[
 {
  "stage": "fixed_batch",
  "batch1_sha256": "d9ce572e4cf770c21e54e31f0cd523d578343475316b862f1453171ae1c3c6eb",
  "batch2_sha256": "3296f064265cd171d6ef153d36aa9aedba44197205d849cfbbbe3e4271481643",
  "t": 1793805596.0907044
 },
 {
  "stage": "get_info",
  "arm": "dimuon_lr1e-3_uninterrupted",
  "after_restore": 0,
  "optimizer": "dimuon",
  "actual_config": {
   "type": "dimuon",
   "version": 1
  },
  "model": "Qwen/Qwen3.5-2B",
  "rank": 32,
  "model_id": "accb828dfecae1fa66fdd1a5cd641b23",
  "t": 1793805604.420229
 },
 {
  "stage": "save",
  "name": "pf-init",
  "path": "tinker://accb828dfecae1fa66fdd1a5cd641b23/weights/pf-init",
  "t": 1793805606.6014175
 },
 {
  "stage": "initial_output",
  "arm": "dimuon_I",
  "output_sha256": "8e848edd4c24d0404a28b7b6eb29f98a8f282b90b7fac71937d9f9f817a9b17f",
  "batch_sha256": "c3bc8df861acd9d7ca3cbddf99aeae9de812ee9d268b5a7a0fbf43d3393842a3",
  "t": 1793805614.8695204
 },
 {
  "stage": "get_info",
  "arm": "grad_recovery_adam",
  "after_restore": 0,
  "optimizer": "adamw",
  "actual_config": {
   "type": "adamw"
  },
  "model": "Qwen/Qwen3.5-2B",
  "rank": 32,
  "model_id": "0acc5b9bd6cd84697ea6d2b3b97c591e",
  "t": 1793805617.7381463
 },
 {
  "stage": "get_info",
  "arm": "grad_recovery_adam",
  "after_restore": 1,
  "optimizer": "adamw",
  "actual_config": {
   "type": "adamw"
  },
  "model": "Qwen/Qwen3.5-2B",
  "rank": 32,
  "model_id": "0acc5b9bd6cd84697ea6d2b3b97c591e",
  "t": 1793805619.139392
 },
 {
  "stage": "initial_output",
  "arm": "grad_recovery_adam",
  "output_sha256": "8e848edd4c24d0404a28b7b6eb29f98a8f282b90b7fac71937d9f9f817a9b17f",
  "batch_sha256": "c3bc8df861acd9d7ca3cbddf99aeae9de812ee9d268b5a7a0fbf43d3393842a3",
  "t": 1793805620.058316
 },
 {
  "stage": "step",
  "arm": "grad_recovery_adam",
  "name": "pf-grad",
  "path": "tinker://0acc5b9bd6cd84697ea6d2b3b97c591e/weights/pf-grad",
  "params": {
   "learning_rate": 1000000.0,
   "beta1": 0.9,
   "beta2": 0.95,
   "eps": 1000000.0,
   "weight_decay": 0.0,
   "grad_clip_norm": 0.0
  },
  "params_type": "AdamParams",
  "fb": {
   "response_length:mean": 294.53125,
   "n_sequences:sum": 32.0,
   "loss:sum": 75.16534215211868,
   "tokens:sum": 9425.0,
   "loss:mean": 0.007975102615609408
  },
  "grad_norm": 36.49832534790039,
  "t": 1793805635.725632
 },
 {
  "stage": "get_info",
  "arm": "adamw_lr1e-4",
  "after_restore": 0,
  "optimizer": "adamw",
  "actual_config": {
   "type": "adamw"
  },
  "model": "Qwen/Qwen3.5-2B",
  "rank": 32,
  "model_id": "dbbab7ce56ee22798836d6cbd3b42e8e",
  "t": 1793805638.4619334
 },
 {
  "stage": "get_info",
  "arm": "adamw_lr1e-4",
  "after_restore": 1,
  "optimizer": "adamw",
  "actual_config": {
   "type": "adamw"
  },
  "model": "Qwen/Qwen3.5-2B",
  "rank": 32,
  "model_id": "dbbab7ce56ee22798836d6cbd3b42e8e",
  "t": 1793805639.6534843
 },
 {
  "stage": "initial_output",
  "arm": "adamw_lr1e-4",
  "output_sha256": "8e848edd4c24d0404a28b7b6eb29f98a8f282b90b7fac71937d9f9f817a9b17f",
  "batch_sha256": "c3bc8df861acd9d7ca3cbddf99aeae9de812ee9d268b5a7a0fbf43d3393842a3",
  "t": 1793805640.486538
 },
 {
  "stage": "step",
  "arm": "adamw_lr1e-4",
  "name": "pf-adam-1e-4",
  "path": "tinker://dbbab7ce56ee22798836d6cbd3b42e8e/weights/pf-adam-1e-4",
  "params": {
   "learning_rate": 0.0001,
   "beta1": 0.9,
   "beta2": 0.95,
   "eps": 1e-08,
   "weight_decay": 0.0,
   "grad_clip_norm": 0.0
  },
  "params_type": "AdamParams",
  "fb": {
   "response_length:mean": 294.53125,
   "n_sequences:sum": 32.0,
   "loss:sum": 75.16534215211868,
   "tokens:sum": 9425.0,
   "loss:mean": 0.007975102615609408
  },
  "grad_norm": 36.497196197509766,
  "t": 1793805651.4068832
 },
 {
  "stage": "get_info",
  "arm": "dimuon_lr0.0001",
  "after_restore": 0,
  "optimizer": "dimuon",
  "actual_config": {
   "type": "dimuon",
   "version": 1
  },
  "model": "Qwen/Qwen3.5-2B",
  "rank": 32,
  "model_id": "0e700d84871de531bdd5e3fce4e023d6",
  "t": 1793805654.5128715
 },
 {
  "stage": "get_info",
  "arm": "dimuon_lr0.0001",
  "after_restore": 1,
  "optimizer": "dimuon",
  "actual_config": {
   "type": "dimuon",
   "version": 1
  },
  "model": "Qwen/Qwen3.5-2B",
  "rank": 32,
  "model_id": "0e700d84871de531bdd5e3fce4e023d6",
  "t": 1793805656.6624954
 },
 {
  "stage": "initial_output",
  "arm": "dimuon_lr0.0001",
  "output_sha256": "8e848edd4c24d0404a28b7b6eb29f98a8f282b90b7fac71937d9f9f817a9b17f",
  "batch_sha256": "c3bc8df861acd9d7ca3cbddf99aeae9de812ee9d268b5a7a0fbf43d3393842a3",
  "t": 1793805657.4938872
 },
 {
  "stage": "step",
  "arm": "dimuon_lr0.0001",
  "name": "pf-dim-0.0001",
  "path": "tinker://0e700d84871de531bdd5e3fce4e023d6/weights/pf-dim-0.0001",
  "params": {
   "type": "dimuon",
   "learning_rate": 0.0001,
   "grad_clip_norm": 0.0,
   "beta1": 0.9
  },
  "params_type": "DimuonParams",
  "fb": {
   "response_length:mean": 294.53125,
   "n_sequences:sum": 32.0,
   "loss:sum": 75.16534215211868,
   "tokens:sum": 9425.0,
   "loss:mean": 0.007975102615609408
  },
  "grad_norm": 36.499351501464844,
  "t": 1793805662.6186361
 },
 {
  "stage": "get_info",
  "arm": "dimuon_lr0.01",
  "after_restore": 0,
  "optimizer": "dimuon",
  "actual_config": {
   "type": "dimuon",
   "version": 1
  },
  "model": "Qwen/Qwen3.5-2B",
  "rank": 32,
  "model_id": "cb87eea43359fa051123ba84dd48776e",
  "t": 1793805665.5273294
 },
 {
  "stage": "get_info",
  "arm": "dimuon_lr0.01",
  "after_restore": 1,
  "optimizer": "dimuon",
  "actual_config": {
   "type": "dimuon",
   "version": 1
  },
  "model": "Qwen/Qwen3.5-2B",
  "rank": 32,
  "model_id": "cb87eea43359fa051123ba84dd48776e",
  "t": 1793805666.702685
 },
 {
  "stage": "initial_output",
  "arm": "dimuon_lr0.01",
  "output_sha256": "8e848edd4c24d0404a28b7b6eb29f98a8f282b90b7fac71937d9f9f817a9b17f",
  "batch_sha256": "c3bc8df861acd9d7ca3cbddf99aeae9de812ee9d268b5a7a0fbf43d3393842a3",
  "t": 1793805667.5277314
 },
 {
  "stage": "step",
  "arm": "dimuon_lr0.01",
  "name": "pf-dim-0.01",
  "path": "tinker://cb87eea43359fa051123ba84dd48776e/weights/pf-dim-0.01",
  "params": {
   "type": "dimuon",
   "learning_rate": 0.01,
   "grad_clip_norm": 0.0,
   "beta1": 0.9
  },
  "params_type": "DimuonParams",
  "fb": {
   "response_length:mean": 294.53125,
   "n_sequences:sum": 32.0,
   "loss:sum": 75.16534215211868,
   "tokens:sum": 9425.0,
   "loss:mean": 0.007975102615609408
  },
  "grad_norm": 36.496673583984375,
  "t": 1793805672.4471462
 },
 {
  "stage": "step",
  "arm": "dimuon_lr1e-3_uninterrupted",
  "name": "pf-dim-0.001",
  "path": "tinker://accb828dfecae1fa66fdd1a5cd641b23/weights/pf-dim-0.001",
  "params": {
   "type": "dimuon",
   "learning_rate": 0.001,
   "grad_clip_norm": 0.0,
   "beta1": 0.9
  },
  "params_type": "DimuonParams",
  "fb": {
   "response_length:mean": 294.53125,
   "n_sequences:sum": 32.0,
   "loss:sum": 75.16534215211868,
   "tokens:sum": 9425.0,
   "loss:mean": 0.007975102615609408
  },
  "grad_norm": 36.495277404785156,
  "t": 1793805677.7789154
 },
 {
  "stage": "step",
  "arm": "dimuon_lr1e-3_uninterrupted",
  "name": "pf-U2",
  "path": "tinker://accb828dfecae1fa66fdd1a5cd641b23/weights/pf-U2",
  "params": {
   "type": "dimuon",
   "learning_rate": 0.001,
   "grad_clip_norm": 0.0,
   "beta1": 0.9
  },
  "params_type": "DimuonParams",
  "fb": {
   "response_length:mean": 302.0,
   "n_sequences:sum": 32.0,
   "loss:sum": 72.37611865997314,
   "tokens:sum": 9664.0,
   "loss:mean": 0.0074892506891528505
  },
  "grad_norm": 55.62613296508789,
  "t": 1793805683.2816603
 },
 {
  "stage": "get_info",
  "arm": "dimuon_restored",
  "after_restore": 0,
  "optimizer": "dimuon",
  "actual_config": {
   "type": "dimuon",
   "version": 1
  },
  "model": "Qwen/Qwen3.5-2B",
  "rank": 32,
  "model_id": "1269a7b3177d539323f3246f0a708173",
  "t": 1793805685.932751
 },
 {
  "stage": "get_info",
  "arm": "dimuon_restored",
  "after_restore": 1,
  "optimizer": "dimuon",
  "actual_config": {
   "type": "dimuon",
   "version": 1
  },
  "model": "Qwen/Qwen3.5-2B",
  "rank": 32,
  "model_id": "1269a7b3177d539323f3246f0a708173",
  "t": 1793805687.584403
 },
 {
  "stage": "step",
  "arm": "dimuon_restored",
  "name": "pf-R2",
  "path": "tinker://1269a7b3177d539323f3246f0a708173/weights/pf-R2",
  "params": {
   "type": "dimuon",
   "learning_rate": 0.001,
   "grad_clip_norm": 0.0,
   "beta1": 0.9
  },
  "params_type": "DimuonParams",
  "fb": {
   "response_length:mean": 302.0,
   "n_sequences:sum": 32.0,
   "loss:sum": 72.37611865997314,
   "tokens:sum": 9664.0,
   "loss:mean": 0.0074892506891528505
  },
  "grad_norm": 55.627845764160156,
  "t": 1793805693.2785964
 },
 {
  "stage": "done",
  "t": 1793805693.2792428
 }
]