MaxView

← Back to run

Log Summary

2026-04-17 13:11:11.004014: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303)
I0417 13:11:11.119732 126697986976896 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag.
I0417 13:11:56.702590 126697986976896 max_utils.py:800] System Information: Jax Version: 0.9.2
I0417 13:11:56.702702 126697986976896 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0417 13:11:56.702740 126697986976896 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Apr 6 2026 20:48:10 (1775533690) cl/895581894
I0417 13:11:56.702764 126697986976896 train_utils.py:335] WARNING: 'dataset_path' might be pointing your local file system
I0417 13:11:56.702785 126697986976896 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0417 13:11:56.702865 126697986976896 train.py:671] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper.
W0417 13:11:56.796317  438423 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0417 13:11:57.198736 126697986976896 maxtext_utils.py:1548] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0417 13:11:57.199076 126697986976896 checkpointing.py:677] Setting up checkpoint logger...
I0417 13:11:57.199122 126697986976896 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0417 13:11:57.199188 126697986976896 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0417 13:11:57.199816 126697986976896 base_pytree_checkpoint_handler.py:411] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x733a95908830>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0417 13:11:59.517939 126697986976896 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0417 13:11:59.518329 126697986976896 checkpoint_manager.py:702] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x73341872c320>}, handler_registry=None
I0417 13:11:59.518852 126697986976896 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x73341872c320>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`.
I0417 13:11:59.518898 126697986976896 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x733417736720>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`.
I0417 13:11:59.518931 126697986976896 composite_checkpoint_handler.py:505] Initialized registry DefaultCheckpointHandlerRegistry({('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x73341872c320>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x73341872c320>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x733417736720>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x733417736720>}).
I0417 13:11:59.519735 126697986976896 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28
I0417 13:11:59.519808 126697986976896 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x733417732160> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0417 13:11:59.635836 126697986976896 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints
I0417 13:11:59.636111 126697986976896 checkpoint_manager.py:921] [process=0][thread=MainThread] CheckpointManager created,  primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_hns=False, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x733417f9c230>
I0417 13:11:59.636211 126697986976896 checkpointing.py:301] Checkpoint manager created!
W0417 13:11:59.655197  438423 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0417 13:12:00.195902 126697986976896 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0417 13:12:00.196008 126697986976896 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0417 13:12:00.279257 126697986976896 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0417 13:12:00.279339 126697986976896 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0417 13:12:00.293457 126697986976896 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0417 13:12:00.293506 126697986976896 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0417 13:12:00.314013 126697986976896 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0417 13:12:00.314077 126697986976896 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0417 13:12:00.328251 126697986976896 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0417 13:12:00.328299 126697986976896 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0417 13:12:00.342574 126697986976896 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0417 13:12:00.342622 126697986976896 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0417 13:12:00.363960 126697986976896 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0417 13:12:00.364016 126697986976896 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0417 13:12:00.382230 126697986976896 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0417 13:12:00.382281 126697986976896 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None).
I0417 13:12:00.571392 126697986976896 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0417 13:12:00.571489 126697986976896 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
W0417 13:12:00.620997  438423 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
[DECOUPLED NO-OP] gcs_storage: using stubs.
[DECOUPLED NO-OP] mldiagnostics: using stub.
[DECOUPLED NO-OP] mldiagnostics: using stub.
[DECOUPLED NO-OP] mldiagnostics: using stub.
[DECOUPLED NO-OP] workload_monitor: using stub.
