MaxView

2026-04-21 16:00:05.902668: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
E0000 00:00:1776787205.913620 1546438 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1776787205.916858 1546438 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1776787205.927954 1546438 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776787205.927966 1546438 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776787205.927968 1546438 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776787205.927969 1546438 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-21 16:00:09.328289: 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)
I0421 16:00:09.444216 132233637768320 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag.
~/maxtext/maxtext_venv/lib/python3.12/site-packages/jax/_src/xla_bridge.py:204: UserWarning: TPU backend initialization is taking more than 60.0 seconds. Did you run your code on all TPU hosts? See https://docs.jax.dev/en/latest/multi_process.html for more information.
  warnings.warn(
I0421 16:01:13.766360 132233637768320 max_utils.py:800] System Information: Jax Version: 0.8.1
I0421 16:01:13.766620 132233637768320 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0421 16:01:13.766671 132233637768320 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709
I0421 16:01:13.766705 132233637768320 train_utils.py:335] WARNING: 'dataset_path' might be pointing your local file system
I0421 16:01:13.766729 132233637768320 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0421 16:01:13.766880 132233637768320 train.py:703] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper.
I0421 16:01:14.132440 132233637768320 maxtext_utils.py:1551] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0421 16:01:14.132814 132233637768320 checkpointing.py:677] Setting up checkpoint logger...
I0421 16:01:14.132861 132233637768320 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0421 16:01:14.132947 132233637768320 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0421 16:01:14.133766 132233637768320 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 0x7843beba7170>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0421 16:01:15.820466 132233637768320 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0421 16:01:15.820706 132233637768320 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 0x784217f40560>}, handler_registry=None
I0421 16:01:15.821100 132233637768320 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x784217f40560>` 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`.
I0421 16:01:15.821145 132233637768320 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x784217f40b00>` 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`.
I0421 16:01:15.821174 132233637768320 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 0x784217f40560>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x784217f40560>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x784217f40b00>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x784217f40b00>}).
I0421 16:01:15.821465 132233637768320 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28
I0421 16:01:15.821526 132233637768320 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x784217f30fe0> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0421 16:01:16.262745 132150003238464 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776787075662816716, 'commit_timestamp_nsecs': 1776787129172331591, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/5/_CHECKPOINT_METADATA
I0421 16:01:16.341856 132150003238464 checkpoint_manager.py:1847] Missing metrics for step 5
I0421 16:01:16.401473 132233637768320 checkpoint_manager.py:1788] Found 3 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints
I0421 16:01:16.401975 132233637768320 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_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7842183592e0>
I0421 16:01:16.402080 132233637768320 checkpointing.py:301] Checkpoint manager created!
I0421 16:01:17.038700 132233637768320 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0421 16:01:17.038791 132233637768320 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0421 16:01:17.335546 132233637768320 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0421 16:01:17.335628 132233637768320 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0421 16:01:17.347568 132233637768320 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0421 16:01:17.347614 132233637768320 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0421 16:01:17.366357 132233637768320 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0421 16:01:17.366411 132233637768320 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0421 16:01:17.378464 132233637768320 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0421 16:01:17.378509 132233637768320 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0421 16:01:17.390433 132233637768320 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0421 16:01:17.390480 132233637768320 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0421 16:01:17.409395 132233637768320 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0421 16:01:17.409450 132233637768320 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0421 16:01:17.424487 132233637768320 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0421 16:01:17.424537 132233637768320 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None).
I0421 16:01:17.571072 132233637768320 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0421 16:01:17.571160 132233637768320 checkpointing.py:581] restoring from this run's directory step 9
I0421 16:01:17.738991 132233637768320 checkpointer.py:304] Restoring checkpoint from gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/9.
I0421 16:01:18.286953 1547950 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json
I0421 16:01:18.287026 1547950 google_auth_provider.cc:156] Using OAuth2 AuthProvider
I0421 16:01:25.207248 132233637768320 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/read/gbytes_per_sec: 1.731 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 7 seconds) (per-host)
I0421 16:01:25.207505 132233637768320 checkpointer.py:318] Finished restoring checkpoint in 7.55 seconds from gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/9.
I0421 16:01:25.207574 132233637768320 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints', 'checkpointer_start_time': 1776787277.6531115, 'checkpointer_duration_secs': 7.5544373989105225, 'checkpoint_manager_start_time': 1776787277.5713594, 'checkpoint_manager_duration_secs': 7.636192321777344}
[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
I0421 16:01:25.531711 132233637768320 maxtext_utils.py:1654]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 16:01:25.531814 132233637768320 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0421 16:01:25.531856 132233637768320 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0421 16:01:25.531900 132233637768320 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   PartitionSpec('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 16:01:25.531947 132233637768320 maxtext_utils.py:1654]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  (None, None)
I0421 16:01:25.531976 132233637768320 maxtext_utils.py:1654]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  (None, None)
I0421 16:01:25.532015 132233637768320 maxtext_utils.py:1654]  params/params/decoder/layers/self_attention/key/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0421 16:01:25.532069 132233637768320 maxtext_utils.py:1654]  params/params/decoder/layers/self_attention/out/kernel
    Shape:     float32[16,16,128,2048]
    Logical:   PartitionSpec('heads', 'layers', 'kv', 'embed')
    Physical:  (None, None, None, 'fsdp')
I0421 16:01:25.532104 132233637768320 maxtext_utils.py:1654]  params/params/decoder/layers/self_attention/query/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None, None)
I0421 16:01:25.532131 132233637768320 maxtext_utils.py:1654]  params/params/decoder/layers/self_attention/value/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0421 16:01:25.532167 132233637768320 maxtext_utils.py:1654]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)
I0421 16:01:25.532203 132233637768320 maxtext_utils.py:1654]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')

I0421 16:01:25.946847 132233637768320 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0421 16:01:25.947005 132233637768320 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None).
I0421 16:01:25.958402 132233637768320 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0421 16:01:25.958450 132233637768320 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None).
I0421 16:01:26.568628 132233637768320 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.
I0421 16:01:26.569268 132233637768320 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter.
I0421 16:01:26.569442 132233637768320 metric_logger.py:301] number parameters: 1.104 billion
I0421 16:01:28.275568 132233637768320 checkpointing.py:772] Waiting for step 10 to finish before checkpoint...
I0421 16:01:28.392723 132233637768320 checkpointing.py:776] Waited 0.11714005470275879 seconds for step 10 to finish before starting checkpointing.
I0421 16:01:28.393288 132233637768320 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0421 16:01:28.393458 132233637768320 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 10
I0421 16:01:28.393873 132233637768320 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/10.
I0421 16:01:28.483443 132233637768320 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient
I0421 16:01:28.567580 132233637768320 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array.
I0421 16:01:28.567668 132233637768320 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
I0421 16:01:28.574874 132149430715968 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/10
I0421 16:01:29.331145 132149420230208 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/10/items
I0421 16:01:29.482144 132149384578624 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776787289239418287, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776787289239418287, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/10/_CHECKPOINT_METADATA
I0421 16:01:35.921989 132233637768320 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 7.354814s
I0421 16:01:35.926411 132233637768320 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.658 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 7 seconds) (per-host)
I0421 16:01:35.926475 132233637768320 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 7.441704s (batch_requests_ready=0.079812s, total_serialization_initiated=7.357583s, others=0.004310s)
I0421 16:01:35.926553 132233637768320 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 7.442639s (all_items=0.000040s, per_item={'items': '0.00004029'}, temp_paths=7.442598)
I0421 16:01:35.928030 132149441201728 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0421 16:01:35.928139 132233637768320 async_checkpointer.py:561] Finished blocking save. Time taken: 7.534628s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/10.
I0421 16:01:35.928415 132233637768320 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0421 16:01:35.928619 132149409744448 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0421 16:01:35.928708 132233637768320 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776787288.3932729, 'wait_for_prev_duration_secs': 4.220008850097656e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776787288.393478, 'checkpointer_blocking_duration_secs': 7.534809112548828, 'get_old_steps_start_time': 1776787295.928314, 'get_old_steps_duration_secs': 4.673004150390625e-05, 'checkpoint_manager_blocking_start_time': 1776787288.393134, 'checkpoint_manager_blocking_duration_secs': 7.5355448722839355}
I0421 16:01:35.928802 132233637768320 checkpointing.py:408] Started an asynchronous checkpoint save for step 10
I0421 16:01:35.928854 132233637768320 max_utils.py:750] 
Memstats: After params initialized:
I0421 16:01:35.928913 132233637768320 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0))
I0421 16:01:35.928941 132233637768320 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0))
I0421 16:01:35.928963 132233637768320 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0))
I0421 16:01:35.928982 132233637768320 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0))
I0421 16:01:35.929001 132233637768320 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0))
I0421 16:01:35.929017 132233637768320 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0))
I0421 16:01:35.929033 132233637768320 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0))
I0421 16:01:35.929070 132233637768320 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0))
I0421 16:01:36.183987 132233637768320 metric_logger.py:196] completed step: 10, seconds: 1.706, TFLOP/s/device: 7.967, Tokens/s/device: 1200.808, total_weights: 16384, loss: 3.602, lm_loss: 3.602, perplexity: 36.686
I0421 16:01:36.300653 132233637768320 metric_logger.py:196] completed step: 11, seconds: 7.907, TFLOP/s/device: 1.718, Tokens/s/device: 259.023, total_weights: 16384, loss: 3.101, lm_loss: 3.101, perplexity: 22.211
I0421 16:01:36.419397 132233637768320 metric_logger.py:196] completed step: 12, seconds: 0.005, TFLOP/s/device: 2494.424, Tokens/s/device: 375986.782, total_weights: 16384, loss: 2.684, lm_loss: 2.684, perplexity: 14.644
I0421 16:01:36.538111 132233637768320 metric_logger.py:196] completed step: 13, seconds: 0.117, TFLOP/s/device: 115.663, Tokens/s/device: 17433.942, total_weights: 16384, loss: 2.348, lm_loss: 2.348, perplexity: 10.461
I0421 16:01:36.542685 132233637768320 checkpointing.py:772] Waiting for step 15 to finish before checkpoint...
I0421 16:01:36.702596 132149399258688 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/10/items/array_metadatas/process_0
I0421 16:01:36.774765 132233637768320 checkpointing.py:776] Waited 0.23205256462097168 seconds for step 15 to finish before starting checkpointing.
I0421 16:01:36.775164 132233637768320 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0421 16:02:11.853257 132149420230208 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 35.926127s (commit=35.496113s, array_metadata_write=0.430014s)
I0421 16:02:11.854388 132149441201728 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 291.374 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 43 seconds) (per-host)
I0421 16:02:11.854445 132149441201728 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 35.926249s.
I0421 16:02:12.325568 132149441201728 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0421 16:02:12.520030 132149384578624 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776787289239418287, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/10/_CHECKPOINT_METADATA
I0421 16:02:12.745015 132149441201728 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0421 16:02:12.745641 132149441201728 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.564230s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/10/items
I0421 16:02:12.746318 132149441201728 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/10/items
I0421 16:02:12.980860 132149441201728 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/10
I0421 16:02:13.681816 132149441201728 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/10`.
I0421 16:02:13.682428 132149441201728 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 45.288935s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/10
I0421 16:02:13.682506 132149441201728 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 37.754311s.
I0421 16:02:13.682620 132149409744448 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0421 16:02:13.682672 132149409744448 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0421 16:02:13.682734 132149409744448 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0421 16:02:13.682777 132149409744448 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0421 16:02:13.682925 132233637768320 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10.
W0421 16:02:13.683062 132233637768320 checkpoint_manager.py:1441] Waiting for previous save to complete took 36.907889 seconds. If this number is high, consider checkpointing less frequently.
I0421 16:02:13.683849 132233637768320 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 15
I0421 16:02:13.684166 132233637768320 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/15.
I0421 16:02:13.856783 132233637768320 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array.
I0421 16:02:13.856918 132149409744448 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/15
I0421 16:02:13.857001 132233637768320 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
I0421 16:02:14.567358 132149374092864 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/15/items
I0421 16:02:15.302501 132233637768320 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.446367s
I0421 16:02:15.303029 132233637768320 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 8.004 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host)
I0421 16:02:15.303096 132233637768320 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.541836s (batch_requests_ready=0.091705s, total_serialization_initiated=1.449701s, others=0.000431s)
I0421 16:02:15.303161 132233637768320 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.542385s (all_items=0.000021s, per_item={'items': '0.00002050'}, temp_paths=1.542365)
I0421 16:02:15.308712 132149441201728 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0421 16:02:15.308806 132233637768320 async_checkpointer.py:561] Finished blocking save. Time taken: 1.624911s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/15.
I0421 16:02:15.309080 132233637768320 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0421 16:02:15.309267 132149420230208 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0421 16:02:15.309369 132233637768320 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776787296.7751358, 'wait_for_prev_duration_secs': 36.907888889312744, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776787333.6838744, 'checkpointer_blocking_duration_secs': 1.6251037120819092, 'get_old_steps_start_time': 1776787335.3089995, 'get_old_steps_duration_secs': 2.47955322265625e-05, 'checkpoint_manager_blocking_start_time': 1776787296.77504, 'checkpoint_manager_blocking_duration_secs': 38.53430461883545}
I0421 16:02:15.309459 132233637768320 checkpointing.py:408] Started an asynchronous checkpoint save for step 15
I0421 16:02:15.310091 132233637768320 metric_logger.py:196] completed step: 14, seconds: 0.119, TFLOP/s/device: 113.886, Tokens/s/device: 17166.087, total_weights: 16384, loss: 2.083, lm_loss: 2.083, perplexity: 8.030
I0421 16:02:15.317569 132233637768320 metric_logger.py:196] completed step: 15, seconds: 0.118, TFLOP/s/device: 115.016, Tokens/s/device: 17336.539, total_weights: 16384, loss: 1.880, lm_loss: 1.880, perplexity: 6.551
I0421 16:02:15.434682 132233637768320 metric_logger.py:196] completed step: 16, seconds: 38.773, TFLOP/s/device: 0.350, Tokens/s/device: 52.820, total_weights: 16384, loss: 1.726, lm_loss: 1.726, perplexity: 5.616
I0421 16:02:15.553651 132233637768320 metric_logger.py:196] completed step: 17, seconds: 0.006, TFLOP/s/device: 2226.668, Tokens/s/device: 335627.663, total_weights: 16384, loss: 1.610, lm_loss: 1.610, perplexity: 5.002
I0421 16:02:15.672492 132233637768320 metric_logger.py:196] completed step: 18, seconds: 0.118, TFLOP/s/device: 115.435, Tokens/s/device: 17399.579, total_weights: 16384, loss: 1.521, lm_loss: 1.521, perplexity: 4.577
I0421 16:02:15.791003 132233637768320 checkpointing.py:772] Waiting for step 19 to finish before checkpoint...
I0421 16:02:15.791805 132233637768320 checkpointing.py:776] Waited 0.0008208751678466797 seconds for step 19 to finish before starting checkpointing.
I0421 16:02:15.792037 132233637768320 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0421 16:02:16.065150 132150024209984 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/15/items/array_metadatas/process_0
I0421 16:03:00.766292 132149374092864 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 45.458383s (commit=44.954510s, array_metadata_write=0.503873s)
I0421 16:03:00.767449 132149441201728 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 268.833 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 47 seconds) (per-host)
I0421 16:03:00.767554 132149441201728 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.458704s.
I0421 16:03:01.214318 132149441201728 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0421 16:03:01.651987 132149441201728 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0421 16:03:01.652608 132149441201728 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.573636s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/15/items
I0421 16:03:01.653272 132149441201728 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/15/items
I0421 16:03:01.902982 132149441201728 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/15
I0421 16:03:02.579482 132149441201728 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/15`.
I0421 16:03:02.580140 132149441201728 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 48.896251s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/15
I0421 16:03:02.580220 132149441201728 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 47.271371s.
I0421 16:03:02.580338 132149420230208 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0421 16:03:02.580390 132149420230208 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0421 16:03:02.580448 132149420230208 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0421 16:03:02.580486 132149420230208 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0421 16:03:02.580637 132233637768320 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15.
W0421 16:03:02.580750 132233637768320 checkpoint_manager.py:1441] Waiting for previous save to complete took 46.788719 seconds. If this number is high, consider checkpointing less frequently.
I0421 16:03:02.580975 132233637768320 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 19
I0421 16:03:02.581279 132233637768320 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/19.
I0421 16:03:02.762236 132149441201728 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/19
I0421 16:03:02.775786 132233637768320 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array.
I0421 16:03:02.775882 132233637768320 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
I0421 16:03:03.488871 132149982266944 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/19/items
I0421 16:03:09.378713 132233637768320 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 6.603249s
I0421 16:03:09.383473 132233637768320 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.840 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host)
I0421 16:03:09.383535 132233637768320 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 6.708760s (batch_requests_ready=0.097111s, total_serialization_initiated=6.607067s, others=0.004581s)
I0421 16:03:09.383610 132233637768320 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 6.709251s (all_items=0.000016s, per_item={'items': '0.00001574'}, temp_paths=6.709235)
I0421 16:03:09.385380 132149950809664 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0421 16:03:09.385475 132233637768320 async_checkpointer.py:561] Finished blocking save. Time taken: 6.804457s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/19.
I0421 16:03:09.385670 132233637768320 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0421 16:03:09.385834 132149420230208 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0421 16:03:09.385945 132233637768320 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776787335.7920146, 'wait_for_prev_duration_secs': 46.788718938827515, 'time_between_consecutive_saves_sec': 2.1092214584350586, 'checkpointer_blocking_start_time': 1776787382.5809982, 'checkpointer_blocking_duration_secs': 6.804581880569458, 'get_old_steps_start_time': 1776787389.3855963, 'get_old_steps_duration_secs': 2.4557113647460938e-05, 'checkpoint_manager_blocking_start_time': 1776787335.7919831, 'checkpoint_manager_blocking_duration_secs': 53.593936920166016}
I0421 16:03:09.386072 132233637768320 checkpointing.py:408] Started an asynchronous checkpoint save for step 19
I0421 16:03:09.386109 132233637768320 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0421 16:03:10.121494 132149409744448 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/19/items/array_metadatas/process_0
I0421 16:03:51.946506 132150024209984 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.562135s (commit=42.117436s, array_metadata_write=0.444699s)
I0421 16:03:51.947543 132149950809664 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 256.466 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 49 seconds) (per-host)
I0421 16:03:51.947642 132149950809664 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.562123s.
I0421 16:03:52.326177 132149950809664 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0421 16:03:52.750290 132149950809664 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0421 16:03:52.751112 132149950809664 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.557000s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/19/items
I0421 16:03:52.751688 132149950809664 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/19/items
I0421 16:03:52.993887 132149950809664 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/19
I0421 16:03:53.685222 132149950809664 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/19`.
I0421 16:03:53.685864 132149950809664 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 51.104854s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_true/checkpoints/19
I0421 16:03:53.685926 132149950809664 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 44.300409s.
I0421 16:03:53.686116 132149420230208 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0421 16:03:53.686218 132149420230208 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0421 16:03:53.686270 132149420230208 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0421 16:03:53.686308 132149420230208 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0421 16:03:53.686671 132233637768320 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19.
I0421 16:03:53.686799 132233637768320 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0421 16:03:53.687704 132233637768320 metric_logger.py:196] completed step: 19, seconds: 0.119, TFLOP/s/device: 114.364, Tokens/s/device: 17238.187, total_weights: 16384, loss: 1.449, lm_loss: 1.449, perplexity: 4.260
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops

2026-04-23 02:30:48.641258: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
E0000 00:00:1776911448.654239 2326285 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1776911448.658060 2326285 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1776911448.669535 2326285 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776911448.669548 2326285 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776911448.669550 2326285 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776911448.669552 2326285 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'}
`rope_scaling`'s factor field must be a float >= 1, got 40
`rope_scaling`'s beta_fast field must be a float, got 32
`rope_scaling`'s beta_slow field must be a float, got 1
Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'}
Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'}
Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'}
2026-04-23 02:30:52.337599: 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)
I0423 02:30:52.473735 139138244671296 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag.
I0423 02:31:52.479078 139138244671296 max_utils.py:800] System Information: Jax Version: 0.8.1
I0423 02:31:52.479344 139138244671296 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0423 02:31:52.479397 139138244671296 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709
I0423 02:31:52.479428 139138244671296 train_utils.py:348] WARNING: 'dataset_path' might be pointing your local file system
I0423 02:31:52.479450 139138244671296 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0423 02:31:52.479621 139138244671296 train.py:703] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper.
I0423 02:31:53.137200 139138244671296 maxtext_utils.py:1604] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0423 02:31:53.137468 139138244671296 checkpointing.py:677] Setting up checkpoint logger...
I0423 02:31:53.137512 139138244671296 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0423 02:31:53.137548 139138244671296 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0423 02:31:53.137994 139138244671296 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 0x7e8b5e1a95e0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0423 02:31:54.781746 139138244671296 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0423 02:31:54.782950 139138244671296 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 0x7e89fed1f680>}, handler_registry=None
I0423 02:31:54.783351 139138244671296 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e89fed1f680>` 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`.
I0423 02:31:54.783400 139138244671296 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e898974a330>` 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`.
I0423 02:31:54.783431 139138244671296 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 0x7e89fed1f680>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e89fed1f680>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e898974a330>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e898974a330>}).
I0423 02:31:54.783715 139138244671296 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28
I0423 02:31:54.783771 139138244671296 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7e8989760900> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0423 02:31:55.216796 139108464723520 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776911339473632212, 'commit_timestamp_nsecs': 1776911375556210504, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/5/_CHECKPOINT_METADATA
I0423 02:31:55.298233 139108464723520 checkpoint_manager.py:1847] Missing metrics for step 5
I0423 02:31:55.373054 139138244671296 checkpoint_manager.py:1788] Found 3 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints
I0423 02:31:55.373656 139138244671296 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_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e8b73b37e60>
I0423 02:31:55.373755 139138244671296 checkpointing.py:301] Checkpoint manager created!
I0423 02:31:55.757609 139138244671296 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0423 02:31:55.757712 139138244671296 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0423 02:31:56.038444 139138244671296 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 02:31:56.038536 139138244671296 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0423 02:31:56.050769 139138244671296 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 02:31:56.050821 139138244671296 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0423 02:31:56.069813 139138244671296 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 02:31:56.069872 139138244671296 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0423 02:31:56.082069 139138244671296 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 02:31:56.082119 139138244671296 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0423 02:31:56.094274 139138244671296 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 02:31:56.094326 139138244671296 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0423 02:31:56.113610 139138244671296 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0423 02:31:56.113670 139138244671296 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0423 02:31:56.129018 139138244671296 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0423 02:31:56.129089 139138244671296 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None).
I0423 02:31:56.278826 139138244671296 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0423 02:31:56.278928 139138244671296 checkpointing.py:581] restoring from this run's directory step 9
I0423 02:31:56.454830 139138244671296 checkpointer.py:304] Restoring checkpoint from gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/9.
I0423 02:31:57.041187 2327800 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json
I0423 02:31:57.041260 2327800 google_auth_provider.cc:156] Using OAuth2 AuthProvider
I0423 02:32:05.322137 139138244671296 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/read/gbytes_per_sec: 1.449 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 8 seconds) (per-host)
I0423 02:32:05.322523 139138244671296 checkpointer.py:318] Finished restoring checkpoint in 8.96 seconds from gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/9.
I0423 02:32:05.322613 139138244671296 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints', 'checkpointer_start_time': 1776911516.3634734, 'checkpointer_duration_secs': 8.959100723266602, 'checkpoint_manager_start_time': 1776911516.2791557, 'checkpoint_manager_duration_secs': 9.04342246055603}
[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
I0423 02:32:05.696537 139138244671296 maxtext_utils.py:1707]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0423 02:32:05.696654 139138244671296 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0423 02:32:05.696696 139138244671296 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0423 02:32:05.696741 139138244671296 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   PartitionSpec('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0423 02:32:05.696783 139138244671296 maxtext_utils.py:1707]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  (None, None)
I0423 02:32:05.696813 139138244671296 maxtext_utils.py:1707]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  (None, None)
I0423 02:32:05.696853 139138244671296 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/key/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0423 02:32:05.696893 139138244671296 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/out/kernel
    Shape:     float32[16,16,128,2048]
    Logical:   PartitionSpec('heads', 'layers', 'kv', 'embed')
    Physical:  (None, None, None, 'fsdp')
I0423 02:32:05.696922 139138244671296 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/query/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None, None)
I0423 02:32:05.696949 139138244671296 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/value/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0423 02:32:05.696992 139138244671296 maxtext_utils.py:1707]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)
I0423 02:32:05.697031 139138244671296 maxtext_utils.py:1707]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')

I0423 02:32:06.098791 139138244671296 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0423 02:32:06.098881 139138244671296 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None).
I0423 02:32:06.110442 139138244671296 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0423 02:32:06.110494 139138244671296 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None).
I0423 02:32:06.928644 139138244671296 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.
I0423 02:32:06.932604 139138244671296 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter.
I0423 02:32:06.932886 139138244671296 metric_logger.py:301] number parameters: 1.104 billion
I0423 02:32:09.198570 139138244671296 checkpointing.py:772] Waiting for step 10 to finish before checkpoint...
I0423 02:32:09.316007 139138244671296 checkpointing.py:776] Waited 0.11741209030151367 seconds for step 10 to finish before starting checkpointing.
I0423 02:32:09.316692 139138244671296 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 02:32:09.316863 139138244671296 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 10
I0423 02:32:09.317201 139138244671296 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/10.
I0423 02:32:09.412326 139138244671296 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient
I0423 02:32:09.502619 139106260616768 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/10
I0423 02:32:09.517823 139138244671296 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array.
I0423 02:32:09.517928 139138244671296 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
I0423 02:32:10.234248 139106250131008 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/10/items
I0423 02:32:10.404625 139106218673728 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776911530144210083, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776911530144210083, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/10/_CHECKPOINT_METADATA
I0423 02:32:12.286269 139138244671296 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 2.768919s
I0423 02:32:12.291143 139138244671296 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.289 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 2 seconds) (per-host)
I0423 02:32:12.291209 139138244671296 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 2.877521s (batch_requests_ready=0.100870s, total_serialization_initiated=2.771891s, others=0.004760s)
I0423 02:32:12.291289 139138244671296 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.878414s (all_items=0.000021s, per_item={'items': '0.00002074'}, temp_paths=2.878393)
I0423 02:32:12.293008 139106271102528 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0423 02:32:12.293144 139138244671296 async_checkpointer.py:561] Finished blocking save. Time taken: 2.976241s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/10.
I0423 02:32:12.293282 139138244671296 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 02:32:12.293473 139106260616768 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0423 02:32:12.293601 139138244671296 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776911529.3166788, 'wait_for_prev_duration_secs': 4.100799560546875e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776911529.316885, 'checkpointer_blocking_duration_secs': 2.9763190746307373, 'get_old_steps_start_time': 1776911532.2932193, 'get_old_steps_duration_secs': 2.6464462280273438e-05, 'checkpoint_manager_blocking_start_time': 1776911529.3165958, 'checkpoint_manager_blocking_duration_secs': 2.976979970932007}
I0423 02:32:12.293676 139138244671296 checkpointing.py:408] Started an asynchronous checkpoint save for step 10
I0423 02:32:12.293713 139138244671296 max_utils.py:750] 
Memstats: After params initialized:
I0423 02:32:12.293767 139138244671296 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0))
I0423 02:32:12.293797 139138244671296 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0))
I0423 02:32:12.293817 139138244671296 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0))
I0423 02:32:12.293834 139138244671296 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0))
I0423 02:32:12.293852 139138244671296 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0))
I0423 02:32:12.293868 139138244671296 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0))
I0423 02:32:12.293885 139138244671296 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0))
I0423 02:32:12.293901 139138244671296 max_utils.py:756] 	Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0))
I0423 02:32:12.539784 139138244671296 metric_logger.py:196] completed step: 10, seconds: 2.265, TFLOP/s/device: 5.999, Tokens/s/device: 904.235, total_weights: 16384, loss: 3.602, lm_loss: 3.602, perplexity: 36.686
I0423 02:32:12.656822 139138244671296 metric_logger.py:196] completed step: 11, seconds: 3.339, TFLOP/s/device: 4.069, Tokens/s/device: 613.331, total_weights: 16384, loss: 3.101, lm_loss: 3.101, perplexity: 22.211
I0423 02:32:12.775943 139138244671296 metric_logger.py:196] completed step: 12, seconds: 0.006, TFLOP/s/device: 2198.565, Tokens/s/device: 331391.586, total_weights: 16384, loss: 2.684, lm_loss: 2.684, perplexity: 14.644
I0423 02:32:12.895107 139138244671296 metric_logger.py:196] completed step: 13, seconds: 0.118, TFLOP/s/device: 115.032, Tokens/s/device: 17338.887, total_weights: 16384, loss: 2.348, lm_loss: 2.348, perplexity: 10.461
I0423 02:32:12.900716 139138244671296 checkpointing.py:772] Waiting for step 15 to finish before checkpoint...
