "cumulative_training_bytes": 904007438, |
| "metrics": { |
| "loss": 0.46135489009074027, |
| "ce_loss": 0.45135489962748343, |
| "lb_loss": 0.999999997118451 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 905000000, |
| "cumulative_training_bytes": 905006302, |
| "metrics": { |
| "loss": 0.4613947179754124, |
| "ce_loss": 0.4513947275121556, |
| "lb_loss": 0.9999999970922208 |
| } |
| }, |
| { |
| "epoch": 19, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.4614244788348828, |
| "ce_loss": 0.45142448837162596, |
| "lb_loss": 0.999999997069566, |
| "training_bytes": 47653396 |
| }, |
| "cumulative_training_bytes": 905414629, |
| "training_bytes_this_epoch": 47653396 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 906000000, |
| "cumulative_training_bytes": 906007139, |
| "metrics": { |
| "loss": 0.4505945521515685, |
| "ce_loss": 0.4405945616883117, |
| "lb_loss": 0.9999999984518274 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 907000000, |
| "cumulative_training_bytes": 907003717, |
| "metrics": { |
| "loss": 0.45137793430383655, |
| "ce_loss": 0.4413779438405797, |
| "lb_loss": 0.9999999956808229 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 908000000, |
| "cumulative_training_bytes": 908005720, |
| "metrics": { |
| "loss": 0.45178184142479527, |
| "ce_loss": 0.44178185096153844, |
| "lb_loss": 0.9999999961204077 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 909000000, |
| "cumulative_training_bytes": 909006677, |
| "metrics": { |
| "loss": 0.45235156962103934, |
| "ce_loss": 0.4423515791577825, |
| "lb_loss": 0.999999996949869 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 910000000, |
| "cumulative_training_bytes": 910005425, |
| "metrics": { |
| "loss": 0.45257681120616167, |
| "ce_loss": 0.44257682074290483, |
| "lb_loss": 0.9999999954226817 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 911000000, |
| "cumulative_training_bytes": 911000290, |
| "metrics": { |
| "loss": 0.45285567890632955, |
| "ce_loss": 0.4428556884430727, |
| "lb_loss": 0.9999999962389388 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 912000000, |
| "cumulative_training_bytes": 912003476, |
| "metrics": { |
| "loss": 0.4531254446783731, |
| "ce_loss": 0.4431254542151163, |
| "lb_loss": 0.9999999967425368 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 913000000, |
| "cumulative_training_bytes": 913000462, |
| "metrics": { |
| "loss": 0.4532784784683073, |
| "ce_loss": 0.4432784880050505, |
| "lb_loss": 0.9999999971100778 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 914000000, |
| "cumulative_training_bytes": 914002700, |
| "metrics": { |
| "loss": 0.4536259382777082, |
| "ce_loss": 0.44362594781445136, |
| "lb_loss": 0.999999997447794 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 915000000, |
| "cumulative_training_bytes": 915005428, |
| "metrics": { |
| "loss": 0.4537899265654933, |
| "ce_loss": 0.44378993610223644, |
| "lb_loss": 0.999999997191155 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 916000000, |
| "cumulative_training_bytes": 916000562, |
| "metrics": { |
| "loss": 0.4539811997613755, |
| "ce_loss": 0.4439812092981187, |
| "lb_loss": 0.999999998059183 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 917000000, |
| "cumulative_training_bytes": 917005115, |
| "metrics": { |
| "loss": 0.45415455713543135, |
| "ce_loss": 0.4441545666721745, |
| "lb_loss": 0.9999999975575096 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 918000000, |
| "cumulative_training_bytes": 918000290, |
| "metrics": { |
| "loss": 0.45433057856429154, |
| "ce_loss": 0.4443305881010347, |
| "lb_loss": 0.9999999979684357 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 919000000, |
| "cumulative_training_bytes": 919001157, |
| "metrics": { |
| "loss": 0.45467502745592875, |
| "ce_loss": 0.4446750369926719, |
| "lb_loss": 0.9999999979168613 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 920000000, |
| "cumulative_training_bytes": 920005392, |
| "metrics": { |
| "loss": 0.4549423707126007, |
| "ce_loss": 0.44494238024934385, |
| "lb_loss": 0.9999999978723801 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 921000000, |
| "cumulative_training_bytes": 921006037, |
| "metrics": { |
| "loss": 0.455245339800192, |
| "ce_loss": 0.44524534933693516, |
