MaxView

Timing Test Reference Updates
‹ 15_ocdbt_true_saveCase: 16_zarr3_false_resume16_zarr3_false_save ›

Metrics: main (586e69205) vs feat/nnx-set-defaults-true (73213e044)

Metricmain  586e69205feat/nnx-set-defaults-true  73213e044Diff (feat/nnx-set-defaults-true − main)
Parameters1.104 billion1.104 billion—
Final loss5.39105.39100
TFLOP/s92.32091.984-0.336
Tok/s13915.413864.8-50.589
Avg s/step8.4728.120-0.352
Memory %1.381.380
JAX0.9.20.8.1—

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

main  ·  586e69205  ·  main_20260423_071538  ·  full log 
XPK Start: Thu Apr 23 08:51:46 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. 
2026-04-23 08:52:11.120910: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303)
I0423 08:52:11.334999 132312705558336 max_utils.py:273] Attempting to initialize the jax distributed system...
I0423 08:52:20.378240 132312705558336 distributed.py:149] Starting JAX distributed service on [::]:8482
I0423 08:52:20.380506 132312705558336 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-rs-w2q9x-slice-job-0-0.mt-16-zarr3-false-rs-w2q9x:8482
I0423 08:52:22.207002 132312705558336 max_utils.py:284] Jax distributed system initialized!
I0423 08:52:28.498000 132312705558336 max_utils.py:800] System Information: Jax Version: 0.9.2
I0423 08:52:28.498121 132312705558336 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0423 08:52:28.498162 132312705558336 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
I0423 08:52:28.498198 132312705558336 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0423 08:52:29.194566 132312705558336 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0423 08:52:29.194851 132312705558336 checkpointing.py:677] Setting up checkpoint logger...
I0423 08:52:29.194905 132312705558336 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=False
I0423 08:52:29.194949 132312705558336 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0423 08:52:29.195310 132312705558336 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 0x7855d2d107a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0423 08:52:30.492290 132312705558336 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0423 08:52:30.492525 132312705558336 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 0x7841b812ae70>}, handler_registry=None
I0423 08:52:30.492758 132312705558336 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7841b812ae70>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`.
I0423 08:52:30.492809 132312705558336 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7841b812ee10>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`.
I0423 08:52:30.492846 132312705558336 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 0x7841b812ae70>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7841b812ae70>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7841b812ee10>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7841b812ee10>}).
I0423 08:52:30.493178 132312705558336 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0423 08:52:30.493254 132312705558336 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7841887a5a80> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0423 08:52:32.436202 132191526819584 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776934084443042511, 'commit_timestamp_nsecs': 1776934131625524951, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/0/_CHECKPOINT_METADATA
I0423 08:52:32.689230 132191526819584 checkpoint_manager.py:1871] Missing metrics for step 0
I0423 08:52:33.512684 132312705558336 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints
I0423 08:52:33.515269 132312705558336 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_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7841b812d6a0>
I0423 08:52:33.515383 132312705558336 checkpointing.py:301] Checkpoint manager created!
I0423 08:52:34.448889 132312705558336 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0423 08:52:34.449011 132312705558336 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 08:52:34.835207 132312705558336 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn').
I0423 08:52:34.835298 132312705558336 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 08:52:34.852027 132312705558336 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn').
I0423 08:52:34.852088 132312705558336 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 08:52:34.876164 132312705558336 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 08:52:34.876233 132312705558336 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 08:52:34.892980 132312705558336 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 08:52:34.893044 132312705558336 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 08:52:34.909823 132312705558336 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 08:52:34.909888 132312705558336 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 08:52:34.937742 132312705558336 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch_attn', 'activation_length_attn', 'activation_heads', 'activation_kv').
I0423 08:52:34.937873 132312705558336 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 08:52:34.964229 132312705558336 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0423 08:52:34.964296 132312705558336 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0423 08:52:35.182181 132312705558336 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0423 08:52:35.182294 132312705558336 checkpointing.py:581] restoring from this run's directory step 9
I0423 08:52:35.864939 132312705558336 event_tracking.py:70] [process=6] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/9.
I0423 08:52:36.980544 132312705558336 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/9.
I0423 08:52:36.980986 132312705558336 event_tracking.py:125] [process=6] [sync] Finished blocking load in 1.12 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/9.
I0423 08:52:39.339218    1454 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0423 08:52:46.305705 132312705558336 jax_array_handlers.py:843] [process=6] /jax/orbax/read/worker/io/requested throughput: 226.649 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.972342014312744 s) (per-host)
I0423 08:52:46.311295 132312705558336 base_pytree_checkpoint_handler.py:130] [process=6] /jax/checkpoint/read/gbytes_per_sec: 202.963 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 7.797441005706787 s) (per-host)
I0423 08:52:46.313577 132312705558336 event_tracking.py:138] [process=6] [sync] Finished load in 10.45 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/9
I0423 08:52:46.313673 132312705558336 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints', 'checkpointer_start_time': 1776934355.8648388, 'checkpointer_duration_secs': 10.448805809020996, 'checkpoint_manager_start_time': 1776934355.1825554, 'checkpoint_manager_duration_secs': 11.131090879440308}
fsdp: 32
I0423 08:52:46.738494 132312705558336 maxtext_utils.py:1707]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0423 08:52:46.738624 132312705558336 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)
I0423 08:52:46.738677 132312705558336 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)
I0423 08:52:46.738736 132312705558336 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')
I0423 08:52:46.738791 132312705558336 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)
I0423 08:52:46.738830 132312705558336 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)
I0423 08:52:46.738882 132312705558336 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)
I0423 08:52:46.738935 132312705558336 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')
I0423 08:52:46.738976 132312705558336 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)
I0423 08:52:46.739012 132312705558336 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)
I0423 08:52:46.739059 132312705558336 maxtext_utils.py:1707]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)
I0423 08:52:46.739117 132312705558336 maxtext_utils.py:1707]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')

I0423 08:52:47.233192 132312705558336 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 08:52:47.233297 132312705558336 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 08:52:47.249214 132312705558336 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 08:52:47.249276 132312705558336 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 08:52:58.309307 132312705558336 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.
I0423 08:52:58.310057 132312705558336 metric_logger.py:301] number parameters: 1.104 billion
I0423 08:53:09.933928 132312705558336 checkpointing.py:772] Waiting for step 10 to finish before checkpoint...
I0423 08:53:10.086792 132312705558336 checkpointing.py:776] Waited 0.15284252166748047 seconds for step 10 to finish before starting checkpointing.
I0423 08:53:10.089290 132312705558336 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 08:53:10.091183 132312705558336 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 10
I0423 08:53:10.092933 132312705558336 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/10.
I0423 08:53:10.851048 132312705558336 signaling_client.py:364] Using JaxDistributedSignalingClient
I0423 08:53:10.852146 132312705558336 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 08:53:10.852202 132312705558336 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
I0423 08:53:11.135125 132312705558336 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.284149s
I0423 08:53:11.135303 132312705558336 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.356 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2880263328552246 s) (per-host)
I0423 08:53:11.135354 132312705558336 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.288089s (batch_requests_ready=0.002183s, total_serialization_initiated=0.285835s, others=0.000071s)
I0423 08:53:11.135444 132312705558336 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.292186s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.292169)
I0423 08:53:11.136127 132312705558336 event_tracking.py:125] [process=6] [async] Finished blocking save in 1.04 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/10.
I0423 08:53:11.136475 132190948017920 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 09:13:11.136435
I0423 08:53:11.138536 132312705558336 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 08:53:11.138881 132187710703360 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0423 08:53:11.139043 132312705558336 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776934390.0892642, 'wait_for_prev_duration_secs': 6.628036499023438e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776934390.0912662, 'checkpointer_blocking_duration_secs': 1.0453650951385498, 'get_old_steps_start_time': 1776934391.1366534, 'get_old_steps_duration_secs': 3.075599670410156e-05, 'checkpoint_manager_blocking_start_time': 1776934390.0873528, 'checkpoint_manager_blocking_duration_secs': 1.0516481399536133}
I0423 08:53:11.139169 132312705558336 checkpointing.py:408] Started an asynchronous checkpoint save for step 10
I0423 08:53:11.139224 132312705558336 max_utils.py:750] 
Memstats: After params initialized:
I0423 08:53:11.139279 132312705558336 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0423 08:53:11.139319 132312705558336 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0423 08:53:11.139348 132312705558336 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0423 08:53:11.139374 132312705558336 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0423 08:53:11.446725 132312705558336 metric_logger.py:196] completed step: 10, seconds: 11.624, TFLOP/s/device: 1.169, Tokens/s/device: 176.191, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869
I0423 08:53:11.626476 132312705558336 metric_logger.py:196] completed step: 11, seconds: 1.511, TFLOP/s/device: 8.991, Tokens/s/device: 1355.214, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659
I0423 08:53:11.774052 132312705558336 metric_logger.py:196] completed step: 12, seconds: 0.033, TFLOP/s/device: 417.911, Tokens/s/device: 62992.126, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068
I0423 08:53:11.921521 132312705558336 metric_logger.py:196] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.667, Tokens/s/device: 13364.918, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001
I0423 08:53:11.925940 132312705558336 checkpointing.py:772] Waiting for step 15 to finish before checkpoint...
I0423 08:53:12.215640 132312705558336 checkpointing.py:776] Waited 0.28966498374938965 seconds for step 15 to finish before starting checkpointing.
I0423 08:53:12.218687 132312705558336 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 08:53:16.580318 132188238485248 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/10/items/array_metadatas/process_6
I0423 08:53:52.823370 132190948017920 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.631 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.97605919837952 s) (per-host)
I0423 08:53:52.823497 132190948017920 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.686909s.
I0423 08:54:00.655679 132190948017920 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 49.519073s.
I0423 08:54:00.655975 132187710703360 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 08:54:00.656087 132187710703360 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0423 08:54:00.656149 132187710703360 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0423 08:54:00.657591 132187710703360 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0423 08:54:00.657766 132312705558336 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.
W0423 08:54:00.657896 132312705558336 checkpoint_manager.py:1452] Waiting for previous save to complete took 48.439211 seconds. If this number is high, consider checkpointing less frequently.
I0423 08:54:00.659872 132312705558336 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 15
I0423 08:54:00.662278 132312705558336 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/15.
I0423 08:54:01.369206 132312705558336 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 08:54:01.369302 132312705558336 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
I0423 08:54:01.410231 132312705558336 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042769s
I0423 08:54:01.410397 132312705558336 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 33.365 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.0462338924407959 s) (per-host)
I0423 08:54:01.410446 132312705558336 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.046291s (batch_requests_ready=0.001773s, total_serialization_initiated=0.044453s, others=0.000065s)
I0423 08:54:01.410536 132312705558336 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.051194s (all_items=0.000012s, per_item={'items': '0.00001192'}, temp_paths=0.051182)
I0423 08:54:01.411245 132312705558336 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/15.
I0423 08:54:01.411548 132187710703360 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 09:14:01.411515
I0423 08:54:01.413603 132312705558336 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 08:54:01.413905 132188238485248 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0423 08:54:01.414071 132312705558336 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776934392.2186553, 'wait_for_prev_duration_secs': 48.43921113014221, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776934440.6599114, 'checkpointer_blocking_duration_secs': 0.751781702041626, 'get_old_steps_start_time': 1776934441.411717, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1776934392.216273, 'checkpoint_manager_blocking_duration_secs': 49.19776153564453}
I0423 08:54:01.414197 132312705558336 checkpointing.py:408] Started an asynchronous checkpoint save for step 15
I0423 08:54:01.414993 132312705558336 metric_logger.py:196] completed step: 14, seconds: 0.148, TFLOP/s/device: 92.052, Tokens/s/device: 13875.057, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193
I0423 08:54:01.421398 132312705558336 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.203, Tokens/s/device: 13897.843, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661
I0423 08:54:06.972141 132187693917952 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/15/items/array_metadatas/process_6
I0423 08:54:22.881310 132312705558336 metric_logger.py:196] completed step: 16, seconds: 49.494, TFLOP/s/device: 0.275, Tokens/s/device: 41.379, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872
I0423 08:54:23.028557 132312705558336 metric_logger.py:196] completed step: 17, seconds: 21.314, TFLOP/s/device: 0.637, Tokens/s/device: 96.088, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975
I0423 08:54:23.175765 132312705558336 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.209, Tokens/s/device: 13446.614, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333
I0423 08:54:23.322369 132312705558336 checkpointing.py:772] Waiting for step 19 to finish before checkpoint...
I0423 08:54:23.323061 132312705558336 checkpointing.py:776] Waited 0.0007121562957763672 seconds for step 19 to finish before starting checkpointing.
I0423 08:54:23.325117 132312705558336 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 08:54:43.045984 132187710703360 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.897 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.68179941177368 s) (per-host)
I0423 08:54:43.046088 132187710703360 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.634428s.
I0423 08:54:50.630109 132187710703360 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 49.218411s.
I0423 08:54:50.630415 132188238485248 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 08:54:50.630543 132188238485248 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0423 08:54:50.630595 132188238485248 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0423 08:54:50.632062 132188238485248 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0423 08:54:50.632259 132312705558336 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.
W0423 08:54:50.632384 132312705558336 checkpoint_manager.py:1452] Waiting for previous save to complete took 27.307285 seconds. If this number is high, consider checkpointing less frequently.
I0423 08:54:50.634263 132312705558336 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 19
I0423 08:54:50.636313 132312705558336 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/19.
I0423 08:54:51.382400 132312705558336 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 08:54:51.382495 132312705558336 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
I0423 08:54:51.421108 132312705558336 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.039537s
I0423 08:54:51.421270 132312705558336 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 35.871 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04300284385681152 s) (per-host)
I0423 08:54:51.421318 132312705558336 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.043060s (batch_requests_ready=0.001620s, total_serialization_initiated=0.041379s, others=0.000061s)
I0423 08:54:51.421404 132312705558336 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.047260s (all_items=0.000011s, per_item={'items': '0.00001073'}, temp_paths=0.047249)
I0423 08:54:51.422070 132312705558336 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.79 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/19.
I0423 08:54:51.422442 132188238485248 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 09:14:51.422405
I0423 08:54:51.424697 132312705558336 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 08:54:51.424922 132187693917952 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0423 08:54:51.425053 132312705558336 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776934463.3250773, 'wait_for_prev_duration_secs': 27.30728530883789, 'time_between_consecutive_saves_sec': 22.667447090148926, 'checkpointer_blocking_start_time': 1776934490.6343026, 'checkpointer_blocking_duration_secs': 0.788287878036499, 'get_old_steps_start_time': 1776934491.4226108, 'get_old_steps_duration_secs': 3.0994415283203125e-05, 'checkpoint_manager_blocking_start_time': 1776934463.3233113, 'checkpoint_manager_blocking_duration_secs': 28.10170555114746}
I0423 08:54:51.425203 132312705558336 checkpointing.py:408] Started an asynchronous checkpoint save for step 19
I0423 08:54:51.425265 132312705558336 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 08:54:56.908008 132186132965120 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/19/items/array_metadatas/process_6
I0423 08:55:33.436022 132188238485248 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.558 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.0577232837677 s) (per-host)
I0423 08:55:33.436140 132188238485248 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.013587s.
I0423 08:55:40.774259 132188238485248 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 49.351690s.
I0423 08:55:40.774498 132187693917952 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 08:55:40.774564 132187693917952 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0423 08:55:40.774603 132187693917952 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0423 08:55:40.777302 132187693917952 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0423 08:55:40.777453 132312705558336 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.
I0423 08:55:40.777605 132312705558336 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 08:55:40.778588 132312705558336 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.320, Tokens/s/device: 13915.407, 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: Thu Apr 23 08:55:51 UTC 2026
EXIT_CODE=0
feat/nnx-set-defaults-true  ·  73213e044  ·  feat_nnx_set_defaults_true_20260423_155242  ·  full log 
XPK Start: Thu Apr 23 17:28:31 UTC 2026
2026-04-23 17:28:35.835151: 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 00:00:1776965315.848266      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1776965315.852017      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1776965315.863795      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776965315.863813      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776965315.863816      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776965315.863818      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-23 17:28:55.092521: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303)
I0423 17:28:55.623805 136793988572992 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-23 17:29:04,666:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0423 17:29:04.666023 136793988572992 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-23 17:29:04,668:jax._src.distributed:157: Connecting to JAX distributed service on mt-16-zarr3-false-rs-y4zpc-slice-job-0-0.mt-16-zarr3-false-rs-y4zpc:8482
I0423 17:29:04.668334 136793988572992 distributed.py:157] Connecting to JAX distributed service on mt-16-zarr3-false-rs-y4zpc-slice-job-0-0.mt-16-zarr3-false-rs-y4zpc:8482
I0423 17:29:05.628062 136793988572992 max_utils.py:284] Jax distributed system initialized!
I0423 17:29:12.608857 136793988572992 max_utils.py:800] System Information: Jax Version: 0.8.1
I0423 17:29:12.608964 136793988572992 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0423 17:29:12.609005 136793988572992 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709
I0423 17:29:12.609041 136793988572992 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0423 17:29:13.224260 136793988572992 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0423 17:29:13.224848 136793988572992 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0423 17:29:13.225029 136793988572992 checkpointing.py:688] Setting up checkpoint logger...
I0423 17:29:13.225076 136793988572992 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=False
I0423 17:29:13.225132 136793988572992 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0423 17:29:13.225527 136793988572992 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 0x7c695b371280>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0423 17:29:14.483281 136793988572992 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0423 17:29:14.483624 136793988572992 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 0x7c54f07ed700>}, handler_registry=None
I0423 17:29:14.483889 136793988572992 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c54f07ed700>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`.
I0423 17:29:14.483944 136793988572992 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c54f07f26f0>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`.
I0423 17:29:14.483981 136793988572992 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 0x7c54f07ed700>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c54f07ed700>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c54f07f26f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c54f07f26f0>}).
I0423 17:29:14.484532 136793988572992 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0423 17:29:14.484614 136793988572992 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7c54f059c180> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0423 17:29:15.994969 136673870456576 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776965140081607379, 'commit_timestamp_nsecs': 1776965186126192949, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints/5/_CHECKPOINT_METADATA
I0423 17:29:16.255421 136673870456576 checkpoint_manager.py:1877] Missing metrics for step 5
I0423 17:29:16.658277 136793988572992 checkpoint_manager.py:1818] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints
I0423 17:29:16.726800 136793988572992 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_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7c54f07ec6b0>
I0423 17:29:16.726976 136793988572992 checkpointing.py:302] Checkpoint manager created!
I0423 17:29:17.621526 136793988572992 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0423 17:29:17.621634 136793988572992 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 17:29:17.994889 136793988572992 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 17:29:17.994980 136793988572992 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 17:29:18.011479 136793988572992 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 17:29:18.011542 136793988572992 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 17:29:18.035996 136793988572992 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 17:29:18.036068 136793988572992 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 17:29:18.052619 136793988572992 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 17:29:18.052682 136793988572992 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 17:29:18.068989 136793988572992 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 17:29:18.069052 136793988572992 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 17:29:18.094129 136793988572992 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0423 17:29:18.094199 136793988572992 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 17:29:18.114002 136793988572992 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0423 17:29:18.114068 136793988572992 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0423 17:29:18.305256 136793988572992 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0423 17:29:18.305363 136793988572992 checkpointing.py:582] restoring from this run's directory step 9
I0423 17:29:18.836466 136793988572992 checkpointer.py:304] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints/9.
I0423 17:29:20.248967    1412 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0423 17:29:26.408687 136793988572992 jax_array_handlers.py:837] [process=6] /jax/orbax/read/worker/io/requested throughput: 256.342 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.16471266746521 s) (per-host)
I0423 17:29:26.414473 136793988572992 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/read/gbytes_per_sec: 240.950 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.568153142929077 s) (per-host)
I0423 17:29:27.102623 136793988572992 checkpointer.py:318] Finished restoring checkpoint in 8.55 seconds from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints/9.
I0423 17:29:27.102781 136793988572992 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints', 'checkpointer_start_time': 1776965358.5543475, 'checkpointer_duration_secs': 8.548403024673462, 'checkpoint_manager_start_time': 1776965358.305607, 'checkpoint_manager_duration_secs': 8.797145366668701}
fsdp: 32
I0423 17:29:27.511279 136793988572992 maxtext_utils.py:1740]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0423 17:29:27.511411 136793988572992 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)
I0423 17:29:27.511467 136793988572992 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)
I0423 17:29:27.511529 136793988572992 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')
I0423 17:29:27.511582 136793988572992 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)
I0423 17:29:27.511627 136793988572992 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)
I0423 17:29:27.511683 136793988572992 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)
I0423 17:29:27.511739 136793988572992 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')
I0423 17:29:27.511781 136793988572992 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)
I0423 17:29:27.511820 136793988572992 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)
I0423 17:29:27.511869 136793988572992 maxtext_utils.py:1740]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed', 'vocab')
    Physical:  ('fsdp', None)
I0423 17:29:27.511917 136793988572992 maxtext_utils.py:1740]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed')
    Physical:  (None, 'fsdp')

I0423 17:29:28.014261 136793988572992 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 17:29:28.014357 136793988572992 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 17:29:28.029865 136793988572992 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 17:29:28.029928 136793988572992 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 17:29:38.714871 136793988572992 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.
I0423 17:29:38.715816 136793988572992 metric_logger.py:289] number parameters: 1.104 billion
I0423 17:29:49.791841 136793988572992 checkpointing.py:794] Waiting for step 10 to finish before checkpoint...
I0423 17:29:50.243954 136793988572992 checkpointing.py:798] Waited 0.45209217071533203 seconds for step 10 to finish before starting checkpointing.
I0423 17:29:50.246499 136793988572992 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 17:29:50.248234 136793988572992 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 10
I0423 17:29:50.250930 136793988572992 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints/10.
I0423 17:29:50.825804 136793988572992 signaling_client.py:364] Using JaxDistributedSignalingClient
I0423 17:29:50.826807 136793988572992 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0423 17:29:50.826862 136793988572992 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
I0423 17:29:51.069969 136793988572992 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.244254s
I0423 17:29:51.070152 136793988572992 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.086 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.4997866153717041 s) (per-host)
I0423 17:29:51.070208 136793988572992 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.499854s (batch_requests_ready=0.253505s, total_serialization_initiated=0.246273s, others=0.000075s)
I0423 17:29:51.070304 136793988572992 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.504407s (all_items=0.000020s, per_item={'items': '0.00001955'}, temp_paths=0.504388)
I0423 17:29:51.071183 136673300047616 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0423 17:29:51.071387 136793988572992 async_checkpointer.py:561] Finished blocking save. Time taken: 0.823075s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints/10.
I0423 17:29:51.337794 136793988572992 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 17:29:51.338231 136668996687616 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0423 17:29:51.338375 136793988572992 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776965390.2464805, 'wait_for_prev_duration_secs': 6.413459777832031e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776965390.2482724, 'checkpointer_blocking_duration_secs': 0.823228120803833, 'get_old_steps_start_time': 1776965391.0715199, 'get_old_steps_duration_secs': 3.552436828613281e-05, 'checkpoint_manager_blocking_start_time': 1776965390.2444658, 'checkpoint_manager_blocking_duration_secs': 1.0938708782196045}
I0423 17:29:51.338539 136793988572992 checkpointing.py:409] Started an asynchronous checkpoint save for step 10
I0423 17:29:51.338603 136793988572992 max_utils.py:750] 
Memstats: After params initialized:
I0423 17:29:51.338664 136793988572992 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0423 17:29:51.338698 136793988572992 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0423 17:29:51.338726 136793988572992 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0423 17:29:51.338752 136793988572992 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0423 17:29:51.604432 136793988572992 metric_logger.py:185] completed step: 10, seconds: 11.076, TFLOP/s/device: 1.227, Tokens/s/device: 184.906, total_weights: 65536, loss: 6.833
I0423 17:29:51.780937 136793988572992 metric_logger.py:185] completed step: 11, seconds: 1.811, TFLOP/s/device: 7.501, Tokens/s/device: 1130.705, total_weights: 65536, loss: 6.528
I0423 17:29:51.928405 136793988572992 metric_logger.py:185] completed step: 12, seconds: 0.030, TFLOP/s/device: 460.222, Tokens/s/device: 69369.644, total_weights: 65536, loss: 6.265
I0423 17:29:52.075695 136793988572992 metric_logger.py:185] completed step: 13, seconds: 0.153, TFLOP/s/device: 89.027, Tokens/s/device: 13419.125, total_weights: 65536, loss: 6.045
I0423 17:29:52.080068 136793988572992 checkpointing.py:794] Waiting for step 15 to finish before checkpoint...
I0423 17:29:52.370385 136793988572992 checkpointing.py:798] Waited 0.290283203125 seconds for step 15 to finish before starting checkpointing.
I0423 17:29:52.373251 136793988572992 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 17:29:56.103137 136669013473024 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_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints/10/items/array_metadatas/process_6
I0423 17:30:32.384148 136673300047616 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 37.777 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.81373476982117 s) (per-host)
I0423 17:30:32.384269 136673300047616 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.312944s.
I0423 17:30:40.814494 136673300047616 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 49.743147s.
I0423 17:30:40.814794 136668996687616 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 17:30:40.814913 136668996687616 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0423 17:30:40.814962 136668996687616 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0423 17:30:40.816361 136668996687616 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0423 17:30:40.816581 136793988572992 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.
W0423 17:30:40.816725 136793988572992 checkpoint_manager.py:1458] Waiting for previous save to complete took 48.443480 seconds. If this number is high, consider checkpointing less frequently.
I0423 17:30:40.818542 136793988572992 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 15
I0423 17:30:40.822819 136793988572992 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints/15.
I0423 17:30:41.790894 136793988572992 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0423 17:30:41.790993 136793988572992 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
I0423 17:30:41.833086 136793988572992 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.043381s
I0423 17:30:41.833267 136793988572992 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.908 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3143167495727539 s) (per-host)
I0423 17:30:41.833316 136793988572992 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.314375s (batch_requests_ready=0.269037s, total_serialization_initiated=0.045273s, others=0.000066s)
I0423 17:30:41.833404 136793988572992 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.319489s (all_items=0.000013s, per_item={'items': '0.00001287'}, temp_paths=0.319476)
I0423 17:30:41.834160 136669013473024 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0423 17:30:41.834261 136793988572992 async_checkpointer.py:561] Finished blocking save. Time taken: 1.015641s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints/15.
I0423 17:30:41.852573 136793988572992 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 17:30:41.852955 136668996687616 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0423 17:30:41.853129 136793988572992 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776965392.3732219, 'wait_for_prev_duration_secs': 48.44347953796387, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776965440.8185887, 'checkpointer_blocking_duration_secs': 1.015791654586792, 'get_old_steps_start_time': 1776965441.8343976, 'get_old_steps_duration_secs': 2.6464462280273438e-05, 'checkpoint_manager_blocking_start_time': 1776965392.3710678, 'checkpoint_manager_blocking_duration_secs': 49.48201775550842}
I0423 17:30:41.853236 136793988572992 checkpointing.py:409] Started an asynchronous checkpoint save for step 15
I0423 17:30:41.853783 136793988572992 metric_logger.py:185] completed step: 14, seconds: 0.148, TFLOP/s/device: 92.079, Tokens/s/device: 13879.100, total_weights: 65536, loss: 5.864
I0423 17:30:41.859904 136793988572992 metric_logger.py:185] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.326, Tokens/s/device: 13916.352, total_weights: 65536, loss: 5.719
I0423 17:30:59.766540 136793988572992 metric_logger.py:185] completed step: 16, seconds: 49.779, TFLOP/s/device: 0.273, Tokens/s/device: 41.142, total_weights: 65536, loss: 5.605
I0423 17:30:59.914164 136793988572992 metric_logger.py:185] completed step: 17, seconds: 17.761, TFLOP/s/device: 0.765, Tokens/s/device: 115.310, total_weights: 65536, loss: 5.517
I0423 17:31:00.061668 136793988572992 metric_logger.py:185] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.280, Tokens/s/device: 13457.217, total_weights: 65536, loss: 5.448
I0423 17:31:00.208561 136793988572992 checkpointing.py:794] Waiting for step 19 to finish before checkpoint...
I0423 17:31:00.209259 136793988572992 checkpointing.py:798] Waited 0.0007181167602539062 seconds for step 19 to finish before starting checkpointing.
I0423 17:31:00.211403 136793988572992 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 17:31:04.217085 136667907667712 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_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints/15/items/array_metadatas/process_6
I0423 17:31:41.103292 136669013473024 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 26.510 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 59.584296226501465 s) (per-host)
I0423 17:31:41.103414 136669013473024 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 59.269194s.
I0423 17:31:50.881987 136669013473024 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 69.047752s.
I0423 17:31:50.882458 136668996687616 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 17:31:50.882612 136668996687616 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0423 17:31:50.882663 136668996687616 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0423 17:31:50.885133 136668996687616 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0423 17:31:50.885336 136793988572992 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.
W0423 17:31:50.885477 136793988572992 checkpoint_manager.py:1458] Waiting for previous save to complete took 50.674087 seconds. If this number is high, consider checkpointing less frequently.
I0423 17:31:50.886967 136793988572992 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 19
I0423 17:31:50.890653 136793988572992 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints/19.
I0423 17:31:51.884657 136793988572992 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0423 17:31:51.884755 136793988572992 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
I0423 17:31:51.930395 136793988572992 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.046764s
I0423 17:31:51.930566 136793988572992 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.045 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.30579328536987305 s) (per-host)
I0423 17:31:51.930636 136793988572992 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.305870s (batch_requests_ready=0.256565s, total_serialization_initiated=0.049223s, others=0.000082s)
I0423 17:31:51.930759 136793988572992 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.309881s (all_items=0.000011s, per_item={'items': '0.00001097'}, temp_paths=0.309870)
I0423 17:31:51.931670 136673291654912 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0423 17:31:51.931822 136793988572992 async_checkpointer.py:561] Finished blocking save. Time taken: 1.044785s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints/19.
I0423 17:31:51.980016 136793988572992 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 17:31:51.980417 136668996687616 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0423 17:31:51.980583 136793988572992 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776965460.2113743, 'wait_for_prev_duration_secs': 50.674086570739746, 'time_between_consecutive_saves_sec': 19.394969940185547, 'checkpointer_blocking_start_time': 1776965510.887006, 'checkpointer_blocking_duration_secs': 1.0449211597442627, 'get_old_steps_start_time': 1776965511.931941, 'get_old_steps_duration_secs': 2.0265579223632812e-05, 'checkpoint_manager_blocking_start_time': 1776965460.209493, 'checkpoint_manager_blocking_duration_secs': 51.77105712890625}
I0423 17:31:51.980690 136793988572992 checkpointing.py:409] Started an asynchronous checkpoint save for step 19
I0423 17:31:51.980733 136793988572992 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 17:31:57.223411 136673300047616 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_20260423_155242/linen_xpk_feat_nnx_set_defaults_true_20260423_155242_16_zarr3_false/checkpoints/19/items/array_metadatas/process_6
I0423 17:32:34.605379 136673291654912 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 36.751 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.9805748462677 s) (per-host)
I0423 17:32:34.605485 136673291654912 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.673708s.
I0423 17:32:40.363245 136673291654912 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 48.431453s.
I0423 17:32:40.363526 136668996687616 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 17:32:40.363653 136668996687616 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0423 17:32:40.363695 136668996687616 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0423 17:32:40.365121 136668996687616 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0423 17:32:40.365310 136793988572992 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.
I0423 17:32:40.365453 136793988572992 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 17:32:40.366044 136793988572992 metric_logger.py:185] completed step: 19, seconds: 0.148, TFLOP/s/device: 91.984, Tokens/s/device: 13864.818, 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: Thu Apr 23 17:32:51 UTC 2026
EXIT_CODE=0