I0423 02:32:13.039817 139106229159488 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/10/items/array_metadatas/process_0
I0423 02:32:13.132140 139138244671296 checkpointing.py:776] Waited 0.23138689994812012 seconds for step 15 to finish before starting checkpointing.
I0423 02:32:13.133166 139138244671296 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 02:32:53.080430 139106250131008 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 40.788490s (commit=40.333812s, array_metadata_write=0.454679s)
I0423 02:32:53.081533 139106271102528 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 289.384 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 43 seconds) (per-host)
I0423 02:32:53.081587 139106271102528 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.788480s.
I0423 02:32:53.572509 139106271102528 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0423 02:32:53.755963 139106218673728 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776911530144210083, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/10/_CHECKPOINT_METADATA
I0423 02:32:53.991172 139106271102528 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0423 02:32:53.991409 139106271102528 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.606405s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/10/items
I0423 02:32:53.992171 139106271102528 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/10/items
I0423 02:32:54.228625 139106271102528 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/10
I0423 02:32:54.939112 139106271102528 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/10`.
I0423 02:32:54.939940 139106271102528 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 45.623043s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/10
I0423 02:32:54.940031 139106271102528 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 42.646924s.
I0423 02:32:54.940250 139106260616768 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 02:32:54.940367 139106260616768 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0423 02:32:54.940418 139106260616768 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0423 02:32:54.940469 139106260616768 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0423 02:32:54.940631 139138244671296 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10.
W0423 02:32:54.940771 139138244671296 checkpoint_manager.py:1441] Waiting for previous save to complete took 41.807607 seconds. If this number is high, consider checkpointing less frequently.
I0423 02:32:54.941472 139138244671296 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 15
I0423 02:32:54.941782 139138244671296 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/15.
I0423 02:32:55.121317 139138244671296 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array.
I0423 02:32:55.121533 139138244671296 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
I0423 02:32:55.122685 139106271102528 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/15
I0423 02:32:55.810657 139106229159488 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/15/items
I0423 02:32:58.967482 139138244671296 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 3.846617s
I0423 02:32:58.968147 139138244671296 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.133 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 3 seconds) (per-host)
I0423 02:32:58.968212 139138244671296 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 3.939504s (batch_requests_ready=0.088853s, total_serialization_initiated=3.850157s, others=0.000494s)
I0423 02:32:58.968292 139138244671296 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 3.940064s (all_items=0.000014s, per_item={'items': '0.00001383'}, temp_paths=3.940050)
I0423 02:32:58.973863 139108273882688 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0423 02:32:58.974059 139138244671296 async_checkpointer.py:561] Finished blocking save. Time taken: 4.032527s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/15.
I0423 02:32:58.974234 139138244671296 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 02:32:58.974482 139108242425408 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0423 02:32:58.974609 139138244671296 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776911533.1331403, 'wait_for_prev_duration_secs': 41.80760717391968, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776911574.9414968, 'checkpointer_blocking_duration_secs': 4.032630443572998, 'get_old_steps_start_time': 1776911578.9741495, 'get_old_steps_duration_secs': 3.4332275390625e-05, 'checkpoint_manager_blocking_start_time': 1776911533.1330748, 'checkpoint_manager_blocking_duration_secs': 45.84150767326355}
I0423 02:32:58.974750 139138244671296 checkpointing.py:408] Started an asynchronous checkpoint save for step 15
I0423 02:32:58.975399 139138244671296 metric_logger.py:196] completed step: 14, seconds: 0.119, TFLOP/s/device: 113.789, Tokens/s/device: 17151.567, total_weights: 16384, loss: 2.083, lm_loss: 2.083, perplexity: 8.030
I0423 02:32:58.984204 139138244671296 metric_logger.py:196] completed step: 15, seconds: 0.119, TFLOP/s/device: 114.096, Tokens/s/device: 17197.800, total_weights: 16384, loss: 1.880, lm_loss: 1.880, perplexity: 6.551
I0423 02:32:59.100393 139138244671296 metric_logger.py:196] completed step: 16, seconds: 46.081, TFLOP/s/device: 0.295, Tokens/s/device: 44.443, total_weights: 16384, loss: 1.726, lm_loss: 1.726, perplexity: 5.616
I0423 02:32:59.219419 139138244671296 metric_logger.py:196] completed step: 17, seconds: 0.007, TFLOP/s/device: 2089.685, Tokens/s/device: 314980.006, total_weights: 16384, loss: 1.610, lm_loss: 1.610, perplexity: 5.002
I0423 02:32:59.338272 139138244671296 metric_logger.py:196] completed step: 18, seconds: 0.118, TFLOP/s/device: 115.526, Tokens/s/device: 17413.337, total_weights: 16384, loss: 1.521, lm_loss: 1.521, perplexity: 4.577
I0423 02:32:59.456724 139138244671296 checkpointing.py:772] Waiting for step 19 to finish before checkpoint...