| "lb_loss": 0.9999999982142027 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 922000000, |
| "cumulative_training_bytes": 922000214, |
| "metrics": { |
| "loss": 0.45546488110268435, |
| "ce_loss": 0.4454648906394275, |
| "lb_loss": 0.9999999985965665 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 923000000, |
| "cumulative_training_bytes": 923002745, |
| "metrics": { |
| "loss": 0.4556951227009582, |
| "ce_loss": 0.44569513223770135, |
| "lb_loss": 0.9999999987804012 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 924000000, |
| "cumulative_training_bytes": 924007204, |
| "metrics": { |
| "loss": 0.4558257086469471, |
| "ce_loss": 0.44582571818369027, |
| "lb_loss": 0.9999999989689476 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 925000000, |
| "cumulative_training_bytes": 925003699, |
| "metrics": { |
| "loss": 0.4559782938252584, |
| "ce_loss": 0.44597830336200156, |
| "lb_loss": 0.9999999988116353 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 926000000, |
| "cumulative_training_bytes": 926005686, |
| "metrics": { |
| "loss": 0.4561811916532903, |
| "ce_loss": 0.44618120119003346, |
| "lb_loss": 0.9999999985813696 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 927000000, |
| "cumulative_training_bytes": 927002982, |
| "metrics": { |
| "loss": 0.4563115703674782, |
| "ce_loss": 0.4463115799042214, |
| "lb_loss": 0.9999999980547615 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 928000000, |
| "cumulative_training_bytes": 928004103, |
| "metrics": { |
| "loss": 0.45646053071749415, |
| "ce_loss": 0.4464605402542373, |
| "lb_loss": 0.999999998282578 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 929000000, |
| "cumulative_training_bytes": 929006434, |
| "metrics": { |
| "loss": 0.4565663954454055, |
| "ce_loss": 0.44656640498214867, |
| "lb_loss": 0.9999999982201794 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 930000000, |
| "cumulative_training_bytes": 930003003, |
| "metrics": { |
| "loss": 0.45671574782077784, |
| "ce_loss": 0.446715757357521, |
| "lb_loss": 0.999999998180861 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 931000000, |
| "cumulative_training_bytes": 931005136, |
| "metrics": { |
| "loss": 0.45679983430676374, |
| "ce_loss": 0.4467998438435069, |
| "lb_loss": 0.9999999983770129 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 932000000, |
| "cumulative_training_bytes": 932004010, |
| "metrics": { |
| "loss": 0.4569366266567495, |
| "ce_loss": 0.44693663619349266, |
| "lb_loss": 0.9999999983867445 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 933000000, |
| "cumulative_training_bytes": 933006103, |
| "metrics": { |
| "loss": 0.4570423392423911, |
| "ce_loss": 0.4470423487791343, |
| "lb_loss": 0.9999999979161528 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 934000000, |
| "cumulative_training_bytes": 934001056, |
| "metrics": { |
| "loss": 0.4571594332524909, |
| "ce_loss": 0.4471594427892341, |
| "lb_loss": 0.9999999978131129 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 935000000, |
| "cumulative_training_bytes": 935003223, |
| "metrics": { |
| "loss": 0.45730806371681443, |
| "ce_loss": 0.4473080732535576, |
| "lb_loss": 0.9999999977175971 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 936000000, |
| "cumulative_training_bytes": 936001763, |
| "metrics": { |
| "loss": 0.45738783855462106, |
| "ce_loss": 0.4473878480913642, |
| "lb_loss": 0.9999999975233114 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 937000000, |
| "cumulative_training_bytes": 937006395, |
| "metrics": { |
| "loss": 0.45747957874488554, |
| "ce_loss": 0.4474795882816287, |
| "lb_loss": 0.9999999975441615 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 938000000, |
| "cumulative_training_bytes": 938001495, |
| "metrics": { |
| "loss": 0.45755059916273993, |
| "ce_loss": 0.4475506086994831, |
| "lb_loss": 0.9999999975771608 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 939000000, |
| "cumulative_training_bytes": 939003006, |
| "metrics": { |
| "loss": 0.45756441776033846, |
| "ce_loss": 0.4475644272970816, |
| "lb_loss": 0.9999999975810244 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 940000000, |
| "cumulative_training_bytes": 940003165, |
| "metrics": { |
| "loss": 0.45759776781046674, |
| "ce_loss": 0.4475977773472099, |
| "lb_loss": 0.9999999973602903 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 941000000, |
| "cumulative_training_bytes": 941001772, |
| "metrics": { |
| "loss": 0.45767297192580525, |
| "ce_loss": 0.4476729814625484, |
| "lb_loss": 0.9999999972288843 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 942000000, |
| "cumulative_training_bytes": 942003089, |
| "metrics": { |
| "loss": 0.45774275527380737, |
| "ce_loss": 0.44774276481055053, |
| "lb_loss": 0.9999999972424898 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 943000000, |
| "cumulative_training_bytes": 943006108, |
| "metrics": { |
| "loss": 0.45780045653905144, |
| "ce_loss": 0.4478004660757946, |
| "lb_loss": 0.9999999973889571 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 944000000, |
| "cumulative_training_bytes": 944003340, |
| "metrics": { |
| "loss": 0.4579041221871552, |
| "ce_loss": 0.44790413172389837, |
| "lb_loss": 0.999999997290698 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 945000000, |
| "cumulative_training_bytes": 945007101, |
| "metrics": { |
| "loss": 0.45798618218312526, |
| "ce_loss": 0.4479861917198684, |
| "lb_loss": 0.9999999974977349 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 946000000, |
| "cumulative_training_bytes": 946006954, |
| "metrics": { |
| "loss": 0.45804723397740776, |
| "ce_loss": 0.44804724351415093, |
| "lb_loss": 0.9999999973683987 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 947000000, |
| "cumulative_training_bytes": 947006874, |
| "metrics": { |
| "loss": 0.45816000358238773, |
| "ce_loss": 0.4481600131191309, |
| "lb_loss": 0.9999999972123771 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 948000000, |
| "cumulative_training_bytes": 948002304, |
| "metrics": { |
| "loss": 0.4582202268191281, |
| "ce_loss": 0.44822023635587127, |
| "lb_loss": 0.9999999972346704 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 949000000, |
| "cumulative_training_bytes": 949000281, |
| "metrics": { |
| "loss": 0.45827323714661655, |
| "ce_loss": 0.4482732466833597, |
| "lb_loss": 0.9999999969312667 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 950000000, |
| "cumulative_training_bytes": 950005460, |
| "metrics": { |
| "loss": 0.4582935053636347, |
| "ce_loss": 0.4482935149003779, |
| "lb_loss": 0.9999999967238945 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 951000000, |
| "cumulative_training_bytes": 951004754, |
| "metrics": { |
| "loss": 0.45839260763107137, |
| "ce_loss": 0.44839261716781453, |
| "lb_loss": 0.9999999967359123 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 952000000, |
| "cumulative_training_bytes": 952003496, |
| "metrics": { |
| "loss": 0.4585031724775566, |
| "ce_loss": 0.4485031820142998, |
| "lb_loss": 0.9999999967180216 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 953000000, |
| "cumulative_training_bytes": 953005419, |
| "metrics": { |
| "loss": 0.4585650799644636, |
| "ce_loss": 0.44856508950120677, |
| "lb_loss": 0.9999999965953904 |
| } |
| }, |
| { |
| "epoch": 20, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.45858579490975426, |
| "ce_loss": 0.4485858044464974, |
| "lb_loss": 0.9999999965715838, |
| "training_bytes": 47653413 |
| }, |
| "cumulative_training_bytes": 953068042, |
| "training_bytes_this_epoch": 47653413 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 954000000, |
| "cumulative_training_bytes": 954000428, |
| "metrics": { |
| "loss": 0.4476760918976831, |
| "ce_loss": 0.43767610143442626, |
| "lb_loss": 1.0000000063513146 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 955000000, |
| "cumulative_training_bytes": 955002200, |
| "metrics": { |
| "loss": 0.44804038077946234, |
| "ce_loss": 0.4380403903162055, |
| "lb_loss": 0.999999999528817 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 956000000, |
| "cumulative_training_bytes": 956005963, |
| "metrics": { |
| "loss": 0.4483036200205485, |
| "ce_loss": 0.4383036295572917, |
| "lb_loss": 0.9999999996895591 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 957000000, |
| "cumulative_training_bytes": 957000742, |
| "metrics": { |
| "loss": 0.448704544208393, |
| "ce_loss": 0.43870455374513617, |
| "lb_loss": 0.9999999988403766 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 958000000, |
| "cumulative_training_bytes": 958002438, |
| "metrics": { |
| "loss": 0.44925628891286923, |
| "ce_loss": 0.4392562984496124, |
| "lb_loss": 0.9999999988910764 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 959000000, |
| "cumulative_training_bytes": 959002429, |
| "metrics": { |
| "loss": 0.44934474852777295, |
| "ce_loss": 0.4393447580645161, |
| "lb_loss": 0.9999999976158143 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 960000000, |
| "cumulative_training_bytes": 960003553, |
| "metrics": { |
| "loss": 0.4496341999003429, |
| "ce_loss": 0.4396342094370861, |
| "lb_loss": 0.9999999972368707 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 961000000, |
| "cumulative_training_bytes": 961003867, |
| "metrics": { |
| "loss": 0.44992774178919703, |
| "ce_loss": 0.4399277513259402, |
| "lb_loss": 0.9999999952868073 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 962000000, |
| "cumulative_training_bytes": 962003623, |
| "metrics": { |
| "loss": 0.45012032169185273, |
| "ce_loss": 0.4401203312285959, |
| "lb_loss": 0.9999999951520194 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 963000000, |
| "cumulative_training_bytes": 963004917, |
| "metrics": { |
| "loss": 0.4505898071491691, |
| "ce_loss": 0.44058981668591224, |
| "lb_loss": 0.9999999956409228 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 964000000, |
| "cumulative_training_bytes": 964007286, |
| "metrics": { |
| "loss": 0.4508954231555645, |
| "ce_loss": 0.4408954326923077, |
| "lb_loss": 0.999999996415385 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 965000000, |
| "cumulative_training_bytes": 965001346, |
| "metrics": { |
| "loss": 0.4511721158638979, |
| "ce_loss": 0.44117212540064105, |
| "lb_loss": 0.9999999975164732 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 966000000, |
| "cumulative_training_bytes": 966000376, |
| "metrics": { |
| "loss": 0.451245598821245, |
| "ce_loss": 0.44124560835798815, |
| "lb_loss": 0.9999999979543968 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 967000000, |
| "cumulative_training_bytes": 967005281, |
| "metrics": { |
| "loss": 0.45132150851652425, |
| "ce_loss": 0.4413215180532674, |
| "lb_loss": 0.9999999973814544 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 968000000, |
| "cumulative_training_bytes": 968007618, |
| "metrics": { |
| "loss": 0.4515903465083388, |
| "ce_loss": 0.44159035604508196, |
| "lb_loss": 0.9999999970075537 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 969000000, |
| "cumulative_training_bytes": 969006307, |
| "metrics": { |
| "loss": 0.4518647513249947, |
| "ce_loss": 0.44186476086173787, |
| "lb_loss": 0.9999999977966598 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 970000000, |
| "cumulative_training_bytes": 970004567, |
| "metrics": { |
| "loss": 0.4520443377858958, |
| "ce_loss": 0.442044347322639, |
| "lb_loss": 0.9999999973604812 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 971000000, |
| "cumulative_training_bytes": 971001385, |
| "metrics": { |
| "loss": 0.4521723189417033, |
| "ce_loss": 0.44217232847844645, |
| "lb_loss": 0.9999999969981442 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 972000000, |
| "cumulative_training_bytes": 972005775, |
| "metrics": { |
| "loss": 0.45231254098872165, |
| "ce_loss": 0.4423125505254648, |
| "lb_loss": 0.9999999970848172 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 973000000, |
| "cumulative_training_bytes": 973005099, |
| "metrics": { |
| "loss": 0.4524783973730457, |
| "ce_loss": 0.44247840690978885, |
| "lb_loss": 0.9999999974831052 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 974000000, |
| "cumulative_training_bytes": 974000735, |
| "metrics": { |
| "loss": 0.4526559548142623, |
| "ce_loss": 0.44265596435100546, |
| "lb_loss": 0.9999999972758389 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 975000000, |
| "cumulative_training_bytes": 975000397, |
| "metrics": { |
| "loss": 0.45285711384950944, |
| "ce_loss": 0.4428571233862526, |
| "lb_loss": 0.9999999974211529 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 976000000, |
| "cumulative_training_bytes": 976004243, |
| "metrics": { |
| "loss": 0.45301266459572265, |
| "ce_loss": 0.4430126741324658, |
| "lb_loss": 0.9999999969372322 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 977000000, |
| "cumulative_training_bytes": 977001859, |
| "metrics": { |
| "loss": 0.4531138975627132, |
| "ce_loss": 0.44311390709945636, |
| "lb_loss": 0.9999999971026844 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 978000000, |
| "cumulative_training_bytes": 978005532, |
| "metrics": { |
| "loss": 0.4532364708576706, |
| "ce_loss": 0.44323648039441377, |
| "lb_loss": 0.9999999968166948 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 979000000, |
| "cumulative_training_bytes": 979005081, |
| "metrics": { |
| "loss": 0.45334783579973853, |
| "ce_loss": 0.4433478453364817, |
| "lb_loss": 0.9999999970971764 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 980000000, |
| "cumulative_training_bytes": 980007499, |
| "metrics": { |
| "loss": 0.4534764752600741, |
| "ce_loss": 0.4434764847968173, |
| "lb_loss": 0.9999999970527967 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 981000000, |
| "cumulative_training_bytes": 981004891, |
| "metrics": { |
| "loss": 0.4536125384216016, |
| "ce_loss": 0.44361254795834476, |
| "lb_loss": 0.999999996733097 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 982000000, |
| "cumulative_training_bytes": 982006800, |
| "metrics": { |
| "loss": 0.45371898847912984, |
| "ce_loss": 0.443718998015873, |
| "lb_loss": 0.9999999963101887 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 983000000, |
| "cumulative_training_bytes": 983004665, |
| "metrics": { |
| "loss": 0.4538384055847402, |
| "ce_loss": 0.44383841512148337, |
| "lb_loss": 0.9999999963109145 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 984000000, |
| "cumulative_training_bytes": 984006393, |
| "metrics": { |
| "loss": 0.4539954239883744, |
| "ce_loss": 0.44399543352511756, |
| "lb_loss": 0.9999999966517559 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 985000000, |
| "cumulative_training_bytes": 985007069, |
| "metrics": { |
| "loss": 0.45418143821014084, |
| "ce_loss": 0.444181447746884, |
| "lb_loss": 0.9999999966140218 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 986000000, |
| "cumulative_training_bytes": 986006787, |
| "metrics": { |
| "loss": 0.45425647727171803, |
| "ce_loss": 0.4442564868084612, |
| "lb_loss": 0.9999999965916452 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 987000000, |
| "cumulative_training_bytes": 987002711, |
| "metrics": { |
| "loss": 0.45432535147408715, |
| "ce_loss": 0.4443253610108303, |
| "lb_loss": 0.9999999966378247 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 988000000, |
| "cumulative_training_bytes": 988007659, |
| "metrics": { |
| "loss": 0.4544178967201054, |
| "ce_loss": 0.44441790625684857, |
| "lb_loss": 0.9999999963816597 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 989000000, |
| "cumulative_training_bytes": 989002984, |
| "metrics": { |
| "loss": 0.45448096611848804, |
| "ce_loss": 0.4444809756552312, |
| "lb_loss": 0.9999999967232052 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 990000000, |
| "cumulative_training_bytes": 990002219, |
| "metrics": { |
| "loss": 0.4545410405099083, |
| "ce_loss": 0.44454105004665145, |
| "lb_loss": 0.999999996885679 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 991000000, |
| "cumulative_training_bytes": 991003464, |
| "metrics": { |
| "loss": 0.45466355147456083, |
| "ce_loss": 0.444663561011304, |
| "lb_loss": 0.9999999970522532 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 992000000, |
| "cumulative_training_bytes": 992005888, |
| "metrics": { |
| "loss": 0.45475785673660984, |
| "ce_loss": 0.444757866273353, |
| "lb_loss": 0.9999999972219664 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 993000000, |
| "cumulative_training_bytes": 993002370, |
| "metrics": { |
| "loss": 0.4548222429800354, |
| "ce_loss": 0.44482225251677854, |
| "lb_loss": 0.9999999970283399 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 994000000, |
| "cumulative_training_bytes": 994003864, |
| "metrics": { |
| "loss": 0.454938013634787, |
| "ce_loss": 0.44493802317153014, |
| "lb_loss": 0.9999999967555272 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 995000000, |
| "cumulative_training_bytes": 995006555, |
| "metrics": { |
| "loss": 0.45501115932851155, |
