{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 200.0,
  "global_step": 35,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "debug/num_lat_loss": 1783.0,
      "debug/num_lat_total": 2241.0,
      "debug/num_tok_loss": 1783.0,
      "debug/num_tok_total": 2241.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.214554786682129,
      "train/diffusion_loss": 0.4518924653530121
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1748.0,
      "debug/num_lat_total": 2156.0,
      "debug/num_tok_loss": 1748.0,
      "debug/num_tok_total": 2156.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.158607482910156,
      "train/diffusion_loss": 0.4541515111923218
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1690.0,
      "debug/num_lat_total": 2009.0,
      "debug/num_tok_loss": 1690.0,
      "debug/num_tok_total": 2009.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.157901763916016,
      "train/diffusion_loss": 0.4803432524204254
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1741.0,
      "debug/num_lat_total": 2238.0,
      "debug/num_tok_loss": 1741.0,
      "debug/num_tok_total": 2238.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.17225170135498,
      "train/diffusion_loss": 0.4446203410625458
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1734.0,
      "debug/num_lat_total": 2185.0,
      "debug/num_tok_loss": 1734.0,
      "debug/num_tok_total": 2185.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.16765022277832,
      "train/diffusion_loss": 0.45456695556640625
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1681.0,
      "debug/num_lat_total": 2020.0,
      "debug/num_tok_loss": 1681.0,
      "debug/num_tok_total": 2020.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.191916465759277,
      "train/diffusion_loss": 0.45108145475387573
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1787.0,
      "debug/num_lat_total": 2334.0,
      "debug/num_tok_loss": 1787.0,
      "debug/num_tok_total": 2334.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.1714506149292,
      "train/diffusion_loss": 0.4591076970100403
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1743.0,
      "debug/num_lat_total": 2334.0,
      "debug/num_tok_loss": 1743.0,
      "debug/num_tok_total": 2334.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.179821014404297,
      "train/diffusion_loss": 0.40789151191711426
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1796.0,
      "debug/num_lat_total": 2434.0,
      "debug/num_tok_loss": 1796.0,
      "debug/num_tok_total": 2434.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.161667823791504,
      "train/diffusion_loss": 0.4303126931190491
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1696.0,
      "debug/num_lat_total": 2107.0,
      "debug/num_tok_loss": 1696.0,
      "debug/num_tok_total": 2107.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.184703826904297,
      "train/diffusion_loss": 0.4528208374977112
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1774.0,
      "debug/num_lat_total": 2139.0,
      "debug/num_tok_loss": 1774.0,
      "debug/num_tok_total": 2139.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.163565635681152,
      "train/diffusion_loss": 0.46860605478286743
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1779.0,
      "debug/num_lat_total": 2193.0,
      "debug/num_tok_loss": 1779.0,
      "debug/num_tok_total": 2193.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.173012733459473,
      "train/diffusion_loss": 0.4421927332878113
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1747.0,
      "debug/num_lat_total": 2251.0,
      "debug/num_tok_loss": 1747.0,
      "debug/num_tok_total": 2251.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.135828018188477,
      "train/diffusion_loss": 0.45780879259109497
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1741.0,
      "debug/num_lat_total": 2298.0,
      "debug/num_tok_loss": 1741.0,
      "debug/num_tok_total": 2298.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.20445728302002,
      "train/diffusion_loss": 0.4124974012374878
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1808.0,
      "debug/num_lat_total": 2252.0,
      "debug/num_tok_loss": 1808.0,
      "debug/num_tok_total": 2252.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.17200756072998,
      "train/diffusion_loss": 0.45158717036247253
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1880.0,
      "debug/num_lat_total": 2461.0,
      "debug/num_tok_loss": 1880.0,
      "debug/num_tok_total": 2461.0,
      "epoch": 0,
      "step": 0
    },
    {
      "epoch": 0,
      "step": 0,
      "train/ce_loss": 12.16403865814209,
      "train/diffusion_loss": 0.43170061707496643
    },
    {
      "epoch": 0,
      "step": 0,
      "train/learning_rate_real": 0.0
    },
    {
      "debug/num_lat_loss": 1743.0,
      "debug/num_lat_total": 2182.0,
      "debug/num_tok_loss": 1743.0,
      "debug/num_tok_total": 2182.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.182713508605957,
      "train/diffusion_loss": 0.45220285654067993
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1716.0,
      "debug/num_lat_total": 2258.0,
      "debug/num_tok_loss": 1716.0,
      "debug/num_tok_total": 2258.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.187424659729004,
      "train/diffusion_loss": 0.4251019060611725
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1710.0,
      "debug/num_lat_total": 2096.0,
      "debug/num_tok_loss": 1710.0,
      "debug/num_tok_total": 2096.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.186864852905273,
      "train/diffusion_loss": 0.47552168369293213
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1742.0,
      "debug/num_lat_total": 2151.0,
      "debug/num_tok_loss": 1742.0,
      "debug/num_tok_total": 2151.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.177823066711426,
      "train/diffusion_loss": 0.47840219736099243
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1786.0,
      "debug/num_lat_total": 2338.0,
      "debug/num_tok_loss": 1786.0,
      "debug/num_tok_total": 2338.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.171921730041504,
      "train/diffusion_loss": 0.43838560581207275
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1778.0,
      "debug/num_lat_total": 2235.0,
      "debug/num_tok_loss": 1778.0,
      "debug/num_tok_total": 2235.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.19234561920166,
      "train/diffusion_loss": 0.4630849361419678
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1743.0,
      "debug/num_lat_total": 2157.0,
      "debug/num_tok_loss": 1743.0,
      "debug/num_tok_total": 2157.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.180907249450684,
      "train/diffusion_loss": 0.4821509122848511
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1655.0,
      "debug/num_lat_total": 2200.0,
      "debug/num_tok_loss": 1655.0,
      "debug/num_tok_total": 2200.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.166112899780273,
      "train/diffusion_loss": 0.4305018186569214
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1735.0,
      "debug/num_lat_total": 2071.0,
      "debug/num_tok_loss": 1735.0,
      "debug/num_tok_total": 2071.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.178923606872559,
      "train/diffusion_loss": 0.48272061347961426
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1682.0,
      "debug/num_lat_total": 2056.0,
      "debug/num_tok_loss": 1682.0,
      "debug/num_tok_total": 2056.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.184454917907715,
      "train/diffusion_loss": 0.48283350467681885
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1805.0,
      "debug/num_lat_total": 2212.0,
      "debug/num_tok_loss": 1805.0,
      "debug/num_tok_total": 2212.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.114720344543457,
      "train/diffusion_loss": 0.48657673597335815
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1695.0,
      "debug/num_lat_total": 2086.0,
      "debug/num_tok_loss": 1695.0,
      "debug/num_tok_total": 2086.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.194059371948242,
      "train/diffusion_loss": 0.4509486258029938
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1706.0,
      "debug/num_lat_total": 2081.0,
      "debug/num_tok_loss": 1706.0,
      "debug/num_tok_total": 2081.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.177267074584961,
      "train/diffusion_loss": 0.46584227681159973
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1719.0,
      "debug/num_lat_total": 2106.0,
      "debug/num_tok_loss": 1719.0,
      "debug/num_tok_total": 2106.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.15927791595459,
      "train/diffusion_loss": 0.4715038239955902
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1726.0,
      "debug/num_lat_total": 2112.0,
      "debug/num_tok_loss": 1726.0,
      "debug/num_tok_total": 2112.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.164405822753906,
      "train/diffusion_loss": 0.48125794529914856
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1695.0,
      "debug/num_lat_total": 2287.0,
      "debug/num_tok_loss": 1695.0,
      "debug/num_tok_total": 2287.0,
      "epoch": 0.14953271028037382,
      "step": 1
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/ce_loss": 12.164380073547363,
      "train/diffusion_loss": 0.43863794207572937
    },
    {
      "epoch": 0.14953271028037382,
      "step": 1,
      "train/learning_rate_real": 1.25e-05
    },
    {
      "debug/num_lat_loss": 1780.0,
      "debug/num_lat_total": 2213.0,
      "debug/num_tok_loss": 1780.0,
      "debug/num_tok_total": 2213.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.215859413146973,
      "train/diffusion_loss": 0.458553284406662
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1716.0,
      "debug/num_lat_total": 2125.0,
      "debug/num_tok_loss": 1716.0,
      "debug/num_tok_total": 2125.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.206511497497559,
      "train/diffusion_loss": 0.4523429274559021
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1770.0,
      "debug/num_lat_total": 2352.0,
      "debug/num_tok_loss": 1770.0,
      "debug/num_tok_total": 2352.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.186681747436523,
      "train/diffusion_loss": 0.4239839017391205
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1761.0,
      "debug/num_lat_total": 2317.0,
      "debug/num_tok_loss": 1761.0,
      "debug/num_tok_total": 2317.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.174463272094727,
      "train/diffusion_loss": 0.41487666964530945
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1730.0,
      "debug/num_lat_total": 2167.0,
      "debug/num_tok_loss": 1730.0,
      "debug/num_tok_total": 2167.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.153246879577637,
      "train/diffusion_loss": 0.46722397208213806
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1732.0,
      "debug/num_lat_total": 2227.0,
      "debug/num_tok_loss": 1732.0,
      "debug/num_tok_total": 2227.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.171008110046387,
      "train/diffusion_loss": 0.44303643703460693
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1838.0,
      "debug/num_lat_total": 2150.0,
      "debug/num_tok_loss": 1838.0,
      "debug/num_tok_total": 2150.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.203962326049805,
      "train/diffusion_loss": 0.4903823733329773
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1723.0,
      "debug/num_lat_total": 2276.0,
      "debug/num_tok_loss": 1723.0,
      "debug/num_tok_total": 2276.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.200774192810059,
      "train/diffusion_loss": 0.4448697865009308
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1743.0,
      "debug/num_lat_total": 2379.0,
      "debug/num_tok_loss": 1743.0,
      "debug/num_tok_total": 2379.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.157685279846191,
      "train/diffusion_loss": 0.40528759360313416
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1707.0,
      "debug/num_lat_total": 2101.0,
      "debug/num_tok_loss": 1707.0,
      "debug/num_tok_total": 2101.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.178916931152344,
      "train/diffusion_loss": 0.4575231075286865
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1670.0,
      "debug/num_lat_total": 2013.0,
      "debug/num_tok_loss": 1670.0,
      "debug/num_tok_total": 2013.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.195891380310059,
      "train/diffusion_loss": 0.44977766275405884
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1706.0,
      "debug/num_lat_total": 2293.0,
      "debug/num_tok_loss": 1706.0,
      "debug/num_tok_total": 2293.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.192096710205078,
      "train/diffusion_loss": 0.41044771671295166
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1767.0,
      "debug/num_lat_total": 2400.0,
      "debug/num_tok_loss": 1767.0,
      "debug/num_tok_total": 2400.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.169843673706055,
      "train/diffusion_loss": 0.40407317876815796
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1656.0,
      "debug/num_lat_total": 2038.0,
      "debug/num_tok_loss": 1656.0,
      "debug/num_tok_total": 2038.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.17802619934082,
      "train/diffusion_loss": 0.470550000667572
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1715.0,
      "debug/num_lat_total": 2040.0,
      "debug/num_tok_loss": 1715.0,
      "debug/num_tok_total": 2040.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.206125259399414,
      "train/diffusion_loss": 0.4850577116012573
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1800.0,
      "debug/num_lat_total": 2091.0,
      "debug/num_tok_loss": 1800.0,
      "debug/num_tok_total": 2091.0,
      "epoch": 0.29906542056074764,
      "step": 2
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/ce_loss": 12.160561561584473,
      "train/diffusion_loss": 0.490349680185318
    },
    {
      "epoch": 0.29906542056074764,
      "step": 2,
      "train/learning_rate_real": 2.5e-05
    },
    {
      "debug/num_lat_loss": 1708.0,
      "debug/num_lat_total": 2022.0,
      "debug/num_tok_loss": 1708.0,
      "debug/num_tok_total": 2022.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.188081741333008,
      "train/diffusion_loss": 0.47481265664100647
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1770.0,
      "debug/num_lat_total": 2267.0,
      "debug/num_tok_loss": 1770.0,
      "debug/num_tok_total": 2267.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.177225112915039,
      "train/diffusion_loss": 0.44798392057418823
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1796.0,
      "debug/num_lat_total": 2222.0,
      "debug/num_tok_loss": 1796.0,
      "debug/num_tok_total": 2222.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.23061752319336,
      "train/diffusion_loss": 0.46705523133277893
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1770.0,
      "debug/num_lat_total": 2040.0,
      "debug/num_tok_loss": 1770.0,
      "debug/num_tok_total": 2040.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.185416221618652,
      "train/diffusion_loss": 0.4840916395187378
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1766.0,
      "debug/num_lat_total": 2299.0,
      "debug/num_tok_loss": 1766.0,
      "debug/num_tok_total": 2299.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.19878101348877,
      "train/diffusion_loss": 0.4340754747390747
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1676.0,
      "debug/num_lat_total": 2198.0,
      "debug/num_tok_loss": 1676.0,
      "debug/num_tok_total": 2198.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.189212799072266,
      "train/diffusion_loss": 0.4262164533138275
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1746.0,
      "debug/num_lat_total": 2169.0,
      "debug/num_tok_loss": 1746.0,
      "debug/num_tok_total": 2169.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.141818046569824,
      "train/diffusion_loss": 0.45898202061653137
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1704.0,
      "debug/num_lat_total": 2325.0,
      "debug/num_tok_loss": 1704.0,
      "debug/num_tok_total": 2325.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.170629501342773,
      "train/diffusion_loss": 0.4151637852191925
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1750.0,
      "debug/num_lat_total": 2158.0,
      "debug/num_tok_loss": 1750.0,
      "debug/num_tok_total": 2158.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.186365127563477,
      "train/diffusion_loss": 0.4442407190799713
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1728.0,
      "debug/num_lat_total": 2162.0,
      "debug/num_tok_loss": 1728.0,
      "debug/num_tok_total": 2162.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.155019760131836,
      "train/diffusion_loss": 0.49728766083717346
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1750.0,
      "debug/num_lat_total": 2237.0,
      "debug/num_tok_loss": 1750.0,
      "debug/num_tok_total": 2237.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.148735046386719,
      "train/diffusion_loss": 0.4589671492576599
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1637.0,
      "debug/num_lat_total": 2085.0,
      "debug/num_tok_loss": 1637.0,
      "debug/num_tok_total": 2085.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.126914024353027,
      "train/diffusion_loss": 0.4655221104621887
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1767.0,
      "debug/num_lat_total": 2346.0,
      "debug/num_tok_loss": 1767.0,
      "debug/num_tok_total": 2346.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.135104179382324,
      "train/diffusion_loss": 0.4546721279621124
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1778.0,
      "debug/num_lat_total": 2127.0,
      "debug/num_tok_loss": 1778.0,
      "debug/num_tok_total": 2127.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.196648597717285,
      "train/diffusion_loss": 0.5008854269981384
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1815.0,
      "debug/num_lat_total": 2265.0,
      "debug/num_tok_loss": 1815.0,
      "debug/num_tok_total": 2265.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.170491218566895,
      "train/diffusion_loss": 0.4510475695133209
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1678.0,
      "debug/num_lat_total": 2084.0,
      "debug/num_tok_loss": 1678.0,
      "debug/num_tok_total": 2084.0,
      "epoch": 0.4485981308411215,
      "step": 3
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/ce_loss": 12.16767406463623,
      "train/diffusion_loss": 0.46289849281311035
    },
    {
      "epoch": 0.4485981308411215,
      "step": 3,
      "train/learning_rate_real": 2.494339903216356e-05
    },
    {
      "debug/num_lat_loss": 1614.0,
      "debug/num_lat_total": 2067.0,
      "debug/num_tok_loss": 1614.0,
      "debug/num_tok_total": 2067.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.186376571655273,
      "train/diffusion_loss": 0.44896766543388367
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1755.0,
      "debug/num_lat_total": 2264.0,
      "debug/num_tok_loss": 1755.0,
      "debug/num_tok_total": 2264.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.194488525390625,
      "train/diffusion_loss": 0.45699965953826904
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1762.0,
      "debug/num_lat_total": 2208.0,
      "debug/num_tok_loss": 1762.0,
      "debug/num_tok_total": 2208.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.146739959716797,
      "train/diffusion_loss": 0.4560658037662506
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1722.0,
      "debug/num_lat_total": 2150.0,
      "debug/num_tok_loss": 1722.0,
      "debug/num_tok_total": 2150.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.168990135192871,
      "train/diffusion_loss": 0.4327857792377472
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1701.0,
      "debug/num_lat_total": 2325.0,
      "debug/num_tok_loss": 1701.0,
      "debug/num_tok_total": 2325.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.165539741516113,
      "train/diffusion_loss": 0.4052616059780121
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1773.0,
      "debug/num_lat_total": 2101.0,
      "debug/num_tok_loss": 1773.0,
      "debug/num_tok_total": 2101.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.168839454650879,
      "train/diffusion_loss": 0.49070754647254944
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1778.0,
      "debug/num_lat_total": 2235.0,
      "debug/num_tok_loss": 1778.0,
      "debug/num_tok_total": 2235.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.132225036621094,
      "train/diffusion_loss": 0.45063555240631104
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1789.0,
      "debug/num_lat_total": 2330.0,
      "debug/num_tok_loss": 1789.0,
      "debug/num_tok_total": 2330.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.154467582702637,
      "train/diffusion_loss": 0.44124725461006165
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1690.0,
      "debug/num_lat_total": 2242.0,
      "debug/num_tok_loss": 1690.0,
      "debug/num_tok_total": 2242.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.173988342285156,
      "train/diffusion_loss": 0.4344332814216614
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1665.0,
      "debug/num_lat_total": 2143.0,
      "debug/num_tok_loss": 1665.0,
      "debug/num_tok_total": 2143.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.202644348144531,
      "train/diffusion_loss": 0.4590062201023102
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1742.0,
      "debug/num_lat_total": 2099.0,
      "debug/num_tok_loss": 1742.0,
      "debug/num_tok_total": 2099.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.214736938476562,
      "train/diffusion_loss": 0.4900093674659729
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1803.0,
      "debug/num_lat_total": 2117.0,
      "debug/num_tok_loss": 1803.0,
      "debug/num_tok_total": 2117.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.18010139465332,
      "train/diffusion_loss": 0.45859843492507935
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1733.0,
      "debug/num_lat_total": 2136.0,
      "debug/num_tok_loss": 1733.0,
      "debug/num_tok_total": 2136.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.151824951171875,
      "train/diffusion_loss": 0.44397464394569397
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1679.0,
      "debug/num_lat_total": 2231.0,
      "debug/num_tok_loss": 1679.0,
      "debug/num_tok_total": 2231.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.15367317199707,
      "train/diffusion_loss": 0.4424894154071808
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1803.0,
      "debug/num_lat_total": 2330.0,
      "debug/num_tok_loss": 1803.0,
      "debug/num_tok_total": 2330.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.184985160827637,
      "train/diffusion_loss": 0.4271906018257141
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1714.0,
      "debug/num_lat_total": 2169.0,
      "debug/num_tok_loss": 1714.0,
      "debug/num_tok_total": 2169.0,
      "epoch": 0.5981308411214953,
      "step": 4
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/ce_loss": 12.191471099853516,
      "train/diffusion_loss": 0.46100130677223206
    },
    {
      "epoch": 0.5981308411214953,
      "step": 4,
      "train/learning_rate_real": 2.4774108715783832e-05
    },
    {
      "debug/num_lat_loss": 1738.0,
      "debug/num_lat_total": 1939.0,
      "debug/num_tok_loss": 1738.0,
      "debug/num_tok_total": 1939.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.195037841796875,
      "train/diffusion_loss": 0.49820560216903687
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1688.0,
      "debug/num_lat_total": 2105.0,
      "debug/num_tok_loss": 1688.0,
      "debug/num_tok_total": 2105.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.162631034851074,
      "train/diffusion_loss": 0.4296850860118866
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1723.0,
      "debug/num_lat_total": 2177.0,
      "debug/num_tok_loss": 1723.0,
      "debug/num_tok_total": 2177.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.171772956848145,
      "train/diffusion_loss": 0.4458027184009552
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1783.0,
      "debug/num_lat_total": 2125.0,
      "debug/num_tok_loss": 1783.0,
      "debug/num_tok_total": 2125.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.18253231048584,
      "train/diffusion_loss": 0.4729151427745819
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1759.0,
      "debug/num_lat_total": 2354.0,
      "debug/num_tok_loss": 1759.0,
      "debug/num_tok_total": 2354.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.203909873962402,
      "train/diffusion_loss": 0.4511413276195526
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1745.0,
      "debug/num_lat_total": 2019.0,
      "debug/num_tok_loss": 1745.0,
      "debug/num_tok_total": 2019.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.175359725952148,
      "train/diffusion_loss": 0.4547876715660095
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1783.0,
      "debug/num_lat_total": 2269.0,
      "debug/num_tok_loss": 1783.0,
      "debug/num_tok_total": 2269.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.203401565551758,
      "train/diffusion_loss": 0.44810980558395386
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1765.0,
      "debug/num_lat_total": 2174.0,
      "debug/num_tok_loss": 1765.0,
      "debug/num_tok_total": 2174.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.165059089660645,
      "train/diffusion_loss": 0.46098846197128296
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1735.0,
      "debug/num_lat_total": 2087.0,
      "debug/num_tok_loss": 1735.0,
      "debug/num_tok_total": 2087.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.127701759338379,
      "train/diffusion_loss": 0.46769240498542786
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1750.0,
      "debug/num_lat_total": 2071.0,
      "debug/num_tok_loss": 1750.0,
      "debug/num_tok_total": 2071.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.178753852844238,
      "train/diffusion_loss": 0.4612983465194702
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1777.0,
      "debug/num_lat_total": 2052.0,
      "debug/num_tok_loss": 1777.0,
      "debug/num_tok_total": 2052.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.192585945129395,
      "train/diffusion_loss": 0.46960434317588806
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1735.0,
      "debug/num_lat_total": 2097.0,
      "debug/num_tok_loss": 1735.0,
      "debug/num_tok_total": 2097.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.17921257019043,
      "train/diffusion_loss": 0.4624210000038147
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1681.0,
      "debug/num_lat_total": 2131.0,
      "debug/num_tok_loss": 1681.0,
      "debug/num_tok_total": 2131.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.16563606262207,
      "train/diffusion_loss": 0.45106709003448486
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1811.0,
      "debug/num_lat_total": 2189.0,
      "debug/num_tok_loss": 1811.0,
      "debug/num_tok_total": 2189.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.170101165771484,
      "train/diffusion_loss": 0.4966118633747101
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1682.0,
      "debug/num_lat_total": 2156.0,
      "debug/num_tok_loss": 1682.0,
      "debug/num_tok_total": 2156.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.169591903686523,
      "train/diffusion_loss": 0.445074200630188
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1752.0,
      "debug/num_lat_total": 2327.0,
      "debug/num_tok_loss": 1752.0,
      "debug/num_tok_total": 2327.0,
      "epoch": 0.7476635514018691,
      "step": 5
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/ce_loss": 12.20793342590332,
      "train/diffusion_loss": 0.42772501707077026
    },
    {
      "epoch": 0.7476635514018691,
      "step": 5,
      "train/learning_rate_real": 2.449366217018122e-05
    },
    {
      "debug/num_lat_loss": 1709.0,
      "debug/num_lat_total": 2042.0,
      "debug/num_tok_loss": 1709.0,
      "debug/num_tok_total": 2042.0,
      "epoch": 0.897196261682243,
      "step": 6
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/ce_loss": 12.166390419006348,
      "train/diffusion_loss": 0.4637673497200012
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/learning_rate_real": 2.4104599162700908e-05
    },
    {
      "debug/num_lat_loss": 1852.0,
      "debug/num_lat_total": 2350.0,
      "debug/num_tok_loss": 1852.0,
      "debug/num_tok_total": 2350.0,
      "epoch": 0.897196261682243,
      "step": 6
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/ce_loss": 12.213371276855469,
      "train/diffusion_loss": 0.4463023245334625
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/learning_rate_real": 2.4104599162700908e-05
    },
    {
      "debug/num_lat_loss": 1697.0,
      "debug/num_lat_total": 2142.0,
      "debug/num_tok_loss": 1697.0,
      "debug/num_tok_total": 2142.0,
      "epoch": 0.897196261682243,
      "step": 6
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/ce_loss": 12.193582534790039,
      "train/diffusion_loss": 0.4670042097568512
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/learning_rate_real": 2.4104599162700908e-05
    },
    {
      "debug/num_lat_loss": 1767.0,
      "debug/num_lat_total": 2228.0,
      "debug/num_tok_loss": 1767.0,
      "debug/num_tok_total": 2228.0,
      "epoch": 0.897196261682243,
      "step": 6
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/ce_loss": 12.215458869934082,
      "train/diffusion_loss": 0.4344768226146698
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/learning_rate_real": 2.4104599162700908e-05
    },
    {
      "debug/num_lat_loss": 1739.0,
      "debug/num_lat_total": 2109.0,
      "debug/num_tok_loss": 1739.0,
      "debug/num_tok_total": 2109.0,
      "epoch": 0.897196261682243,
      "step": 6
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/ce_loss": 12.17676067352295,
      "train/diffusion_loss": 0.4460485279560089
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/learning_rate_real": 2.4104599162700908e-05
    },
    {
      "debug/num_lat_loss": 1736.0,
      "debug/num_lat_total": 2020.0,
      "debug/num_tok_loss": 1736.0,
      "debug/num_tok_total": 2020.0,
      "epoch": 0.897196261682243,
      "step": 6
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/ce_loss": 12.166215896606445,
      "train/diffusion_loss": 0.44714999198913574
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/learning_rate_real": 2.4104599162700908e-05
    },
    {
      "debug/num_lat_loss": 1769.0,
      "debug/num_lat_total": 2321.0,
      "debug/num_tok_loss": 1769.0,
      "debug/num_tok_total": 2321.0,
      "epoch": 0.897196261682243,
      "step": 6
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/ce_loss": 12.184131622314453,
      "train/diffusion_loss": 0.4322231411933899
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/learning_rate_real": 2.4104599162700908e-05
    },
    {
      "debug/num_lat_loss": 1661.0,
      "debug/num_lat_total": 2082.0,
      "debug/num_tok_loss": 1661.0,
      "debug/num_tok_total": 2082.0,
      "epoch": 0.897196261682243,
      "step": 6
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/ce_loss": 12.15831470489502,
      "train/diffusion_loss": 0.44721469283103943
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/learning_rate_real": 2.4104599162700908e-05
    },
    {
      "debug/num_lat_loss": 1731.0,
      "debug/num_lat_total": 2121.0,
      "debug/num_tok_loss": 1731.0,
      "debug/num_tok_total": 2121.0,
      "epoch": 0.897196261682243,
      "step": 6
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/ce_loss": 12.174907684326172,
      "train/diffusion_loss": 0.43733832240104675
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/learning_rate_real": 2.4104599162700908e-05
    },
    {
      "debug/num_lat_loss": 1783.0,
      "debug/num_lat_total": 2077.0,
      "debug/num_tok_loss": 1783.0,
      "debug/num_tok_total": 2077.0,
      "epoch": 0.897196261682243,
      "step": 6
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/ce_loss": 12.199905395507812,
      "train/diffusion_loss": 0.45236700773239136
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/learning_rate_real": 2.4104599162700908e-05
    },
    {
      "debug/num_lat_loss": 229.0,
      "debug/num_lat_total": 229.0,
      "debug/num_tok_loss": 229.0,
      "debug/num_tok_total": 229.0,
      "epoch": 0.897196261682243,
      "step": 6
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/ce_loss": 12.282599449157715,
      "train/diffusion_loss": 0.5522232055664062
    },
    {
      "epoch": 0.897196261682243,
      "step": 6,
      "train/learning_rate_real": 2.4104599162700908e-05
    },
    {
      "debug/num_lat_loss": 1699.0,
      "debug/num_lat_total": 2101.0,
      "debug/num_tok_loss": 1699.0,
      "debug/num_tok_total": 2101.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.172179222106934,
      "train/diffusion_loss": 0.4296949505805969
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1688.0,
      "debug/num_lat_total": 2086.0,
      "debug/num_tok_loss": 1688.0,
      "debug/num_tok_total": 2086.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.194772720336914,
      "train/diffusion_loss": 0.4478878676891327
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1729.0,
      "debug/num_lat_total": 2374.0,
      "debug/num_tok_loss": 1729.0,
      "debug/num_tok_total": 2374.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.205586433410645,
      "train/diffusion_loss": 0.4039336144924164
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1692.0,
      "debug/num_lat_total": 2233.0,
      "debug/num_tok_loss": 1692.0,
      "debug/num_tok_total": 2233.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.20322322845459,
      "train/diffusion_loss": 0.4448912441730499
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1747.0,
      "debug/num_lat_total": 2252.0,
      "debug/num_tok_loss": 1747.0,
      "debug/num_tok_total": 2252.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.161945343017578,
      "train/diffusion_loss": 0.4232192933559418
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1782.0,
      "debug/num_lat_total": 2246.0,
      "debug/num_tok_loss": 1782.0,
      "debug/num_tok_total": 2246.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.166516304016113,
      "train/diffusion_loss": 0.44215527176856995
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1684.0,
      "debug/num_lat_total": 2101.0,
      "debug/num_tok_loss": 1684.0,
      "debug/num_tok_total": 2101.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.172343254089355,
      "train/diffusion_loss": 0.46213802695274353
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1741.0,
      "debug/num_lat_total": 2286.0,
      "debug/num_tok_loss": 1741.0,
      "debug/num_tok_total": 2286.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.176877975463867,
      "train/diffusion_loss": 0.43911412358283997
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1646.0,
      "debug/num_lat_total": 1761.0,
      "debug/num_tok_loss": 1646.0,
      "debug/num_tok_total": 1761.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.188158988952637,
      "train/diffusion_loss": 0.4860377311706543
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1823.0,
      "debug/num_lat_total": 2333.0,
      "debug/num_tok_loss": 1823.0,
      "debug/num_tok_total": 2333.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.187177658081055,
      "train/diffusion_loss": 0.43619203567504883
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1817.0,
      "debug/num_lat_total": 2273.0,
      "debug/num_tok_loss": 1817.0,
      "debug/num_tok_total": 2273.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.17230224609375,
      "train/diffusion_loss": 0.4542900025844574
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1701.0,
      "debug/num_lat_total": 2256.0,
      "debug/num_tok_loss": 1701.0,
      "debug/num_tok_total": 2256.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.149282455444336,
      "train/diffusion_loss": 0.43968647718429565
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1708.0,
      "debug/num_lat_total": 2278.0,
      "debug/num_tok_loss": 1708.0,
      "debug/num_tok_total": 2278.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.181864738464355,
      "train/diffusion_loss": 0.433378666639328
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1808.0,
      "debug/num_lat_total": 2109.0,
      "debug/num_tok_loss": 1808.0,
      "debug/num_tok_total": 2109.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.208248138427734,
      "train/diffusion_loss": 0.45877185463905334
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1655.0,
      "debug/num_lat_total": 2171.0,
      "debug/num_tok_loss": 1655.0,
      "debug/num_tok_total": 2171.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.172933578491211,
      "train/diffusion_loss": 0.4297572076320648
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1760.0,
      "debug/num_lat_total": 2230.0,
      "debug/num_tok_loss": 1760.0,
      "debug/num_tok_total": 2230.0,
      "epoch": 1.0,
      "step": 7
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/ce_loss": 12.13405990600586,
      "train/diffusion_loss": 0.4668017029762268
    },
    {
      "epoch": 1.0,
      "step": 7,
      "train/learning_rate_real": 2.3610443108186546e-05
    },
    {
      "debug/num_lat_loss": 1678.0,
      "debug/num_lat_total": 2256.0,
      "debug/num_tok_loss": 1678.0,
      "debug/num_tok_total": 2256.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.185417175292969,
      "train/diffusion_loss": 0.4371003806591034
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1747.0,
      "debug/num_lat_total": 2139.0,
      "debug/num_tok_loss": 1747.0,
      "debug/num_tok_total": 2139.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.221545219421387,
      "train/diffusion_loss": 0.4711552560329437
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1767.0,
      "debug/num_lat_total": 2145.0,
      "debug/num_tok_loss": 1767.0,
      "debug/num_tok_total": 2145.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.16722583770752,
      "train/diffusion_loss": 0.46221664547920227
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1871.0,
      "debug/num_lat_total": 2412.0,
      "debug/num_tok_loss": 1871.0,
      "debug/num_tok_total": 2412.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.183208465576172,
      "train/diffusion_loss": 0.42312583327293396
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1770.0,
      "debug/num_lat_total": 2357.0,
      "debug/num_tok_loss": 1770.0,
      "debug/num_tok_total": 2357.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.181570053100586,
      "train/diffusion_loss": 0.4198269248008728
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1691.0,
      "debug/num_lat_total": 2096.0,
      "debug/num_tok_loss": 1691.0,
      "debug/num_tok_total": 2096.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.1736421585083,
      "train/diffusion_loss": 0.4673723876476288
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1703.0,
      "debug/num_lat_total": 2182.0,
      "debug/num_tok_loss": 1703.0,
      "debug/num_tok_total": 2182.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.155488014221191,
      "train/diffusion_loss": 0.4332273304462433
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1732.0,
      "debug/num_lat_total": 2212.0,
      "debug/num_tok_loss": 1732.0,
      "debug/num_tok_total": 2212.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.16242504119873,
      "train/diffusion_loss": 0.44073769450187683
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1707.0,
      "debug/num_lat_total": 2084.0,
      "debug/num_tok_loss": 1707.0,
      "debug/num_tok_total": 2084.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.175156593322754,
      "train/diffusion_loss": 0.4498469829559326
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1635.0,
      "debug/num_lat_total": 1945.0,
      "debug/num_tok_loss": 1635.0,
      "debug/num_tok_total": 1945.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.206686019897461,
      "train/diffusion_loss": 0.46417516469955444
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1735.0,
      "debug/num_lat_total": 2205.0,
      "debug/num_tok_loss": 1735.0,
      "debug/num_tok_total": 2205.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.191603660583496,
      "train/diffusion_loss": 0.4583619236946106
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1767.0,
      "debug/num_lat_total": 2142.0,
      "debug/num_tok_loss": 1767.0,
      "debug/num_tok_total": 2142.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.173551559448242,
      "train/diffusion_loss": 0.47600993514060974
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1746.0,
      "debug/num_lat_total": 2214.0,
      "debug/num_tok_loss": 1746.0,
      "debug/num_tok_total": 2214.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.174034118652344,
      "train/diffusion_loss": 0.4496922194957733
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1743.0,
      "debug/num_lat_total": 2097.0,
      "debug/num_tok_loss": 1743.0,
      "debug/num_tok_total": 2097.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.169427871704102,
      "train/diffusion_loss": 0.46032634377479553
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1720.0,
      "debug/num_lat_total": 2218.0,
      "debug/num_tok_loss": 1720.0,
      "debug/num_tok_total": 2218.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.195882797241211,
      "train/diffusion_loss": 0.4113229513168335
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1775.0,
      "debug/num_lat_total": 2160.0,
      "debug/num_tok_loss": 1775.0,
      "debug/num_tok_total": 2160.0,
      "epoch": 1.1495327102803738,
      "step": 8
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/ce_loss": 12.156371116638184,
      "train/diffusion_loss": 0.4266815781593323
    },
    {
      "epoch": 1.1495327102803738,
      "step": 8,
      "train/learning_rate_real": 2.3015669160389767e-05
    },
    {
      "debug/num_lat_loss": 1731.0,
      "debug/num_lat_total": 2141.0,
      "debug/num_tok_loss": 1731.0,
      "debug/num_tok_total": 2141.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.152505874633789,
      "train/diffusion_loss": 0.46556028723716736
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1801.0,
      "debug/num_lat_total": 2340.0,
      "debug/num_tok_loss": 1801.0,
      "debug/num_tok_total": 2340.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.16299819946289,
      "train/diffusion_loss": 0.42512306571006775
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1731.0,
      "debug/num_lat_total": 2226.0,
      "debug/num_tok_loss": 1731.0,
      "debug/num_tok_total": 2226.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.169912338256836,
      "train/diffusion_loss": 0.4332532584667206
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1699.0,
      "debug/num_lat_total": 2058.0,
      "debug/num_tok_loss": 1699.0,
      "debug/num_tok_total": 2058.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.193463325500488,
      "train/diffusion_loss": 0.466909795999527
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1696.0,
      "debug/num_lat_total": 2215.0,
      "debug/num_tok_loss": 1696.0,
      "debug/num_tok_total": 2215.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.178363800048828,
      "train/diffusion_loss": 0.4238511621952057
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1704.0,
      "debug/num_lat_total": 2075.0,
      "debug/num_tok_loss": 1704.0,
      "debug/num_tok_total": 2075.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.177104949951172,
      "train/diffusion_loss": 0.4522017538547516
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1743.0,
      "debug/num_lat_total": 2164.0,
      "debug/num_tok_loss": 1743.0,
      "debug/num_tok_total": 2164.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.171281814575195,
      "train/diffusion_loss": 0.45279401540756226
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1785.0,
      "debug/num_lat_total": 2090.0,
      "debug/num_tok_loss": 1785.0,
      "debug/num_tok_total": 2090.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.184865951538086,
      "train/diffusion_loss": 0.47387492656707764
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1716.0,
      "debug/num_lat_total": 2273.0,
      "debug/num_tok_loss": 1716.0,
      "debug/num_tok_total": 2273.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.163240432739258,
      "train/diffusion_loss": 0.4024707078933716
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1820.0,
      "debug/num_lat_total": 2160.0,
      "debug/num_tok_loss": 1820.0,
      "debug/num_tok_total": 2160.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.185773849487305,
      "train/diffusion_loss": 0.4599778354167938
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1792.0,
      "debug/num_lat_total": 2318.0,
      "debug/num_tok_loss": 1792.0,
      "debug/num_tok_total": 2318.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.170687675476074,
      "train/diffusion_loss": 0.4315679967403412
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1769.0,
      "debug/num_lat_total": 2129.0,
      "debug/num_tok_loss": 1769.0,
      "debug/num_tok_total": 2129.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.157472610473633,
      "train/diffusion_loss": 0.4691459834575653
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1743.0,
      "debug/num_lat_total": 1996.0,
      "debug/num_tok_loss": 1743.0,
      "debug/num_tok_total": 1996.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.206344604492188,
      "train/diffusion_loss": 0.47081127762794495
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1730.0,
      "debug/num_lat_total": 2146.0,
      "debug/num_tok_loss": 1730.0,
      "debug/num_tok_total": 2146.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.170424461364746,
      "train/diffusion_loss": 0.4682280719280243
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1785.0,
      "debug/num_lat_total": 2218.0,
      "debug/num_tok_loss": 1785.0,
      "debug/num_tok_total": 2218.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.194655418395996,
      "train/diffusion_loss": 0.4655971825122833
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "debug/num_lat_loss": 1816.0,
      "debug/num_lat_total": 2312.0,
      "debug/num_tok_loss": 1816.0,
      "debug/num_tok_total": 2312.0,
      "epoch": 1.2990654205607477,
      "step": 9
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/ce_loss": 12.17812728881836,
      "train/diffusion_loss": 0.4462972581386566
    },
    {
      "epoch": 1.2990654205607477,
      "step": 9,
      "train/learning_rate_real": 2.2325663684284844e-05
    },
    {
      "epoch": 1.4485981308411215,
      "grad_norm": 0.10791015625,
      "learning_rate": 2.2325663684284844e-05,
      "loss": 1.1205,
      "step": 10
    },
    {
      "debug/num_lat_loss": 1693.0,
      "debug/num_lat_total": 2185.0,
      "debug/num_tok_loss": 1693.0,
      "debug/num_tok_total": 2185.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.169471740722656,
      "train/diffusion_loss": 0.4354258179664612
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1766.0,
      "debug/num_lat_total": 2349.0,
      "debug/num_tok_loss": 1766.0,
      "debug/num_tok_total": 2349.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.206490516662598,
      "train/diffusion_loss": 0.39133599400520325
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1719.0,
      "debug/num_lat_total": 2088.0,
      "debug/num_tok_loss": 1719.0,
      "debug/num_tok_total": 2088.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.198739051818848,
      "train/diffusion_loss": 0.4691934883594513
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1711.0,
      "debug/num_lat_total": 2242.0,
      "debug/num_tok_loss": 1711.0,
      "debug/num_tok_total": 2242.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.150611877441406,
      "train/diffusion_loss": 0.4412599205970764
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1795.0,
      "debug/num_lat_total": 2284.0,
      "debug/num_tok_loss": 1795.0,
      "debug/num_tok_total": 2284.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.177535057067871,
      "train/diffusion_loss": 0.41242048144340515
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1715.0,
      "debug/num_lat_total": 2231.0,
      "debug/num_tok_loss": 1715.0,
      "debug/num_tok_total": 2231.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.17111873626709,
      "train/diffusion_loss": 0.43765971064567566
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1726.0,
      "debug/num_lat_total": 2153.0,
      "debug/num_tok_loss": 1726.0,
      "debug/num_tok_total": 2153.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.200366973876953,
      "train/diffusion_loss": 0.4417407512664795
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1805.0,
      "debug/num_lat_total": 2263.0,
      "debug/num_tok_loss": 1805.0,
      "debug/num_tok_total": 2263.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.149072647094727,
      "train/diffusion_loss": 0.4602048397064209
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1811.0,
      "debug/num_lat_total": 2264.0,
      "debug/num_tok_loss": 1811.0,
      "debug/num_tok_total": 2264.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.151084899902344,
      "train/diffusion_loss": 0.4598139822483063
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1795.0,
      "debug/num_lat_total": 2296.0,
      "debug/num_tok_loss": 1795.0,
      "debug/num_tok_total": 2296.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.177088737487793,
      "train/diffusion_loss": 0.4397522211074829
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1757.0,
      "debug/num_lat_total": 2316.0,
      "debug/num_tok_loss": 1757.0,
      "debug/num_tok_total": 2316.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.180797576904297,
      "train/diffusion_loss": 0.46918240189552307
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1710.0,
      "debug/num_lat_total": 2171.0,
      "debug/num_tok_loss": 1710.0,
      "debug/num_tok_total": 2171.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.182649612426758,
      "train/diffusion_loss": 0.44089797139167786
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1830.0,
      "debug/num_lat_total": 2244.0,
      "debug/num_tok_loss": 1830.0,
      "debug/num_tok_total": 2244.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.196844100952148,
      "train/diffusion_loss": 0.43193933367729187
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1700.0,
      "debug/num_lat_total": 2234.0,
      "debug/num_tok_loss": 1700.0,
      "debug/num_tok_total": 2234.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.192126274108887,
      "train/diffusion_loss": 0.44308194518089294
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1675.0,
      "debug/num_lat_total": 2107.0,
      "debug/num_tok_loss": 1675.0,
      "debug/num_tok_total": 2107.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.138769149780273,
      "train/diffusion_loss": 0.43758758902549744
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1729.0,
      "debug/num_lat_total": 2221.0,
      "debug/num_tok_loss": 1729.0,
      "debug/num_tok_total": 2221.0,
      "epoch": 1.4485981308411215,
      "step": 10
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/ce_loss": 12.164264678955078,
      "train/diffusion_loss": 0.4479159712791443
    },
    {
      "epoch": 1.4485981308411215,
      "step": 10,
      "train/learning_rate_real": 2.154667547631338e-05
    },
    {
      "debug/num_lat_loss": 1773.0,
      "debug/num_lat_total": 2226.0,
      "debug/num_tok_loss": 1773.0,
      "debug/num_tok_total": 2226.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.166282653808594,
      "train/diffusion_loss": 0.45015960931777954
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1780.0,
      "debug/num_lat_total": 2155.0,
      "debug/num_tok_loss": 1780.0,
      "debug/num_tok_total": 2155.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.18748950958252,
      "train/diffusion_loss": 0.46807411313056946
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1712.0,
      "debug/num_lat_total": 2075.0,
      "debug/num_tok_loss": 1712.0,
      "debug/num_tok_total": 2075.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.184382438659668,
      "train/diffusion_loss": 0.45887696743011475
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1835.0,
      "debug/num_lat_total": 2278.0,
      "debug/num_tok_loss": 1835.0,
      "debug/num_tok_total": 2278.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.13071060180664,
      "train/diffusion_loss": 0.4673099219799042
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1728.0,
      "debug/num_lat_total": 2127.0,
      "debug/num_tok_loss": 1728.0,
      "debug/num_tok_total": 2127.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.159421920776367,
      "train/diffusion_loss": 0.44304782152175903
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1729.0,
      "debug/num_lat_total": 2149.0,
      "debug/num_tok_loss": 1729.0,
      "debug/num_tok_total": 2149.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.162511825561523,
      "train/diffusion_loss": 0.4454917013645172
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1803.0,
      "debug/num_lat_total": 2170.0,
      "debug/num_tok_loss": 1803.0,
      "debug/num_tok_total": 2170.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.184347152709961,
      "train/diffusion_loss": 0.4569142460823059
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1758.0,
      "debug/num_lat_total": 2278.0,
      "debug/num_tok_loss": 1758.0,
      "debug/num_tok_total": 2278.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.191046714782715,
      "train/diffusion_loss": 0.4136866629123688
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1779.0,
      "debug/num_lat_total": 2387.0,
      "debug/num_tok_loss": 1779.0,
      "debug/num_tok_total": 2387.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.171055793762207,
      "train/diffusion_loss": 0.4274556338787079
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1735.0,
      "debug/num_lat_total": 2052.0,
      "debug/num_tok_loss": 1735.0,
      "debug/num_tok_total": 2052.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.187844276428223,
      "train/diffusion_loss": 0.4536263048648834
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1751.0,
      "debug/num_lat_total": 2068.0,
      "debug/num_tok_loss": 1751.0,
      "debug/num_tok_total": 2068.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.170848846435547,
      "train/diffusion_loss": 0.4488384425640106
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1728.0,
      "debug/num_lat_total": 2149.0,
      "debug/num_tok_loss": 1728.0,
      "debug/num_tok_total": 2149.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.222294807434082,
      "train/diffusion_loss": 0.4231165945529938
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1672.0,
      "debug/num_lat_total": 2265.0,
      "debug/num_tok_loss": 1672.0,
      "debug/num_tok_total": 2265.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.197094917297363,
      "train/diffusion_loss": 0.42200806736946106
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1809.0,
      "debug/num_lat_total": 2274.0,
      "debug/num_tok_loss": 1809.0,
      "debug/num_tok_total": 2274.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.177701950073242,
      "train/diffusion_loss": 0.4552193582057953
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1671.0,
      "debug/num_lat_total": 2181.0,
      "debug/num_tok_loss": 1671.0,
      "debug/num_tok_total": 2181.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.13666820526123,
      "train/diffusion_loss": 0.419909805059433
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1770.0,
      "debug/num_lat_total": 2254.0,
      "debug/num_tok_loss": 1770.0,
      "debug/num_tok_total": 2254.0,
      "epoch": 1.5981308411214954,
      "step": 11
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/ce_loss": 12.172820091247559,
      "train/diffusion_loss": 0.43515339493751526
    },
    {
      "epoch": 1.5981308411214954,
      "step": 11,
      "train/learning_rate_real": 2.0685759174316066e-05
    },
    {
      "debug/num_lat_loss": 1760.0,
      "debug/num_lat_total": 2127.0,
      "debug/num_tok_loss": 1760.0,
      "debug/num_tok_total": 2127.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.22167682647705,
      "train/diffusion_loss": 0.4806464910507202
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1802.0,
      "debug/num_lat_total": 2147.0,
      "debug/num_tok_loss": 1802.0,
      "debug/num_tok_total": 2147.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.198952674865723,
      "train/diffusion_loss": 0.4417625069618225
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1730.0,
      "debug/num_lat_total": 2123.0,
      "debug/num_tok_loss": 1730.0,
      "debug/num_tok_total": 2123.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.184418678283691,
      "train/diffusion_loss": 0.45918792486190796
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1656.0,
      "debug/num_lat_total": 1983.0,
      "debug/num_tok_loss": 1656.0,
      "debug/num_tok_total": 1983.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.202133178710938,
      "train/diffusion_loss": 0.4891001284122467
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1718.0,
      "debug/num_lat_total": 2212.0,
      "debug/num_tok_loss": 1718.0,
      "debug/num_tok_total": 2212.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.170212745666504,
      "train/diffusion_loss": 0.4450775682926178
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1697.0,
      "debug/num_lat_total": 2134.0,
      "debug/num_tok_loss": 1697.0,
      "debug/num_tok_total": 2134.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.188855171203613,
      "train/diffusion_loss": 0.4286879301071167
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1689.0,
      "debug/num_lat_total": 2269.0,
      "debug/num_tok_loss": 1689.0,
      "debug/num_tok_total": 2269.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.180878639221191,
      "train/diffusion_loss": 0.3961124122142792
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1748.0,
      "debug/num_lat_total": 2199.0,
      "debug/num_tok_loss": 1748.0,
      "debug/num_tok_total": 2199.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.146210670471191,
      "train/diffusion_loss": 0.458207905292511
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1738.0,
      "debug/num_lat_total": 2336.0,
      "debug/num_tok_loss": 1738.0,
      "debug/num_tok_total": 2336.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.17103099822998,
      "train/diffusion_loss": 0.41296547651290894
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1691.0,
      "debug/num_lat_total": 2167.0,
      "debug/num_tok_loss": 1691.0,
      "debug/num_tok_total": 2167.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.197969436645508,
      "train/diffusion_loss": 0.44860365986824036
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1751.0,
      "debug/num_lat_total": 2084.0,
      "debug/num_tok_loss": 1751.0,
      "debug/num_tok_total": 2084.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.17561149597168,
      "train/diffusion_loss": 0.4686024487018585
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1750.0,
      "debug/num_lat_total": 2188.0,
      "debug/num_tok_loss": 1750.0,
      "debug/num_tok_total": 2188.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.174299240112305,
      "train/diffusion_loss": 0.45362716913223267
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1728.0,
      "debug/num_lat_total": 2347.0,
      "debug/num_tok_loss": 1728.0,
      "debug/num_tok_total": 2347.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.163989067077637,
      "train/diffusion_loss": 0.40822833776474
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1764.0,
      "debug/num_lat_total": 2213.0,
      "debug/num_tok_loss": 1764.0,
      "debug/num_tok_total": 2213.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.16737174987793,
      "train/diffusion_loss": 0.437155544757843
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1747.0,
      "debug/num_lat_total": 2335.0,
      "debug/num_tok_loss": 1747.0,
      "debug/num_tok_total": 2335.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.159972190856934,
      "train/diffusion_loss": 0.4289741516113281
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1694.0,
      "debug/num_lat_total": 2192.0,
      "debug/num_tok_loss": 1694.0,
      "debug/num_tok_total": 2192.0,
      "epoch": 1.7476635514018692,
      "step": 12
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/ce_loss": 12.163084030151367,
      "train/diffusion_loss": 0.4414047300815582
    },
    {
      "epoch": 1.7476635514018692,
      "step": 12,
      "train/learning_rate_real": 1.9750711369639978e-05
    },
    {
      "debug/num_lat_loss": 1726.0,
      "debug/num_lat_total": 2200.0,
      "debug/num_tok_loss": 1726.0,
      "debug/num_tok_total": 2200.0,
      "epoch": 1.897196261682243,
      "step": 13
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/ce_loss": 12.161957740783691,
      "train/diffusion_loss": 0.4361889362335205
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/learning_rate_real": 1.8750000000000002e-05
    },
    {
      "debug/num_lat_loss": 1617.0,
      "debug/num_lat_total": 2198.0,
      "debug/num_tok_loss": 1617.0,
      "debug/num_tok_total": 2198.0,
      "epoch": 1.897196261682243,
      "step": 13
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/ce_loss": 12.193390846252441,
      "train/diffusion_loss": 0.4090780019760132
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/learning_rate_real": 1.8750000000000002e-05
    },
    {
      "debug/num_lat_loss": 1813.0,
      "debug/num_lat_total": 2269.0,
      "debug/num_tok_loss": 1813.0,
      "debug/num_tok_total": 2269.0,
      "epoch": 1.897196261682243,
      "step": 13
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/ce_loss": 12.186473846435547,
      "train/diffusion_loss": 0.45172038674354553
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/learning_rate_real": 1.8750000000000002e-05
    },
    {
      "debug/num_lat_loss": 1796.0,
      "debug/num_lat_total": 2229.0,
      "debug/num_tok_loss": 1796.0,
      "debug/num_tok_total": 2229.0,
      "epoch": 1.897196261682243,
      "step": 13
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/ce_loss": 12.181390762329102,
      "train/diffusion_loss": 0.43670597672462463
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/learning_rate_real": 1.8750000000000002e-05
    },
    {
      "debug/num_lat_loss": 1712.0,
      "debug/num_lat_total": 2176.0,
      "debug/num_tok_loss": 1712.0,
      "debug/num_tok_total": 2176.0,
      "epoch": 1.897196261682243,
      "step": 13
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/ce_loss": 12.185544967651367,
      "train/diffusion_loss": 0.4369116425514221
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/learning_rate_real": 1.8750000000000002e-05
    },
    {
      "debug/num_lat_loss": 1784.0,
      "debug/num_lat_total": 2337.0,
      "debug/num_tok_loss": 1784.0,
      "debug/num_tok_total": 2337.0,
      "epoch": 1.897196261682243,
      "step": 13
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/ce_loss": 12.142572402954102,
      "train/diffusion_loss": 0.4199765920639038
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/learning_rate_real": 1.8750000000000002e-05
    },
    {
      "debug/num_lat_loss": 1742.0,
      "debug/num_lat_total": 2196.0,
      "debug/num_tok_loss": 1742.0,
      "debug/num_tok_total": 2196.0,
      "epoch": 1.897196261682243,
      "step": 13
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/ce_loss": 12.183996200561523,
      "train/diffusion_loss": 0.4249250292778015
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/learning_rate_real": 1.8750000000000002e-05
    },
    {
      "debug/num_lat_loss": 1725.0,
      "debug/num_lat_total": 2215.0,
      "debug/num_tok_loss": 1725.0,
      "debug/num_tok_total": 2215.0,
      "epoch": 1.897196261682243,
      "step": 13
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/ce_loss": 12.17967700958252,
      "train/diffusion_loss": 0.4066171944141388
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/learning_rate_real": 1.8750000000000002e-05
    },
    {
      "debug/num_lat_loss": 1690.0,
      "debug/num_lat_total": 2162.0,
      "debug/num_tok_loss": 1690.0,
      "debug/num_tok_total": 2162.0,
      "epoch": 1.897196261682243,
      "step": 13
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/ce_loss": 12.194439888000488,
      "train/diffusion_loss": 0.45021936297416687
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/learning_rate_real": 1.8750000000000002e-05
    },
    {
      "debug/num_lat_loss": 1711.0,
      "debug/num_lat_total": 2092.0,
      "debug/num_tok_loss": 1711.0,
      "debug/num_tok_total": 2092.0,
      "epoch": 1.897196261682243,
      "step": 13
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/ce_loss": 12.181828498840332,
      "train/diffusion_loss": 0.4418796896934509
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/learning_rate_real": 1.8750000000000002e-05
    },
    {
      "debug/num_lat_loss": 243.0,
      "debug/num_lat_total": 341.0,
      "debug/num_tok_loss": 243.0,
      "debug/num_tok_total": 341.0,
      "epoch": 1.897196261682243,
      "step": 13
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/ce_loss": 12.189529418945312,
      "train/diffusion_loss": 0.4245636463165283
    },
    {
      "epoch": 1.897196261682243,
      "step": 13,
      "train/learning_rate_real": 1.8750000000000002e-05
    },
    {
      "debug/num_lat_loss": 1713.0,
      "debug/num_lat_total": 2182.0,
      "debug/num_tok_loss": 1713.0,
      "debug/num_tok_total": 2182.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.153608322143555,
      "train/diffusion_loss": 0.4352942705154419
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1792.0,
      "debug/num_lat_total": 2291.0,
      "debug/num_tok_loss": 1792.0,
      "debug/num_tok_total": 2291.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.172529220581055,
      "train/diffusion_loss": 0.41673797369003296
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1818.0,
      "debug/num_lat_total": 2261.0,
      "debug/num_tok_loss": 1818.0,
      "debug/num_tok_total": 2261.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.160377502441406,
      "train/diffusion_loss": 0.44119855761528015
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1688.0,
      "debug/num_lat_total": 2162.0,
      "debug/num_tok_loss": 1688.0,
      "debug/num_tok_total": 2162.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.186813354492188,
      "train/diffusion_loss": 0.41701310873031616
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1789.0,
      "debug/num_lat_total": 2174.0,
      "debug/num_tok_loss": 1789.0,
      "debug/num_tok_total": 2174.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.174489974975586,
      "train/diffusion_loss": 0.45595914125442505
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1731.0,
      "debug/num_lat_total": 2278.0,
      "debug/num_tok_loss": 1731.0,
      "debug/num_tok_total": 2278.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.182331085205078,
      "train/diffusion_loss": 0.4117718040943146
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1755.0,
      "debug/num_lat_total": 2268.0,
      "debug/num_tok_loss": 1755.0,
      "debug/num_tok_total": 2268.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.161623001098633,
      "train/diffusion_loss": 0.4075234830379486
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1648.0,
      "debug/num_lat_total": 2062.0,
      "debug/num_tok_loss": 1648.0,
      "debug/num_tok_total": 2062.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.20829963684082,
      "train/diffusion_loss": 0.4243350625038147
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1684.0,
      "debug/num_lat_total": 2065.0,
      "debug/num_tok_loss": 1684.0,
      "debug/num_tok_total": 2065.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.194366455078125,
      "train/diffusion_loss": 0.4790455102920532
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1738.0,
      "debug/num_lat_total": 2258.0,
      "debug/num_tok_loss": 1738.0,
      "debug/num_tok_total": 2258.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.151714324951172,
      "train/diffusion_loss": 0.4375750422477722
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1714.0,
      "debug/num_lat_total": 2035.0,
      "debug/num_tok_loss": 1714.0,
      "debug/num_tok_total": 2035.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.17888355255127,
      "train/diffusion_loss": 0.48229870200157166
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1791.0,
      "debug/num_lat_total": 2202.0,
      "debug/num_tok_loss": 1791.0,
      "debug/num_tok_total": 2202.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.186660766601562,
      "train/diffusion_loss": 0.44068536162376404
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1718.0,
      "debug/num_lat_total": 2154.0,
      "debug/num_tok_loss": 1718.0,
      "debug/num_tok_total": 2154.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.159738540649414,
      "train/diffusion_loss": 0.4153040647506714
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1741.0,
      "debug/num_lat_total": 2189.0,
      "debug/num_tok_loss": 1741.0,
      "debug/num_tok_total": 2189.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.197320938110352,
      "train/diffusion_loss": 0.438213586807251
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1695.0,
      "debug/num_lat_total": 2245.0,
      "debug/num_tok_loss": 1695.0,
      "debug/num_tok_total": 2245.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.16823673248291,
      "train/diffusion_loss": 0.41748422384262085
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1803.0,
      "debug/num_lat_total": 2096.0,
      "debug/num_tok_loss": 1803.0,
      "debug/num_tok_total": 2096.0,
      "epoch": 2.0,
      "step": 14
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/ce_loss": 12.14641284942627,
      "train/diffusion_loss": 0.4777478277683258
    },
    {
      "epoch": 2.0,
      "step": 14,
      "train/learning_rate_real": 1.7692687662523583e-05
    },
    {
      "debug/num_lat_loss": 1804.0,
      "debug/num_lat_total": 2239.0,
      "debug/num_tok_loss": 1804.0,
      "debug/num_tok_total": 2239.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.182928085327148,
      "train/diffusion_loss": 0.4475792646408081
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1797.0,
      "debug/num_lat_total": 2358.0,
      "debug/num_tok_loss": 1797.0,
      "debug/num_tok_total": 2358.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.155641555786133,
      "train/diffusion_loss": 0.41666021943092346
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1727.0,
      "debug/num_lat_total": 2232.0,
      "debug/num_tok_loss": 1727.0,
      "debug/num_tok_total": 2232.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.173539161682129,
      "train/diffusion_loss": 0.45217418670654297
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1698.0,
      "debug/num_lat_total": 2181.0,
      "debug/num_tok_loss": 1698.0,
      "debug/num_tok_total": 2181.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.21281623840332,
      "train/diffusion_loss": 0.44602900743484497
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1759.0,
      "debug/num_lat_total": 2135.0,
      "debug/num_tok_loss": 1759.0,
      "debug/num_tok_total": 2135.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.174015998840332,
      "train/diffusion_loss": 0.4635976552963257
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1727.0,
      "debug/num_lat_total": 2196.0,
      "debug/num_tok_loss": 1727.0,
      "debug/num_tok_total": 2196.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.1640625,
      "train/diffusion_loss": 0.4420984983444214
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1758.0,
      "debug/num_lat_total": 2018.0,
      "debug/num_tok_loss": 1758.0,
      "debug/num_tok_total": 2018.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.15977668762207,
      "train/diffusion_loss": 0.4814409911632538
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1678.0,
      "debug/num_lat_total": 2058.0,
      "debug/num_tok_loss": 1678.0,
      "debug/num_tok_total": 2058.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.213717460632324,
      "train/diffusion_loss": 0.4840865135192871
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1840.0,
      "debug/num_lat_total": 2459.0,
      "debug/num_tok_loss": 1840.0,
      "debug/num_tok_total": 2459.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.171710968017578,
      "train/diffusion_loss": 0.43757733702659607
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1760.0,
      "debug/num_lat_total": 2225.0,
      "debug/num_tok_loss": 1760.0,
      "debug/num_tok_total": 2225.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.18978214263916,
      "train/diffusion_loss": 0.44430798292160034
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1721.0,
      "debug/num_lat_total": 2157.0,
      "debug/num_tok_loss": 1721.0,
      "debug/num_tok_total": 2157.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.193751335144043,
      "train/diffusion_loss": 0.4408707916736603
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1709.0,
      "debug/num_lat_total": 2245.0,
      "debug/num_tok_loss": 1709.0,
      "debug/num_tok_total": 2245.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.183303833007812,
      "train/diffusion_loss": 0.4226411283016205
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1795.0,
      "debug/num_lat_total": 2263.0,
      "debug/num_tok_loss": 1795.0,
      "debug/num_tok_total": 2263.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.184673309326172,
      "train/diffusion_loss": 0.4556589722633362
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1673.0,
      "debug/num_lat_total": 2082.0,
      "debug/num_tok_loss": 1673.0,
      "debug/num_tok_total": 2082.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.158052444458008,
      "train/diffusion_loss": 0.4358730912208557
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1708.0,
      "debug/num_lat_total": 2268.0,
      "debug/num_tok_loss": 1708.0,
      "debug/num_tok_total": 2268.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.191332817077637,
      "train/diffusion_loss": 0.4384799599647522
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1706.0,
      "debug/num_lat_total": 2108.0,
      "debug/num_tok_loss": 1706.0,
      "debug/num_tok_total": 2108.0,
      "epoch": 2.149532710280374,
      "step": 15
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/ce_loss": 12.224422454833984,
      "train/diffusion_loss": 0.44930553436279297
    },
    {
      "epoch": 2.149532710280374,
      "step": 15,
      "train/learning_rate_real": 1.6588349541467772e-05
    },
    {
      "debug/num_lat_loss": 1746.0,
      "debug/num_lat_total": 2080.0,
      "debug/num_tok_loss": 1746.0,
      "debug/num_tok_total": 2080.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.148018836975098,
      "train/diffusion_loss": 0.48873013257980347
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1709.0,
      "debug/num_lat_total": 2070.0,
      "debug/num_tok_loss": 1709.0,
      "debug/num_tok_total": 2070.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.192659378051758,
      "train/diffusion_loss": 0.453331857919693
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1769.0,
      "debug/num_lat_total": 2275.0,
      "debug/num_tok_loss": 1769.0,
      "debug/num_tok_total": 2275.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.193303108215332,
      "train/diffusion_loss": 0.43189382553100586
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1780.0,
      "debug/num_lat_total": 2346.0,
      "debug/num_tok_loss": 1780.0,
      "debug/num_tok_total": 2346.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.181623458862305,
      "train/diffusion_loss": 0.4424782991409302
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1762.0,
      "debug/num_lat_total": 2180.0,
      "debug/num_tok_loss": 1762.0,
      "debug/num_tok_total": 2180.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.204513549804688,
      "train/diffusion_loss": 0.4519851803779602
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1695.0,
      "debug/num_lat_total": 2054.0,
      "debug/num_tok_loss": 1695.0,
      "debug/num_tok_total": 2054.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.146162033081055,
      "train/diffusion_loss": 0.4398154020309448
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1761.0,
      "debug/num_lat_total": 2290.0,
      "debug/num_tok_loss": 1761.0,
      "debug/num_tok_total": 2290.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.200496673583984,
      "train/diffusion_loss": 0.4298996031284332
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1779.0,
      "debug/num_lat_total": 2133.0,
      "debug/num_tok_loss": 1779.0,
      "debug/num_tok_total": 2133.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.160360336303711,
      "train/diffusion_loss": 0.4252331554889679
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1761.0,
      "debug/num_lat_total": 2142.0,
      "debug/num_tok_loss": 1761.0,
      "debug/num_tok_total": 2142.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.186250686645508,
      "train/diffusion_loss": 0.4651545584201813
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1748.0,
      "debug/num_lat_total": 2243.0,
      "debug/num_tok_loss": 1748.0,
      "debug/num_tok_total": 2243.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.174160957336426,
      "train/diffusion_loss": 0.4075007140636444
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1712.0,
      "debug/num_lat_total": 2172.0,
      "debug/num_tok_loss": 1712.0,
      "debug/num_tok_total": 2172.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.175134658813477,
      "train/diffusion_loss": 0.42257148027420044
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1671.0,
      "debug/num_lat_total": 2062.0,
      "debug/num_tok_loss": 1671.0,
      "debug/num_tok_total": 2062.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.141824722290039,
      "train/diffusion_loss": 0.470732718706131
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1715.0,
      "debug/num_lat_total": 2233.0,
      "debug/num_tok_loss": 1715.0,
      "debug/num_tok_total": 2233.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.187322616577148,
      "train/diffusion_loss": 0.4473858177661896
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1748.0,
      "debug/num_lat_total": 2375.0,
      "debug/num_tok_loss": 1748.0,
      "debug/num_tok_total": 2375.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.16384506225586,
      "train/diffusion_loss": 0.4119928479194641
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1770.0,
      "debug/num_lat_total": 2268.0,
      "debug/num_tok_loss": 1770.0,
      "debug/num_tok_total": 2268.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.176751136779785,
      "train/diffusion_loss": 0.4215977191925049
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1763.0,
      "debug/num_lat_total": 2374.0,
      "debug/num_tok_loss": 1763.0,
      "debug/num_tok_total": 2374.0,
      "epoch": 2.2990654205607477,
      "step": 16
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/ce_loss": 12.189051628112793,
      "train/diffusion_loss": 0.40282106399536133
    },
    {
      "epoch": 2.2990654205607477,
      "step": 16,
      "train/learning_rate_real": 1.5446986693867843e-05
    },
    {
      "debug/num_lat_loss": 1764.0,
      "debug/num_lat_total": 2167.0,
      "debug/num_tok_loss": 1764.0,
      "debug/num_tok_total": 2167.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.176045417785645,
      "train/diffusion_loss": 0.45012861490249634
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1688.0,
      "debug/num_lat_total": 1927.0,
      "debug/num_tok_loss": 1688.0,
      "debug/num_tok_total": 1927.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.193452835083008,
      "train/diffusion_loss": 0.4965061545372009
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1691.0,
      "debug/num_lat_total": 2287.0,
      "debug/num_tok_loss": 1691.0,
      "debug/num_tok_total": 2287.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.18310260772705,
      "train/diffusion_loss": 0.42198505997657776
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1744.0,
      "debug/num_lat_total": 2118.0,
      "debug/num_tok_loss": 1744.0,
      "debug/num_tok_total": 2118.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.154280662536621,
      "train/diffusion_loss": 0.4763759672641754
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1775.0,
      "debug/num_lat_total": 2223.0,
      "debug/num_tok_loss": 1775.0,
      "debug/num_tok_total": 2223.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.169533729553223,
      "train/diffusion_loss": 0.4356708228588104
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1777.0,
      "debug/num_lat_total": 2186.0,
      "debug/num_tok_loss": 1777.0,
      "debug/num_tok_total": 2186.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.186701774597168,
      "train/diffusion_loss": 0.44108399748802185
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1747.0,
      "debug/num_lat_total": 2174.0,
      "debug/num_tok_loss": 1747.0,
      "debug/num_tok_total": 2174.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.183830261230469,
      "train/diffusion_loss": 0.4281693398952484
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1764.0,
      "debug/num_lat_total": 2347.0,
      "debug/num_tok_loss": 1764.0,
      "debug/num_tok_total": 2347.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.176342964172363,
      "train/diffusion_loss": 0.43779289722442627
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1774.0,
      "debug/num_lat_total": 2054.0,
      "debug/num_tok_loss": 1774.0,
      "debug/num_tok_total": 2054.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.179503440856934,
      "train/diffusion_loss": 0.4560769498348236
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1712.0,
      "debug/num_lat_total": 2106.0,
      "debug/num_tok_loss": 1712.0,
      "debug/num_tok_total": 2106.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.209382057189941,
      "train/diffusion_loss": 0.442518413066864
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1709.0,
      "debug/num_lat_total": 1990.0,
      "debug/num_tok_loss": 1709.0,
      "debug/num_tok_total": 1990.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.172338485717773,
      "train/diffusion_loss": 0.4518572986125946
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1696.0,
      "debug/num_lat_total": 2036.0,
      "debug/num_tok_loss": 1696.0,
      "debug/num_tok_total": 2036.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.185693740844727,
      "train/diffusion_loss": 0.43408462405204773
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1798.0,
      "debug/num_lat_total": 2274.0,
      "debug/num_tok_loss": 1798.0,
      "debug/num_tok_total": 2274.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.175325393676758,
      "train/diffusion_loss": 0.4459254741668701
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1704.0,
      "debug/num_lat_total": 2057.0,
      "debug/num_tok_loss": 1704.0,
      "debug/num_tok_total": 2057.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.171283721923828,
      "train/diffusion_loss": 0.444389671087265
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1762.0,
      "debug/num_lat_total": 2073.0,
      "debug/num_tok_loss": 1762.0,
      "debug/num_tok_total": 2073.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.175067901611328,
      "train/diffusion_loss": 0.5100155472755432
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1728.0,
      "debug/num_lat_total": 2309.0,
      "debug/num_tok_loss": 1728.0,
      "debug/num_tok_total": 2309.0,
      "epoch": 2.4485981308411215,
      "step": 17
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/ce_loss": 12.165181159973145,
      "train/diffusion_loss": 0.4261295795440674
    },
    {
      "epoch": 2.4485981308411215,
      "step": 17,
      "train/learning_rate_real": 1.4278935478416066e-05
    },
    {
      "debug/num_lat_loss": 1790.0,
      "debug/num_lat_total": 2235.0,
      "debug/num_tok_loss": 1790.0,
      "debug/num_tok_total": 2235.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.180500030517578,
      "train/diffusion_loss": 0.4695003032684326
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1726.0,
      "debug/num_lat_total": 2267.0,
      "debug/num_tok_loss": 1726.0,
      "debug/num_tok_total": 2267.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.168024063110352,
      "train/diffusion_loss": 0.44884082674980164
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1786.0,
      "debug/num_lat_total": 2243.0,
      "debug/num_tok_loss": 1786.0,
      "debug/num_tok_total": 2243.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.177437782287598,
      "train/diffusion_loss": 0.43809157609939575
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1728.0,
      "debug/num_lat_total": 2247.0,
      "debug/num_tok_loss": 1728.0,
      "debug/num_tok_total": 2247.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.172117233276367,
      "train/diffusion_loss": 0.45602065324783325
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1767.0,
      "debug/num_lat_total": 2294.0,
      "debug/num_tok_loss": 1767.0,
      "debug/num_tok_total": 2294.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.180341720581055,
      "train/diffusion_loss": 0.41540735960006714
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1786.0,
      "debug/num_lat_total": 2094.0,
      "debug/num_tok_loss": 1786.0,
      "debug/num_tok_total": 2094.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.180994033813477,
      "train/diffusion_loss": 0.44258132576942444
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1732.0,
      "debug/num_lat_total": 2205.0,
      "debug/num_tok_loss": 1732.0,
      "debug/num_tok_total": 2205.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.193119049072266,
      "train/diffusion_loss": 0.4508376717567444
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1663.0,
      "debug/num_lat_total": 2067.0,
      "debug/num_tok_loss": 1663.0,
      "debug/num_tok_total": 2067.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.170428276062012,
      "train/diffusion_loss": 0.4963611662387848
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1732.0,
      "debug/num_lat_total": 2262.0,
      "debug/num_tok_loss": 1732.0,
      "debug/num_tok_total": 2262.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.17246150970459,
      "train/diffusion_loss": 0.4106256365776062
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1689.0,
      "debug/num_lat_total": 2285.0,
      "debug/num_tok_loss": 1689.0,
      "debug/num_tok_total": 2285.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.178051948547363,
      "train/diffusion_loss": 0.4075928330421448
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1792.0,
      "debug/num_lat_total": 2185.0,
      "debug/num_tok_loss": 1792.0,
      "debug/num_tok_total": 2185.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.185931205749512,
      "train/diffusion_loss": 0.43419307470321655
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1733.0,
      "debug/num_lat_total": 2252.0,
      "debug/num_tok_loss": 1733.0,
      "debug/num_tok_total": 2252.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.168244361877441,
      "train/diffusion_loss": 0.4210459291934967
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1751.0,
      "debug/num_lat_total": 2186.0,
      "debug/num_tok_loss": 1751.0,
      "debug/num_tok_total": 2186.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.163827896118164,
      "train/diffusion_loss": 0.43450531363487244
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1805.0,
      "debug/num_lat_total": 2312.0,
      "debug/num_tok_loss": 1805.0,
      "debug/num_tok_total": 2312.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.179530143737793,
      "train/diffusion_loss": 0.44190558791160583
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1823.0,
      "debug/num_lat_total": 2209.0,
      "debug/num_tok_loss": 1823.0,
      "debug/num_tok_total": 2209.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.182696342468262,
      "train/diffusion_loss": 0.49013906717300415
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1720.0,
      "debug/num_lat_total": 2257.0,
      "debug/num_tok_loss": 1720.0,
      "debug/num_tok_total": 2257.0,
      "epoch": 2.5981308411214954,
      "step": 18
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/ce_loss": 12.155743598937988,
      "train/diffusion_loss": 0.4087804853916168
    },
    {
      "epoch": 2.5981308411214954,
      "step": 18,
      "train/learning_rate_real": 1.3094773947796781e-05
    },
    {
      "debug/num_lat_loss": 1648.0,
      "debug/num_lat_total": 2245.0,
      "debug/num_tok_loss": 1648.0,
      "debug/num_tok_total": 2245.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.173835754394531,
      "train/diffusion_loss": 0.40596216917037964
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1770.0,
      "debug/num_lat_total": 2270.0,
      "debug/num_tok_loss": 1770.0,
      "debug/num_tok_total": 2270.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.166971206665039,
      "train/diffusion_loss": 0.41651174426078796
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1732.0,
      "debug/num_lat_total": 2314.0,
      "debug/num_tok_loss": 1732.0,
      "debug/num_tok_total": 2314.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.187857627868652,
      "train/diffusion_loss": 0.41560783982276917
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1764.0,
      "debug/num_lat_total": 2279.0,
      "debug/num_tok_loss": 1764.0,
      "debug/num_tok_total": 2279.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.170059204101562,
      "train/diffusion_loss": 0.4451139569282532
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1793.0,
      "debug/num_lat_total": 2246.0,
      "debug/num_tok_loss": 1793.0,
      "debug/num_tok_total": 2246.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.177019119262695,
      "train/diffusion_loss": 0.45224013924598694
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1628.0,
      "debug/num_lat_total": 2159.0,
      "debug/num_tok_loss": 1628.0,
      "debug/num_tok_total": 2159.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.135967254638672,
      "train/diffusion_loss": 0.4413244128227234
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1774.0,
      "debug/num_lat_total": 2277.0,
      "debug/num_tok_loss": 1774.0,
      "debug/num_tok_total": 2277.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.178704261779785,
      "train/diffusion_loss": 0.45805883407592773
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1695.0,
      "debug/num_lat_total": 2213.0,
      "debug/num_tok_loss": 1695.0,
      "debug/num_tok_total": 2213.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.163392066955566,
      "train/diffusion_loss": 0.4596419930458069
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1688.0,
      "debug/num_lat_total": 2197.0,
      "debug/num_tok_loss": 1688.0,
      "debug/num_tok_total": 2197.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.184797286987305,
      "train/diffusion_loss": 0.42093706130981445
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1703.0,
      "debug/num_lat_total": 2192.0,
      "debug/num_tok_loss": 1703.0,
      "debug/num_tok_total": 2192.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.184639930725098,
      "train/diffusion_loss": 0.40613827109336853
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1821.0,
      "debug/num_lat_total": 2138.0,
      "debug/num_tok_loss": 1821.0,
      "debug/num_tok_total": 2138.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.184473991394043,
      "train/diffusion_loss": 0.45566800236701965
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1733.0,
      "debug/num_lat_total": 2259.0,
      "debug/num_tok_loss": 1733.0,
      "debug/num_tok_total": 2259.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.151446342468262,
      "train/diffusion_loss": 0.44823265075683594
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1692.0,
      "debug/num_lat_total": 2014.0,
      "debug/num_tok_loss": 1692.0,
      "debug/num_tok_total": 2014.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.180910110473633,
      "train/diffusion_loss": 0.4702076315879822
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1776.0,
      "debug/num_lat_total": 2414.0,
      "debug/num_tok_loss": 1776.0,
      "debug/num_tok_total": 2414.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.170491218566895,
      "train/diffusion_loss": 0.3913854658603668
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1720.0,
      "debug/num_lat_total": 2192.0,
      "debug/num_tok_loss": 1720.0,
      "debug/num_tok_total": 2192.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.153451919555664,
      "train/diffusion_loss": 0.4243754744529724
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "debug/num_lat_loss": 1739.0,
      "debug/num_lat_total": 2216.0,
      "debug/num_tok_loss": 1739.0,
      "debug/num_tok_total": 2216.0,
      "epoch": 2.7476635514018692,
      "step": 19
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/ce_loss": 12.181028366088867,
      "train/diffusion_loss": 0.4464891254901886
    },
    {
      "epoch": 2.7476635514018692,
      "step": 19,
      "train/learning_rate_real": 1.1905226052203222e-05
    },
    {
      "epoch": 2.897196261682243,
      "grad_norm": 0.080078125,
      "learning_rate": 1.1905226052203222e-05,
      "loss": 1.1042,
      "step": 20
    },
    {
      "debug/num_lat_loss": 1720.0,
      "debug/num_lat_total": 1958.0,
      "debug/num_tok_loss": 1720.0,
      "debug/num_tok_total": 1958.0,
      "epoch": 2.897196261682243,
      "step": 20
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/ce_loss": 12.18447494506836,
      "train/diffusion_loss": 0.4715176522731781
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/learning_rate_real": 1.0721064521583937e-05
    },
    {
      "debug/num_lat_loss": 1738.0,
      "debug/num_lat_total": 2334.0,
      "debug/num_tok_loss": 1738.0,
      "debug/num_tok_total": 2334.0,
      "epoch": 2.897196261682243,
      "step": 20
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/ce_loss": 12.158917427062988,
      "train/diffusion_loss": 0.4330300986766815
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/learning_rate_real": 1.0721064521583937e-05
    },
    {
      "debug/num_lat_loss": 1704.0,
      "debug/num_lat_total": 2079.0,
      "debug/num_tok_loss": 1704.0,
      "debug/num_tok_total": 2079.0,
      "epoch": 2.897196261682243,
      "step": 20
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/ce_loss": 12.178497314453125,
      "train/diffusion_loss": 0.4486521780490875
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/learning_rate_real": 1.0721064521583937e-05
    },
    {
      "debug/num_lat_loss": 1756.0,
      "debug/num_lat_total": 2360.0,
      "debug/num_tok_loss": 1756.0,
      "debug/num_tok_total": 2360.0,
      "epoch": 2.897196261682243,
      "step": 20
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/ce_loss": 12.15247917175293,
      "train/diffusion_loss": 0.401742160320282
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/learning_rate_real": 1.0721064521583937e-05
    },
    {
      "debug/num_lat_loss": 1790.0,
      "debug/num_lat_total": 2236.0,
      "debug/num_tok_loss": 1790.0,
      "debug/num_tok_total": 2236.0,
      "epoch": 2.897196261682243,
      "step": 20
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/ce_loss": 12.17627239227295,
      "train/diffusion_loss": 0.4607166051864624
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/learning_rate_real": 1.0721064521583937e-05
    },
    {
      "debug/num_lat_loss": 1718.0,
      "debug/num_lat_total": 2358.0,
      "debug/num_tok_loss": 1718.0,
      "debug/num_tok_total": 2358.0,
      "epoch": 2.897196261682243,
      "step": 20
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/ce_loss": 12.17782974243164,
      "train/diffusion_loss": 0.41307130455970764
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/learning_rate_real": 1.0721064521583937e-05
    },
    {
      "debug/num_lat_loss": 1711.0,
      "debug/num_lat_total": 2199.0,
      "debug/num_tok_loss": 1711.0,
      "debug/num_tok_total": 2199.0,
      "epoch": 2.897196261682243,
      "step": 20
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/ce_loss": 12.14721393585205,
      "train/diffusion_loss": 0.4494905471801758
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/learning_rate_real": 1.0721064521583937e-05
    },
    {
      "debug/num_lat_loss": 1699.0,
      "debug/num_lat_total": 2313.0,
      "debug/num_tok_loss": 1699.0,
      "debug/num_tok_total": 2313.0,
      "epoch": 2.897196261682243,
      "step": 20
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/ce_loss": 12.170693397521973,
      "train/diffusion_loss": 0.41251811385154724
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/learning_rate_real": 1.0721064521583937e-05
    },
    {
      "debug/num_lat_loss": 1776.0,
      "debug/num_lat_total": 2357.0,
      "debug/num_tok_loss": 1776.0,
      "debug/num_tok_total": 2357.0,
      "epoch": 2.897196261682243,
      "step": 20
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/ce_loss": 12.161653518676758,
      "train/diffusion_loss": 0.42305028438568115
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/learning_rate_real": 1.0721064521583937e-05
    },
    {
      "debug/num_lat_loss": 1811.0,
      "debug/num_lat_total": 2179.0,
      "debug/num_tok_loss": 1811.0,
      "debug/num_tok_total": 2179.0,
      "epoch": 2.897196261682243,
      "step": 20
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/ce_loss": 12.158014297485352,
      "train/diffusion_loss": 0.4795665144920349
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/learning_rate_real": 1.0721064521583937e-05
    },
    {
      "debug/num_lat_loss": 198.0,
      "debug/num_lat_total": 285.0,
      "debug/num_tok_loss": 198.0,
      "debug/num_tok_total": 285.0,
      "epoch": 2.897196261682243,
      "step": 20
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/ce_loss": 12.160079956054688,
      "train/diffusion_loss": 0.43429556488990784
    },
    {
      "epoch": 2.897196261682243,
      "step": 20,
      "train/learning_rate_real": 1.0721064521583937e-05
    },
    {
      "debug/num_lat_loss": 1731.0,
      "debug/num_lat_total": 2029.0,
      "debug/num_tok_loss": 1731.0,
      "debug/num_tok_total": 2029.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.185141563415527,
      "train/diffusion_loss": 0.47871869802474976
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1705.0,
      "debug/num_lat_total": 2217.0,
      "debug/num_tok_loss": 1705.0,
      "debug/num_tok_total": 2217.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.161572456359863,
      "train/diffusion_loss": 0.42034080624580383
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1708.0,
      "debug/num_lat_total": 2155.0,
      "debug/num_tok_loss": 1708.0,
      "debug/num_tok_total": 2155.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.176295280456543,
      "train/diffusion_loss": 0.43198931217193604
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1724.0,
      "debug/num_lat_total": 2164.0,
      "debug/num_tok_loss": 1724.0,
      "debug/num_tok_total": 2164.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.164573669433594,
      "train/diffusion_loss": 0.49672797322273254
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1785.0,
      "debug/num_lat_total": 2239.0,
      "debug/num_tok_loss": 1785.0,
      "debug/num_tok_total": 2239.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.175557136535645,
      "train/diffusion_loss": 0.43289849162101746
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1715.0,
      "debug/num_lat_total": 2117.0,
      "debug/num_tok_loss": 1715.0,
      "debug/num_tok_total": 2117.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.159709930419922,
      "train/diffusion_loss": 0.4435068964958191
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1713.0,
      "debug/num_lat_total": 1979.0,
      "debug/num_tok_loss": 1713.0,
      "debug/num_tok_total": 1979.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.173222541809082,
      "train/diffusion_loss": 0.4639737010002136
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1692.0,
      "debug/num_lat_total": 2159.0,
      "debug/num_tok_loss": 1692.0,
      "debug/num_tok_total": 2159.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.170083045959473,
      "train/diffusion_loss": 0.4349198639392853
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1748.0,
      "debug/num_lat_total": 2184.0,
      "debug/num_tok_loss": 1748.0,
      "debug/num_tok_total": 2184.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.192635536193848,
      "train/diffusion_loss": 0.4608837068080902
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1773.0,
      "debug/num_lat_total": 2210.0,
      "debug/num_tok_loss": 1773.0,
      "debug/num_tok_total": 2210.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.185585975646973,
      "train/diffusion_loss": 0.44979315996170044
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1739.0,
      "debug/num_lat_total": 2259.0,
      "debug/num_tok_loss": 1739.0,
      "debug/num_tok_total": 2259.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.171687126159668,
      "train/diffusion_loss": 0.4464608430862427
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1703.0,
      "debug/num_lat_total": 2137.0,
      "debug/num_tok_loss": 1703.0,
      "debug/num_tok_total": 2137.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.194785118103027,
      "train/diffusion_loss": 0.4453159272670746
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1741.0,
      "debug/num_lat_total": 2147.0,
      "debug/num_tok_loss": 1741.0,
      "debug/num_tok_total": 2147.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.135725021362305,
      "train/diffusion_loss": 0.4403951168060303
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1685.0,
      "debug/num_lat_total": 2181.0,
      "debug/num_tok_loss": 1685.0,
      "debug/num_tok_total": 2181.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.180353164672852,
      "train/diffusion_loss": 0.4038219153881073
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1712.0,
      "debug/num_lat_total": 2147.0,
      "debug/num_tok_loss": 1712.0,
      "debug/num_tok_total": 2147.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.182561874389648,
      "train/diffusion_loss": 0.44825613498687744
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1772.0,
      "debug/num_lat_total": 2275.0,
      "debug/num_tok_loss": 1772.0,
      "debug/num_tok_total": 2275.0,
      "epoch": 3.0,
      "step": 21
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/ce_loss": 12.186759948730469,
      "train/diffusion_loss": 0.42617136240005493
    },
    {
      "epoch": 3.0,
      "step": 21,
      "train/learning_rate_real": 9.553013306132158e-06
    },
    {
      "debug/num_lat_loss": 1688.0,
      "debug/num_lat_total": 1989.0,
      "debug/num_tok_loss": 1688.0,
      "debug/num_tok_total": 1989.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.166876792907715,
      "train/diffusion_loss": 0.4436281621456146
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1663.0,
      "debug/num_lat_total": 2146.0,
      "debug/num_tok_loss": 1663.0,
      "debug/num_tok_total": 2146.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.181292533874512,
      "train/diffusion_loss": 0.43562567234039307
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1733.0,
      "debug/num_lat_total": 2199.0,
      "debug/num_tok_loss": 1733.0,
      "debug/num_tok_total": 2199.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.206137657165527,
      "train/diffusion_loss": 0.4738239347934723
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1697.0,
      "debug/num_lat_total": 2169.0,
      "debug/num_tok_loss": 1697.0,
      "debug/num_tok_total": 2169.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.14698314666748,
      "train/diffusion_loss": 0.41713082790374756
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1868.0,
      "debug/num_lat_total": 2287.0,
      "debug/num_tok_loss": 1868.0,
      "debug/num_tok_total": 2287.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.143259048461914,
      "train/diffusion_loss": 0.4566551744937897
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1811.0,
      "debug/num_lat_total": 2433.0,
      "debug/num_tok_loss": 1811.0,
      "debug/num_tok_total": 2433.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.159565925598145,
      "train/diffusion_loss": 0.41677698493003845
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1716.0,
      "debug/num_lat_total": 2281.0,
      "debug/num_tok_loss": 1716.0,
      "debug/num_tok_total": 2281.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.1969575881958,
      "train/diffusion_loss": 0.45901116728782654
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1741.0,
      "debug/num_lat_total": 2128.0,
      "debug/num_tok_loss": 1741.0,
      "debug/num_tok_total": 2128.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.142516136169434,
      "train/diffusion_loss": 0.4245382249355316
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1732.0,
      "debug/num_lat_total": 2290.0,
      "debug/num_tok_loss": 1732.0,
      "debug/num_tok_total": 2290.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.163461685180664,
      "train/diffusion_loss": 0.41974490880966187
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1673.0,
      "debug/num_lat_total": 2143.0,
      "debug/num_tok_loss": 1673.0,
      "debug/num_tok_total": 2143.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.204856872558594,
      "train/diffusion_loss": 0.4260376989841461
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1832.0,
      "debug/num_lat_total": 2448.0,
      "debug/num_tok_loss": 1832.0,
      "debug/num_tok_total": 2448.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.169397354125977,
      "train/diffusion_loss": 0.444273442029953
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1746.0,
      "debug/num_lat_total": 2292.0,
      "debug/num_tok_loss": 1746.0,
      "debug/num_tok_total": 2292.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.207819938659668,
      "train/diffusion_loss": 0.42534464597702026
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1724.0,
      "debug/num_lat_total": 2197.0,
      "debug/num_tok_loss": 1724.0,
      "debug/num_tok_total": 2197.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.198328971862793,
      "train/diffusion_loss": 0.42757922410964966
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1890.0,
      "debug/num_lat_total": 2202.0,
      "debug/num_tok_loss": 1890.0,
      "debug/num_tok_total": 2202.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.175276756286621,
      "train/diffusion_loss": 0.44557538628578186
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1775.0,
      "debug/num_lat_total": 2346.0,
      "debug/num_tok_loss": 1775.0,
      "debug/num_tok_total": 2346.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.125227928161621,
      "train/diffusion_loss": 0.4252835214138031
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1765.0,
      "debug/num_lat_total": 2221.0,
      "debug/num_tok_loss": 1765.0,
      "debug/num_tok_total": 2221.0,
      "epoch": 3.149532710280374,
      "step": 22
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/ce_loss": 12.168408393859863,
      "train/diffusion_loss": 0.4143877625465393
    },
    {
      "epoch": 3.149532710280374,
      "step": 22,
      "train/learning_rate_real": 8.41165045853223e-06
    },
    {
      "debug/num_lat_loss": 1774.0,
      "debug/num_lat_total": 2312.0,
      "debug/num_tok_loss": 1774.0,
      "debug/num_tok_total": 2312.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.194767951965332,
      "train/diffusion_loss": 0.40749505162239075
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1748.0,
      "debug/num_lat_total": 2337.0,
      "debug/num_tok_loss": 1748.0,
      "debug/num_tok_total": 2337.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.180062294006348,
      "train/diffusion_loss": 0.45502886176109314
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1764.0,
      "debug/num_lat_total": 2187.0,
      "debug/num_tok_loss": 1764.0,
      "debug/num_tok_total": 2187.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.19076919555664,
      "train/diffusion_loss": 0.4535031318664551
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1728.0,
      "debug/num_lat_total": 2437.0,
      "debug/num_tok_loss": 1728.0,
      "debug/num_tok_total": 2437.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.163100242614746,
      "train/diffusion_loss": 0.4055667221546173
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1738.0,
      "debug/num_lat_total": 2215.0,
      "debug/num_tok_loss": 1738.0,
      "debug/num_tok_total": 2215.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.180434226989746,
      "train/diffusion_loss": 0.42948290705680847
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1705.0,
      "debug/num_lat_total": 2159.0,
      "debug/num_tok_loss": 1705.0,
      "debug/num_tok_total": 2159.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.177447319030762,
      "train/diffusion_loss": 0.4368229806423187
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1677.0,
      "debug/num_lat_total": 1909.0,
      "debug/num_tok_loss": 1677.0,
      "debug/num_tok_total": 1909.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.212713241577148,
      "train/diffusion_loss": 0.4799862802028656
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1807.0,
      "debug/num_lat_total": 2175.0,
      "debug/num_tok_loss": 1807.0,
      "debug/num_tok_total": 2175.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.173916816711426,
      "train/diffusion_loss": 0.47795355319976807
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1762.0,
      "debug/num_lat_total": 2179.0,
      "debug/num_tok_loss": 1762.0,
      "debug/num_tok_total": 2179.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.16403865814209,
      "train/diffusion_loss": 0.43208783864974976
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1737.0,
      "debug/num_lat_total": 2147.0,
      "debug/num_tok_loss": 1737.0,
      "debug/num_tok_total": 2147.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.195341110229492,
      "train/diffusion_loss": 0.42608267068862915
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1694.0,
      "debug/num_lat_total": 2095.0,
      "debug/num_tok_loss": 1694.0,
      "debug/num_tok_total": 2095.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.224993705749512,
      "train/diffusion_loss": 0.4705503582954407
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1715.0,
      "debug/num_lat_total": 2151.0,
      "debug/num_tok_loss": 1715.0,
      "debug/num_tok_total": 2151.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.196223258972168,
      "train/diffusion_loss": 0.44091811776161194
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1663.0,
      "debug/num_lat_total": 2059.0,
      "debug/num_tok_loss": 1663.0,
      "debug/num_tok_total": 2059.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.171528816223145,
      "train/diffusion_loss": 0.47177937626838684
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1700.0,
      "debug/num_lat_total": 2117.0,
      "debug/num_tok_loss": 1700.0,
      "debug/num_tok_total": 2117.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.132240295410156,
      "train/diffusion_loss": 0.4833270311355591
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1800.0,
      "debug/num_lat_total": 2179.0,
      "debug/num_tok_loss": 1800.0,
      "debug/num_tok_total": 2179.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.171632766723633,
      "train/diffusion_loss": 0.4560210704803467
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1743.0,
      "debug/num_lat_total": 2176.0,
      "debug/num_tok_loss": 1743.0,
      "debug/num_tok_total": 2176.0,
      "epoch": 3.2990654205607477,
      "step": 23
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/ce_loss": 12.158903121948242,
      "train/diffusion_loss": 0.43445152044296265
    },
    {
      "epoch": 3.2990654205607477,
      "step": 23,
      "train/learning_rate_real": 7.307312337476421e-06
    },
    {
      "debug/num_lat_loss": 1815.0,
      "debug/num_lat_total": 2252.0,
      "debug/num_tok_loss": 1815.0,
      "debug/num_tok_total": 2252.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.207298278808594,
      "train/diffusion_loss": 0.4564967453479767
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1737.0,
      "debug/num_lat_total": 2213.0,
      "debug/num_tok_loss": 1737.0,
      "debug/num_tok_total": 2213.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.177847862243652,
      "train/diffusion_loss": 0.4450188875198364
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1676.0,
      "debug/num_lat_total": 2186.0,
      "debug/num_tok_loss": 1676.0,
      "debug/num_tok_total": 2186.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.1792631149292,
      "train/diffusion_loss": 0.4131785035133362
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1822.0,
      "debug/num_lat_total": 2243.0,
      "debug/num_tok_loss": 1822.0,
      "debug/num_tok_total": 2243.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.178991317749023,
      "train/diffusion_loss": 0.46926724910736084
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1668.0,
      "debug/num_lat_total": 2087.0,
      "debug/num_tok_loss": 1668.0,
      "debug/num_tok_total": 2087.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.16328239440918,
      "train/diffusion_loss": 0.46612855792045593
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1724.0,
      "debug/num_lat_total": 2304.0,
      "debug/num_tok_loss": 1724.0,
      "debug/num_tok_total": 2304.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.163469314575195,
      "train/diffusion_loss": 0.40386247634887695
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1803.0,
      "debug/num_lat_total": 2119.0,
      "debug/num_tok_loss": 1803.0,
      "debug/num_tok_total": 2119.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.176862716674805,
      "train/diffusion_loss": 0.4733399748802185
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1701.0,
      "debug/num_lat_total": 2097.0,
      "debug/num_tok_loss": 1701.0,
      "debug/num_tok_total": 2097.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.180752754211426,
      "train/diffusion_loss": 0.4555293917655945
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1705.0,
      "debug/num_lat_total": 2041.0,
      "debug/num_tok_loss": 1705.0,
      "debug/num_tok_total": 2041.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.19352912902832,
      "train/diffusion_loss": 0.4499252736568451
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1749.0,
      "debug/num_lat_total": 2035.0,
      "debug/num_tok_loss": 1749.0,
      "debug/num_tok_total": 2035.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.191561698913574,
      "train/diffusion_loss": 0.438742995262146
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1681.0,
      "debug/num_lat_total": 2014.0,
      "debug/num_tok_loss": 1681.0,
      "debug/num_tok_total": 2014.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.201146125793457,
      "train/diffusion_loss": 0.4450303614139557
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1753.0,
      "debug/num_lat_total": 2150.0,
      "debug/num_tok_loss": 1753.0,
      "debug/num_tok_total": 2150.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.183883666992188,
      "train/diffusion_loss": 0.43293264508247375
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1733.0,
      "debug/num_lat_total": 2146.0,
      "debug/num_tok_loss": 1733.0,
      "debug/num_tok_total": 2146.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.142557144165039,
      "train/diffusion_loss": 0.43829670548439026
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1722.0,
      "debug/num_lat_total": 2148.0,
      "debug/num_tok_loss": 1722.0,
      "debug/num_tok_total": 2148.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.163311004638672,
      "train/diffusion_loss": 0.45310938358306885
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1698.0,
      "debug/num_lat_total": 2301.0,
      "debug/num_tok_loss": 1698.0,
      "debug/num_tok_total": 2301.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.16663932800293,
      "train/diffusion_loss": 0.3920009434223175
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1745.0,
      "debug/num_lat_total": 2103.0,
      "debug/num_tok_loss": 1745.0,
      "debug/num_tok_total": 2103.0,
      "epoch": 3.4485981308411215,
      "step": 24
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/ce_loss": 12.189055442810059,
      "train/diffusion_loss": 0.45429617166519165
    },
    {
      "epoch": 3.4485981308411215,
      "step": 24,
      "train/learning_rate_real": 6.250000000000003e-06
    },
    {
      "debug/num_lat_loss": 1793.0,
      "debug/num_lat_total": 2362.0,
      "debug/num_tok_loss": 1793.0,
      "debug/num_tok_total": 2362.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.192378044128418,
      "train/diffusion_loss": 0.4308478832244873
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1769.0,
      "debug/num_lat_total": 2132.0,
      "debug/num_tok_loss": 1769.0,
      "debug/num_tok_total": 2132.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.206249237060547,
      "train/diffusion_loss": 0.476319283246994
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1816.0,
      "debug/num_lat_total": 2395.0,
      "debug/num_tok_loss": 1816.0,
      "debug/num_tok_total": 2395.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.17150592803955,
      "train/diffusion_loss": 0.41233769059181213
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1775.0,
      "debug/num_lat_total": 2214.0,
      "debug/num_tok_loss": 1775.0,
      "debug/num_tok_total": 2214.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.180362701416016,
      "train/diffusion_loss": 0.45229336619377136
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1725.0,
      "debug/num_lat_total": 2213.0,
      "debug/num_tok_loss": 1725.0,
      "debug/num_tok_total": 2213.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.19162654876709,
      "train/diffusion_loss": 0.41457587480545044
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1696.0,
      "debug/num_lat_total": 2192.0,
      "debug/num_tok_loss": 1696.0,
      "debug/num_tok_total": 2192.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.148582458496094,
      "train/diffusion_loss": 0.4307917654514313
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1758.0,
      "debug/num_lat_total": 2080.0,
      "debug/num_tok_loss": 1758.0,
      "debug/num_tok_total": 2080.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.170329093933105,
      "train/diffusion_loss": 0.4580996632575989
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1674.0,
      "debug/num_lat_total": 2183.0,
      "debug/num_tok_loss": 1674.0,
      "debug/num_tok_total": 2183.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.19109058380127,
      "train/diffusion_loss": 0.42841607332229614
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1698.0,
      "debug/num_lat_total": 2182.0,
      "debug/num_tok_loss": 1698.0,
      "debug/num_tok_total": 2182.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.186971664428711,
      "train/diffusion_loss": 0.42583712935447693
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1726.0,
      "debug/num_lat_total": 2186.0,
      "debug/num_tok_loss": 1726.0,
      "debug/num_tok_total": 2186.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.18460750579834,
      "train/diffusion_loss": 0.4379256069660187
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1706.0,
      "debug/num_lat_total": 2120.0,
      "debug/num_tok_loss": 1706.0,
      "debug/num_tok_total": 2120.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.156734466552734,
      "train/diffusion_loss": 0.42288535833358765
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1783.0,
      "debug/num_lat_total": 2324.0,
      "debug/num_tok_loss": 1783.0,
      "debug/num_tok_total": 2324.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.204289436340332,
      "train/diffusion_loss": 0.41219431161880493
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1789.0,
      "debug/num_lat_total": 2248.0,
      "debug/num_tok_loss": 1789.0,
      "debug/num_tok_total": 2248.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.165375709533691,
      "train/diffusion_loss": 0.4418776035308838
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1701.0,
      "debug/num_lat_total": 2123.0,
      "debug/num_tok_loss": 1701.0,
      "debug/num_tok_total": 2123.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.188913345336914,
      "train/diffusion_loss": 0.4265272319316864
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1703.0,
      "debug/num_lat_total": 2134.0,
      "debug/num_tok_loss": 1703.0,
      "debug/num_tok_total": 2134.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.143837928771973,
      "train/diffusion_loss": 0.4750172793865204
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1757.0,
      "debug/num_lat_total": 2405.0,
      "debug/num_tok_loss": 1757.0,
      "debug/num_tok_total": 2405.0,
      "epoch": 3.5981308411214954,
      "step": 25
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/ce_loss": 12.149174690246582,
      "train/diffusion_loss": 0.43947383761405945
    },
    {
      "epoch": 3.5981308411214954,
      "step": 25,
      "train/learning_rate_real": 5.249288630360025e-06
    },
    {
      "debug/num_lat_loss": 1717.0,
      "debug/num_lat_total": 2117.0,
      "debug/num_tok_loss": 1717.0,
      "debug/num_tok_total": 2117.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.182785987854004,
      "train/diffusion_loss": 0.43974003195762634
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1714.0,
      "debug/num_lat_total": 2123.0,
      "debug/num_tok_loss": 1714.0,
      "debug/num_tok_total": 2123.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.178577423095703,
      "train/diffusion_loss": 0.4370800256729126
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1674.0,
      "debug/num_lat_total": 2086.0,
      "debug/num_tok_loss": 1674.0,
      "debug/num_tok_total": 2086.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.184208869934082,
      "train/diffusion_loss": 0.4335319697856903
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1707.0,
      "debug/num_lat_total": 2047.0,
      "debug/num_tok_loss": 1707.0,
      "debug/num_tok_total": 2047.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.191585540771484,
      "train/diffusion_loss": 0.4503075182437897
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1789.0,
      "debug/num_lat_total": 2288.0,
      "debug/num_tok_loss": 1789.0,
      "debug/num_tok_total": 2288.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.189672470092773,
      "train/diffusion_loss": 0.4307839572429657
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1781.0,
      "debug/num_lat_total": 2118.0,
      "debug/num_tok_loss": 1781.0,
      "debug/num_tok_total": 2118.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.171289443969727,
      "train/diffusion_loss": 0.45949694514274597
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1749.0,
      "debug/num_lat_total": 2243.0,
      "debug/num_tok_loss": 1749.0,
      "debug/num_tok_total": 2243.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.190430641174316,
      "train/diffusion_loss": 0.4002898335456848
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1762.0,
      "debug/num_lat_total": 1911.0,
      "debug/num_tok_loss": 1762.0,
      "debug/num_tok_total": 1911.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.152243614196777,
      "train/diffusion_loss": 0.48624396324157715
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1709.0,
      "debug/num_lat_total": 2202.0,
      "debug/num_tok_loss": 1709.0,
      "debug/num_tok_total": 2202.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.171443939208984,
      "train/diffusion_loss": 0.43149277567863464
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1801.0,
      "debug/num_lat_total": 2253.0,
      "debug/num_tok_loss": 1801.0,
      "debug/num_tok_total": 2253.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.186466217041016,
      "train/diffusion_loss": 0.4318007826805115
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1727.0,
      "debug/num_lat_total": 2297.0,
      "debug/num_tok_loss": 1727.0,
      "debug/num_tok_total": 2297.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.161050796508789,
      "train/diffusion_loss": 0.39592570066452026
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1728.0,
      "debug/num_lat_total": 2203.0,
      "debug/num_tok_loss": 1728.0,
      "debug/num_tok_total": 2203.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.172682762145996,
      "train/diffusion_loss": 0.4314633011817932
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1729.0,
      "debug/num_lat_total": 2218.0,
      "debug/num_tok_loss": 1729.0,
      "debug/num_tok_total": 2218.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.163350105285645,
      "train/diffusion_loss": 0.4142179489135742
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1807.0,
      "debug/num_lat_total": 2361.0,
      "debug/num_tok_loss": 1807.0,
      "debug/num_tok_total": 2361.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.161543846130371,
      "train/diffusion_loss": 0.43518614768981934
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1745.0,
      "debug/num_lat_total": 2152.0,
      "debug/num_tok_loss": 1745.0,
      "debug/num_tok_total": 2152.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.140242576599121,
      "train/diffusion_loss": 0.43826156854629517
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1799.0,
      "debug/num_lat_total": 2113.0,
      "debug/num_tok_loss": 1799.0,
      "debug/num_tok_total": 2113.0,
      "epoch": 3.7476635514018692,
      "step": 26
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/ce_loss": 12.205531120300293,
      "train/diffusion_loss": 0.4470859467983246
    },
    {
      "epoch": 3.7476635514018692,
      "step": 26,
      "train/learning_rate_real": 4.314240825683938e-06
    },
    {
      "debug/num_lat_loss": 1771.0,
      "debug/num_lat_total": 2168.0,
      "debug/num_tok_loss": 1771.0,
      "debug/num_tok_total": 2168.0,
      "epoch": 3.897196261682243,
      "step": 27
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/ce_loss": 12.175880432128906,
      "train/diffusion_loss": 0.4578666090965271
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/learning_rate_real": 3.453324523686623e-06
    },
    {
      "debug/num_lat_loss": 1747.0,
      "debug/num_lat_total": 2183.0,
      "debug/num_tok_loss": 1747.0,
      "debug/num_tok_total": 2183.0,
      "epoch": 3.897196261682243,
      "step": 27
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/ce_loss": 12.155753135681152,
      "train/diffusion_loss": 0.4481011629104614
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/learning_rate_real": 3.453324523686623e-06
    },
    {
      "debug/num_lat_loss": 1747.0,
      "debug/num_lat_total": 2105.0,
      "debug/num_tok_loss": 1747.0,
      "debug/num_tok_total": 2105.0,
      "epoch": 3.897196261682243,
      "step": 27
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/ce_loss": 12.197052955627441,
      "train/diffusion_loss": 0.43395644426345825
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/learning_rate_real": 3.453324523686623e-06
    },
    {
      "debug/num_lat_loss": 1786.0,
      "debug/num_lat_total": 2220.0,
      "debug/num_tok_loss": 1786.0,
      "debug/num_tok_total": 2220.0,
      "epoch": 3.897196261682243,
      "step": 27
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/ce_loss": 12.178627014160156,
      "train/diffusion_loss": 0.44644054770469666
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/learning_rate_real": 3.453324523686623e-06
    },
    {
      "debug/num_lat_loss": 1617.0,
      "debug/num_lat_total": 1821.0,
      "debug/num_tok_loss": 1617.0,
      "debug/num_tok_total": 1821.0,
      "epoch": 3.897196261682243,
      "step": 27
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/ce_loss": 12.175630569458008,
      "train/diffusion_loss": 0.45170465111732483
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/learning_rate_real": 3.453324523686623e-06
    },
    {
      "debug/num_lat_loss": 1779.0,
      "debug/num_lat_total": 2263.0,
      "debug/num_tok_loss": 1779.0,
      "debug/num_tok_total": 2263.0,
      "epoch": 3.897196261682243,
      "step": 27
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/ce_loss": 12.189778327941895,
      "train/diffusion_loss": 0.4389124810695648
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/learning_rate_real": 3.453324523686623e-06
    },
    {
      "debug/num_lat_loss": 1782.0,
      "debug/num_lat_total": 2202.0,
      "debug/num_tok_loss": 1782.0,
      "debug/num_tok_total": 2202.0,
      "epoch": 3.897196261682243,
      "step": 27
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/ce_loss": 12.168231964111328,
      "train/diffusion_loss": 0.4539552628993988
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/learning_rate_real": 3.453324523686623e-06
    },
    {
      "debug/num_lat_loss": 1722.0,
      "debug/num_lat_total": 2049.0,
      "debug/num_tok_loss": 1722.0,
      "debug/num_tok_total": 2049.0,
      "epoch": 3.897196261682243,
      "step": 27
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/ce_loss": 12.237095832824707,
      "train/diffusion_loss": 0.45853713154792786
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/learning_rate_real": 3.453324523686623e-06
    },
    {
      "debug/num_lat_loss": 1898.0,
      "debug/num_lat_total": 2291.0,
      "debug/num_tok_loss": 1898.0,
      "debug/num_tok_total": 2291.0,
      "epoch": 3.897196261682243,
      "step": 27
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/ce_loss": 12.171428680419922,
      "train/diffusion_loss": 0.44604358077049255
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/learning_rate_real": 3.453324523686623e-06
    },
    {
      "debug/num_lat_loss": 1667.0,
      "debug/num_lat_total": 2193.0,
      "debug/num_tok_loss": 1667.0,
      "debug/num_tok_total": 2193.0,
      "epoch": 3.897196261682243,
      "step": 27
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/ce_loss": 12.138124465942383,
      "train/diffusion_loss": 0.4041489362716675
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/learning_rate_real": 3.453324523686623e-06
    },
    {
      "debug/num_lat_loss": 210.0,
      "debug/num_lat_total": 260.0,
      "debug/num_tok_loss": 210.0,
      "debug/num_tok_total": 260.0,
      "epoch": 3.897196261682243,
      "step": 27
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/ce_loss": 12.142877578735352,
      "train/diffusion_loss": 0.5250693559646606
    },
    {
      "epoch": 3.897196261682243,
      "step": 27,
      "train/learning_rate_real": 3.453324523686623e-06
    },
    {
      "debug/num_lat_loss": 1679.0,
      "debug/num_lat_total": 2145.0,
      "debug/num_tok_loss": 1679.0,
      "debug/num_tok_total": 2145.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.175436019897461,
      "train/diffusion_loss": 0.41446301341056824
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1853.0,
      "debug/num_lat_total": 2411.0,
      "debug/num_tok_loss": 1853.0,
      "debug/num_tok_total": 2411.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.186768531799316,
      "train/diffusion_loss": 0.41826552152633667
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1724.0,
      "debug/num_lat_total": 2208.0,
      "debug/num_tok_loss": 1724.0,
      "debug/num_tok_total": 2208.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.172440528869629,
      "train/diffusion_loss": 0.4254646599292755
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1725.0,
      "debug/num_lat_total": 2244.0,
      "debug/num_tok_loss": 1725.0,
      "debug/num_tok_total": 2244.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.202057838439941,
      "train/diffusion_loss": 0.4286242425441742
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1759.0,
      "debug/num_lat_total": 2305.0,
      "debug/num_tok_loss": 1759.0,
      "debug/num_tok_total": 2305.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.16329288482666,
      "train/diffusion_loss": 0.4380427598953247
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1662.0,
      "debug/num_lat_total": 2193.0,
      "debug/num_tok_loss": 1662.0,
      "debug/num_tok_total": 2193.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.1865816116333,
      "train/diffusion_loss": 0.4253368377685547
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1799.0,
      "debug/num_lat_total": 2314.0,
      "debug/num_tok_loss": 1799.0,
      "debug/num_tok_total": 2314.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.162942886352539,
      "train/diffusion_loss": 0.4381959140300751
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1733.0,
      "debug/num_lat_total": 2279.0,
      "debug/num_tok_loss": 1733.0,
      "debug/num_tok_total": 2279.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.188685417175293,
      "train/diffusion_loss": 0.4371460974216461
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1711.0,
      "debug/num_lat_total": 2178.0,
      "debug/num_tok_loss": 1711.0,
      "debug/num_tok_total": 2178.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.155284881591797,
      "train/diffusion_loss": 0.464201956987381
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1752.0,
      "debug/num_lat_total": 2227.0,
      "debug/num_tok_loss": 1752.0,
      "debug/num_tok_total": 2227.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.149553298950195,
      "train/diffusion_loss": 0.4439617693424225
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1704.0,
      "debug/num_lat_total": 2086.0,
      "debug/num_tok_loss": 1704.0,
      "debug/num_tok_total": 2086.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.204044342041016,
      "train/diffusion_loss": 0.44241979718208313
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1713.0,
      "debug/num_lat_total": 2119.0,
      "debug/num_tok_loss": 1713.0,
      "debug/num_tok_total": 2119.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.202098846435547,
      "train/diffusion_loss": 0.44500264525413513
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1720.0,
      "debug/num_lat_total": 2300.0,
      "debug/num_tok_loss": 1720.0,
      "debug/num_tok_total": 2300.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.154370307922363,
      "train/diffusion_loss": 0.4311031103134155
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1731.0,
      "debug/num_lat_total": 2194.0,
      "debug/num_tok_loss": 1731.0,
      "debug/num_tok_total": 2194.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.157872200012207,
      "train/diffusion_loss": 0.44460704922676086
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1806.0,
      "debug/num_lat_total": 2261.0,
      "debug/num_tok_loss": 1806.0,
      "debug/num_tok_total": 2261.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.186095237731934,
      "train/diffusion_loss": 0.4375755488872528
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1728.0,
      "debug/num_lat_total": 2022.0,
      "debug/num_tok_loss": 1728.0,
      "debug/num_tok_total": 2022.0,
      "epoch": 4.0,
      "step": 28
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/ce_loss": 12.171619415283203,
      "train/diffusion_loss": 0.4596668481826782
    },
    {
      "epoch": 4.0,
      "step": 28,
      "train/learning_rate_real": 2.674336315715159e-06
    },
    {
      "debug/num_lat_loss": 1674.0,
      "debug/num_lat_total": 2120.0,
      "debug/num_tok_loss": 1674.0,
      "debug/num_tok_total": 2120.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.189408302307129,
      "train/diffusion_loss": 0.4574336111545563
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1790.0,
      "debug/num_lat_total": 2237.0,
      "debug/num_tok_loss": 1790.0,
      "debug/num_tok_total": 2237.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.166314125061035,
      "train/diffusion_loss": 0.4323076903820038
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1812.0,
      "debug/num_lat_total": 2238.0,
      "debug/num_tok_loss": 1812.0,
      "debug/num_tok_total": 2238.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.176814079284668,
      "train/diffusion_loss": 0.42043188214302063
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1780.0,
      "debug/num_lat_total": 2107.0,
      "debug/num_tok_loss": 1780.0,
      "debug/num_tok_total": 2107.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.157387733459473,
      "train/diffusion_loss": 0.4889250695705414
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1831.0,
      "debug/num_lat_total": 2428.0,
      "debug/num_tok_loss": 1831.0,
      "debug/num_tok_total": 2428.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.166900634765625,
      "train/diffusion_loss": 0.4233861565589905
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1804.0,
      "debug/num_lat_total": 2144.0,
      "debug/num_tok_loss": 1804.0,
      "debug/num_tok_total": 2144.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.19821834564209,
      "train/diffusion_loss": 0.4677843153476715
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1752.0,
      "debug/num_lat_total": 2197.0,
      "debug/num_tok_loss": 1752.0,
      "debug/num_tok_total": 2197.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.170734405517578,
      "train/diffusion_loss": 0.4539377987384796
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1783.0,
      "debug/num_lat_total": 2446.0,
      "debug/num_tok_loss": 1783.0,
      "debug/num_tok_total": 2446.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.198660850524902,
      "train/diffusion_loss": 0.40065276622772217
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1713.0,
      "debug/num_lat_total": 2128.0,
      "debug/num_tok_loss": 1713.0,
      "debug/num_tok_total": 2128.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.169668197631836,
      "train/diffusion_loss": 0.4404948949813843
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1655.0,
      "debug/num_lat_total": 2195.0,
      "debug/num_tok_loss": 1655.0,
      "debug/num_tok_total": 2195.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.177712440490723,
      "train/diffusion_loss": 0.3918401896953583
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1730.0,
      "debug/num_lat_total": 2218.0,
      "debug/num_tok_loss": 1730.0,
      "debug/num_tok_total": 2218.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.167113304138184,
      "train/diffusion_loss": 0.4331023395061493
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1738.0,
      "debug/num_lat_total": 2119.0,
      "debug/num_tok_loss": 1738.0,
      "debug/num_tok_total": 2119.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.170269966125488,
      "train/diffusion_loss": 0.43736377358436584
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1736.0,
      "debug/num_lat_total": 2199.0,
      "debug/num_tok_loss": 1736.0,
      "debug/num_tok_total": 2199.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.155172348022461,
      "train/diffusion_loss": 0.42950916290283203
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1665.0,
      "debug/num_lat_total": 2159.0,
      "debug/num_tok_loss": 1665.0,
      "debug/num_tok_total": 2159.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.158082962036133,
      "train/diffusion_loss": 0.4130508303642273
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1736.0,
      "debug/num_lat_total": 2237.0,
      "debug/num_tok_loss": 1736.0,
      "debug/num_tok_total": 2237.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.169498443603516,
      "train/diffusion_loss": 0.453546404838562
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "debug/num_lat_loss": 1689.0,
      "debug/num_lat_total": 2065.0,
      "debug/num_tok_loss": 1689.0,
      "debug/num_tok_total": 2065.0,
      "epoch": 4.149532710280374,
      "step": 29
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/ce_loss": 12.140535354614258,
      "train/diffusion_loss": 0.46471476554870605
    },
    {
      "epoch": 4.149532710280374,
      "step": 29,
      "train/learning_rate_real": 1.984330839610234e-06
    },
    {
      "epoch": 4.299065420560748,
      "grad_norm": 0.07275390625,
      "learning_rate": 1.984330839610234e-06,
      "loss": 1.1041,
      "step": 30
    },
    {
      "debug/num_lat_loss": 1752.0,
      "debug/num_lat_total": 2173.0,
      "debug/num_tok_loss": 1752.0,
      "debug/num_tok_total": 2173.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.145671844482422,
      "train/diffusion_loss": 0.4525708556175232
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1755.0,
      "debug/num_lat_total": 1993.0,
      "debug/num_tok_loss": 1755.0,
      "debug/num_tok_total": 1993.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.190377235412598,
      "train/diffusion_loss": 0.46880972385406494
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1746.0,
      "debug/num_lat_total": 2238.0,
      "debug/num_tok_loss": 1746.0,
      "debug/num_tok_total": 2238.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.153491020202637,
      "train/diffusion_loss": 0.44733402132987976
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1749.0,
      "debug/num_lat_total": 2220.0,
      "debug/num_tok_loss": 1749.0,
      "debug/num_tok_total": 2220.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.16679859161377,
      "train/diffusion_loss": 0.4347849488258362
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1705.0,
      "debug/num_lat_total": 2243.0,
      "debug/num_tok_loss": 1705.0,
      "debug/num_tok_total": 2243.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.193336486816406,
      "train/diffusion_loss": 0.42132002115249634
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1671.0,
      "debug/num_lat_total": 2218.0,
      "debug/num_tok_loss": 1671.0,
      "debug/num_tok_total": 2218.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.164517402648926,
      "train/diffusion_loss": 0.41408732533454895
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1732.0,
      "debug/num_lat_total": 2149.0,
      "debug/num_tok_loss": 1732.0,
      "debug/num_tok_total": 2149.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.206441879272461,
      "train/diffusion_loss": 0.4473426342010498
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1762.0,
      "debug/num_lat_total": 2352.0,
      "debug/num_tok_loss": 1762.0,
      "debug/num_tok_total": 2352.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.163651466369629,
      "train/diffusion_loss": 0.39565032720565796
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1795.0,
      "debug/num_lat_total": 2295.0,
      "debug/num_tok_loss": 1795.0,
      "debug/num_tok_total": 2295.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.172779083251953,
      "train/diffusion_loss": 0.4185372292995453
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1687.0,
      "debug/num_lat_total": 2142.0,
      "debug/num_tok_loss": 1687.0,
      "debug/num_tok_total": 2142.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.184844017028809,
      "train/diffusion_loss": 0.4559231698513031
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1717.0,
      "debug/num_lat_total": 2221.0,
      "debug/num_tok_loss": 1717.0,
      "debug/num_tok_total": 2221.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.17270565032959,
      "train/diffusion_loss": 0.44331198930740356
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1717.0,
      "debug/num_lat_total": 2179.0,
      "debug/num_tok_loss": 1717.0,
      "debug/num_tok_total": 2179.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.174978256225586,
      "train/diffusion_loss": 0.42323049902915955
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1840.0,
      "debug/num_lat_total": 2161.0,
      "debug/num_tok_loss": 1840.0,
      "debug/num_tok_total": 2161.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.170124053955078,
      "train/diffusion_loss": 0.48068806529045105
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1775.0,
      "debug/num_lat_total": 2192.0,
      "debug/num_tok_loss": 1775.0,
      "debug/num_tok_total": 2192.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.172566413879395,
      "train/diffusion_loss": 0.4287210702896118
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1728.0,
      "debug/num_lat_total": 2038.0,
      "debug/num_tok_loss": 1728.0,
      "debug/num_tok_total": 2038.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.167974472045898,
      "train/diffusion_loss": 0.44199228286743164
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1700.0,
      "debug/num_lat_total": 2257.0,
      "debug/num_tok_loss": 1700.0,
      "debug/num_tok_total": 2257.0,
      "epoch": 4.299065420560748,
      "step": 30
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/ce_loss": 12.186758995056152,
      "train/diffusion_loss": 0.4000075161457062
    },
    {
      "epoch": 4.299065420560748,
      "step": 30,
      "train/learning_rate_real": 1.389556891813458e-06
    },
    {
      "debug/num_lat_loss": 1806.0,
      "debug/num_lat_total": 2203.0,
      "debug/num_tok_loss": 1806.0,
      "debug/num_tok_total": 2203.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.169504165649414,
      "train/diffusion_loss": 0.47231945395469666
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1767.0,
      "debug/num_lat_total": 2238.0,
      "debug/num_tok_loss": 1767.0,
      "debug/num_tok_total": 2238.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.156598091125488,
      "train/diffusion_loss": 0.4360898733139038
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1719.0,
      "debug/num_lat_total": 2105.0,
      "debug/num_tok_loss": 1719.0,
      "debug/num_tok_total": 2105.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.160431861877441,
      "train/diffusion_loss": 0.43545302748680115
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1952.0,
      "debug/num_lat_total": 2575.0,
      "debug/num_tok_loss": 1952.0,
      "debug/num_tok_total": 2575.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.136270523071289,
      "train/diffusion_loss": 0.4197411835193634
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1746.0,
      "debug/num_lat_total": 2267.0,
      "debug/num_tok_loss": 1746.0,
      "debug/num_tok_total": 2267.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.166632652282715,
      "train/diffusion_loss": 0.466814786195755
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1723.0,
      "debug/num_lat_total": 2326.0,
      "debug/num_tok_loss": 1723.0,
      "debug/num_tok_total": 2326.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.176079750061035,
      "train/diffusion_loss": 0.40206533670425415
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1705.0,
      "debug/num_lat_total": 2229.0,
      "debug/num_tok_loss": 1705.0,
      "debug/num_tok_total": 2229.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.191839218139648,
      "train/diffusion_loss": 0.41346275806427
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1785.0,
      "debug/num_lat_total": 2381.0,
      "debug/num_tok_loss": 1785.0,
      "debug/num_tok_total": 2381.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.170215606689453,
      "train/diffusion_loss": 0.4004969894886017
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1740.0,
      "debug/num_lat_total": 2062.0,
      "debug/num_tok_loss": 1740.0,
      "debug/num_tok_total": 2062.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.164392471313477,
      "train/diffusion_loss": 0.46367114782333374
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1706.0,
      "debug/num_lat_total": 2305.0,
      "debug/num_tok_loss": 1706.0,
      "debug/num_tok_total": 2305.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.16750717163086,
      "train/diffusion_loss": 0.4109986424446106
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1668.0,
      "debug/num_lat_total": 2229.0,
      "debug/num_tok_loss": 1668.0,
      "debug/num_tok_total": 2229.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.20853328704834,
      "train/diffusion_loss": 0.4263615310192108
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1804.0,
      "debug/num_lat_total": 2314.0,
      "debug/num_tok_loss": 1804.0,
      "debug/num_tok_total": 2314.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.14916706085205,
      "train/diffusion_loss": 0.4520219564437866
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1679.0,
      "debug/num_lat_total": 2148.0,
      "debug/num_tok_loss": 1679.0,
      "debug/num_tok_total": 2148.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.153385162353516,
      "train/diffusion_loss": 0.42382287979125977
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1724.0,
      "debug/num_lat_total": 2220.0,
      "debug/num_tok_loss": 1724.0,
      "debug/num_tok_total": 2220.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.130288124084473,
      "train/diffusion_loss": 0.4364231526851654
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1774.0,
      "debug/num_lat_total": 2283.0,
      "debug/num_tok_loss": 1774.0,
      "debug/num_tok_total": 2283.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.175270080566406,
      "train/diffusion_loss": 0.4409050941467285
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1835.0,
      "debug/num_lat_total": 2207.0,
      "debug/num_tok_loss": 1835.0,
      "debug/num_tok_total": 2207.0,
      "epoch": 4.4485981308411215,
      "step": 31
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/ce_loss": 12.18620491027832,
      "train/diffusion_loss": 0.44913163781166077
    },
    {
      "epoch": 4.4485981308411215,
      "step": 31,
      "train/learning_rate_real": 8.95400837299093e-07
    },
    {
      "debug/num_lat_loss": 1719.0,
      "debug/num_lat_total": 2323.0,
      "debug/num_tok_loss": 1719.0,
      "debug/num_tok_total": 2323.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.183679580688477,
      "train/diffusion_loss": 0.41472896933555603
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1646.0,
      "debug/num_lat_total": 2072.0,
      "debug/num_tok_loss": 1646.0,
      "debug/num_tok_total": 2072.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.173807144165039,
      "train/diffusion_loss": 0.4390676021575928
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1762.0,
      "debug/num_lat_total": 2427.0,
      "debug/num_tok_loss": 1762.0,
      "debug/num_tok_total": 2427.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.151962280273438,
      "train/diffusion_loss": 0.41235706210136414
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1758.0,
      "debug/num_lat_total": 2354.0,
      "debug/num_tok_loss": 1758.0,
      "debug/num_tok_total": 2354.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.149380683898926,
      "train/diffusion_loss": 0.4324599802494049
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1862.0,
      "debug/num_lat_total": 2345.0,
      "debug/num_tok_loss": 1862.0,
      "debug/num_tok_total": 2345.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.148853302001953,
      "train/diffusion_loss": 0.420383483171463
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1795.0,
      "debug/num_lat_total": 2269.0,
      "debug/num_tok_loss": 1795.0,
      "debug/num_tok_total": 2269.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.170392036437988,
      "train/diffusion_loss": 0.4294658899307251
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1760.0,
      "debug/num_lat_total": 2156.0,
      "debug/num_tok_loss": 1760.0,
      "debug/num_tok_total": 2156.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.155299186706543,
      "train/diffusion_loss": 0.44193097949028015
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1774.0,
      "debug/num_lat_total": 2170.0,
      "debug/num_tok_loss": 1774.0,
      "debug/num_tok_total": 2170.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.195367813110352,
      "train/diffusion_loss": 0.44668319821357727
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1744.0,
      "debug/num_lat_total": 2061.0,
      "debug/num_tok_loss": 1744.0,
      "debug/num_tok_total": 2061.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.192215919494629,
      "train/diffusion_loss": 0.4522995352745056
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1809.0,
      "debug/num_lat_total": 2322.0,
      "debug/num_tok_loss": 1809.0,
      "debug/num_tok_total": 2322.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.157405853271484,
      "train/diffusion_loss": 0.416936457157135
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1643.0,
      "debug/num_lat_total": 2054.0,
      "debug/num_tok_loss": 1643.0,
      "debug/num_tok_total": 2054.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.180978775024414,
      "train/diffusion_loss": 0.42913898825645447
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1742.0,
      "debug/num_lat_total": 2406.0,
      "debug/num_tok_loss": 1742.0,
      "debug/num_tok_total": 2406.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.150097846984863,
      "train/diffusion_loss": 0.40735092759132385
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1646.0,
      "debug/num_lat_total": 1997.0,
      "debug/num_tok_loss": 1646.0,
      "debug/num_tok_total": 1997.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.196208000183105,
      "train/diffusion_loss": 0.43271905183792114
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1702.0,
      "debug/num_lat_total": 2094.0,
      "debug/num_tok_loss": 1702.0,
      "debug/num_tok_total": 2094.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.162835121154785,
      "train/diffusion_loss": 0.4518500566482544
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1741.0,
      "debug/num_lat_total": 2201.0,
      "debug/num_tok_loss": 1741.0,
      "debug/num_tok_total": 2201.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.16962718963623,
      "train/diffusion_loss": 0.442145437002182
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1756.0,
      "debug/num_lat_total": 2102.0,
      "debug/num_tok_loss": 1756.0,
      "debug/num_tok_total": 2102.0,
      "epoch": 4.598130841121495,
      "step": 32
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/ce_loss": 12.188155174255371,
      "train/diffusion_loss": 0.4411594271659851
    },
    {
      "epoch": 4.598130841121495,
      "step": 32,
      "train/learning_rate_real": 5.063378298187843e-07
    },
    {
      "debug/num_lat_loss": 1730.0,
      "debug/num_lat_total": 2254.0,
      "debug/num_tok_loss": 1730.0,
      "debug/num_tok_total": 2254.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.188690185546875,
      "train/diffusion_loss": 0.44849783182144165
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1723.0,
      "debug/num_lat_total": 2360.0,
      "debug/num_tok_loss": 1723.0,
      "debug/num_tok_total": 2360.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.205450057983398,
      "train/diffusion_loss": 0.400063157081604
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1678.0,
      "debug/num_lat_total": 2055.0,
      "debug/num_tok_loss": 1678.0,
      "debug/num_tok_total": 2055.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.187387466430664,
      "train/diffusion_loss": 0.46367162466049194
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1666.0,
      "debug/num_lat_total": 2027.0,
      "debug/num_tok_loss": 1666.0,
      "debug/num_tok_total": 2027.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.189556121826172,
      "train/diffusion_loss": 0.5164213180541992
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1640.0,
      "debug/num_lat_total": 2208.0,
      "debug/num_tok_loss": 1640.0,
      "debug/num_tok_total": 2208.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.173091888427734,
      "train/diffusion_loss": 0.41635480523109436
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1715.0,
      "debug/num_lat_total": 2113.0,
      "debug/num_tok_loss": 1715.0,
      "debug/num_tok_total": 2113.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.206910133361816,
      "train/diffusion_loss": 0.4815617799758911
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1737.0,
      "debug/num_lat_total": 2221.0,
      "debug/num_tok_loss": 1737.0,
      "debug/num_tok_total": 2221.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.145868301391602,
      "train/diffusion_loss": 0.4430966377258301
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1707.0,
      "debug/num_lat_total": 2296.0,
      "debug/num_tok_loss": 1707.0,
      "debug/num_tok_total": 2296.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.134885787963867,
      "train/diffusion_loss": 0.4013361930847168
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1736.0,
      "debug/num_lat_total": 2219.0,
      "debug/num_tok_loss": 1736.0,
      "debug/num_tok_total": 2219.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.147263526916504,
      "train/diffusion_loss": 0.43085482716560364
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1685.0,
      "debug/num_lat_total": 2033.0,
      "debug/num_tok_loss": 1685.0,
      "debug/num_tok_total": 2033.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.185604095458984,
      "train/diffusion_loss": 0.4586755037307739
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1719.0,
      "debug/num_lat_total": 2199.0,
      "debug/num_tok_loss": 1719.0,
      "debug/num_tok_total": 2199.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.170775413513184,
      "train/diffusion_loss": 0.4456702172756195
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1749.0,
      "debug/num_lat_total": 2183.0,
      "debug/num_tok_loss": 1749.0,
      "debug/num_tok_total": 2183.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.213212966918945,
      "train/diffusion_loss": 0.4502941966056824
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1718.0,
      "debug/num_lat_total": 2255.0,
      "debug/num_tok_loss": 1718.0,
      "debug/num_tok_total": 2255.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.188613891601562,
      "train/diffusion_loss": 0.428547739982605
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1810.0,
      "debug/num_lat_total": 2148.0,
      "debug/num_tok_loss": 1810.0,
      "debug/num_tok_total": 2148.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.191283226013184,
      "train/diffusion_loss": 0.4674947261810303
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1694.0,
      "debug/num_lat_total": 2097.0,
      "debug/num_tok_loss": 1694.0,
      "debug/num_tok_total": 2097.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.186917304992676,
      "train/diffusion_loss": 0.4404945969581604
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1772.0,
      "debug/num_lat_total": 2293.0,
      "debug/num_tok_loss": 1772.0,
      "debug/num_tok_total": 2293.0,
      "epoch": 4.747663551401869,
      "step": 33
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/ce_loss": 12.178369522094727,
      "train/diffusion_loss": 0.45209217071533203
    },
    {
      "epoch": 4.747663551401869,
      "step": 33,
      "train/learning_rate_real": 2.258912842161662e-07
    },
    {
      "debug/num_lat_loss": 1762.0,
      "debug/num_lat_total": 2343.0,
      "debug/num_tok_loss": 1762.0,
      "debug/num_tok_total": 2343.0,
      "epoch": 4.897196261682243,
      "step": 34
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/ce_loss": 12.175055503845215,
      "train/diffusion_loss": 0.44022753834724426
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/learning_rate_real": 5.6600967836442644e-08
    },
    {
      "debug/num_lat_loss": 1747.0,
      "debug/num_lat_total": 2348.0,
      "debug/num_tok_loss": 1747.0,
      "debug/num_tok_total": 2348.0,
      "epoch": 4.897196261682243,
      "step": 34
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/ce_loss": 12.170550346374512,
      "train/diffusion_loss": 0.42128288745880127
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/learning_rate_real": 5.6600967836442644e-08
    },
    {
      "debug/num_lat_loss": 1708.0,
      "debug/num_lat_total": 2088.0,
      "debug/num_tok_loss": 1708.0,
      "debug/num_tok_total": 2088.0,
      "epoch": 4.897196261682243,
      "step": 34
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/ce_loss": 12.167820930480957,
      "train/diffusion_loss": 0.44098177552223206
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/learning_rate_real": 5.6600967836442644e-08
    },
    {
      "debug/num_lat_loss": 1723.0,
      "debug/num_lat_total": 2002.0,
      "debug/num_tok_loss": 1723.0,
      "debug/num_tok_total": 2002.0,
      "epoch": 4.897196261682243,
      "step": 34
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/ce_loss": 12.243925094604492,
      "train/diffusion_loss": 0.44518256187438965
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/learning_rate_real": 5.6600967836442644e-08
    },
    {
      "debug/num_lat_loss": 1705.0,
      "debug/num_lat_total": 2094.0,
      "debug/num_tok_loss": 1705.0,
      "debug/num_tok_total": 2094.0,
      "epoch": 4.897196261682243,
      "step": 34
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/ce_loss": 12.187417030334473,
      "train/diffusion_loss": 0.41937023401260376
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/learning_rate_real": 5.6600967836442644e-08
    },
    {
      "debug/num_lat_loss": 1765.0,
      "debug/num_lat_total": 2297.0,
      "debug/num_tok_loss": 1765.0,
      "debug/num_tok_total": 2297.0,
      "epoch": 4.897196261682243,
      "step": 34
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/ce_loss": 12.190154075622559,
      "train/diffusion_loss": 0.41703900694847107
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/learning_rate_real": 5.6600967836442644e-08
    },
    {
      "debug/num_lat_loss": 1782.0,
      "debug/num_lat_total": 2235.0,
      "debug/num_tok_loss": 1782.0,
      "debug/num_tok_total": 2235.0,
      "epoch": 4.897196261682243,
      "step": 34
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/ce_loss": 12.209789276123047,
      "train/diffusion_loss": 0.4473593533039093
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/learning_rate_real": 5.6600967836442644e-08
    },
    {
      "debug/num_lat_loss": 1774.0,
      "debug/num_lat_total": 2079.0,
      "debug/num_tok_loss": 1774.0,
      "debug/num_tok_total": 2079.0,
      "epoch": 4.897196261682243,
      "step": 34
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/ce_loss": 12.186965942382812,
      "train/diffusion_loss": 0.47279080748558044
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/learning_rate_real": 5.6600967836442644e-08
    },
    {
      "debug/num_lat_loss": 1723.0,
      "debug/num_lat_total": 2052.0,
      "debug/num_tok_loss": 1723.0,
      "debug/num_tok_total": 2052.0,
      "epoch": 4.897196261682243,
      "step": 34
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/ce_loss": 12.189580917358398,
      "train/diffusion_loss": 0.45107120275497437
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/learning_rate_real": 5.6600967836442644e-08
    },
    {
      "debug/num_lat_loss": 1821.0,
      "debug/num_lat_total": 2294.0,
      "debug/num_tok_loss": 1821.0,
      "debug/num_tok_total": 2294.0,
      "epoch": 4.897196261682243,
      "step": 34
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/ce_loss": 12.185731887817383,
      "train/diffusion_loss": 0.4438938796520233
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/learning_rate_real": 5.6600967836442644e-08
    },
    {
      "debug/num_lat_loss": 221.0,
      "debug/num_lat_total": 275.0,
      "debug/num_tok_loss": 221.0,
      "debug/num_tok_total": 275.0,
      "epoch": 4.897196261682243,
      "step": 34
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/ce_loss": 12.213896751403809,
      "train/diffusion_loss": 0.4825238287448883
    },
    {
      "epoch": 4.897196261682243,
      "step": 34,
      "train/learning_rate_real": 5.6600967836442644e-08
    }
  ],
  "logging_steps": 10,
  "max_steps": 35,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 200,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 9.708115641912502e+16,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}