MaxView

Timing Test Reference Updates
‹ 15_ocdbt_false_saveCase: 15_ocdbt_true_resume15_ocdbt_true_save ›

Metrics: main (574ad3fb9) vs test/pipeline-scan-nnx (e27fc1e97)

Metricmain  574ad3fb9test/pipeline-scan-nnx  e27fc1e97Diff (test/pipeline-scan-nnx − main)
Parameters1.104 billion1.104 billion—
Final loss5.39105.39100
TFLOP/s91.90591.836-0.069
Tok/s13852.913842.5-10.394
Avg s/step8.4858.298-0.187
Memory %1.381.380
JAX0.9.20.8.1—

Diff = branch value − main value. Green = branch improved. Red = branch regressed.

main  ·  574ad3fb9  ·  main_20260420_060028  ·  full log 
XPK Start: Mon Apr 20 07:04:10 UTC 2026
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
2026-04-20 07:04:34.674009: 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)
I0420 07:04:34.855295 139502449792832 max_utils.py:273] Attempting to initialize the jax distributed system...
I0420 07:04:43.895489 139502449792832 distributed.py:149] Starting JAX distributed service on [::]:8482
I0420 07:04:43.897724 139502449792832 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-true-rs-3ikq6-slice-job-0-0.mt-15-ocdbt-true-rs-3ikq6:8482
I0420 07:04:45.234313 139502449792832 max_utils.py:284] Jax distributed system initialized!
I0420 07:04:51.723086 139502449792832 max_utils.py:800] System Information: Jax Version: 0.9.2
I0420 07:04:51.723188 139502449792832 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0420 07:04:51.723231 139502449792832 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365
I0420 07:04:51.723273 139502449792832 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0420 07:04:52.426061 139502449792832 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0420 07:04:52.426362 139502449792832 checkpointing.py:677] Setting up checkpoint logger...
I0420 07:04:52.426421 139502449792832 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0420 07:04:52.426483 139502449792832 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0420 07:04:52.426878 139502449792832 base_pytree_checkpoint_handler.py:441] 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 0x7edfcef1cc50>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0420 07:04:53.794626 139502449792832 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0420 07:04:53.794863 139502449792832 checkpoint_manager.py:708] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7eca7c505370>}, handler_registry=None
I0420 07:04:53.795136 139502449792832 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7eca7c505370>` 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`.
I0420 07:04:53.795195 139502449792832 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7edeadd3cc80>` 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`.
I0420 07:04:53.795241 139502449792832 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 0x7eca7c505370>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7eca7c505370>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7edeadd3cc80>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7edeadd3cc80>}).
I0420 07:04:53.795622 139502449792832 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0420 07:04:53.795696 139502449792832 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7eca7c1b7d80> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0420 07:04:55.692311 139379842463488 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776668529904500213, 'commit_timestamp_nsecs': 1776668577814483129, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/9/_CHECKPOINT_METADATA
I0420 07:04:55.944305 139379842463488 checkpoint_manager.py:1871] Missing metrics for step 9
I0420 07:04:55.988883 139502449792832 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints
I0420 07:04:56.544683 139502449792832 checkpoint_manager.py:929] [process=4][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_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, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7eca7c59e450>
I0420 07:04:56.544853 139502449792832 checkpointing.py:301] Checkpoint manager created!
I0420 07:04:57.462141 139502449792832 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0420 07:04:57.462253 139502449792832 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0420 07:04:57.840832 139502449792832 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0420 07:04:57.840924 139502449792832 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0420 07:04:57.857082 139502449792832 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0420 07:04:57.857145 139502449792832 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0420 07:04:57.881572 139502449792832 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0420 07:04:57.881640 139502449792832 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0420 07:04:57.897986 139502449792832 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0420 07:04:57.898044 139502449792832 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0420 07:04:57.914187 139502449792832 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0420 07:04:57.914250 139502449792832 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0420 07:04:57.939591 139502449792832 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0420 07:04:57.939670 139502449792832 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0420 07:04:57.960627 139502449792832 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0420 07:04:57.960687 139502449792832 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0420 07:04:58.179750 139502449792832 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0420 07:04:58.179864 139502449792832 checkpointing.py:581] restoring from this run's directory step 9
I0420 07:04:58.430176 139502449792832 event_tracking.py:70] [process=4] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/9.
I0420 07:04:58.685937 139502449792832 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/9.
I0420 07:04:58.686375 139502449792832 event_tracking.py:125] [process=4] [sync] Finished blocking load in 0.26 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/9.
I0420 07:05:00.088348    1469 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0420 07:05:05.660292 139502449792832 jax_array_handlers.py:843] [process=4] /jax/orbax/read/worker/io/requested throughput: 283.317 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 5.57775616645813 s) (per-host)
I0420 07:05:05.664913 139502449792832 base_pytree_checkpoint_handler.py:130] [process=4] /jax/checkpoint/read/gbytes_per_sec: 264.773 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 5.977170944213867 s) (per-host)
I0420 07:05:08.687083 139502449792832 event_tracking.py:138] [process=4] [sync] Finished load in 10.26 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/9
I0420 07:05:08.687236 139502449792832 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints', 'checkpointer_start_time': 1776668698.430079, 'checkpointer_duration_secs': 10.25712537765503, 'checkpoint_manager_start_time': 1776668698.1801052, 'checkpoint_manager_duration_secs': 10.50710153579712}
fsdp: 32
I0420 07:05:09.098513 139502449792832 maxtext_utils.py:1654]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0420 07:05:09.098638 139502449792832 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0420 07:05:09.098691 139502449792832 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0420 07:05:09.098748 139502449792832 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   P('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0420 07:05:09.098800 139502449792832 maxtext_utils.py:1654]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0420 07:05:09.098843 139502449792832 maxtext_utils.py:1654]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0420 07:05:09.098902 139502449792832 maxtext_utils.py:1654]  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)
I0420 07:05:09.098955 139502449792832 maxtext_utils.py:1654]  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')
I0420 07:05:09.098996 139502449792832 maxtext_utils.py:1654]  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)
I0420 07:05:09.099034 139502449792832 maxtext_utils.py:1654]  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)
I0420 07:05:09.099081 139502449792832 maxtext_utils.py:1654]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)
I0420 07:05:09.099125 139502449792832 maxtext_utils.py:1654]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')

I0420 07:05:09.580388 139502449792832 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0420 07:05:09.580478 139502449792832 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0420 07:05:09.595714 139502449792832 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0420 07:05:09.595777 139502449792832 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0420 07:05:20.478641 139502449792832 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0420 07:05:20.479428 139502449792832 metric_logger.py:301] number parameters: 1.104 billion
I0420 07:05:32.002796 139502449792832 checkpointing.py:772] Waiting for step 10 to finish before checkpoint...
I0420 07:05:32.156572 139502449792832 checkpointing.py:776] Waited 0.15375709533691406 seconds for step 10 to finish before starting checkpointing.
I0420 07:05:32.158885 139502449792832 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0420 07:05:32.160897 139502449792832 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 10
I0420 07:05:32.162382 139502449792832 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/10.
I0420 07:05:32.499304 139502449792832 signaling_client.py:364] Using JaxDistributedSignalingClient
I0420 07:05:32.500390 139502449792832 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0420 07:05:32.500448 139502449792832 replica_slices.py:424] 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
I0420 07:05:32.777740 139502449792832 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278517s
I0420 07:05:32.777910 139502449792832 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 5.463 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28238344192504883 s) (per-host)
I0420 07:05:32.777966 139502449792832 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.282448s (batch_requests_ready=0.002156s, total_serialization_initiated=0.280219s, others=0.000073s)
I0420 07:05:32.778060 139502449792832 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.286598s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=0.286581)
I0420 07:05:32.778827 139502449792832 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.62 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/10.
I0420 07:05:32.779159 139379272054528 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:25:32.779123
I0420 07:05:32.787134 139502449792832 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0420 07:05:32.787389 139375513958144 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save.
I0420 07:05:32.787538 139502449792832 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776668732.15886, 'wait_for_prev_duration_secs': 6.127357482910156e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776668732.160936, 'checkpointer_blocking_duration_secs': 0.6183724403381348, 'get_old_steps_start_time': 1776668732.7793412, 'get_old_steps_duration_secs': 3.123283386230469e-05, 'checkpoint_manager_blocking_start_time': 1776668732.1569564, 'checkpoint_manager_blocking_duration_secs': 0.6305477619171143}
I0420 07:05:32.787649 139502449792832 checkpointing.py:408] Started an asynchronous checkpoint save for step 10
I0420 07:05:32.787701 139502449792832 max_utils.py:750] 
Memstats: After params initialized:
I0420 07:05:32.787753 139502449792832 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0))
I0420 07:05:32.787786 139502449792832 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0))
I0420 07:05:32.787821 139502449792832 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0))
I0420 07:05:32.787861 139502449792832 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0))
I0420 07:05:33.100154 139502449792832 metric_logger.py:196] completed step: 10, seconds: 11.523, TFLOP/s/device: 1.179, Tokens/s/device: 177.727, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869
I0420 07:05:33.273315 139502449792832 metric_logger.py:196] completed step: 11, seconds: 1.096, TFLOP/s/device: 12.400, Tokens/s/device: 1869.033, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659
I0420 07:05:33.420740 139502449792832 metric_logger.py:196] completed step: 12, seconds: 0.026, TFLOP/s/device: 529.816, Tokens/s/device: 79859.622, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068
I0420 07:05:33.568119 139502449792832 metric_logger.py:196] completed step: 13, seconds: 0.154, TFLOP/s/device: 88.471, Tokens/s/device: 13335.330, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001
I0420 07:05:33.572522 139502449792832 checkpointing.py:772] Waiting for step 15 to finish before checkpoint...
I0420 07:05:33.862511 139502449792832 checkpointing.py:776] Waited 0.28995633125305176 seconds for step 15 to finish before starting checkpointing.
I0420 07:05:33.865554 139502449792832 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0420 07:05:38.115698 139376428304128 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/10/items/array_metadatas/process_4
I0420 07:06:14.514570 139379272054528 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 37.592 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.01900768280029 s) (per-host)
I0420 07:06:14.514678 139379272054528 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.735412s.
I0420 07:06:23.639806 139379272054528 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.860524s.
I0420 07:06:23.640084 139375513958144 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save.
I0420 07:06:23.640201 139375513958144 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save.
I0420 07:06:23.640250 139375513958144 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0420 07:06:23.642159 139375513958144 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0420 07:06:23.642356 139502449792832 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10.
W0420 07:06:23.642576 139502449792832 checkpoint_manager.py:1452] Waiting for previous save to complete took 49.776946 seconds. If this number is high, consider checkpointing less frequently.
I0420 07:06:23.644462 139502449792832 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 15
I0420 07:06:23.646746 139502449792832 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/15.
I0420 07:06:24.358293 139502449792832 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0420 07:06:24.358402 139502449792832 replica_slices.py:424] 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
I0420 07:06:24.384248 139502449792832 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027188s
I0420 07:06:24.384394 139502449792832 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 50.397 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030608654022216797 s) (per-host)
I0420 07:06:24.384444 139502449792832 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030667s (batch_requests_ready=0.001763s, total_serialization_initiated=0.028840s, others=0.000065s)
I0420 07:06:24.384523 139502449792832 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.035573s (all_items=0.000013s, per_item={'items': '0.00001311'}, temp_paths=0.035560)
I0420 07:06:24.385167 139502449792832 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.74 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/15.
I0420 07:06:24.385429 139375513958144 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:26:24.385406
I0420 07:06:24.398324 139502449792832 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0420 07:06:24.398622 139376428304128 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save.
I0420 07:06:24.398806 139502449792832 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776668733.865512, 'wait_for_prev_duration_secs': 49.77694630622864, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776668783.6445036, 'checkpointer_blocking_duration_secs': 0.7410242557525635, 'get_old_steps_start_time': 1776668784.3855507, 'get_old_steps_duration_secs': 3.0279159545898438e-05, 'checkpoint_manager_blocking_start_time': 1776668733.8631551, 'checkpoint_manager_blocking_duration_secs': 50.53560209274292}
I0420 07:06:24.398936 139502449792832 checkpointing.py:408] Started an asynchronous checkpoint save for step 15
I0420 07:06:24.399791 139502449792832 metric_logger.py:196] completed step: 14, seconds: 0.147, TFLOP/s/device: 92.192, Tokens/s/device: 13896.145, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193
I0420 07:06:24.405954 139502449792832 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.263, Tokens/s/device: 13906.903, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661
I0420 07:06:29.655711 139373366474496 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/15/items/array_metadatas/process_4
I0420 07:06:45.173888 139502449792832 metric_logger.py:196] completed step: 16, seconds: 50.832, TFLOP/s/device: 0.267, Tokens/s/device: 40.290, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872
I0420 07:06:45.321469 139502449792832 metric_logger.py:196] completed step: 17, seconds: 20.622, TFLOP/s/device: 0.659, Tokens/s/device: 99.313, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975
I0420 07:06:45.469204 139502449792832 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.097, Tokens/s/device: 13429.596, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333
I0420 07:06:45.616024 139502449792832 checkpointing.py:772] Waiting for step 19 to finish before checkpoint...
I0420 07:06:45.616745 139502449792832 checkpointing.py:776] Waited 0.0007398128509521484 seconds for step 19 to finish before starting checkpointing.
I0420 07:06:45.618909 139502449792832 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0420 07:07:05.988435 139375513958144 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 37.940 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.634613275527954 s) (per-host)
I0420 07:07:05.988548 139375513958144 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.603045s.
I0420 07:07:14.840983 139375513958144 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.455464s.
I0420 07:07:14.841271 139376428304128 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save.
I0420 07:07:14.841418 139376428304128 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save.
I0420 07:07:14.841469 139376428304128 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0420 07:07:14.843204 139376428304128 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0420 07:07:14.843389 139502449792832 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15.
W0420 07:07:14.843512 139502449792832 checkpoint_manager.py:1452] Waiting for previous save to complete took 29.224610 seconds. If this number is high, consider checkpointing less frequently.
I0420 07:07:14.844974 139502449792832 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 19
I0420 07:07:14.846941 139502449792832 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/19.
I0420 07:07:15.560478 139502449792832 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0420 07:07:15.560569 139502449792832 replica_slices.py:424] 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
I0420 07:07:15.586678 139502449792832 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027126s
I0420 07:07:15.586809 139502449792832 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 50.501 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03054523468017578 s) (per-host)
I0420 07:07:15.586855 139502449792832 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030602s (batch_requests_ready=0.001640s, total_serialization_initiated=0.028901s, others=0.000061s)
I0420 07:07:15.586923 139502449792832 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034831s (all_items=0.000010s, per_item={'items': '0.00001049'}, temp_paths=0.034820)
I0420 07:07:15.587598 139502449792832 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.74 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/19.
I0420 07:07:15.587943 139376428304128 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:27:15.587904
I0420 07:07:15.591105 139502449792832 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0420 07:07:15.591329 139373894952704 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save.
I0420 07:07:15.591474 139502449792832 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776668805.618878, 'wait_for_prev_duration_secs': 29.224609851837158, 'time_between_consecutive_saves_sec': 21.97667908668518, 'checkpointer_blocking_start_time': 1776668834.8450115, 'checkpointer_blocking_duration_secs': 0.7430789470672607, 'get_old_steps_start_time': 1776668835.588108, 'get_old_steps_duration_secs': 2.5272369384765625e-05, 'checkpoint_manager_blocking_start_time': 1776668805.6169841, 'checkpoint_manager_blocking_duration_secs': 29.974452257156372}
I0420 07:07:15.591618 139502449792832 checkpointing.py:408] Started an asynchronous checkpoint save for step 19
I0420 07:07:15.591665 139502449792832 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0420 07:07:20.890455 139373324510976 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_true/checkpoints/19/items/array_metadatas/process_4
I0420 07:07:56.881030 139376428304128 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 38.224 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.324729204177856 s) (per-host)
I0420 07:07:56.881147 139376428304128 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.293087s.
I0420 07:08:05.629621 139376428304128 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.041543s.
I0420 07:08:05.629898 139373894952704 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save.
I0420 07:08:05.630024 139373894952704 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save.
I0420 07:08:05.630065 139373894952704 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0420 07:08:05.631787 139373894952704 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0420 07:08:05.631958 139502449792832 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19.
I0420 07:08:05.632101 139502449792832 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0420 07:08:05.633024 139502449792832 metric_logger.py:196] completed step: 19, seconds: 0.148, TFLOP/s/device: 91.905, Tokens/s/device: 13852.908, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Mon Apr 20 07:08:16 UTC 2026
EXIT_CODE=0
test/pipeline-scan-nnx  ·  e27fc1e97  ·  test_pipeline_scan_nnx_20260421_005327  ·  full log 
XPK Start: Tue Apr 21 01:58:49 UTC 2026
2026-04-21 01:58:53.887627: 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 00:00:1776736733.900413      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1776736733.904079      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1776736733.915354      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776736733.915375      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776736733.915378      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776736733.915379      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-21 01:59:13.017518: 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 01:59:13.535750 136415047964480 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-21 01:59:22,575:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0421 01:59:22.575042 136415047964480 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-21 01:59:22,577:jax._src.distributed:157: Connecting to JAX distributed service on mt-15-ocdbt-true-rs-olo4d-slice-job-0-0.mt-15-ocdbt-true-rs-olo4d:8482
I0421 01:59:22.577385 136415047964480 distributed.py:157] Connecting to JAX distributed service on mt-15-ocdbt-true-rs-olo4d-slice-job-0-0.mt-15-ocdbt-true-rs-olo4d:8482
I0421 01:59:23.676991 136415047964480 max_utils.py:284] Jax distributed system initialized!
I0421 01:59:30.500114 136415047964480 max_utils.py:800] System Information: Jax Version: 0.8.1
I0421 01:59:30.500218 136415047964480 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0421 01:59:30.500258 136415047964480 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 01:59:30.500293 136415047964480 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0421 01:59:31.117085 136415047964480 maxtext_utils.py:1398] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0421 01:59:31.117364 136415047964480 checkpointing.py:677] Setting up checkpoint logger...
I0421 01:59:31.117433 136415047964480 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0421 01:59:31.117479 136415047964480 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0421 01:59:31.117815 136415047964480 base_pytree_checkpoint_handler.py:415] 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 0x7c1127979100>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0421 01:59:32.785326 136415047964480 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0421 01:59:32.785706 136415047964480 checkpoint_manager.py:709] [process=1][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bfe64d341d0>}, handler_registry=None
I0421 01:59:32.785951 136415047964480 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bfe64d341d0>` 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 01:59:32.786005 136415047964480 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bfe64d37c80>` 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 01:59:32.786042 136415047964480 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 0x7bfe64d341d0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bfe64d341d0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bfe64d37c80>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bfe64d37c80>}).
I0421 01:59:32.788667 136415047964480 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0421 01:59:32.788752 136415047964480 async_checkpointer.py:177] [process=1][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7c11135fb2e0> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0421 01:59:35.169545 136294915094272 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776736554305353126, 'commit_timestamp_nsecs': 1776736603314739446, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/5/_CHECKPOINT_METADATA
I0421 01:59:35.407330 136294915094272 checkpoint_manager.py:1877] Missing metrics for step 5
I0421 01:59:36.230365 136415047964480 checkpoint_manager.py:1818] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints
I0421 01:59:37.943028 136415047964480 checkpoint_manager.py:929] [process=1][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_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, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7bfe64d30680>
I0421 01:59:37.943216 136415047964480 checkpointing.py:301] Checkpoint manager created!
I0421 01:59:38.821515 136415047964480 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0421 01:59:38.821617 136415047964480 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0421 01:59:39.194870 136415047964480 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0421 01:59:39.194964 136415047964480 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0421 01:59:39.211518 136415047964480 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0421 01:59:39.211578 136415047964480 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0421 01:59:39.235459 136415047964480 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0421 01:59:39.235526 136415047964480 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0421 01:59:39.251863 136415047964480 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0421 01:59:39.251928 136415047964480 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0421 01:59:39.268262 136415047964480 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0421 01:59:39.268323 136415047964480 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0421 01:59:39.293034 136415047964480 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0421 01:59:39.293106 136415047964480 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0421 01:59:39.312988 136415047964480 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0421 01:59:39.313055 136415047964480 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0421 01:59:39.505183 136415047964480 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0421 01:59:39.505289 136415047964480 checkpointing.py:581] restoring from this run's directory step 9
I0421 01:59:41.081570 136415047964480 checkpointer.py:304] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/9.
I0421 01:59:42.999445    1416 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0421 01:59:49.307948 136415047964480 jax_array_handlers.py:837] [process=1] /jax/orbax/read/worker/io/requested throughput: 250.291 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.313756942749023 s) (per-host)
I0421 01:59:49.312874 136415047964480 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/read/gbytes_per_sec: 235.896 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.70885443687439 s) (per-host)
I0421 01:59:50.634961 136415047964480 checkpointer.py:318] Finished restoring checkpoint in 10.89 seconds from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/9.
I0421 01:59:50.635120 136415047964480 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints', 'checkpointer_start_time': 1776736779.7418056, 'checkpointer_duration_secs': 10.893283128738403, 'checkpoint_manager_start_time': 1776736779.5055494, 'checkpoint_manager_duration_secs': 11.129541158676147}
fsdp: 32
I0421 01:59:51.039909 136415047964480 maxtext_utils.py:1501]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0421 01:59:51.040040 136415047964480 maxtext_utils.py:1501]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0421 01:59:51.040094 136415047964480 maxtext_utils.py:1501]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0421 01:59:51.040154 136415047964480 maxtext_utils.py:1501]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   PartitionSpec('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0421 01:59:51.040215 136415047964480 maxtext_utils.py:1501]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  (None, None)
I0421 01:59:51.040256 136415047964480 maxtext_utils.py:1501]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  (None, None)
I0421 01:59:51.040310 136415047964480 maxtext_utils.py:1501]  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 01:59:51.040363 136415047964480 maxtext_utils.py:1501]  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 01:59:51.040418 136415047964480 maxtext_utils.py:1501]  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 01:59:51.040460 136415047964480 maxtext_utils.py:1501]  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 01:59:51.040510 136415047964480 maxtext_utils.py:1501]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed', 'vocab')
    Physical:  ('fsdp', None)

I0421 01:59:51.040559 136415047964480 maxtext_utils.py:1501]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed')
    Physical:  (None, 'fsdp')
I0421 01:59:51.539448 136415047964480 train.py:156] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0421 01:59:51.539543 136415047964480 train.py:156] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0421 01:59:51.554541 136415047964480 train.py:163] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0421 01:59:51.554601 136415047964480 train.py:163] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0421 02:00:02.291095 136415047964480 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0421 02:00:02.291983 136415047964480 metric_logger.py:289] number parameters: 1.104 billion
I0421 02:00:13.185941 136415047964480 checkpointing.py:772] Waiting for step 10 to finish before checkpoint...
I0421 02:00:13.632779 136415047964480 checkpointing.py:776] Waited 0.44681715965270996 seconds for step 10 to finish before starting checkpointing.
I0421 02:00:13.635287 136415047964480 checkpoint_manager.py:2013] [process=1][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0421 02:00:13.637154 136415047964480 checkpoint_manager.py:1518] [process=1] Saving checkpoint at step 10
I0421 02:00:13.639799 136415047964480 async_checkpointer.py:452] [process=1] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/10.
I0421 02:00:14.661252 136415047964480 signaling_client.py:364] Using JaxDistributedSignalingClient
I0421 02:00:14.662227 136415047964480 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0421 02:00:14.662284 136415047964480 replica_slices.py:424] 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 02:00:14.915899 136415047964480 base_pytree_checkpoint_handler.py:153] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.254738s
I0421 02:00:14.916068 136415047964480 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.974 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5186583995819092 s) (per-host)
I0421 02:00:14.916120 136415047964480 base_pytree_checkpoint_handler.py:737] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.518720s (batch_requests_ready=0.261935s, total_serialization_initiated=0.256715s, others=0.000069s)
I0421 02:00:14.916212 136415047964480 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.522896s (all_items=0.000020s, per_item={'items': '0.00001955'}, temp_paths=0.522877)
I0421 02:00:14.917129 136294327899904 async_checkpointer.py:79] [process=1][thread=async_save] Background save thread started.
I0421 02:00:14.917298 136415047964480 async_checkpointer.py:561] Finished blocking save. Time taken: 1.280071s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/10.
I0421 02:00:15.348317 136415047964480 checkpoint_manager.py:1566] [process=1][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0421 02:00:15.348702 136289998546688 async_checkpointer.py:265] [process=1][thread=save_finalize] Waiting for background save thread=async_save.
I0421 02:00:15.348864 136415047964480 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776736813.6352692, 'wait_for_prev_duration_secs': 6.270408630371094e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776736813.6371925, 'checkpointer_blocking_duration_secs': 1.2802221775054932, 'get_old_steps_start_time': 1776736814.917435, 'get_old_steps_duration_secs': 2.6464462280273438e-05, 'checkpoint_manager_blocking_start_time': 1776736813.63333, 'checkpoint_manager_blocking_duration_secs': 1.7154946327209473}
I0421 02:00:15.348975 136415047964480 checkpointing.py:408] Started an asynchronous checkpoint save for step 10
I0421 02:00:15.349027 136415047964480 max_utils.py:750] 
Memstats: After params initialized:
I0421 02:00:15.349083 136415047964480 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_2(process=1,(2,0,0,0))
I0421 02:00:15.349117 136415047964480 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_3(process=1,(3,0,0,0))
I0421 02:00:15.349148 136415047964480 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_6(process=1,(2,1,0,0))
I0421 02:00:15.349174 136415047964480 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_7(process=1,(3,1,0,0))
I0421 02:00:15.610551 136415047964480 metric_logger.py:185] completed step: 10, seconds: 10.894, TFLOP/s/device: 1.247, Tokens/s/device: 187.996, total_weights: 65536, loss: 6.833
I0421 02:00:15.773159 136415047964480 metric_logger.py:185] completed step: 11, seconds: 2.423, TFLOP/s/device: 5.607, Tokens/s/device: 845.133, total_weights: 65536, loss: 6.528
I0421 02:00:15.920936 136415047964480 metric_logger.py:185] completed step: 12, seconds: 0.016, TFLOP/s/device: 874.783, Tokens/s/device: 131856.812, total_weights: 65536, loss: 6.265
I0421 02:00:16.068460 136415047964480 metric_logger.py:185] completed step: 13, seconds: 0.153, TFLOP/s/device: 89.026, Tokens/s/device: 13419.037, total_weights: 65536, loss: 6.045
I0421 02:00:16.073483 136415047964480 checkpointing.py:772] Waiting for step 15 to finish before checkpoint...
I0421 02:00:16.363074 136415047964480 checkpointing.py:776] Waited 0.28957200050354004 seconds for step 15 to finish before starting checkpointing.
I0421 02:00:16.365876 136415047964480 checkpoint_manager.py:2024] [process=1][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0421 02:00:21.059700 136290536232704 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/10/items/array_metadatas/process_1
I0421 02:00:56.520648 136294327899904 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/gbytes_per_sec: 37.499 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.123204708099365 s) (per-host)
I0421 02:00:56.520750 136294327899904 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.603512s.
I0421 02:01:05.725008 136294327899904 async_checkpointer.py:144] [process=1][thread=async_save] Background save thread done. Time taken: 50.807752s.
I0421 02:01:05.725288 136289998546688 async_checkpointer.py:273] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save.
I0421 02:01:05.725436 136289998546688 async_checkpointer.py:283] [process=1][thread=save_finalize] No errors found in background save thread=async_save.
I0421 02:01:05.725488 136289998546688 checkpoint_manager.py:2133] [process=1][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0421 02:01:05.726908 136289998546688 checkpoint_manager.py:2142] [process=1][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0421 02:01:05.727102 136415047964480 checkpoint_manager.py:2036] [process=1][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10.
W0421 02:01:05.727232 136415047964480 checkpoint_manager.py:1458] Waiting for previous save to complete took 49.361360 seconds. If this number is high, consider checkpointing less frequently.
I0421 02:01:05.729221 136415047964480 checkpoint_manager.py:1518] [process=1] Saving checkpoint at step 15
I0421 02:01:05.733538 136415047964480 async_checkpointer.py:452] [process=1] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/15.
I0421 02:01:06.688373 136415047964480 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0421 02:01:06.688485 136415047964480 replica_slices.py:424] 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 02:01:06.731339 136415047964480 base_pytree_checkpoint_handler.py:153] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044182s
I0421 02:01:06.731522 136415047964480 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.307 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2906501293182373 s) (per-host)
I0421 02:01:06.731576 136415047964480 base_pytree_checkpoint_handler.py:737] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.290713s (batch_requests_ready=0.244569s, total_serialization_initiated=0.046072s, others=0.000072s)
I0421 02:01:06.731671 136415047964480 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.295869s (all_items=0.000013s, per_item={'items': '0.00001335'}, temp_paths=0.295856)
I0421 02:01:06.732546 136290553018112 async_checkpointer.py:79] [process=1][thread=async_save] Background save thread started.
I0421 02:01:06.732709 136415047964480 async_checkpointer.py:561] Finished blocking save. Time taken: 1.003415s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/15.
I0421 02:01:07.182352 136415047964480 checkpoint_manager.py:1566] [process=1][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0421 02:01:07.182733 136289998546688 async_checkpointer.py:265] [process=1][thread=save_finalize] Waiting for background save thread=async_save.
I0421 02:01:07.182904 136415047964480 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776736816.3658478, 'wait_for_prev_duration_secs': 49.3613600730896, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776736865.729259, 'checkpointer_blocking_duration_secs': 1.0035557746887207, 'get_old_steps_start_time': 1776736866.7328348, 'get_old_steps_duration_secs': 2.7179718017578125e-05, 'checkpoint_manager_blocking_start_time': 1776736816.3637512, 'checkpoint_manager_blocking_duration_secs': 50.81911754608154}
I0421 02:01:07.183014 136415047964480 checkpointing.py:408] Started an asynchronous checkpoint save for step 15
I0421 02:01:07.183637 136415047964480 metric_logger.py:185] completed step: 14, seconds: 0.148, TFLOP/s/device: 91.963, Tokens/s/device: 13861.627, total_weights: 65536, loss: 5.864
I0421 02:01:07.189082 136415047964480 metric_logger.py:185] completed step: 15, seconds: 0.148, TFLOP/s/device: 91.672, Tokens/s/device: 13817.765, total_weights: 65536, loss: 5.719
I0421 02:01:25.115704 136415047964480 metric_logger.py:185] completed step: 16, seconds: 51.115, TFLOP/s/device: 0.266, Tokens/s/device: 40.067, total_weights: 65536, loss: 5.605
I0421 02:01:25.263075 136415047964480 metric_logger.py:185] completed step: 17, seconds: 17.780, TFLOP/s/device: 0.764, Tokens/s/device: 115.184, total_weights: 65536, loss: 5.517
I0421 02:01:25.410682 136415047964480 metric_logger.py:185] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.444, Tokens/s/device: 13482.022, total_weights: 65536, loss: 5.448
I0421 02:01:25.557975 136415047964480 checkpointing.py:772] Waiting for step 19 to finish before checkpoint...
I0421 02:01:25.558639 136415047964480 checkpointing.py:776] Waited 0.0006844997406005859 seconds for step 19 to finish before starting checkpointing.
I0421 02:01:25.560729 136415047964480 checkpoint_manager.py:2024] [process=1][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0421 02:01:33.511958 136290536232704 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/15/items/array_metadatas/process_1
I0421 02:02:07.916672 136290553018112 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/gbytes_per_sec: 25.695 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 61.475759983062744 s) (per-host)
I0421 02:02:07.916790 136290553018112 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 61.184132s.
I0421 02:02:18.751256 136290553018112 async_checkpointer.py:144] [process=1][thread=async_save] Background save thread done. Time taken: 72.018580s.
I0421 02:02:18.751570 136289998546688 async_checkpointer.py:273] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save.
I0421 02:02:18.751710 136289998546688 async_checkpointer.py:283] [process=1][thread=save_finalize] No errors found in background save thread=async_save.
I0421 02:02:18.751760 136289998546688 checkpoint_manager.py:2133] [process=1][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0421 02:02:18.753339 136289998546688 checkpoint_manager.py:2142] [process=1][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0421 02:02:18.753520 136415047964480 checkpoint_manager.py:2036] [process=1][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15.
W0421 02:02:18.753656 136415047964480 checkpoint_manager.py:1458] Waiting for previous save to complete took 53.192938 seconds. If this number is high, consider checkpointing less frequently.
I0421 02:02:18.755276 136415047964480 checkpoint_manager.py:1518] [process=1] Saving checkpoint at step 19
I0421 02:02:18.759050 136415047964480 async_checkpointer.py:452] [process=1] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/19.
I0421 02:02:19.735848 136415047964480 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0421 02:02:19.735943 136415047964480 replica_slices.py:424] 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 02:02:19.777934 136415047964480 base_pytree_checkpoint_handler.py:153] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.043253s
I0421 02:02:19.778102 136415047964480 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.313 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.29036521911621094 s) (per-host)
I0421 02:02:19.778152 136415047964480 base_pytree_checkpoint_handler.py:737] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.290425s (batch_requests_ready=0.245073s, total_serialization_initiated=0.045287s, others=0.000065s)
I0421 02:02:19.778242 136415047964480 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.294468s (all_items=0.000013s, per_item={'items': '0.00001287'}, temp_paths=0.294455)
I0421 02:02:19.779147 136294319507200 async_checkpointer.py:79] [process=1][thread=async_save] Background save thread started.
I0421 02:02:19.779305 136415047964480 async_checkpointer.py:561] Finished blocking save. Time taken: 1.023958s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/19.
I0421 02:02:20.230687 136415047964480 checkpoint_manager.py:1566] [process=1][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0421 02:02:20.231037 136289998546688 async_checkpointer.py:265] [process=1][thread=save_finalize] Waiting for background save thread=async_save.
I0421 02:02:20.231207 136415047964480 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776736885.5607028, 'wait_for_prev_duration_secs': 53.19293761253357, 'time_between_consecutive_saves_sec': 19.833755016326904, 'checkpointer_blocking_start_time': 1776736938.7553158, 'checkpointer_blocking_duration_secs': 1.0241153240203857, 'get_old_steps_start_time': 1776736939.779445, 'get_old_steps_duration_secs': 2.0742416381835938e-05, 'checkpoint_manager_blocking_start_time': 1776736885.5588627, 'checkpoint_manager_blocking_duration_secs': 54.67231011390686}
I0421 02:02:20.231314 136415047964480 checkpointing.py:408] Started an asynchronous checkpoint save for step 19
I0421 02:02:20.231357 136415047964480 checkpoint_manager.py:2024] [process=1][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0421 02:02:27.956300 136294327899904 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_true/checkpoints/19/items/array_metadatas/process_1
I0421 02:03:02.887104 136294319507200 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/gbytes_per_sec: 36.397 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.399333477020264 s) (per-host)
I0421 02:03:02.887210 136294319507200 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.107953s.
I0421 02:03:12.688557 136294319507200 async_checkpointer.py:144] [process=1][thread=async_save] Background save thread done. Time taken: 52.909284s.
I0421 02:03:12.688833 136289998546688 async_checkpointer.py:273] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save.
I0421 02:03:12.688957 136289998546688 async_checkpointer.py:283] [process=1][thread=save_finalize] No errors found in background save thread=async_save.
I0421 02:03:12.689000 136289998546688 checkpoint_manager.py:2133] [process=1][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0421 02:03:12.690648 136289998546688 checkpoint_manager.py:2142] [process=1][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0421 02:03:12.690822 136415047964480 checkpoint_manager.py:2036] [process=1][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19.
I0421 02:03:12.690961 136415047964480 checkpoint_manager.py:2013] [process=1][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0421 02:03:12.691662 136415047964480 metric_logger.py:185] completed step: 19, seconds: 0.148, TFLOP/s/device: 91.836, Tokens/s/device: 13842.514, total_weights: 65536, loss: 5.391
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Tue Apr 21 02:03:25 UTC 2026
EXIT_CODE=0