[DECOUPLED NO-OP] vertex_tensorboard: using stub.
fsdp: 8
I0417 13:12:01.001326 126697986976896 maxtext_utils.py:1651]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0417 13:12:01.001434 126697986976896 maxtext_utils.py:1651]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0417 13:12:01.001475 126697986976896 maxtext_utils.py:1651]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0417 13:12:01.001520 126697986976896 maxtext_utils.py:1651]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   P('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0417 13:12:01.001559 126697986976896 maxtext_utils.py:1651]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0417 13:12:01.001590 126697986976896 maxtext_utils.py:1651]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0417 13:12:01.001629 126697986976896 maxtext_utils.py:1651]  params/params/decoder/layers/self_attention/key/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0417 13:12:01.001674 126697986976896 maxtext_utils.py:1651]  params/params/decoder/layers/self_attention/out/kernel
    Shape:     float32[16,16,128,2048]
    Logical:   P('heads', 'layers', 'kv', 'embed')
    Physical:  (None, None, None, 'fsdp')
I0417 13:12:01.001703 126697986976896 maxtext_utils.py:1651]  params/params/decoder/layers/self_attention/query/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None, None)
I0417 13:12:01.001728 126697986976896 maxtext_utils.py:1651]  params/params/decoder/layers/self_attention/value/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0417 13:12:01.001762 126697986976896 maxtext_utils.py:1651]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)
I0417 13:12:01.001797 126697986976896 maxtext_utils.py:1651]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')

I0417 13:12:01.421331 126697986976896 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0417 13:12:01.421412 126697986976896 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None).
I0417 13:12:01.434849 126697986976896 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0417 13:12:01.434897 126697986976896 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None).
I0417 13:12:09.316201 126697986976896 max_utils.py:791] Total memory size: 3.6 GB, Output size: 1.5 GB, Temp size: 2.0 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB.
I0417 13:12:09.317103 126697986976896 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter.
I0417 13:12:09.317287 126697986976896 metric_logger.py:301] number parameters: 1.104 billion
I0417 13:12:17.135916 126697986976896 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0417 13:12:17.262432 126697986976896 checkpointing.py:776] Waited 0.12650680541992188 seconds for step 0 to finish before starting checkpointing.
I0417 13:12:17.262977 126697986976896 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0417 13:12:17.263163 126697986976896 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0
I0417 13:12:17.263591 126697986976896 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/0.
I0417 13:12:17.349205 126697986976896 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient
I0417 13:12:17.461823 126697986976896 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array.
I0417 13:12:17.461912 126697986976896 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0417 13:12:17.468986 126590352623168 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/0
W0417 13:12:17.776253  438423 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0417 13:12:17.790490  438423 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0417 13:12:17.807192  438423 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0417 13:12:17.817142  438423 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0417 13:12:17.822821  438423 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0417 13:12:17.827763  438423 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0417 13:12:17.832640  438423 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0417 13:12:17.837666  438423 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0417 13:12:18.021736 126697986976896 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.560353s
I0417 13:12:18.022406 126697986976896 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 18.360 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 672 milliseconds) (per-host)
I0417 13:12:18.022536 126697986976896 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.672365s (batch_requests_ready=0.107662s, total_serialization_initiated=0.564129s, others=0.000574s)
I0417 13:12:18.022662 126697986976896 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.672988s (all_items=0.000026s, per_item={'items': '0.00002599'}, temp_paths=0.672962)
I0417 13:12:18.023593 126590300194368 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0417 13:12:18.023685 126697986976896 async_checkpointer.py:561] Finished blocking save. Time taken: 0.760477s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/0.