| "ce_loss": 0.4450111688652547, |
| "lb_loss": 0.9999999965828267 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 996000000, |
| "cumulative_training_bytes": 996000608, |
| "metrics": { |
| "loss": 0.4550933865529661, |
| "ce_loss": 0.44509339608970927, |
| "lb_loss": 0.9999999965132292 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 997000000, |
| "cumulative_training_bytes": 997007326, |
| "metrics": { |
| "loss": 0.4551511483144245, |
| "ce_loss": 0.44515115785116766, |
| "lb_loss": 0.9999999964058545 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 998000000, |
| "cumulative_training_bytes": 998006581, |
| "metrics": { |
| "loss": 0.4552532544349726, |
| "ce_loss": 0.4452532639717158, |
| "lb_loss": 0.9999999961610913 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 999000000, |
| "cumulative_training_bytes": 999000135, |
| "metrics": { |
| "loss": 0.4552769968612228, |
| "ce_loss": 0.44527700639796597, |
| "lb_loss": 0.9999999962337178 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1000000000, |
| "cumulative_training_bytes": 1000002883, |
| "metrics": { |
| "loss": 0.4553679389458804, |
| "ce_loss": 0.44536794848262357, |
| "lb_loss": 0.9999999961586172 |
| } |
| }, |
| { |
| "epoch": 21, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.4554653137079548, |
| "ce_loss": 0.44546532324469795, |
| "lb_loss": 0.9999999964087819, |
| "training_bytes": 47653407 |
| }, |
| "cumulative_training_bytes": 1000721449, |
| "training_bytes_this_epoch": 47653407 |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1001000000, |
| "cumulative_training_bytes": 1001004765, |
| "metrics": { |
| "loss": 0.4417461897875812, |
| "ce_loss": 0.43174619932432434, |
| "lb_loss": 0.9999999967781273 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1002000000, |
| "cumulative_training_bytes": 1002000171, |
| "metrics": { |
| "loss": 0.4442837666608616, |
| "ce_loss": 0.4342837761976048, |
| "lb_loss": 1.0000000014276562 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1003000000, |
| "cumulative_training_bytes": 1003000999, |
| "metrics": { |
| "loss": 0.4452322681478206, |
| "ce_loss": 0.43523227768456374, |
| "lb_loss": 0.999999996599735 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1004000000, |
| "cumulative_training_bytes": 1004003831, |
| "metrics": { |
| "loss": 0.4486882553233967, |
| "ce_loss": 0.43868826486013984, |
| "lb_loss": 0.9999999962486588 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1005000000, |
| "cumulative_training_bytes": 1005001152, |
| "metrics": { |
| "loss": 0.44896745468508154, |
| "ce_loss": 0.4389674642218247, |
| "lb_loss": 0.999999996054791 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1006000000, |
| "cumulative_training_bytes": 1006005315, |
| "metrics": { |
| "loss": 0.44883037995601044, |
| "ce_loss": 0.4388303894927536, |
| "lb_loss": 0.9999999951625216 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1007000000, |
| "cumulative_training_bytes": 1007001983, |
| "metrics": { |
| "loss": 0.44876475799374466, |
| "ce_loss": 0.4387647675304878, |
| "lb_loss": 0.9999999939668469 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1008000000, |
| "cumulative_training_bytes": 1008002750, |
| "metrics": { |
| "loss": 0.448804127555791, |
| "ce_loss": 0.4388041370925342, |
| "lb_loss": 0.9999999941084788 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1009000000, |
| "cumulative_training_bytes": 1009004840, |
| "metrics": { |
| "loss": 0.4488484048578964, |
| "ce_loss": 0.43884841439463956, |
| "lb_loss": 0.9999999951523024 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1010000000, |
| "cumulative_training_bytes": 1010007632, |
| "metrics": { |
| "loss": 0.44896201643192957, |
| "ce_loss": 0.43896202596867273, |
| "lb_loss": 0.999999996216358 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1011000000, |
| "cumulative_training_bytes": 1011001073, |
| "metrics": { |
| "loss": 0.44882147891333135, |
| "ce_loss": 0.4388214884500745, |
| "lb_loss": 0.999999996446817 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1012000000, |
| "cumulative_training_bytes": 1012002820, |
| "metrics": { |
| "loss": 0.44899433661057525, |
| "ce_loss": 0.4389943461473184, |
| "lb_loss": 0.9999999968032811 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1013000000, |
| "cumulative_training_bytes": 1013000900, |
| "metrics": { |
| "loss": 0.4489474717483473, |
| "ce_loss": 0.4389474812850905, |
| "lb_loss": 0.9999999971368948 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1014000000, |
| "cumulative_training_bytes": 1014005352, |
| "metrics": { |
| "loss": 0.449001441227386, |
| "ce_loss": 0.43900145076412916, |
| "lb_loss": 0.9999999966657148 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1015000000, |
| "cumulative_training_bytes": 1015002241, |
| "metrics": { |
| "loss": 0.4491251497309607, |
| "ce_loss": 0.43912515926770385, |
| "lb_loss": 0.9999999973139537 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1016000000, |
| "cumulative_training_bytes": 1016006378, |
| "metrics": { |
| "loss": 0.4492504603880689, |
| "ce_loss": 0.43925046992481204, |
| "lb_loss": 0.9999999975799618 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1017000000, |
| "cumulative_training_bytes": 1017004553, |
| "metrics": { |
| "loss": 0.44930973311031563, |
| "ce_loss": 0.4393097426470588, |
| "lb_loss": 0.9999999975597157 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1018000000, |
| "cumulative_training_bytes": 1018005638, |
| "metrics": { |
| "loss": 0.4494938055674235, |
| "ce_loss": 0.4394938151041667, |
| "lb_loss": 0.9999999973579501 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1019000000, |
| "cumulative_training_bytes": 1019003819, |
| "metrics": { |
| "loss": 0.44971085923158527, |
| "ce_loss": 0.43971086876832843, |
| "lb_loss": 0.9999999975029474 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1020000000, |
| "cumulative_training_bytes": 1020001116, |
| "metrics": { |
| "loss": 0.44981704697895014, |
| "ce_loss": 0.4398170565156933, |
| "lb_loss": 0.9999999973714281 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1021000000, |
| "cumulative_training_bytes": 1021000064, |
| "metrics": { |
| "loss": 0.4500520344587251, |
| "ce_loss": 0.44005204399546827, |
| "lb_loss": 0.9999999973439018 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1022000000, |
| "cumulative_training_bytes": 1022006171, |
| "metrics": { |
| "loss": 0.4501109510156138, |
| "ce_loss": 0.440110960552357, |
| "lb_loss": 0.999999997533453 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1023000000, |
| "cumulative_training_bytes": 1023007389, |
| "metrics": { |
| "loss": 0.4502652396041503, |
| "ce_loss": 0.4402652491408935, |
| "lb_loss": 0.9999999975420765 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1024000000, |
| "cumulative_training_bytes": 1024005719, |
| "metrics": { |
| "loss": 0.45038213478891476, |
| "ce_loss": 0.4403821443256579, |
| "lb_loss": 0.9999999971178017 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1025000000, |
| "cumulative_training_bytes": 1025000620, |
| "metrics": { |
| "loss": 0.45051531972193193, |
| "ce_loss": 0.4405153292586751, |
| "lb_loss": 0.9999999971607882 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1026000000, |
| "cumulative_training_bytes": 1026003600, |
| "metrics": { |
| "loss": 0.45060333509367184, |
| "ce_loss": 0.440603344630415, |
| "lb_loss": 0.9999999973637449 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1027000000, |
| "cumulative_training_bytes": 1027002396, |
| "metrics": { |
| "loss": 0.45068954434696423, |
| "ce_loss": 0.4406895538837074, |
| "lb_loss": 0.9999999975852388 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1028000000, |
| "cumulative_training_bytes": 1028006189, |
| "metrics": { |
| "loss": 0.45083050176308836, |
| "ce_loss": 0.44083051129983153, |
| "lb_loss": 0.9999999974063111 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1029000000, |
| "cumulative_training_bytes": 1029000329, |
| "metrics": { |
| "loss": 0.45092642346027745, |
| "ce_loss": 0.4409264329970206, |
| "lb_loss": 0.999999997368484 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1030000000, |
| "cumulative_training_bytes": 1030002716, |
| "metrics": { |
| "loss": 0.45103073774941954, |
| "ce_loss": 0.4410307472861627, |
| "lb_loss": 0.9999999974430652 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1031000000, |
| "cumulative_training_bytes": 1031006184, |
| "metrics": { |
| "loss": 0.4510980082933024, |
| "ce_loss": 0.44109801783004554, |
| "lb_loss": 0.9999999974674809 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1032000000, |
| "cumulative_training_bytes": 1032004553, |
| "metrics": { |
| "loss": 0.45115468714543117, |
| "ce_loss": 0.44115469668217433, |
| "lb_loss": 0.9999999971978227 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1033000000, |
| "cumulative_training_bytes": 1033007149, |
| "metrics": { |
| "loss": 0.4512519366376341, |
| "ce_loss": 0.44125194617437724, |
| "lb_loss": 0.9999999973839006 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1034000000, |
| "cumulative_training_bytes": 1034001070, |
| "metrics": { |
| "loss": 0.4513257780064099, |
| "ce_loss": 0.44132578754315305, |
| "lb_loss": 0.9999999975444807 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1035000000, |
| "cumulative_training_bytes": 1035006816, |
| "metrics": { |
| "loss": 0.4513804955179932, |
| "ce_loss": 0.44138050505473636, |
| "lb_loss": 0.9999999975364479 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1036000000, |
| "cumulative_training_bytes": 1036006156, |
| "metrics": { |
| "loss": 0.4515215541163934, |
| "ce_loss": 0.44152156365313655, |
| "lb_loss": 0.9999999976582503 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1037000000, |
| "cumulative_training_bytes": 1037000394, |
| "metrics": { |
| "loss": 0.45160786134672737, |
| "ce_loss": 0.44160787088347053, |
| "lb_loss": 0.9999999976470195 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1038000000, |
| "cumulative_training_bytes": 1038006020, |
| "metrics": { |
| "loss": 0.45169471558137514, |
| "ce_loss": 0.4416947251181183, |
| "lb_loss": 0.9999999977225834 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1039000000, |
| "cumulative_training_bytes": 1039004839, |
| "metrics": { |
| "loss": 0.45180788329182253, |
| "ce_loss": 0.4418078928285657, |
| "lb_loss": 0.9999999978418802 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1040000000, |
| "cumulative_training_bytes": 1040005527, |
| "metrics": { |
| "loss": 0.4519148905607227, |
| "ce_loss": 0.44191490009746587, |
| "lb_loss": 0.999999997618138 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1041000000, |
| "cumulative_training_bytes": 1041000015, |
| "metrics": { |
| "loss": 0.4519806671505192, |
| "ce_loss": 0.4419806766872624, |
| "lb_loss": 0.9999999976656736 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1042000000, |
| "cumulative_training_bytes": 1042000408, |
| "metrics": { |
| "loss": 0.4521246687465067, |
| "ce_loss": 0.44212467828324986, |
| "lb_loss": 0.9999999977776788 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1043000000, |
| "cumulative_training_bytes": 1043006070, |
| "metrics": { |
| "loss": 0.4522623903138815, |
| "ce_loss": 0.44226239985062465, |
| "lb_loss": 0.9999999981221785 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1044000000, |
| "cumulative_training_bytes": 1044005567, |
| "metrics": { |
| "loss": 0.4523166392559333, |
| "ce_loss": 0.44231664879267646, |
| "lb_loss": 0.9999999979228524 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1045000000, |
| "cumulative_training_bytes": 1045001996, |
| "metrics": { |
| "loss": 0.4523748661765544, |
| "ce_loss": 0.4423748757132976, |
| "lb_loss": 0.9999999980416941 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1046000000, |
| "cumulative_training_bytes": 1046006987, |
| "metrics": { |
| "loss": 0.45245876149386216, |
| "ce_loss": 0.4424587710306053, |
| "lb_loss": 0.9999999979238152 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1047000000, |
| "cumulative_training_bytes": 1047002167, |
| "metrics": { |
| "loss": 0.452550329069543, |
| "ce_loss": 0.4425503386062862, |
| "lb_loss": 0.9999999979293672 |
| } |
| }, |
| { |
| "checkpoint_type": "bytes", |
| "bytes_threshold": 1048000000, |
| "cumulative_training_bytes": 1048003428, |
| "metrics": { |
| "loss": 0.4525869717385605, |
| "ce_loss": 0.44258698127530366, |
| "lb_loss": 0.9999999980694851 |
| } |
| }, |
| { |
| "epoch": 22, |
| "checkpoint_type": "epoch", |
| "metrics": { |
| "loss": 0.4525974458777813, |
| "ce_loss": 0.44259745541452444, |
| "lb_loss": 0.9999999981517197, |
| "training_bytes": 47653407 |
| }, |
| "cumulative_training_bytes": 1048374856, |
| "training_bytes_this_epoch": 47653407 |
| } |
| ] |
| } |
