MaxView

Timing Test Reference Updates
‹ 16_zarr3_false_saveCase: 16_zarr3_true_resume16_zarr3_true_save ›

Metrics: main (59e0f1759) vs feat/nnx-set-defaults-true (73213e044)

Metricmain  59e0f1759feat/nnx-set-defaults-true  73213e044Diff (feat/nnx-set-defaults-true − main)
Parameters1.104 billion1.104 billion—
Final loss5.39105.39100
TFLOP/s92.19592.191-0.004
Tok/s13896.613896.0-0.66
Avg s/step8.4068.082-0.324
Memory %1.381.380
JAX0.9.20.8.1—

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

main  ·  59e0f1759  ·  main_20260425_071456  ·  full log 
XPK Start: Sat Apr 25 08:29:24 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. 
2026-04-25 08:29:49.180829: 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)
I0425 08:29:49.384943 138468346181440 max_utils.py:273] Attempting to initialize the jax distributed system...
I0425 08:29:58.425435 138468346181440 distributed.py:149] Starting JAX distributed service on [::]:8482
I0425 08:29:58.428014 138468346181440 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-true-rs-uln7e-slice-job-0-0.mt-16-zarr3-true-rs-uln7e:8482
I0425 08:29:59.480324 138468346181440 max_utils.py:284] Jax distributed system initialized!
I0425 08:30:04.900868 138468346181440 max_utils.py:800] System Information: Jax Version: 0.9.2
I0425 08:30:04.900977 138468346181440 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0425 08:30:04.901017 138468346181440 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
I0425 08:30:04.901052 138468346181440 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0425 08:30:05.595171 138468346181440 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1)
I0425 08:30:05.595451 138468346181440 checkpointing.py:677] Setting up checkpoint logger...
I0425 08:30:05.595506 138468346181440 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0425 08:30:05.595555 138468346181440 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0425 08:30:05.595902 138468346181440 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 0x7def0b9193d0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0425 08:30:06.899710 138468346181440 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0425 08:30:06.899947 138468346181440 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 0x7ddb5c0bc1d0>}, handler_registry=None
I0425 08:30:06.900207 138468346181440 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ddb5c0bc1d0>` 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`.
I0425 08:30:06.900258 138468346181440 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ddb5c0be3c0>` 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`.
I0425 08:30:06.900295 138468346181440 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 0x7ddb5c0bc1d0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ddb5c0bc1d0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ddb5c0be3c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ddb5c0be3c0>}).
I0425 08:30:06.900635 138468346181440 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0425 08:30:06.900705 138468346181440 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7ddb3c4cda80> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0425 08:30:08.438500 138349251639040 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1777105598494766318, 'commit_timestamp_nsecs': 1777105644277983871, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints/5/_CHECKPOINT_METADATA
I0425 08:30:08.692321 138349251639040 checkpoint_manager.py:1871] Missing metrics for step 5
I0425 08:30:08.713970 138468346181440 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints
I0425 08:30:09.376235 138468346181440 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_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7ddb5c0bb230>
I0425 08:30:09.376409 138468346181440 checkpointing.py:301] Checkpoint manager created!
I0425 08:30:10.305882 138468346181440 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0425 08:30:10.306004 138468346181440 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0425 08:30:10.688783 138468346181440 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn').
I0425 08:30:10.688897 138468346181440 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0425 08:30:10.705762 138468346181440 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn').
I0425 08:30:10.705822 138468346181440 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0425 08:30:10.729940 138468346181440 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0425 08:30:10.730011 138468346181440 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0425 08:30:10.746773 138468346181440 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0425 08:30:10.746836 138468346181440 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0425 08:30:10.763336 138468346181440 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0425 08:30:10.763395 138468346181440 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0425 08:30:10.788407 138468346181440 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch_attn', 'activation_length_attn', 'activation_heads', 'activation_kv').
I0425 08:30:10.788479 138468346181440 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0425 08:30:10.809831 138468346181440 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0425 08:30:10.809898 138468346181440 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0425 08:30:11.023308 138468346181440 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0425 08:30:11.023417 138468346181440 checkpointing.py:581] restoring from this run's directory step 9
I0425 08:30:11.275697 138468346181440 event_tracking.py:70] [process=6] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints/9.
I0425 08:30:11.517190 138468346181440 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints/9.
I0425 08:30:11.517635 138468346181440 event_tracking.py:125] [process=6] [sync] Finished blocking load in 0.24 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints/9.
I0425 08:30:12.931375    1416 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0425 08:30:18.554903 138468346181440 jax_array_handlers.py:843] [process=6] /jax/orbax/read/worker/io/requested throughput: 280.723 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 5.629295587539673 s) (per-host)
I0425 08:30:18.572116 138468346181440 base_pytree_checkpoint_handler.py:130] [process=6] /jax/checkpoint/read/gbytes_per_sec: 261.052 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.062373161315918 s) (per-host)
I0425 08:30:19.911445 138468346181440 event_tracking.py:138] [process=6] [sync] Finished load in 8.64 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints/9
I0425 08:30:19.911616 138468346181440 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints', 'checkpointer_start_time': 1777105811.2755961, 'checkpointer_duration_secs': 8.635987281799316, 'checkpoint_manager_start_time': 1777105811.0236742, 'checkpoint_manager_duration_secs': 8.887911319732666}
fsdp: 32
I0425 08:30:20.332105 138468346181440 maxtext_utils.py:1707]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0425 08:30:20.332238 138468346181440 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)
I0425 08:30:20.332293 138468346181440 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)
I0425 08:30:20.332353 138468346181440 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')
I0425 08:30:20.332407 138468346181440 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)
I0425 08:30:20.332447 138468346181440 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)
I0425 08:30:20.332499 138468346181440 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)
I0425 08:30:20.332556 138468346181440 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')
I0425 08:30:20.332597 138468346181440 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)
I0425 08:30:20.332632 138468346181440 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)
I0425 08:30:20.332677 138468346181440 maxtext_utils.py:1707]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)
I0425 08:30:20.332723 138468346181440 maxtext_utils.py:1707]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')

I0425 08:30:20.828829 138468346181440 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0425 08:30:20.828927 138468346181440 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0425 08:30:20.844599 138468346181440 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0425 08:30:20.844663 138468346181440 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0425 08:30:31.765109 138468346181440 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.
I0425 08:30:31.765895 138468346181440 metric_logger.py:301] number parameters: 1.104 billion
I0425 08:30:43.441798 138468346181440 checkpointing.py:772] Waiting for step 10 to finish before checkpoint...
I0425 08:30:43.598000 138468346181440 checkpointing.py:776] Waited 0.15617918968200684 seconds for step 10 to finish before starting checkpointing.
I0425 08:30:43.600456 138468346181440 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0425 08:30:43.602251 138468346181440 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 10
I0425 08:30:43.604153 138468346181440 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints/10.
I0425 08:30:43.926631 138468346181440 signaling_client.py:364] Using JaxDistributedSignalingClient
I0425 08:30:43.927575 138468346181440 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0425 08:30:43.927631 138468346181440 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
I0425 08:30:44.201082 138468346181440 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.274530s
I0425 08:30:44.201272 138468346181440 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.540 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2784237861633301 s) (per-host)
I0425 08:30:44.201326 138468346181440 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.278488s (batch_requests_ready=0.002156s, total_serialization_initiated=0.276259s, others=0.000073s)
I0425 08:30:44.201421 138468346181440 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.282604s (all_items=0.000017s, per_item={'items': '0.00001669'}, temp_paths=0.282588)
I0425 08:30:44.202249 138468346181440 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.60 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints/10.
I0425 08:30:44.202604 138348748338944 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 08:50:44.202565
I0425 08:30:44.221442 138468346181440 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0425 08:30:44.221721 138345460004608 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0425 08:30:44.221874 138468346181440 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777105843.6004372, 'wait_for_prev_duration_secs': 6.341934204101562e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777105843.602292, 'checkpointer_blocking_duration_secs': 0.6004681587219238, 'get_old_steps_start_time': 1777105844.2027829, 'get_old_steps_duration_secs': 3.266334533691406e-05, 'checkpoint_manager_blocking_start_time': 1777105843.5986075, 'checkpoint_manager_blocking_duration_secs': 0.6232302188873291}
I0425 08:30:44.221979 138468346181440 checkpointing.py:408] Started an asynchronous checkpoint save for step 10
I0425 08:30:44.222029 138468346181440 max_utils.py:750] 
Memstats: After params initialized:
I0425 08:30:44.222077 138468346181440 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0425 08:30:44.222127 138468346181440 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0425 08:30:44.222162 138468346181440 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0425 08:30:44.222186 138468346181440 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0425 08:30:44.532853 138468346181440 metric_logger.py:196] completed step: 10, seconds: 11.676, TFLOP/s/device: 1.164, Tokens/s/device: 175.406, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869
I0425 08:30:44.715744 138468346181440 metric_logger.py:196] completed step: 11, seconds: 1.090, TFLOP/s/device: 12.471, Tokens/s/device: 1879.715, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659
I0425 08:30:44.863286 138468346181440 metric_logger.py:196] completed step: 12, seconds: 0.036, TFLOP/s/device: 381.618, Tokens/s/device: 57521.627, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068
I0425 08:30:45.010869 138468346181440 metric_logger.py:196] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.603, Tokens/s/device: 13355.244, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001
I0425 08:30:45.016428 138468346181440 checkpointing.py:772] Waiting for step 15 to finish before checkpoint...
I0425 08:30:45.304942 138468346181440 checkpointing.py:776] Waited 0.28847765922546387 seconds for step 15 to finish before starting checkpointing.
I0425 08:30:45.307965 138468346181440 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0425 08:30:49.172086 138345946519296 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints/10/items/array_metadatas/process_6
I0425 08:31:26.407128 138348748338944 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.181 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.48422956466675 s) (per-host)
I0425 08:31:26.407242 138348748338944 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.204525s.
I0425 08:31:34.359701 138348748338944 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.156968s.
I0425 08:31:34.360067 138345460004608 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0425 08:31:34.360204 138345460004608 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0425 08:31:34.360253 138345460004608 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0425 08:31:34.362588 138345460004608 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0425 08:31:34.362768 138468346181440 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.
W0425 08:31:34.362898 138468346181440 checkpoint_manager.py:1452] Waiting for previous save to complete took 49.054935 seconds. If this number is high, consider checkpointing less frequently.
I0425 08:31:34.364768 138468346181440 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 15
I0425 08:31:34.367168 138468346181440 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints/15.
I0425 08:31:34.652347 138468346181440 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0425 08:31:34.652441 138468346181440 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
I0425 08:31:34.678921 138468346181440 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027735s
I0425 08:31:34.679117 138468346181440 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 49.409 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.031220674514770508 s) (per-host)
I0425 08:31:34.679177 138468346181440 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.031311s (batch_requests_ready=0.001791s, total_serialization_initiated=0.029421s, others=0.000098s)
I0425 08:31:34.679278 138468346181440 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.036250s (all_items=0.000013s, per_item={'items': '0.00001335'}, temp_paths=0.036237)
I0425 08:31:34.679880 138468346181440 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.32 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints/15.
I0425 08:31:34.680176 138342758864640 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 08:51:34.680142
I0425 08:31:34.691077 138468346181440 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0425 08:31:34.691375 138345460004608 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0425 08:31:34.691548 138468346181440 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777105845.3079357, 'wait_for_prev_duration_secs': 49.054935455322266, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777105894.3648095, 'checkpointer_blocking_duration_secs': 0.3155231475830078, 'get_old_steps_start_time': 1777105894.680353, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1777105845.3055995, 'checkpoint_manager_blocking_duration_secs': 49.38591170310974}
I0425 08:31:34.691659 138468346181440 checkpointing.py:408] Started an asynchronous checkpoint save for step 15
I0425 08:31:34.692573 138468346181440 metric_logger.py:196] completed step: 14, seconds: 0.147, TFLOP/s/device: 92.147, Tokens/s/device: 13889.454, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193
I0425 08:31:34.699242 138468346181440 metric_logger.py:196] completed step: 15, seconds: 0.149, TFLOP/s/device: 91.462, Tokens/s/device: 13786.140, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661
I0425 08:31:40.412063 138345946519296 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints/15/items/array_metadatas/process_6
I0425 08:31:55.670522 138468346181440 metric_logger.py:196] completed step: 16, seconds: 49.681, TFLOP/s/device: 0.273, Tokens/s/device: 41.223, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872
I0425 08:31:55.817884 138468346181440 metric_logger.py:196] completed step: 17, seconds: 20.825, TFLOP/s/device: 0.652, Tokens/s/device: 98.342, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975
I0425 08:31:55.965236 138468346181440 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.310, Tokens/s/device: 13461.728, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333
I0425 08:31:56.111902 138468346181440 checkpointing.py:772] Waiting for step 19 to finish before checkpoint...
I0425 08:31:56.112648 138468346181440 checkpointing.py:776] Waited 0.0007672309875488281 seconds for step 19 to finish before starting checkpointing.
I0425 08:31:56.114691 138468346181440 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0425 08:32:17.750363 138342758864640 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 36.647 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.1024489402771 s) (per-host)
I0425 08:32:17.750494 138342758864640 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.070200s.
I0425 08:32:25.185799 138342758864640 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.505488s.
I0425 08:32:25.186161 138345460004608 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0425 08:32:25.186312 138345460004608 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0425 08:32:25.186360 138345460004608 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0425 08:32:25.187811 138345460004608 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0425 08:32:25.187979 138468346181440 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.
W0425 08:32:25.188123 138468346181440 checkpoint_manager.py:1452] Waiting for previous save to complete took 29.073422 seconds. If this number is high, consider checkpointing less frequently.
I0425 08:32:25.189564 138468346181440 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 19
I0425 08:32:25.191682 138468346181440 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints/19.
I0425 08:32:25.583365 138468346181440 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0425 08:32:25.583456 138468346181440 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
I0425 08:32:25.609001 138468346181440 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026422s
I0425 08:32:25.609194 138468346181440 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 51.555 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.02992105484008789 s) (per-host)
I0425 08:32:25.609250 138468346181440 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.029987s (batch_requests_ready=0.001639s, total_serialization_initiated=0.028277s, others=0.000071s)
I0425 08:32:25.609349 138468346181440 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034090s (all_items=0.000010s, per_item={'items': '0.00001049'}, temp_paths=0.034079)
I0425 08:32:25.610085 138468346181440 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.42 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints/19.
I0425 08:32:25.610437 138343312520960 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 08:52:25.610400
I0425 08:32:25.613342 138468346181440 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0425 08:32:25.613589 138345460004608 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0425 08:32:25.613727 138468346181440 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777105916.114662, 'wait_for_prev_duration_secs': 29.0734224319458, 'time_between_consecutive_saves_sec': 21.752031803131104, 'checkpointer_blocking_start_time': 1777105945.189603, 'checkpointer_blocking_duration_secs': 0.4209866523742676, 'get_old_steps_start_time': 1777105945.6106083, 'get_old_steps_duration_secs': 2.6464462280273438e-05, 'checkpoint_manager_blocking_start_time': 1777105916.1128864, 'checkpoint_manager_blocking_duration_secs': 29.500807523727417}
I0425 08:32:25.613839 138468346181440 checkpointing.py:408] Started an asynchronous checkpoint save for step 19
I0425 08:32:25.613882 138468346181440 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0425 08:32:31.217622 138345946519296 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_16_zarr3_true/checkpoints/19/items/array_metadatas/process_6
I0425 08:33:07.425915 138343312520960 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.747 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.84660291671753 s) (per-host)
I0425 08:33:07.426041 138343312520960 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.815485s.
I0425 08:33:15.610469 138343312520960 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 49.999897s.
I0425 08:33:15.610760 138345460004608 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0425 08:33:15.610874 138345460004608 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0425 08:33:15.610915 138345460004608 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0425 08:33:15.612477 138345460004608 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0425 08:33:15.612645 138468346181440 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.
I0425 08:33:15.612795 138468346181440 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0425 08:33:15.613795 138468346181440 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.195, Tokens/s/device: 13896.617, 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: Sat Apr 25 08:33:29 UTC 2026
EXIT_CODE=0
feat/nnx-set-defaults-true  ·  73213e044  ·  feat_nnx_set_defaults_true_20260425_150412  ·  full log 
XPK Start: Sat Apr 25 16:19:45 UTC 2026
2026-04-25 16:19:49.375014: 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 00:00:1777133989.387815      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1777133989.391533      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1777133989.403020      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777133989.403040      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777133989.403042      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777133989.403044      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-25 16:20:08.551112: 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)
I0425 16:20:09.082034 132860367845184 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-25 16:20:18,122:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0425 16:20:18.122858 132860367845184 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-25 16:20:18,125:jax._src.distributed:157: Connecting to JAX distributed service on mt-16-zarr3-true-rs-1bwv1-slice-job-0-0.mt-16-zarr3-true-rs-1bwv1:8482
I0425 16:20:18.125195 132860367845184 distributed.py:157] Connecting to JAX distributed service on mt-16-zarr3-true-rs-1bwv1-slice-job-0-0.mt-16-zarr3-true-rs-1bwv1:8482
I0425 16:20:19.229493 132860367845184 max_utils.py:284] Jax distributed system initialized!
I0425 16:20:25.902996 132860367845184 max_utils.py:800] System Information: Jax Version: 0.8.1
I0425 16:20:25.903101 132860367845184 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0425 16:20:25.903141 132860367845184 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
I0425 16:20:25.903176 132860367845184 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0425 16:20:26.539967 132860367845184 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0425 16:20:26.540574 132860367845184 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0425 16:20:26.540756 132860367845184 checkpointing.py:688] Setting up checkpoint logger...
I0425 16:20:26.540808 132860367845184 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0425 16:20:26.540852 132860367845184 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0425 16:20:26.541257 132860367845184 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 0x78d58c1b0e00>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0425 16:20:27.853000 132860367845184 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0425 16:20:27.853362 132860367845184 checkpoint_manager.py:709] [process=3][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78c399b91eb0>}, handler_registry=None
I0425 16:20:27.853610 132860367845184 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78c399b91eb0>` 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`.
I0425 16:20:27.853662 132860367845184 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78c399b96060>` 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`.
I0425 16:20:27.853697 132860367845184 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 0x78c399b91eb0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78c399b91eb0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78c399b96060>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78c399b96060>}).
I0425 16:20:27.854223 132860367845184 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0425 16:20:27.854304 132860367845184 async_checkpointer.py:177] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x78c2a36a8180> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0425 16:20:29.971604 132744134776576 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1777133812385284716, 'commit_timestamp_nsecs': 1777133858837150802, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints/5/_CHECKPOINT_METADATA
I0425 16:20:30.216129 132744134776576 checkpoint_manager.py:1877] Missing metrics for step 5
I0425 16:20:30.437340 132860367845184 checkpoint_manager.py:1818] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints
I0425 16:20:30.439768 132860367845184 checkpoint_manager.py:929] [process=3][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_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x78c399b91820>
I0425 16:20:30.439884 132860367845184 checkpointing.py:302] Checkpoint manager created!
I0425 16:20:31.335180 132860367845184 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0425 16:20:31.335289 132860367845184 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0425 16:20:31.709408 132860367845184 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0425 16:20:31.709503 132860367845184 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0425 16:20:31.725409 132860367845184 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0425 16:20:31.725470 132860367845184 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0425 16:20:31.748299 132860367845184 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0425 16:20:31.748373 132860367845184 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0425 16:20:31.764291 132860367845184 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0425 16:20:31.764353 132860367845184 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0425 16:20:31.780195 132860367845184 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0425 16:20:31.780256 132860367845184 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0425 16:20:31.805422 132860367845184 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0425 16:20:31.805499 132860367845184 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0425 16:20:31.825052 132860367845184 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0425 16:20:31.825119 132860367845184 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0425 16:20:32.015770 132860367845184 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0425 16:20:32.015882 132860367845184 checkpointing.py:582] restoring from this run's directory step 9
I0425 16:20:32.503964 132860367845184 checkpointer.py:304] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints/9.
I0425 16:20:33.945486    1375 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0425 16:20:39.364862 132860367845184 jax_array_handlers.py:837] [process=3] /jax/orbax/read/worker/io/requested throughput: 291.322 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 5.424484491348267 s) (per-host)
I0425 16:20:39.380159 132860367845184 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/read/gbytes_per_sec: 268.947 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 5.88440728187561 s) (per-host)
I0425 16:20:40.747984 132860367845184 checkpointer.py:318] Finished restoring checkpoint in 8.48 seconds from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints/9.
I0425 16:20:40.748143 132860367845184 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints', 'checkpointer_start_time': 1777134032.2652192, 'checkpointer_duration_secs': 8.48289155960083, 'checkpoint_manager_start_time': 1777134032.0161495, 'checkpoint_manager_duration_secs': 8.731963396072388}
fsdp: 32
I0425 16:20:41.153750 132860367845184 maxtext_utils.py:1740]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0425 16:20:41.153883 132860367845184 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)
I0425 16:20:41.153960 132860367845184 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)
I0425 16:20:41.154036 132860367845184 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')
I0425 16:20:41.154093 132860367845184 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)
I0425 16:20:41.154136 132860367845184 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)
I0425 16:20:41.154192 132860367845184 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)
I0425 16:20:41.154249 132860367845184 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')
I0425 16:20:41.154290 132860367845184 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)
I0425 16:20:41.154330 132860367845184 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)
I0425 16:20:41.154378 132860367845184 maxtext_utils.py:1740]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed', 'vocab')
    Physical:  ('fsdp', None)

I0425 16:20:41.154429 132860367845184 maxtext_utils.py:1740]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed')
    Physical:  (None, 'fsdp')
I0425 16:20:41.653068 132860367845184 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0425 16:20:41.653164 132860367845184 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0425 16:20:41.668302 132860367845184 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0425 16:20:41.668366 132860367845184 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0425 16:20:52.596914 132860367845184 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.
I0425 16:20:52.597812 132860367845184 metric_logger.py:289] number parameters: 1.104 billion
I0425 16:21:04.056746 132860367845184 checkpointing.py:794] Waiting for step 10 to finish before checkpoint...
I0425 16:21:04.210584 132860367845184 checkpointing.py:798] Waited 0.1538245677947998 seconds for step 10 to finish before starting checkpointing.
I0425 16:21:04.213235 132860367845184 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0425 16:21:04.215449 132860367845184 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 10
I0425 16:21:04.218086 132860367845184 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints/10.
I0425 16:21:04.810092 132860367845184 signaling_client.py:364] Using JaxDistributedSignalingClient
I0425 16:21:04.811074 132860367845184 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0425 16:21:04.811137 132860367845184 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
I0425 16:21:05.056739 132860367845184 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.246742s
I0425 16:21:05.056912 132860367845184 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.041 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5072276592254639 s) (per-host)
I0425 16:21:05.056991 132860367845184 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.507314s (batch_requests_ready=0.258446s, total_serialization_initiated=0.248774s, others=0.000094s)
I0425 16:21:05.057114 132860367845184 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.511485s (all_items=0.000019s, per_item={'items': '0.00001860'}, temp_paths=0.511467)
I0425 16:21:05.058134 132744109598464 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started.
I0425 16:21:05.058317 132860367845184 async_checkpointer.py:561] Finished blocking save. Time taken: 0.842794s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints/10.
I0425 16:21:05.098836 132860367845184 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0425 16:21:05.099168 132736530503424 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save.
I0425 16:21:05.099328 132860367845184 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777134064.2132168, 'wait_for_prev_duration_secs': 7.414817810058594e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777134064.2154872, 'checkpointer_blocking_duration_secs': 0.8429470062255859, 'get_old_steps_start_time': 1777134065.058454, 'get_old_steps_duration_secs': 3.409385681152344e-05, 'checkpoint_manager_blocking_start_time': 1777134064.2110884, 'checkpoint_manager_blocking_duration_secs': 0.8882043361663818}
I0425 16:21:05.099448 132860367845184 checkpointing.py:409] Started an asynchronous checkpoint save for step 10
I0425 16:21:05.099508 132860367845184 max_utils.py:750] 
Memstats: After params initialized:
I0425 16:21:05.099575 132860367845184 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_10(process=3,(2,2,0,0))
I0425 16:21:05.099619 132860367845184 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_11(process=3,(3,2,0,0))
I0425 16:21:05.099655 132860367845184 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_14(process=3,(2,3,0,0))
I0425 16:21:05.099692 132860367845184 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_15(process=3,(3,3,0,0))
I0425 16:21:05.363865 132860367845184 metric_logger.py:185] completed step: 10, seconds: 11.459, TFLOP/s/device: 1.186, Tokens/s/device: 178.727, total_weights: 65536, loss: 6.833
I0425 16:21:05.538920 132860367845184 metric_logger.py:185] completed step: 11, seconds: 1.306, TFLOP/s/device: 10.406, Tokens/s/device: 1568.459, total_weights: 65536, loss: 6.528
I0425 16:21:05.686717 132860367845184 metric_logger.py:185] completed step: 12, seconds: 0.028, TFLOP/s/device: 482.429, Tokens/s/device: 72716.944, total_weights: 65536, loss: 6.265
I0425 16:21:05.834284 132860367845184 metric_logger.py:185] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.868, Tokens/s/device: 13395.164, total_weights: 65536, loss: 6.045
I0425 16:21:05.838962 132860367845184 checkpointing.py:794] Waiting for step 15 to finish before checkpoint...
I0425 16:21:06.129120 132860367845184 checkpointing.py:798] Waited 0.29012131690979004 seconds for step 15 to finish before starting checkpointing.
I0425 16:21:06.132089 132860367845184 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0425 16:21:09.528487 132738120816384 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints/10/items/array_metadatas/process_3
I0425 16:21:46.010972 132744109598464 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 38.098 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.461238384246826 s) (per-host)
I0425 16:21:46.011091 132744109598464 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.952843s.
I0425 16:21:53.862899 132744109598464 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 48.804634s.
I0425 16:21:53.863208 132736530503424 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save.
I0425 16:21:53.863337 132736530503424 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save.
I0425 16:21:53.863385 132736530503424 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0425 16:21:53.864835 132736530503424 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0425 16:21:53.865029 132860367845184 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10.
W0425 16:21:53.865164 132860367845184 checkpoint_manager.py:1458] Waiting for previous save to complete took 47.733080 seconds. If this number is high, consider checkpointing less frequently.
I0425 16:21:53.867191 132860367845184 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 15
I0425 16:21:53.871588 132860367845184 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints/15.
I0425 16:21:54.840453 132860367845184 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0425 16:21:54.840559 132860367845184 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
I0425 16:21:54.885076 132860367845184 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.045888s
I0425 16:21:54.885249 132860367845184 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.843 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3185079097747803 s) (per-host)
I0425 16:21:54.885303 132860367845184 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.318572s (batch_requests_ready=0.270766s, total_serialization_initiated=0.047735s, others=0.000072s)
I0425 16:21:54.885395 132860367845184 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.323301s (all_items=0.000017s, per_item={'items': '0.00001717'}, temp_paths=0.323284)
I0425 16:21:54.886366 132737088345856 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started.
I0425 16:21:54.886533 132860367845184 async_checkpointer.py:561] Finished blocking save. Time taken: 1.019268s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints/15.
I0425 16:21:55.322154 132860367845184 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0425 16:21:55.322533 132736530503424 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save.
I0425 16:21:55.322722 132860367845184 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777134066.1320596, 'wait_for_prev_duration_secs': 47.73307967185974, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777134113.8672318, 'checkpointer_blocking_duration_secs': 1.0194125175476074, 'get_old_steps_start_time': 1777134114.886665, 'get_old_steps_duration_secs': 2.6464462280273438e-05, 'checkpoint_manager_blocking_start_time': 1777134066.129761, 'checkpoint_manager_blocking_duration_secs': 49.19292402267456}
I0425 16:21:55.322848 132860367845184 checkpointing.py:409] Started an asynchronous checkpoint save for step 15
I0425 16:21:55.323519 132860367845184 metric_logger.py:185] completed step: 14, seconds: 0.148, TFLOP/s/device: 91.990, Tokens/s/device: 13865.663, total_weights: 65536, loss: 5.864
I0425 16:21:55.329247 132860367845184 metric_logger.py:185] completed step: 15, seconds: 0.148, TFLOP/s/device: 92.056, Tokens/s/device: 13875.715, total_weights: 65536, loss: 5.719
I0425 16:22:13.262196 132860367845184 metric_logger.py:185] completed step: 16, seconds: 49.489, TFLOP/s/device: 0.275, Tokens/s/device: 41.383, total_weights: 65536, loss: 5.605
I0425 16:22:13.409675 132860367845184 metric_logger.py:185] completed step: 17, seconds: 17.787, TFLOP/s/device: 0.764, Tokens/s/device: 115.141, total_weights: 65536, loss: 5.517
I0425 16:22:13.557179 132860367845184 metric_logger.py:185] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.435, Tokens/s/device: 13480.602, total_weights: 65536, loss: 5.448
I0425 16:22:13.704332 132860367845184 checkpointing.py:794] Waiting for step 19 to finish before checkpoint...
I0425 16:22:13.705017 132860367845184 checkpointing.py:798] Waited 0.0007050037384033203 seconds for step 19 to finish before starting checkpointing.
I0425 16:22:13.707112 132860367845184 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0425 16:22:16.643142 132738120816384 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints/15/items/array_metadatas/process_3
I0425 16:22:52.340522 132737088345856 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 27.341 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 57.77374029159546 s) (per-host)
I0425 16:22:52.340647 132737088345856 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 57.454169s.
I0425 16:23:00.784475 132737088345856 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 65.897982s.
I0425 16:23:00.784765 132736530503424 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save.
I0425 16:23:00.784885 132736530503424 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save.
I0425 16:23:00.784987 132736530503424 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0425 16:23:00.786478 132736530503424 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0425 16:23:00.786671 132860367845184 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15.
W0425 16:23:00.786824 132860367845184 checkpoint_manager.py:1458] Waiting for previous save to complete took 47.079722 seconds. If this number is high, consider checkpointing less frequently.
I0425 16:23:00.791576 132860367845184 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 19
I0425 16:23:00.794995 132860367845184 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints/19.
I0425 16:23:01.326882 132860367845184 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0425 16:23:01.326991 132860367845184 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
I0425 16:23:01.369647 132860367845184 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.043793s
I0425 16:23:01.369817 132860367845184 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.302 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.29091477394104004 s) (per-host)
I0425 16:23:01.369866 132860367845184 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.290973s (batch_requests_ready=0.245097s, total_serialization_initiated=0.045813s, others=0.000063s)
I0425 16:23:01.369969 132860367845184 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.295053s (all_items=0.000011s, per_item={'items': '0.00001073'}, temp_paths=0.295043)
I0425 16:23:01.370818 132744101205760 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started.
I0425 16:23:01.370988 132860367845184 async_checkpointer.py:561] Finished blocking save. Time taken: 0.579340s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints/19.
I0425 16:23:01.411150 132860367845184 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0425 16:23:01.411484 132736530503424 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save.
I0425 16:23:01.411648 132860367845184 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777134133.707085, 'wait_for_prev_duration_secs': 47.07972192764282, 'time_between_consecutive_saves_sec': 19.842210054397583, 'checkpointer_blocking_start_time': 1777134180.791617, 'checkpointer_blocking_duration_secs': 0.5794825553894043, 'get_old_steps_start_time': 1777134181.3711143, 'get_old_steps_duration_secs': 2.2172927856445312e-05, 'checkpoint_manager_blocking_start_time': 1777134133.7052407, 'checkpoint_manager_blocking_duration_secs': 47.706374645233154}
I0425 16:23:01.411759 132860367845184 checkpointing.py:409] Started an asynchronous checkpoint save for step 19
I0425 16:23:01.411803 132860367845184 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0425 16:23:06.086716 132744109598464 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_16_zarr3_true/checkpoints/19/items/array_metadatas/process_3
I0425 16:23:41.885864 132744101205760 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 38.709 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 40.80693864822388 s) (per-host)
I0425 16:23:41.885982 132744101205760 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.515052s.
I0425 16:23:50.568149 132744101205760 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 49.197205s.
I0425 16:23:50.568417 132736530503424 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save.
I0425 16:23:50.568542 132736530503424 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save.
I0425 16:23:50.568583 132736530503424 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0425 16:23:50.570051 132736530503424 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0425 16:23:50.570250 132860367845184 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19.
I0425 16:23:50.570406 132860367845184 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0425 16:23:50.571081 132860367845184 metric_logger.py:185] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.191, Tokens/s/device: 13895.957, 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: Sat Apr 25 16:24:00 UTC 2026
EXIT_CODE=0