I0417 13:12:18.023871 126697986976896 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0417 13:12:18.024058 126590289708608 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0417 13:12:18.024145 126697986976896 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776431537.2629623, 'wait_for_prev_duration_secs': 4.4345855712890625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776431537.2631862, 'checkpointer_blocking_duration_secs': 0.7606022357940674, 'get_old_steps_start_time': 1776431538.0238066, 'get_old_steps_duration_secs': 2.8133392333984375e-05, 'checkpoint_manager_blocking_start_time': 1776431537.2628667, 'checkpoint_manager_blocking_duration_secs': 0.7612428665161133}
I0417 13:12:18.024239 126697986976896 checkpointing.py:408] Started an asynchronous checkpoint save for step 0
I0417 13:12:18.024280 126697986976896 max_utils.py:750] 
Memstats: After params initialized:
I0417 13:12:18.024383 126697986976896 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0))
I0417 13:12:18.024412 126697986976896 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0))
I0417 13:12:18.024434 126697986976896 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0))
I0417 13:12:18.024453 126697986976896 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0))
I0417 13:12:18.024471 126697986976896 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0))
I0417 13:12:18.024491 126697986976896 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0))
I0417 13:12:18.024509 126697986976896 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0))
I0417 13:12:18.024527 126697986976896 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0))
W0417 13:12:18.031253  438423 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0417 13:12:18.037331  438423 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0417 13:12:18.042831  438423 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0417 13:12:18.232019 126590342137408 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/0/items
I0417 13:12:18.252874 126697986976896 metric_logger.py:196] completed step: 0, seconds: 7.818, TFLOP/s/device: 1.738, Tokens/s/device: 261.971, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941
I0417 13:12:18.253765 126697986976896 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/tensorboard/'
I0417 13:12:18.368389 126697986976896 metric_logger.py:196] completed step: 1, seconds: 1.115, TFLOP/s/device: 12.191, Tokens/s/device: 1837.572, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941
I0417 13:12:18.386811 126590363108928 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776431538118742465, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776431538118742465, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/0/_CHECKPOINT_METADATA
I0417 13:12:18.486518 126697986976896 metric_logger.py:196] completed step: 2, seconds: 0.007, TFLOP/s/device: 2055.231, Tokens/s/device: 309786.719, total_weights: 16384, loss: 10.276, lm_loss: 10.276, perplexity: 29019.973
I0417 13:12:18.605884 126697986976896 metric_logger.py:196] completed step: 3, seconds: 0.116, TFLOP/s/device: 116.974, Tokens/s/device: 17631.613, total_weights: 16384, loss: 9.282, lm_loss: 9.282, perplexity: 10739.224
I0417 13:12:18.610782 126697986976896 checkpointing.py:772] Waiting for step 5 to finish before checkpoint...
I0417 13:12:18.693583  442649 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json
I0417 13:12:18.693656  442649 google_auth_provider.cc:156] Using OAuth2 AuthProvider
I0417 13:12:18.842068 126697986976896 checkpointing.py:776] Waited 0.23124194145202637 seconds for step 5 to finish before starting checkpointing.
I0417 13:12:18.842456 126697986976896 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0417 13:12:19.424345 126590394566208 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/0/items/array_metadatas/process_0
I0417 13:12:45.053477 126590310680128 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 26.358233s (commit=25.906141s, array_metadata_write=0.452093s)
I0417 13:12:45.054536 126590300194368 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 456.130 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 27 seconds) (per-host)
I0417 13:12:45.054595 126590300194368 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 27.030866s.
I0417 13:12:45.324632 126590300194368 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776431538118742465, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/0/_CHECKPOINT_METADATA
I0417 13:12:45.523117 126590300194368 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0417 13:12:45.701143 126590363108928 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776431538118742465, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/0/_CHECKPOINT_METADATA
I0417 13:12:45.946012 126590300194368 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0417 13:12:45.946736 126590300194368 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.576583s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/0/items
I0417 13:12:45.947392 126590300194368 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/0/items
I0417 13:12:46.214022 126590300194368 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/0
I0417 13:12:46.905894 126590300194368 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/0`.
I0417 13:12:46.906522 126590300194368 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 29.643322s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/0
I0417 13:12:46.906599 126590300194368 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 28.882871s.
I0417 13:12:46.906714 126590289708608 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0417 13:12:46.906766 126590289708608 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0417 13:12:46.906827 126590289708608 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0417 13:12:46.906867 126590289708608 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0417 13:12:46.906956 126697986976896 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0417 13:12:46.907022 126697986976896 checkpoint_manager.py:1441] Waiting for previous save to complete took 28.064568 seconds. If this number is high, consider checkpointing less frequently.
I0417 13:12:46.907762 126697986976896 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 5
I0417 13:12:46.908004 126697986976896 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/5.
I0417 13:12:47.089227 126590300194368 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/5
I0417 13:12:47.115522 126697986976896 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array.