I0423 02:32:59.457781 139138244671296 checkpointing.py:776] Waited 0.0010912418365478516 seconds for step 19 to finish before starting checkpointing.
I0423 02:32:59.458071 139138244671296 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 02:32:59.711601 139108284368448 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/15/items/array_metadatas/process_0
I0423 02:33:41.112714 139106229159488 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.139571s (commit=41.688033s, array_metadata_write=0.451539s)
I0423 02:33:41.113960 139108273882688 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 274.205 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 46 seconds) (per-host)
I0423 02:33:41.114099 139108273882688 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.140141s.
I0423 02:33:41.597220 139108273882688 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0423 02:33:42.036329 139108273882688 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0423 02:33:42.036558 139108273882688 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.603591s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/15/items
I0423 02:33:42.037285 139108273882688 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/15/items
I0423 02:33:42.271153 139108273882688 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/15
I0423 02:33:42.998947 139108273882688 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/15`.
I0423 02:33:42.999604 139108273882688 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 48.058094s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/15
I0423 02:33:42.999677 139108273882688 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 44.025723s.
I0423 02:33:42.999830 139108242425408 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 02:33:42.999945 139108242425408 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0423 02:33:43.000009 139108242425408 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0423 02:33:43.000067 139108242425408 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0423 02:33:43.000274 139138244671296 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15.
W0423 02:33:43.000404 139138244671296 checkpoint_manager.py:1441] Waiting for previous save to complete took 43.542357 seconds. If this number is high, consider checkpointing less frequently.
I0423 02:33:43.000803 139138244671296 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 19
I0423 02:33:43.001142 139138244671296 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/19.
I0423 02:33:43.170872 139138244671296 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array.
I0423 02:33:43.170978 139138244671296 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
I0423 02:33:43.195584 139108273882688 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/19
I0423 02:33:43.951565 139108284368448 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/19/items
I0423 02:33:46.633905 139138244671296 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 3.463428s
I0423 02:33:46.638070 139138244671296 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.470 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 3 seconds) (per-host)
I0423 02:33:46.638130 139138244671296 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 3.556309s (batch_requests_ready=0.084945s, total_serialization_initiated=3.467383s, others=0.003980s)
I0423 02:33:46.638204 139138244671296 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 3.556882s (all_items=0.000015s, per_item={'items': '0.00001526'}, temp_paths=3.556867)
I0423 02:33:46.639809 139106229159488 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0423 02:33:46.639892 139138244671296 async_checkpointer.py:561] Finished blocking save. Time taken: 3.639046s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/19.
I0423 02:33:46.640081 139138244671296 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 02:33:46.640236 139108242425408 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0423 02:33:46.640350 139138244671296 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776911579.458031, 'wait_for_prev_duration_secs': 43.542356729507446, 'time_between_consecutive_saves_sec': 4.517547369003296, 'checkpointer_blocking_start_time': 1776911623.0008278, 'checkpointer_blocking_duration_secs': 3.6391658782958984, 'get_old_steps_start_time': 1776911626.6400056, 'get_old_steps_duration_secs': 2.2649765014648438e-05, 'checkpoint_manager_blocking_start_time': 1776911579.457999, 'checkpoint_manager_blocking_duration_secs': 47.182329177856445}
I0423 02:33:46.640440 139138244671296 checkpointing.py:408] Started an asynchronous checkpoint save for step 19
I0423 02:33:46.640472 139138244671296 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 02:33:47.366662 139108231939648 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/19/items/array_metadatas/process_0
I0423 02:34:20.958642 139108273882688 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 34.319855s (commit=33.851048s, array_metadata_write=0.468807s)
I0423 02:34:20.959766 139106229159488 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 333.619 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 37 seconds) (per-host)
I0423 02:34:20.959873 139106229159488 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 34.319937s.
I0423 02:34:21.474884 139106229159488 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0423 02:34:21.890297 139106229159488 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0423 02:34:21.890478 139106229159488 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.583814s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/19/items
I0423 02:34:21.891362 139106229159488 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/19/items
I0423 02:34:22.126785 139106229159488 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/19
I0423 02:34:22.831663 139106229159488 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/19`.
I0423 02:34:22.832431 139106229159488 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 39.831591s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_16_zarr3_true/checkpoints/19
I0423 02:34:22.832492 139106229159488 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 36.192558s.
I0423 02:34:22.832703 139108242425408 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 02:34:22.832823 139108242425408 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0423 02:34:22.832883 139108242425408 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0423 02:34:22.832923 139108242425408 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0423 02:34:22.833093 139138244671296 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19.
I0423 02:34:22.833261 139138244671296 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 02:34:22.834218 139138244671296 metric_logger.py:196] completed step: 19, seconds: 0.119, TFLOP/s/device: 113.986, Tokens/s/device: 17181.208, total_weights: 16384, loss: 1.449, lm_loss: 1.449, perplexity: 4.260
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops