MaxView

Timing Test Reference Updates
‹ 15_ocdbt_true_saveCase: 16_zarr3_false_resume16_zarr3_false_save ›

Metrics: main (b117f50cf) vs feat/nnx-set-defaults-true (73213e044)

Metricmain  b117f50cffeat/nnx-set-defaults-true  73213e044Diff (feat/nnx-set-defaults-true − main)
Parameters1.104 billion1.104 billion—
Final loss5.39105.39100
TFLOP/s92.30692.216-0.09
Tok/s13913.313899.7-13.598
Avg s/step8.4578.043-0.414
Memory %1.381.380
JAX0.9.20.8.1—

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

main  ·  b117f50cf  ·  main_20260424_070227  ·  full log 
XPK Start: Fri Apr 24 08:31:19 UTC 2026
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
`rope_parameters`'s factor field must be a float >= 1, got 40
`rope_parameters`'s beta_fast field must be a float, got 32
`rope_parameters`'s beta_slow field must be a float, got 1
DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 
2026-04-24 08:31:44.203429: 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)
I0424 08:31:44.410198 132465606420288 max_utils.py:273] Attempting to initialize the jax distributed system...
I0424 08:31:53.452249 132465606420288 distributed.py:149] Starting JAX distributed service on [::]:8482
I0424 08:31:53.454657 132465606420288 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-rs-l6luk-slice-job-0-0.mt-16-zarr3-false-rs-l6luk:8482
I0424 08:31:54.560960 132465606420288 max_utils.py:284] Jax distributed system initialized!
I0424 08:32:00.603538 132465606420288 max_utils.py:800] System Information: Jax Version: 0.9.2
I0424 08:32:00.603644 132465606420288 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0424 08:32:00.603683 132465606420288 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
I0424 08:32:00.603717 132465606420288 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 08:32:01.295442 132465606420288 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 08:32:01.295726 132465606420288 checkpointing.py:677] Setting up checkpoint logger...
I0424 08:32:01.295780 132465606420288 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=False
I0424 08:32:01.295825 132465606420288 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0424 08:32:01.296182 132465606420288 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x78796c914080>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 08:32:02.558325 132465606420288 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0424 08:32:02.558573 132465606420288 checkpoint_manager.py:708] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x786574490200>}, handler_registry=None
I0424 08:32:02.558816 132465606420288 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x786574490200>` 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`.
I0424 08:32:02.558865 132465606420288 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7865744908f0>` 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`.
I0424 08:32:02.558899 132465606420288 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 0x786574490200>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x786574490200>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7865744908f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7865744908f0>}).
I0424 08:32:02.559237 132465606420288 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0424 08:32:02.559308 132465606420288 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7865742bd800> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0424 08:32:04.273732 132344711210752 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1777019254974743517, 'commit_timestamp_nsecs': 1777019301770798699, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints/0/_CHECKPOINT_METADATA
I0424 08:32:04.521251 132344711210752 checkpoint_manager.py:1871] Missing metrics for step 0
I0424 08:32:04.649308 132465606420288 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints
I0424 08:32:05.275205 132465606420288 checkpoint_manager.py:929] [process=6][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_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7865744932c0>
I0424 08:32:05.275390 132465606420288 checkpointing.py:301] Checkpoint manager created!
I0424 08:32:06.203835 132465606420288 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 08:32:06.203952 132465606420288 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 08:32:06.585666 132465606420288 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn').
I0424 08:32:06.585761 132465606420288 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 08:32:06.602289 132465606420288 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn').
I0424 08:32:06.602350 132465606420288 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 08:32:06.626208 132465606420288 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 08:32:06.626291 132465606420288 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 08:32:06.642946 132465606420288 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 08:32:06.643010 132465606420288 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 08:32:06.659577 132465606420288 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 08:32:06.659637 132465606420288 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 08:32:06.684462 132465606420288 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch_attn', 'activation_length_attn', 'activation_heads', 'activation_kv').
I0424 08:32:06.684540 132465606420288 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 08:32:06.706444 132465606420288 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 08:32:06.706510 132465606420288 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 08:32:06.915524 132465606420288 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 08:32:06.915632 132465606420288 checkpointing.py:581] restoring from this run's directory step 9
I0424 08:32:07.169829 132465606420288 event_tracking.py:70] [process=6] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints/9.
I0424 08:32:07.413200 132465606420288 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints/9.
I0424 08:32:07.413651 132465606420288 event_tracking.py:125] [process=6] [sync] Finished blocking load in 0.24 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints/9.
I0424 08:32:09.114791    1445 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 08:32:14.682622 132465606420288 jax_array_handlers.py:843] [process=6] /jax/orbax/read/worker/io/requested throughput: 283.524 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 5.573685884475708 s) (per-host)
I0424 08:32:14.687364 132465606420288 base_pytree_checkpoint_handler.py:130] [process=6] /jax/checkpoint/read/gbytes_per_sec: 264.713 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 5.978528261184692 s) (per-host)
I0424 08:32:15.338886 132465606420288 event_tracking.py:138] [process=6] [sync] Finished load in 8.17 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints/9
I0424 08:32:15.339039 132465606420288 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints', 'checkpointer_start_time': 1777019527.1697273, 'checkpointer_duration_secs': 8.169278383255005, 'checkpoint_manager_start_time': 1777019526.9158754, 'checkpoint_manager_duration_secs': 8.423132419586182}
fsdp: 32
I0424 08:32:15.763562 132465606420288 maxtext_utils.py:1707]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0424 08:32:15.763692 132465606420288 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0424 08:32:15.763746 132465606420288 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0424 08:32:15.763803 132465606420288 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   P('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0424 08:32:15.763855 132465606420288 maxtext_utils.py:1707]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0424 08:32:15.763892 132465606420288 maxtext_utils.py:1707]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0424 08:32:15.763942 132465606420288 maxtext_utils.py:1707]  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)
I0424 08:32:15.763991 132465606420288 maxtext_utils.py:1707]  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')
I0424 08:32:15.764029 132465606420288 maxtext_utils.py:1707]  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)
I0424 08:32:15.764065 132465606420288 maxtext_utils.py:1707]  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)
I0424 08:32:15.764122 132465606420288 maxtext_utils.py:1707]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)

I0424 08:32:15.764169 132465606420288 maxtext_utils.py:1707]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')
I0424 08:32:16.257295 132465606420288 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 08:32:16.257392 132465606420288 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 08:32:16.273017 132465606420288 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 08:32:16.273078 132465606420288 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 08:32:27.229995 132465606420288 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.
I0424 08:32:27.230792 132465606420288 metric_logger.py:301] number parameters: 1.104 billion
I0424 08:32:38.430070 132465606420288 checkpointing.py:772] Waiting for step 10 to finish before checkpoint...
I0424 08:32:39.190154 132465606420288 checkpointing.py:776] Waited 0.7600383758544922 seconds for step 10 to finish before starting checkpointing.
I0424 08:32:39.192616 132465606420288 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 08:32:39.194676 132465606420288 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 10
I0424 08:32:39.196548 132465606420288 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints/10.
I0424 08:32:39.520389 132465606420288 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 08:32:39.521240 132465606420288 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 08:32:39.521298 132465606420288 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
I0424 08:32:39.797171 132465606420288 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.276862s
I0424 08:32:39.797348 132465606420288 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.495 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2807450294494629 s) (per-host)
I0424 08:32:39.797403 132465606420288 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.280817s (batch_requests_ready=0.002168s, total_serialization_initiated=0.278569s, others=0.000080s)
I0424 08:32:39.797492 132465606420288 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.285096s (all_items=0.000017s, per_item={'items': '0.00001740'}, temp_paths=0.285079)
I0424 08:32:39.798235 132465606420288 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.60 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints/10.
I0424 08:32:39.798505 132344669247232 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 08:52:39.798481
I0424 08:32:39.800492 132465606420288 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 08:32:39.800781 132340869220096 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 08:32:39.800935 132465606420288 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777019559.1925962, 'wait_for_prev_duration_secs': 6.127357482910156e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777019559.1947165, 'checkpointer_blocking_duration_secs': 0.6039035320281982, 'get_old_steps_start_time': 1777019559.798645, 'get_old_steps_duration_secs': 3.0994415283203125e-05, 'checkpoint_manager_blocking_start_time': 1777019559.190644, 'checkpoint_manager_blocking_duration_secs': 0.6102547645568848}
I0424 08:32:39.801038 132465606420288 checkpointing.py:408] Started an asynchronous checkpoint save for step 10
I0424 08:32:39.801086 132465606420288 max_utils.py:750] 
Memstats: After params initialized:
I0424 08:32:39.801159 132465606420288 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0424 08:32:39.801196 132465606420288 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0424 08:32:39.801221 132465606420288 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0424 08:32:39.801246 132465606420288 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0424 08:32:40.111648 132465606420288 metric_logger.py:196] completed step: 10, seconds: 11.199, TFLOP/s/device: 1.213, Tokens/s/device: 182.871, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869
I0424 08:32:40.295436 132465606420288 metric_logger.py:196] completed step: 11, seconds: 1.680, TFLOP/s/device: 8.088, Tokens/s/device: 1219.056, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659
I0424 08:32:40.443029 132465606420288 metric_logger.py:196] completed step: 12, seconds: 0.036, TFLOP/s/device: 373.437, Tokens/s/device: 56288.478, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068
I0424 08:32:40.590398 132465606420288 metric_logger.py:196] completed step: 13, seconds: 0.154, TFLOP/s/device: 88.494, Tokens/s/device: 13338.804, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001
I0424 08:32:40.595078 132465606420288 checkpointing.py:772] Waiting for step 15 to finish before checkpoint...
I0424 08:32:40.884347 132465606420288 checkpointing.py:776] Waited 0.28921961784362793 seconds for step 15 to finish before starting checkpointing.
I0424 08:32:40.887401 132465606420288 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 08:32:44.771718 132340877612800 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints/10/items/array_metadatas/process_6
I0424 08:33:23.629943 132344669247232 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 35.808 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 44.11332321166992 s) (per-host)
I0424 08:33:23.630050 132344669247232 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.831467s.
I0424 08:33:30.117366 132344669247232 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.318756s.
I0424 08:33:30.117663 132340869220096 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 08:33:30.117777 132340869220096 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 08:33:30.117824 132340869220096 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0424 08:33:30.120133 132340869220096 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0424 08:33:30.120321 132465606420288 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10.
W0424 08:33:30.120456 132465606420288 checkpoint_manager.py:1452] Waiting for previous save to complete took 49.233058 seconds. If this number is high, consider checkpointing less frequently.
I0424 08:33:30.122353 132465606420288 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 15
I0424 08:33:30.124646 132465606420288 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints/15.
I0424 08:33:30.781718 132465606420288 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 08:33:30.781815 132465606420288 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
I0424 08:33:30.807692 132465606420288 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027181s
I0424 08:33:30.807865 132465606420288 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 50.400 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030606746673583984 s) (per-host)
I0424 08:33:30.807930 132465606420288 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030683s (batch_requests_ready=0.001747s, total_serialization_initiated=0.028854s, others=0.000082s)
I0424 08:33:30.808037 132465606420288 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.035583s (all_items=0.000013s, per_item={'items': '0.00001311'}, temp_paths=0.035570)
I0424 08:33:30.808843 132465606420288 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.69 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints/15.
I0424 08:33:30.809186 132340869220096 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 08:53:30.809148
I0424 08:33:30.822069 132465606420288 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 08:33:30.822335 132338159687424 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 08:33:30.822469 132465606420288 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777019560.8873708, 'wait_for_prev_duration_secs': 49.23305821418762, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777019610.1223917, 'checkpointer_blocking_duration_secs': 0.6869401931762695, 'get_old_steps_start_time': 1777019610.8093581, 'get_old_steps_duration_secs': 3.0279159545898438e-05, 'checkpoint_manager_blocking_start_time': 1777019560.8850012, 'checkpoint_manager_blocking_duration_secs': 49.93743395805359}
I0424 08:33:30.822576 132465606420288 checkpointing.py:408] Started an asynchronous checkpoint save for step 15
I0424 08:33:30.823380 132465606420288 metric_logger.py:196] completed step: 14, seconds: 0.148, TFLOP/s/device: 92.042, Tokens/s/device: 13873.647, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193
I0424 08:33:30.829704 132465606420288 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.194, Tokens/s/device: 13896.522, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661
I0424 08:33:36.160636 132340877612800 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints/15/items/array_metadatas/process_6
I0424 08:33:51.650159 132465606420288 metric_logger.py:196] completed step: 16, seconds: 50.233, TFLOP/s/device: 0.270, Tokens/s/device: 40.770, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872
I0424 08:33:51.797506 132465606420288 metric_logger.py:196] completed step: 17, seconds: 20.674, TFLOP/s/device: 0.657, Tokens/s/device: 99.061, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975
I0424 08:33:51.944701 132465606420288 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.132, Tokens/s/device: 13434.882, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333
I0424 08:33:52.091475 132465606420288 checkpointing.py:772] Waiting for step 19 to finish before checkpoint...
I0424 08:33:52.092180 132465606420288 checkpointing.py:776] Waited 0.0007259845733642578 seconds for step 19 to finish before starting checkpointing.
I0424 08:33:52.094376 132465606420288 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 08:34:13.901948 132340869220096 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 36.629 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.12465310096741 s) (per-host)
I0424 08:34:13.902074 132340869220096 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.092778s.
I0424 08:34:20.437817 132340869220096 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 49.628505s.
I0424 08:34:20.438160 132338159687424 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 08:34:20.438282 132338159687424 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 08:34:20.438342 132338159687424 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0424 08:34:20.439809 132338159687424 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0424 08:34:20.439998 132465606420288 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15.
W0424 08:34:20.440138 132465606420288 checkpoint_manager.py:1452] Waiting for previous save to complete took 28.345769 seconds. If this number is high, consider checkpointing less frequently.
I0424 08:34:20.441556 132465606420288 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 19
I0424 08:34:20.443657 132465606420288 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints/19.
I0424 08:34:20.776228 132465606420288 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 08:34:20.776324 132465606420288 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
I0424 08:34:20.802274 132465606420288 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026981s
I0424 08:34:20.802419 132465606420288 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 50.647 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030457496643066406 s) (per-host)
I0424 08:34:20.802464 132465606420288 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030513s (batch_requests_ready=0.001639s, total_serialization_initiated=0.028814s, others=0.000059s)
I0424 08:34:20.802546 132465606420288 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034634s (all_items=0.000011s, per_item={'items': '0.00001144'}, temp_paths=0.034622)
I0424 08:34:20.803194 132465606420288 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.36 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints/19.
I0424 08:34:20.803507 132342505010944 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 08:54:20.803465
I0424 08:34:20.807675 132465606420288 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 08:34:20.807917 132338159687424 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 08:34:20.808033 132465606420288 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777019632.0943458, 'wait_for_prev_duration_secs': 28.345768928527832, 'time_between_consecutive_saves_sec': 21.97417187690735, 'checkpointer_blocking_start_time': 1777019660.4415958, 'checkpointer_blocking_duration_secs': 0.36205220222473145, 'get_old_steps_start_time': 1777019660.8036656, 'get_old_steps_duration_secs': 2.6226043701171875e-05, 'checkpoint_manager_blocking_start_time': 1777019632.092423, 'checkpoint_manager_blocking_duration_secs': 28.715579509735107}
I0424 08:34:20.808152 132465606420288 checkpointing.py:408] Started an asynchronous checkpoint save for step 19
I0424 08:34:20.808195 132465606420288 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 08:34:25.917613 132340877612800 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_16_zarr3_false/checkpoints/19/items/array_metadatas/process_6
I0424 08:35:02.960242 132342505010944 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.442 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.18822717666626 s) (per-host)
I0424 08:35:02.960364 132342505010944 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.156747s.
I0424 08:35:08.946881 132342505010944 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 48.143250s.
I0424 08:35:08.947189 132338159687424 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 08:35:08.947314 132338159687424 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 08:35:08.947357 132338159687424 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0424 08:35:08.950397 132338159687424 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0424 08:35:08.950556 132465606420288 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19.
I0424 08:35:08.950709 132465606420288 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 08:35:08.951725 132465606420288 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.306, Tokens/s/device: 13913.327, 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: Fri Apr 24 08:35:19 UTC 2026
EXIT_CODE=0
feat/nnx-set-defaults-true  ·  73213e044  ·  feat_nnx_set_defaults_true_20260424_145744  ·  full log 
XPK Start: Fri Apr 24 16:26:33 UTC 2026
2026-04-24 16:26:37.530000: 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 00:00:1777047997.543114      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1777047997.546841      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1777047997.558503      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777047997.558523      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777047997.558526      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777047997.558528      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-24 16:26:56.761766: 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)
I0424 16:26:57.291924 132043629082432 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-24 16:27:06,334:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0424 16:27:06.334558 132043629082432 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-24 16:27:06,340:jax._src.distributed:157: Connecting to JAX distributed service on mt-16-zarr3-false-rs-fno2p-slice-job-0-0.mt-16-zarr3-false-rs-fno2p:8482
I0424 16:27:06.340120 132043629082432 distributed.py:157] Connecting to JAX distributed service on mt-16-zarr3-false-rs-fno2p-slice-job-0-0.mt-16-zarr3-false-rs-fno2p:8482
I0424 16:27:07.322661 132043629082432 max_utils.py:284] Jax distributed system initialized!
I0424 16:27:14.430495 132043629082432 max_utils.py:800] System Information: Jax Version: 0.8.1
I0424 16:27:14.430600 132043629082432 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0424 16:27:14.430641 132043629082432 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
I0424 16:27:14.430675 132043629082432 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 16:27:15.051450 132043629082432 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 16:27:15.052041 132043629082432 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 16:27:15.052253 132043629082432 checkpointing.py:688] Setting up checkpoint logger...
I0424 16:27:15.052306 132043629082432 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=False
I0424 16:27:15.052349 132043629082432 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0424 16:27:15.052704 132043629082432 base_pytree_checkpoint_handler.py:415] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x78174eb79580>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 16:27:16.328484 132043629082432 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0424 16:27:16.328727 132043629082432 checkpoint_manager.py:709] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78029c1a1f10>}, handler_registry=None
I0424 16:27:16.328998 132043629082432 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78029c1a1f10>` 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`.
I0424 16:27:16.329066 132043629082432 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78029c309dc0>` 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`.
I0424 16:27:16.329126 132043629082432 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 0x78029c1a1f10>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78029c1a1f10>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78029c309dc0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78029c309dc0>}).
I0424 16:27:16.329669 132043629082432 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0424 16:27:16.329750 132043629082432 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x78028439c180> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0424 16:27:18.184970 131922705495808 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1777047748963202567, 'commit_timestamp_nsecs': 1777047796298196305, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints/0/_CHECKPOINT_METADATA
I0424 16:27:18.439201 131922697103104 checkpoint_manager.py:1877] Missing metrics for step 5
I0424 16:27:18.457726 132043629082432 checkpoint_manager.py:1818] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints
I0424 16:27:18.753084 132043629082432 checkpoint_manager.py:929] [process=6][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_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x78029c35d730>
I0424 16:27:18.753298 132043629082432 checkpointing.py:302] Checkpoint manager created!
I0424 16:27:19.650034 132043629082432 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 16:27:19.650151 132043629082432 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 16:27:20.024652 132043629082432 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 16:27:20.024745 132043629082432 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 16:27:20.040703 132043629082432 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 16:27:20.040763 132043629082432 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 16:27:20.063723 132043629082432 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 16:27:20.063791 132043629082432 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:27:20.079982 132043629082432 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 16:27:20.080047 132043629082432 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:27:20.095937 132043629082432 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 16:27:20.095999 132043629082432 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:27:20.120785 132043629082432 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0424 16:27:20.120855 132043629082432 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:27:20.140476 132043629082432 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 16:27:20.140542 132043629082432 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 16:27:20.330333 132043629082432 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 16:27:20.330445 132043629082432 checkpointing.py:582] restoring from this run's directory step 9
I0424 16:27:20.833318 132043629082432 checkpointer.py:304] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints/9.
I0424 16:27:22.232784    1431 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 16:27:28.213722 132043629082432 jax_array_handlers.py:837] [process=6] /jax/orbax/read/worker/io/requested throughput: 263.987 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 5.986175298690796 s) (per-host)
I0424 16:27:28.229941 132043629082432 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/read/gbytes_per_sec: 247.269 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.400290250778198 s) (per-host)
I0424 16:27:28.560680 132043629082432 checkpointer.py:318] Finished restoring checkpoint in 7.97 seconds from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints/9.
I0424 16:27:28.560850 132043629082432 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints', 'checkpointer_start_time': 1777048040.5861528, 'checkpointer_duration_secs': 7.974658727645874, 'checkpoint_manager_start_time': 1777048040.3306918, 'checkpoint_manager_duration_secs': 8.230122566223145}
fsdp: 32
I0424 16:27:28.972890 132043629082432 maxtext_utils.py:1740]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0424 16:27:28.973019 132043629082432 maxtext_utils.py:1740]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0424 16:27:28.973074 132043629082432 maxtext_utils.py:1740]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0424 16:27:28.973166 132043629082432 maxtext_utils.py:1740]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   PartitionSpec('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0424 16:27:28.973235 132043629082432 maxtext_utils.py:1740]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  (None, None)
I0424 16:27:28.973278 132043629082432 maxtext_utils.py:1740]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  (None, None)
I0424 16:27:28.973335 132043629082432 maxtext_utils.py:1740]  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)
I0424 16:27:28.973390 132043629082432 maxtext_utils.py:1740]  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')
I0424 16:27:28.973433 132043629082432 maxtext_utils.py:1740]  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)
I0424 16:27:28.973474 132043629082432 maxtext_utils.py:1740]  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)
I0424 16:27:28.973525 132043629082432 maxtext_utils.py:1740]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed', 'vocab')
    Physical:  ('fsdp', None)
I0424 16:27:28.973575 132043629082432 maxtext_utils.py:1740]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed')

    Physical:  (None, 'fsdp')
I0424 16:27:29.473422 132043629082432 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 16:27:29.473517 132043629082432 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 16:27:29.488494 132043629082432 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 16:27:29.488559 132043629082432 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 16:27:40.466253 132043629082432 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.
I0424 16:27:40.467076 132043629082432 metric_logger.py:289] number parameters: 1.104 billion
I0424 16:27:51.793229 132043629082432 checkpointing.py:794] Waiting for step 10 to finish before checkpoint...
I0424 16:27:51.946742 132043629082432 checkpointing.py:798] Waited 0.15349745750427246 seconds for step 10 to finish before starting checkpointing.
I0424 16:27:51.949054 132043629082432 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 16:27:51.950992 132043629082432 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 10
I0424 16:27:51.953619 132043629082432 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints/10.
I0424 16:27:52.534330 132043629082432 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 16:27:52.536528 132043629082432 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0424 16:27:52.536588 132043629082432 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
I0424 16:27:52.790844 132043629082432 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.256611s
I0424 16:27:52.791018 132043629082432 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.022 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5104806423187256 s) (per-host)
I0424 16:27:52.791073 132043629082432 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.510545s (batch_requests_ready=0.251841s, total_serialization_initiated=0.258632s, others=0.000072s)
I0424 16:27:52.791174 132043629082432 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.514854s (all_items=0.000019s, per_item={'items': '0.00001907'}, temp_paths=0.514835)
I0424 16:27:52.792181 131922109908736 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0424 16:27:52.792359 132043629082432 async_checkpointer.py:561] Finished blocking save. Time taken: 0.841294s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints/10.
I0424 16:27:52.812215 132043629082432 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 16:27:52.812546 131917303248640 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 16:27:52.812707 132043629082432 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777048071.9490361, 'wait_for_prev_duration_secs': 8.440017700195312e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777048071.951032, 'checkpointer_blocking_duration_secs': 0.841437816619873, 'get_old_steps_start_time': 1777048072.7924895, 'get_old_steps_duration_secs': 3.4332275390625e-05, 'checkpoint_manager_blocking_start_time': 1777048071.9472282, 'checkpoint_manager_blocking_duration_secs': 0.8654425144195557}
I0424 16:27:52.812814 132043629082432 checkpointing.py:409] Started an asynchronous checkpoint save for step 10
I0424 16:27:52.812868 132043629082432 max_utils.py:750] 
Memstats: After params initialized:
I0424 16:27:52.812916 132043629082432 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0424 16:27:52.812948 132043629082432 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0424 16:27:52.812975 132043629082432 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0424 16:27:52.812999 132043629082432 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0424 16:27:53.080972 132043629082432 metric_logger.py:185] completed step: 10, seconds: 11.326, TFLOP/s/device: 1.200, Tokens/s/device: 180.823, total_weights: 65536, loss: 6.833
I0424 16:27:53.252820 132043629082432 metric_logger.py:185] completed step: 11, seconds: 1.286, TFLOP/s/device: 10.563, Tokens/s/device: 1592.125, total_weights: 65536, loss: 6.528
I0424 16:27:53.400733 132043629082432 metric_logger.py:185] completed step: 12, seconds: 0.025, TFLOP/s/device: 548.487, Tokens/s/device: 82673.987, total_weights: 65536, loss: 6.265
I0424 16:27:53.548124 132043629082432 metric_logger.py:185] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.671, Tokens/s/device: 13365.529, total_weights: 65536, loss: 6.045
I0424 16:27:53.552753 132043629082432 checkpointing.py:794] Waiting for step 15 to finish before checkpoint...
I0424 16:27:53.842568 132043629082432 checkpointing.py:798] Waited 0.2897803783416748 seconds for step 15 to finish before starting checkpointing.
I0424 16:27:53.845473 132043629082432 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 16:27:57.471590 131918217594624 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints/10/items/array_metadatas/process_6
I0424 16:28:33.724615 131922109908736 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 38.114 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.44404101371765 s) (per-host)
I0424 16:28:33.724718 131922109908736 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.932422s.
I0424 16:28:42.423590 131922109908736 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 49.631276s.
I0424 16:28:42.423896 131917303248640 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 16:28:42.424017 131917303248640 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 16:28:42.424062 131917303248640 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0424 16:28:42.426799 131917303248640 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0424 16:28:42.426964 132043629082432 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10.
W0424 16:28:42.427111 132043629082432 checkpoint_manager.py:1458] Waiting for previous save to complete took 48.581631 seconds. If this number is high, consider checkpointing less frequently.
I0424 16:28:42.429064 132043629082432 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 15
I0424 16:28:42.433022 132043629082432 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints/15.
I0424 16:28:42.993986 132043629082432 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0424 16:28:42.994089 132043629082432 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
I0424 16:28:43.036982 132043629082432 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044125s
I0424 16:28:43.037163 132043629082432 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.078 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3038036823272705 s) (per-host)
I0424 16:28:43.037216 132043629082432 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.303866s (batch_requests_ready=0.257672s, total_serialization_initiated=0.046123s, others=0.000071s)
I0424 16:28:43.037309 132043629082432 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.308697s (all_items=0.000015s, per_item={'items': '0.00001454'}, temp_paths=0.308683)
I0424 16:28:43.039885 131918335026944 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0424 16:28:43.040047 132043629082432 async_checkpointer.py:561] Finished blocking save. Time taken: 0.610892s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints/15.
I0424 16:28:43.050253 132043629082432 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 16:28:43.050513 131917303248640 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 16:28:43.050649 132043629082432 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777048073.8454413, 'wait_for_prev_duration_secs': 48.58163070678711, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777048122.429122, 'checkpointer_blocking_duration_secs': 0.6110496520996094, 'get_old_steps_start_time': 1777048123.0401907, 'get_old_steps_duration_secs': 2.8371810913085938e-05, 'checkpoint_manager_blocking_start_time': 1777048073.8432171, 'checkpoint_manager_blocking_duration_secs': 49.2073974609375}
I0424 16:28:43.050757 132043629082432 checkpointing.py:409] Started an asynchronous checkpoint save for step 15
I0424 16:28:43.051332 132043629082432 metric_logger.py:185] completed step: 14, seconds: 0.148, TFLOP/s/device: 91.896, Tokens/s/device: 13851.596, total_weights: 65536, loss: 5.864
I0424 16:28:43.057382 132043629082432 metric_logger.py:185] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.284, Tokens/s/device: 13910.020, total_weights: 65536, loss: 5.719
I0424 16:29:00.744416 132043629082432 metric_logger.py:185] completed step: 16, seconds: 49.503, TFLOP/s/device: 0.274, Tokens/s/device: 41.371, total_weights: 65536, loss: 5.605
I0424 16:29:00.891836 132043629082432 metric_logger.py:185] completed step: 17, seconds: 17.541, TFLOP/s/device: 0.775, Tokens/s/device: 116.755, total_weights: 65536, loss: 5.517
I0424 16:29:01.039148 132043629082432 metric_logger.py:185] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.268, Tokens/s/device: 13455.449, total_weights: 65536, loss: 5.448
I0424 16:29:01.186244 132043629082432 checkpointing.py:794] Waiting for step 19 to finish before checkpoint...
I0424 16:29:01.186924 132043629082432 checkpointing.py:798] Waited 0.0007007122039794922 seconds for step 19 to finish before starting checkpointing.
I0424 16:29:01.188911 132043629082432 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 16:29:05.218654 131918217594624 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints/15/items/array_metadatas/process_6
I0424 16:29:42.412534 131918335026944 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 26.468 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 59.67913746833801 s) (per-host)
I0424 16:29:42.412648 131918335026944 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 59.372651s.
I0424 16:29:50.218060 131918335026944 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 67.178048s.
I0424 16:29:50.218396 131917303248640 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 16:29:50.218531 131917303248640 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 16:29:50.218596 131917303248640 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0424 16:29:50.220690 131917303248640 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0424 16:29:50.220884 132043629082432 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15.
W0424 16:29:50.221016 132043629082432 checkpoint_manager.py:1458] Waiting for previous save to complete took 49.032118 seconds. If this number is high, consider checkpointing less frequently.
I0424 16:29:50.222630 132043629082432 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 19
I0424 16:29:50.225961 132043629082432 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints/19.
I0424 16:29:51.179083 132043629082432 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0424 16:29:51.179193 132043629082432 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
I0424 16:29:51.223041 132043629082432 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.045012s
I0424 16:29:51.223224 132043629082432 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.213 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.29588985443115234 s) (per-host)
I0424 16:29:51.223278 132043629082432 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.295953s (batch_requests_ready=0.248778s, total_serialization_initiated=0.047107s, others=0.000068s)
I0424 16:29:51.223367 132043629082432 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.300003s (all_items=0.000010s, per_item={'items': '0.00001001'}, temp_paths=0.299993)
I0424 16:29:51.224271 131922101516032 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0424 16:29:51.224431 132043629082432 async_checkpointer.py:561] Finished blocking save. Time taken: 1.001732s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints/19.
I0424 16:29:51.249436 132043629082432 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 16:29:51.249725 131917303248640 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 16:29:51.249882 132043629082432 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777048141.1888824, 'wait_for_prev_duration_secs': 49.03211784362793, 'time_between_consecutive_saves_sec': 18.76203966140747, 'checkpointer_blocking_start_time': 1777048190.2226677, 'checkpointer_blocking_duration_secs': 1.0018763542175293, 'get_old_steps_start_time': 1777048191.2245586, 'get_old_steps_duration_secs': 2.288818359375e-05, 'checkpoint_manager_blocking_start_time': 1777048141.1871622, 'checkpoint_manager_blocking_duration_secs': 50.062690019607544}
I0424 16:29:51.249988 132043629082432 checkpointing.py:409] Started an asynchronous checkpoint save for step 19
I0424 16:29:51.250029 132043629082432 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 16:29:55.944678 131922109908736 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_16_zarr3_false/checkpoints/19/items/array_metadatas/process_6
I0424 16:30:33.081395 131922101516032 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 37.472 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.154030084609985 s) (per-host)
I0424 16:30:33.081499 131922101516032 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.857114s.
I0424 16:30:40.974906 131922101516032 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 49.750505s.
I0424 16:30:40.975216 131917303248640 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 16:30:40.975351 131917303248640 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 16:30:40.975393 131917303248640 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0424 16:30:40.976872 131917303248640 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0424 16:30:40.977030 132043629082432 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19.
I0424 16:30:40.977197 132043629082432 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 16:30:40.977833 132043629082432 metric_logger.py:185] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.216, Tokens/s/device: 13899.729, 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: Fri Apr 24 16:30:59 UTC 2026
EXIT_CODE=0