I0417 13:12:47.115618 126697986976896 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0417 13:12:47.820492 126590310680128 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/5/items
I0417 13:12:53.274181 126697986976896 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 6.159022s
I0417 13:12:53.278801 126697986976896 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.967 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host)
I0417 13:12:53.278872 126697986976896 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 6.273800s (batch_requests_ready=0.106800s, total_serialization_initiated=6.162499s, others=0.004501s)
I0417 13:12:53.278975 126697986976896 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 6.274354s (all_items=0.000019s, per_item={'items': '0.00001884'}, temp_paths=6.274335)
I0417 13:12:53.280385 126590394566208 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0417 13:12:53.280479 126697986976896 async_checkpointer.py:561] Finished blocking save. Time taken: 6.372673s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/5.
I0417 13:12:53.280774 126697986976896 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize
I0417 13:12:53.280925 126590289708608 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0417 13:12:53.281013 126697986976896 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776431538.8424337, 'wait_for_prev_duration_secs': 28.064567804336548, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776431566.9077852, 'checkpointer_blocking_duration_secs': 6.372886896133423, 'get_old_steps_start_time': 1776431573.2806964, 'get_old_steps_duration_secs': 3.361701965332031e-05, 'checkpoint_manager_blocking_start_time': 1776431538.8423724, 'checkpoint_manager_blocking_duration_secs': 34.43860936164856}
I0417 13:12:53.281184 126697986976896 checkpointing.py:408] Started an asynchronous checkpoint save for step 5
I0417 13:12:53.281953 126697986976896 metric_logger.py:196] completed step: 4, seconds: 0.118, TFLOP/s/device: 114.896, Tokens/s/device: 17318.360, total_weights: 16384, loss: 8.327, lm_loss: 8.327, perplexity: 4134.812
I0417 13:12:53.290370 126697986976896 metric_logger.py:196] completed step: 5, seconds: 0.119, TFLOP/s/device: 114.037, Tokens/s/device: 17188.851, total_weights: 16384, loss: 7.367, lm_loss: 7.367, perplexity: 1583.522
I0417 13:12:53.405690 126697986976896 metric_logger.py:196] completed step: 6, seconds: 34.677, TFLOP/s/device: 0.392, Tokens/s/device: 59.059, total_weights: 16384, loss: 6.459, lm_loss: 6.459, perplexity: 638.363
I0417 13:12:53.524318 126697986976896 metric_logger.py:196] completed step: 7, seconds: 0.006, TFLOP/s/device: 2170.815, Tokens/s/device: 327208.819, total_weights: 16384, loss: 5.622, lm_loss: 5.622, perplexity: 276.577
I0417 13:12:53.642707 126697986976896 metric_logger.py:196] completed step: 8, seconds: 0.116, TFLOP/s/device: 117.040, Tokens/s/device: 17641.485, total_weights: 16384, loss: 4.864, lm_loss: 4.864, perplexity: 129.523
I0417 13:12:53.760447 126697986976896 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0417 13:12:53.761455 126697986976896 checkpointing.py:776] Waited 0.0010285377502441406 seconds for step 9 to finish before starting checkpointing.
I0417 13:12:53.761769 126697986976896 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0417 13:12:54.041894 126590243571264 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/5/items/array_metadatas/process_0
I0417 13:13:39.098925 126590310680128 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 45.819329s (commit=45.348548s, array_metadata_write=0.470781s)
I0417 13:13:39.099954 126590394566208 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 242.573 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 52 seconds) (per-host)
I0417 13:13:39.100108 126590394566208 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.819578s.
I0417 13:13:39.596260 126590394566208 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0417 13:13:40.001708 126590394566208 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0417 13:13:40.002515 126590394566208 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.590676s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/5/items
I0417 13:13:40.003210 126590394566208 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/5/items
I0417 13:13:40.254032 126590394566208 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/5
I0417 13:13:40.952226 126590394566208 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/5`.
I0417 13:13:40.952852 126590394566208 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 54.045053s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/5
I0417 13:13:40.952928 126590394566208 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 47.672405s.
I0417 13:13:40.953058 126590289708608 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0417 13:13:40.953106 126590289708608 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0417 13:13:40.953158 126590289708608 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts...
I0417 13:13:40.953199 126590289708608 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts.
I0417 13:13:40.953320 126697986976896 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5.
W0417 13:13:40.953429 126697986976896 checkpoint_manager.py:1441] Waiting for previous save to complete took 47.191669 seconds. If this number is high, consider checkpointing less frequently.
I0417 13:13:40.953641 126697986976896 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9
I0417 13:13:40.953908 126697986976896 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/9.
I0417 13:13:41.124448 126697986976896 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array.
I0417 13:13:41.124532 126697986976896 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0417 13:13:41.173639 126590394566208 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/9
I0417 13:13:41.873139 126590264542784 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/9/items
I0417 13:13:46.549316 126697986976896 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 5.425216s
I0417 13:13:46.553409 126697986976896 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.236 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 5 seconds) (per-host)
I0417 13:13:46.553472 126697986976896 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 5.519675s (batch_requests_ready=0.087313s, total_serialization_initiated=5.428365s, others=0.003997s)
I0417 13:13:46.553545 126697986976896 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 5.520132s (all_items=0.000011s, per_item={'items': '0.00001121'}, temp_paths=5.520120)
I0417 13:13:46.555211 126590824482368 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0417 13:13:46.555325 126697986976896 async_checkpointer.py:561] Finished blocking save. Time taken: 5.601641s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/9.
I0417 13:13:46.555493 126697986976896 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0417 13:13:46.555644 126590289708608 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0417 13:13:46.555739 126697986976896 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776431573.7617452, 'wait_for_prev_duration_secs': 47.19166946411133, 'time_between_consecutive_saves_sec': 6.854863882064819, 'checkpointer_blocking_start_time': 1776431620.9536648, 'checkpointer_blocking_duration_secs': 5.601768255233765, 'get_old_steps_start_time': 1776431626.5554447, 'get_old_steps_duration_secs': 1.5735626220703125e-05, 'checkpoint_manager_blocking_start_time': 1776431573.7617126, 'checkpoint_manager_blocking_duration_secs': 52.79400300979614}
I0417 13:13:46.555849 126697986976896 checkpointing.py:408] Started an asynchronous checkpoint save for step 9
I0417 13:13:46.555879 126697986976896 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0417 13:13:47.293642 126590300194368 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/9/items/array_metadatas/process_0
I0417 13:14:19.552266 126590310680128 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 32.997980s (commit=32.563102s, array_metadata_write=0.434878s)
I0417 13:14:19.553499 126590824482368 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 328.061 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 38 seconds) (per-host)
I0417 13:14:19.553612 126590824482368 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 32.998244s.
I0417 13:14:19.946799 126590824482368 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0417 13:14:20.362359 126590824482368 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0417 13:14:20.363075 126590824482368 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.560252s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/9/items
I0417 13:14:20.363705 126590824482368 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/9/items
I0417 13:14:20.592260 126590824482368 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/9
I0417 13:14:21.278249 126590824482368 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/9`.
I0417 13:14:21.278863 126590824482368 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 40.325186s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_08_checkpoint_async_true/checkpoints/9
I0417 13:14:21.278928 126590824482368 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 34.723562s.
I0417 13:14:21.279080 126590289708608 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0417 13:14:21.279181 126590289708608 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0417 13:14:21.279235 126590289708608 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0417 13:14:21.279274 126590289708608 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0417 13:14:21.279412 126697986976896 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0417 13:14:21.279553 126697986976896 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0417 13:14:21.280748 126697986976896 metric_logger.py:196] completed step: 9, seconds: 0.119, TFLOP/s/device: 114.339, Tokens/s/device: 17234.415, total_weights: 16384, loss: 4.189, lm_loss: 4.189, perplexity: 65.979
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops