MaxView

Timing Test Reference Updates
‹ 16_zarr3_false_saveCase: 16_zarr3_true_resume16_zarr3_true_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/s92.16992.214+0.045
Tok/s13892.713899.5+6.883
Avg s/step8.4447.942-0.502
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 08:41:42 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 08:42:07.517640: 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 08:42:07.728953 140378241681216 max_utils.py:273] Attempting to initialize the jax distributed system...
I0423 08:42:16.770580 140378241681216 distributed.py:149] Starting JAX distributed service on [::]:8482
I0423 08:42:16.772898 140378241681216 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-true-rs-bc7xa-slice-job-0-0.mt-16-zarr3-true-rs-bc7xa:8482
I0423 08:42:17.542819 140378241681216 max_utils.py:284] Jax distributed system initialized!
I0423 08:42:23.543408 140378241681216 max_utils.py:800] System Information: Jax Version: 0.9.2
I0423 08:42:23.543515 140378241681216 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0423 08:42:23.543559 140378241681216 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 08:42:23.543595 140378241681216 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0423 08:42:24.235146 140378241681216 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0423 08:42:24.235429 140378241681216 checkpointing.py:677] Setting up checkpoint logger...
I0423 08:42:24.235483 140378241681216 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0423 08:42:24.235532 140378241681216 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0423 08:42:24.235879 140378241681216 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 0x7fabc6b18f20>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0423 08:42:25.498023 140378241681216 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0423 08:42:25.498288 140378241681216 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 0x7faa97947c80>}, handler_registry=None
I0423 08:42:25.498526 140378241681216 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7faa97947c80>` 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 08:42:25.498574 140378241681216 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f972450a540>` 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 08:42:25.498609 140378241681216 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 0x7faa97947c80>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7faa97947c80>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f972450a540>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f972450a540>}).
I0423 08:42:25.498924 140378241681216 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0423 08:42:25.498991 140378241681216 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f963c2e5800> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0423 08:42:27.414942 140255512491776 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776933531957309842, 'commit_timestamp_nsecs': 1776933578974848004, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_true/checkpoints/5/_CHECKPOINT_METADATA
I0423 08:42:27.658209 140255512491776 checkpoint_manager.py:1871] Missing metrics for step 5
I0423 08:42:28.538253 140378241681216 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_true/checkpoints
I0423 08:42:28.938981 140378241681216 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_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f9724507470>
I0423 08:42:28.939187 140378241681216 checkpointing.py:301] Checkpoint manager created!
I0423 08:42:29.868432 140378241681216 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0423 08:42:29.868555 140378241681216 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 08:42:30.246138 140378241681216 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn').
I0423 08:42:30.246240 140378241681216 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 08:42:30.262816 140378241681216 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn').
I0423 08:42:30.262880 140378241681216 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 08:42:30.286940 140378241681216 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 08:42:30.287013 140378241681216 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 08:42:30.303734 140378241681216 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 08:42:30.303807 140378241681216 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 08:42:30.320456 140378241681216 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 08:42:30.320520 140378241681216 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 08:42:30.345524 140378241681216 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch_attn', 'activation_length_attn', 'activation_heads', 'activation_kv').
I0423 08:42:30.345601 140378241681216 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 08:42:30.366940 140378241681216 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0423 08:42:30.367011 140378241681216 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0423 08:42:30.581654 140378241681216 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0423 08:42:30.581777 140378241681216 checkpointing.py:581] restoring from this run's directory step 9
I0423 08:42:30.832695 140378241681216 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_true/checkpoints/9.
I0423 08:42:31.078580 140378241681216 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_true/checkpoints/9.
I0423 08:42:31.079019 140378241681216 event_tracking.py:125] [process=6] [sync] Finished blocking load in 0.25 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_true/checkpoints/9.
I0423 08:42:33.351878    1469 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0423 08:42:39.637445 140378241681216 jax_array_handlers.py:843] [process=6] /jax/orbax/read/worker/io/requested throughput: 251.175 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.291522741317749 s) (per-host)
I0423 08:42:39.642892 140378241681216 base_pytree_checkpoint_handler.py:130] [process=6] /jax/checkpoint/read/gbytes_per_sec: 221.999 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 7.128828525543213 s) (per-host)
I0423 08:42:39.800992 140378241681216 event_tracking.py:138] [process=6] [sync] Finished load in 8.97 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_true/checkpoints/9
I0423 08:42:39.801186 140378241681216 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_true/checkpoints', 'checkpointer_start_time': 1776933750.8325965, 'checkpointer_duration_secs': 8.968554735183716, 'checkpoint_manager_start_time': 1776933750.5820434, 'checkpoint_manager_duration_secs': 9.21911072731018}
fsdp: 32
I0423 08:42:40.222033 140378241681216 maxtext_utils.py:1707]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0423 08:42:40.222186 140378241681216 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 08:42:40.222244 140378241681216 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 08:42:40.222305 140378241681216 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 08:42:40.222359 140378241681216 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 08:42:40.222397 140378241681216 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 08:42:40.222449 140378241681216 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 08:42:40.222507 140378241681216 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 08:42:40.222548 140378241681216 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 08:42:40.222583 140378241681216 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 08:42:40.222627 140378241681216 maxtext_utils.py:1707]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)
I0423 08:42:40.222671 140378241681216 maxtext_utils.py:1707]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]

    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')
I0423 08:42:40.709024 140378241681216 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 08:42:40.709136 140378241681216 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 08:42:40.724694 140378241681216 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 08:42:40.724755 140378241681216 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 08:42:51.610398 140378241681216 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 08:42:51.612698 140378241681216 metric_logger.py:301] number parameters: 1.104 billion
I0423 08:43:02.848384 140378241681216 checkpointing.py:772] Waiting for step 10 to finish before checkpoint...
I0423 08:43:03.174909 140378241681216 checkpointing.py:776] Waited 0.32650208473205566 seconds for step 10 to finish before starting checkpointing.
I0423 08:43:03.177304 140378241681216 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 08:43:03.179173 140378241681216 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 10
I0423 08:43:03.180654 140378241681216 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_true/checkpoints/10.
I0423 08:43:03.936449 140378241681216 signaling_client.py:364] Using JaxDistributedSignalingClient
I0423 08:43:03.937448 140378241681216 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 08:43:03.937505 140378241681216 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 08:43:04.216307 140378241681216 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.279929s
I0423 08:43:04.216478 140378241681216 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.435 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28380537033081055 s) (per-host)
I0423 08:43:04.216529 140378241681216 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.283867s (batch_requests_ready=0.002160s, total_serialization_initiated=0.281637s, others=0.000070s)
I0423 08:43:04.216618 140378241681216 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.288105s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=0.288087)
I0423 08:43:04.217272 140378241681216 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_true/checkpoints/10.
I0423 08:43:04.217637 140254321284864 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 09:03:04.217595
I0423 08:43:04.219607 140378241681216 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 08:43:04.219887 140255462135552 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0423 08:43:04.220018 140378241681216 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_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776933783.177285, 'wait_for_prev_duration_secs': 6.127357482910156e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776933783.17921, 'checkpointer_blocking_duration_secs': 1.0385887622833252, 'get_old_steps_start_time': 1776933784.2178228, 'get_old_steps_duration_secs': 3.170967102050781e-05, 'checkpoint_manager_blocking_start_time': 1776933783.175459, 'checkpoint_manager_blocking_duration_secs': 1.0445261001586914}
I0423 08:43:04.220186 140378241681216 checkpointing.py:408] Started an asynchronous checkpoint save for step 10
I0423 08:43:04.220234 140378241681216 max_utils.py:750] 
Memstats: After params initialized:
I0423 08:43:04.220282 140378241681216 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0423 08:43:04.220313 140378241681216 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0423 08:43:04.220338 140378241681216 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0423 08:43:04.220360 140378241681216 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0423 08:43:04.532129 140378241681216 metric_logger.py:196] completed step: 10, seconds: 11.236, TFLOP/s/device: 1.209, Tokens/s/device: 182.278, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869
I0423 08:43:04.712384 140378241681216 metric_logger.py:196] completed step: 11, seconds: 1.682, TFLOP/s/device: 8.077, Tokens/s/device: 1217.476, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659
I0423 08:43:04.859809 140378241681216 metric_logger.py:196] completed step: 12, seconds: 0.033, TFLOP/s/device: 410.524, Tokens/s/device: 61878.720, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068
I0423 08:43:05.007048 140378241681216 metric_logger.py:196] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.647, Tokens/s/device: 13361.779, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001
I0423 08:43:05.012259 140378241681216 checkpointing.py:772] Waiting for step 15 to finish before checkpoint...
I0423 08:43:05.301485 140378241681216 checkpointing.py:776] Waited 0.28918933868408203 seconds for step 15 to finish before starting checkpointing.
I0423 08:43:05.304533 140378241681216 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 08:43:09.529210 140254908512000 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_true/checkpoints/10/items/array_metadatas/process_6
I0423 08:43:46.161821 140254321284864 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.405 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.22911739349365 s) (per-host)
I0423 08:43:46.161931 140254321284864 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.944176s.
I0423 08:43:54.379284 140254321284864 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.161512s.
I0423 08:43:54.379564 140255462135552 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 08:43:54.379682 140255462135552 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0423 08:43:54.379730 140255462135552 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0423 08:43:54.381414 140255462135552 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0423 08:43:54.381574 140378241681216 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 08:43:54.381713 140378241681216 checkpoint_manager.py:1452] Waiting for previous save to complete took 49.077182 seconds. If this number is high, consider checkpointing less frequently.
I0423 08:43:54.383739 140378241681216 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 15
I0423 08:43:54.386198 140378241681216 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_true/checkpoints/15.
I0423 08:43:55.101116 140378241681216 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 08:43:55.101213 140378241681216 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 08:43:55.129231 140378241681216 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.029245s
I0423 08:43:55.129370 140378241681216 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 47.284 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.0326237678527832 s) (per-host)
I0423 08:43:55.129416 140378241681216 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.032679s (batch_requests_ready=0.001772s, total_serialization_initiated=0.030846s, others=0.000061s)
I0423 08:43:55.129496 140378241681216 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.037530s (all_items=0.000013s, per_item={'items': '0.00001287'}, temp_paths=0.037517)
I0423 08:43:55.130226 140378241681216 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_true/checkpoints/15.
I0423 08:43:55.130566 140255462135552 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 09:03:55.130527
I0423 08:43:55.140584 140378241681216 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 08:43:55.140825 140253297899264 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0423 08:43:55.140958 140378241681216 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_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776933785.3045015, 'wait_for_prev_duration_secs': 49.077181816101074, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776933834.3837783, 'checkpointer_blocking_duration_secs': 0.7469298839569092, 'get_old_steps_start_time': 1776933835.1307323, 'get_old_steps_duration_secs': 3.1948089599609375e-05, 'checkpoint_manager_blocking_start_time': 1776933785.3020546, 'checkpoint_manager_blocking_duration_secs': 49.83886647224426}
I0423 08:43:55.141077 140378241681216 checkpointing.py:408] Started an asynchronous checkpoint save for step 15
I0423 08:43:55.141931 140378241681216 metric_logger.py:196] completed step: 14, seconds: 0.147, TFLOP/s/device: 92.142, Tokens/s/device: 13888.606, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193
I0423 08:43:55.148441 140378241681216 metric_logger.py:196] completed step: 15, seconds: 0.148, TFLOP/s/device: 91.953, Tokens/s/device: 13860.126, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661
I0423 08:44:00.206448 140254908512000 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_true/checkpoints/15/items/array_metadatas/process_6
I0423 08:44:15.897934 140378241681216 metric_logger.py:196] completed step: 16, seconds: 50.135, TFLOP/s/device: 0.271, Tokens/s/device: 40.850, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872
I0423 08:44:16.045328 140378241681216 metric_logger.py:196] completed step: 17, seconds: 20.603, TFLOP/s/device: 0.659, Tokens/s/device: 99.401, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975
I0423 08:44:16.192697 140378241681216 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.136, Tokens/s/device: 13435.499, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333
I0423 08:44:16.339261 140378241681216 checkpointing.py:772] Waiting for step 19 to finish before checkpoint...
I0423 08:44:16.340013 140378241681216 checkpointing.py:776] Waited 0.0007801055908203125 seconds for step 19 to finish before starting checkpointing.
I0423 08:44:16.342024 140378241681216 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 08:44:37.000404 140255462135552 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.696 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.903631925582886 s) (per-host)
I0423 08:44:37.000508 140255462135552 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.869829s.
I0423 08:44:45.226682 140255462135552 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.095988s.
I0423 08:44:45.227003 140253297899264 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 08:44:45.227170 140253297899264 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0423 08:44:45.227220 140253297899264 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0423 08:44:45.232422 140253297899264 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0423 08:44:45.232591 140378241681216 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 08:44:45.232716 140378241681216 checkpoint_manager.py:1452] Waiting for previous save to complete took 28.890698 seconds. If this number is high, consider checkpointing less frequently.
I0423 08:44:45.234285 140378241681216 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 19
I0423 08:44:45.236294 140378241681216 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_true/checkpoints/19.
I0423 08:44:45.948458 140378241681216 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 08:44:45.948547 140378241681216 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 08:44:45.974276 140378241681216 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026733s
I0423 08:44:45.974462 140378241681216 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 51.104 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030185222625732422 s) (per-host)
I0423 08:44:45.974523 140378241681216 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030262s (batch_requests_ready=0.001618s, total_serialization_initiated=0.028561s, others=0.000083s)
I0423 08:44:45.974628 140378241681216 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034509s (all_items=0.000010s, per_item={'items': '0.00001049'}, temp_paths=0.034499)
I0423 08:44:45.975335 140378241681216 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.74 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_true/checkpoints/19.
I0423 08:44:45.975605 140253297899264 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 09:04:45.975582
I0423 08:44:45.978743 140378241681216 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 08:44:45.978963 140254908512000 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0423 08:44:45.979080 140378241681216 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_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776933856.3419929, 'wait_for_prev_duration_secs': 28.890697717666626, 'time_between_consecutive_saves_sec': 21.960540294647217, 'checkpointer_blocking_start_time': 1776933885.234324, 'checkpointer_blocking_duration_secs': 0.741379976272583, 'get_old_steps_start_time': 1776933885.9757233, 'get_old_steps_duration_secs': 2.574920654296875e-05, 'checkpoint_manager_blocking_start_time': 1776933856.340288, 'checkpoint_manager_blocking_duration_secs': 29.638758897781372}
I0423 08:44:45.979209 140378241681216 checkpointing.py:408] Started an asynchronous checkpoint save for step 19
I0423 08:44:45.979252 140378241681216 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 08:44:52.823704 140254321284864 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_true/checkpoints/19/items/array_metadatas/process_6
I0423 08:45:28.229225 140253297899264 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.356 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.28491544723511 s) (per-host)
I0423 08:45:28.229344 140253297899264 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.253665s.
I0423 08:45:36.648293 140253297899264 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.672598s.
I0423 08:45:36.648514 140254908512000 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 08:45:36.648571 140254908512000 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0423 08:45:36.648611 140254908512000 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0423 08:45:36.650444 140254908512000 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0423 08:45:36.650639 140378241681216 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 08:45:36.650799 140378241681216 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 08:45:36.651797 140378241681216 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.169, Tokens/s/device: 13892.658, 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 08:45:47 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 17:17:59 UTC 2026
2026-04-23 17:18:03.766791: 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 00:00:1776964683.780270      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1776964683.784020      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1776964683.795679      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776964683.795699      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776964683.795701      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776964683.795703      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-23 17:18:22.831476: 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 17:18:23.364759 138370872182592 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-23 17:18:32,405:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0423 17:18:32.405601 138370872182592 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-23 17:18:32,407:jax._src.distributed:157: Connecting to JAX distributed service on mt-16-zarr3-true-rs-hzwns-slice-job-0-0.mt-16-zarr3-true-rs-hzwns:8482
I0423 17:18:32.407994 138370872182592 distributed.py:157] Connecting to JAX distributed service on mt-16-zarr3-true-rs-hzwns-slice-job-0-0.mt-16-zarr3-true-rs-hzwns:8482
I0423 17:18:33.685057 138370872182592 max_utils.py:284] Jax distributed system initialized!
I0423 17:18:40.561052 138370872182592 max_utils.py:800] System Information: Jax Version: 0.8.1
I0423 17:18:40.561179 138370872182592 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0423 17:18:40.561223 138370872182592 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 17:18:40.561259 138370872182592 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0423 17:18:41.238361 138370872182592 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0423 17:18:41.238954 138370872182592 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0423 17:18:41.239157 138370872182592 checkpointing.py:688] Setting up checkpoint logger...
I0423 17:18:41.239208 138370872182592 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0423 17:18:41.239252 138370872182592 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0423 17:18:41.239657 138370872182592 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 0x7dd86f575610>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0423 17:18:42.547949 138370872182592 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0423 17:18:42.548211 138370872182592 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 0x7dd73573e7b0>}, handler_registry=None
I0423 17:18:42.548449 138370872182592 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dd73573e7b0>` 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 17:18:42.548558 138370872182592 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc3f830aba0>` 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 17:18:42.548598 138370872182592 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 0x7dd73573e7b0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dd73573e7b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc3f830aba0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dc3f830aba0>}).
I0423 17:18:42.549129 138370872182592 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0423 17:18:42.549213 138370872182592 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7dc3b86fc180> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0423 17:18:44.174314 138249997629184 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776964560544046702, 'commit_timestamp_nsecs': 1776964606358704134, '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_true/checkpoints/9/_CHECKPOINT_METADATA
I0423 17:18:44.427183 138249997629184 checkpoint_manager.py:1877] Missing metrics for step 9
I0423 17:18:45.242194 138370872182592 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_true/checkpoints
I0423 17:18:45.244707 138370872182592 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_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7dc3b87dcef0>
I0423 17:18:45.244823 138370872182592 checkpointing.py:302] Checkpoint manager created!
I0423 17:18:46.147294 138370872182592 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0423 17:18:46.147403 138370872182592 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 17:18:46.513576 138370872182592 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 17:18:46.513669 138370872182592 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 17:18:46.529461 138370872182592 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 17:18:46.529529 138370872182592 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 17:18:46.552121 138370872182592 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 17:18:46.552186 138370872182592 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 17:18:46.568017 138370872182592 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 17:18:46.568081 138370872182592 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 17:18:46.583833 138370872182592 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 17:18:46.583892 138370872182592 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 17:18:46.608493 138370872182592 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0423 17:18:46.608566 138370872182592 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 17:18:46.628077 138370872182592 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0423 17:18:46.628156 138370872182592 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0423 17:18:46.817675 138370872182592 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0423 17:18:46.817784 138370872182592 checkpointing.py:582] restoring from this run's directory step 9
I0423 17:18:47.314365 138370872182592 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_true/checkpoints/9.
I0423 17:18:48.731772    1427 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0423 17:18:55.097516 138370872182592 jax_array_handlers.py:837] [process=6] /jax/orbax/read/worker/io/requested throughput: 248.042 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.3709800243377686 s) (per-host)
I0423 17:18:55.111090 138370872182592 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/read/gbytes_per_sec: 233.212 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.786073923110962 s) (per-host)
I0423 17:18:55.185784 138370872182592 checkpointer.py:318] Finished restoring checkpoint in 8.13 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_true/checkpoints/9.
I0423 17:18:55.185934 138370872182592 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_true/checkpoints', 'checkpointer_start_time': 1776964727.0594966, 'checkpointer_duration_secs': 8.126405954360962, 'checkpoint_manager_start_time': 1776964726.8180299, 'checkpoint_manager_duration_secs': 8.36787486076355}
fsdp: 32
I0423 17:18:55.589030 138370872182592 maxtext_utils.py:1740]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0423 17:18:55.589182 138370872182592 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 17:18:55.589244 138370872182592 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 17:18:55.589307 138370872182592 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 17:18:55.589361 138370872182592 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 17:18:55.589401 138370872182592 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 17:18:55.589456 138370872182592 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 17:18:55.589510 138370872182592 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 17:18:55.589551 138370872182592 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 17:18:55.589590 138370872182592 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 17:18:55.589637 138370872182592 maxtext_utils.py:1740]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed', 'vocab')
    Physical:  ('fsdp', None)

I0423 17:18:55.589685 138370872182592 maxtext_utils.py:1740]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed')
    Physical:  (None, 'fsdp')
I0423 17:18:56.083471 138370872182592 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 17:18:56.083566 138370872182592 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 17:18:56.098418 138370872182592 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 17:18:56.098478 138370872182592 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 17:19:06.941836 138370872182592 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 17:19:06.942729 138370872182592 metric_logger.py:289] number parameters: 1.104 billion
I0423 17:19:17.862987 138370872182592 checkpointing.py:794] Waiting for step 10 to finish before checkpoint...
I0423 17:19:18.221888 138370872182592 checkpointing.py:798] Waited 0.35888028144836426 seconds for step 10 to finish before starting checkpointing.
I0423 17:19:18.224403 138370872182592 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 17:19:18.226284 138370872182592 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 10
I0423 17:19:18.229274 138370872182592 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_true/checkpoints/10.
I0423 17:19:18.829573 138370872182592 signaling_client.py:364] Using JaxDistributedSignalingClient
I0423 17:19:18.830647 138370872182592 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0423 17:19:18.830704 138370872182592 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 17:19:19.083142 138370872182592 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.253661s
I0423 17:19:19.083337 138370872182592 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.012 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5121040344238281 s) (per-host)
I0423 17:19:19.083402 138370872182592 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.512184s (batch_requests_ready=0.256405s, total_serialization_initiated=0.255688s, others=0.000090s)
I0423 17:19:19.083516 138370872182592 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.516288s (all_items=0.000020s, per_item={'items': '0.00001979'}, temp_paths=0.516268)
I0423 17:19:19.084459 138249485936384 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0423 17:19:19.084594 138370872182592 async_checkpointer.py:561] Finished blocking save. Time taken: 0.858237s. 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_true/checkpoints/10.
I0423 17:19:19.109684 138370872182592 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 17:19:19.110000 138245199361792 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0423 17:19:19.110185 138370872182592 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_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776964758.2243793, 'wait_for_prev_duration_secs': 0.00011324882507324219, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776964758.2263243, 'checkpointer_blocking_duration_secs': 0.8584539890289307, 'get_old_steps_start_time': 1776964759.0847971, 'get_old_steps_duration_secs': 2.6464462280273438e-05, 'checkpoint_manager_blocking_start_time': 1776964758.222633, 'checkpoint_manager_blocking_duration_secs': 0.8875133991241455}
I0423 17:19:19.110289 138370872182592 checkpointing.py:409] Started an asynchronous checkpoint save for step 10
I0423 17:19:19.110341 138370872182592 max_utils.py:750] 
Memstats: After params initialized:
I0423 17:19:19.110395 138370872182592 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0423 17:19:19.110428 138370872182592 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0423 17:19:19.110455 138370872182592 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0423 17:19:19.110480 138370872182592 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0423 17:19:19.376579 138370872182592 metric_logger.py:185] completed step: 10, seconds: 10.920, TFLOP/s/device: 1.244, Tokens/s/device: 187.543, total_weights: 65536, loss: 6.833
I0423 17:19:19.543596 138370872182592 metric_logger.py:185] completed step: 11, seconds: 1.512, TFLOP/s/device: 8.984, Tokens/s/device: 1354.241, total_weights: 65536, loss: 6.528
I0423 17:19:19.691157 138370872182592 metric_logger.py:185] completed step: 12, seconds: 0.020, TFLOP/s/device: 687.434, Tokens/s/device: 103617.506, total_weights: 65536, loss: 6.265
I0423 17:19:19.838603 138370872182592 metric_logger.py:185] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.802, Tokens/s/device: 13385.183, total_weights: 65536, loss: 6.045
I0423 17:19:19.843233 138370872182592 checkpointing.py:794] Waiting for step 15 to finish before checkpoint...
I0423 17:19:20.133374 138370872182592 checkpointing.py:798] Waited 0.2901036739349365 seconds for step 15 to finish before starting checkpointing.
I0423 17:19:20.136383 138370872182592 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 17:19:23.743610 138245694269184 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_true/checkpoints/10/items/array_metadatas/process_6
I0423 17:20:00.427506 138249485936384 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 37.739 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.856244802474976 s) (per-host)
I0423 17:20:00.427626 138249485936384 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.342950s.
I0423 17:20:06.865222 138249485936384 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 47.780530s.
I0423 17:20:06.865523 138245199361792 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 17:20:06.865638 138245199361792 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0423 17:20:06.865686 138245199361792 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0423 17:20:06.867401 138245199361792 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0423 17:20:06.867639 138370872182592 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 17:20:06.867771 138370872182592 checkpoint_manager.py:1458] Waiting for previous save to complete took 46.731394 seconds. If this number is high, consider checkpointing less frequently.
I0423 17:20:06.869672 138370872182592 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 15
I0423 17:20:06.873814 138370872182592 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_true/checkpoints/15.
I0423 17:20:07.424213 138370872182592 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0423 17:20:07.424317 138370872182592 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 17:20:07.466780 138370872182592 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.043784s
I0423 17:20:07.466947 138370872182592 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.987 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3093407154083252 s) (per-host)
I0423 17:20:07.466998 138370872182592 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.309401s (batch_requests_ready=0.263689s, total_serialization_initiated=0.045646s, others=0.000065s)
I0423 17:20:07.467086 138370872182592 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.314231s (all_items=0.000013s, per_item={'items': '0.00001311'}, temp_paths=0.314218)
I0423 17:20:07.468042 138243563570944 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0423 17:20:07.468192 138370872182592 async_checkpointer.py:561] Finished blocking save. Time taken: 0.598441s. 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_true/checkpoints/15.
I0423 17:20:07.735659 138370872182592 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 17:20:07.736028 138245199361792 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0423 17:20:07.736211 138370872182592 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_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776964760.136353, 'wait_for_prev_duration_secs': 46.731393814086914, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776964806.8697114, 'checkpointer_blocking_duration_secs': 0.5986838340759277, 'get_old_steps_start_time': 1776964807.4684193, 'get_old_steps_duration_secs': 3.218650817871094e-05, 'checkpoint_manager_blocking_start_time': 1776964760.1340475, 'checkpoint_manager_blocking_duration_secs': 47.60212850570679}
I0423 17:20:07.736317 138370872182592 checkpointing.py:409] Started an asynchronous checkpoint save for step 15
I0423 17:20:07.736950 138370872182592 metric_logger.py:185] completed step: 14, seconds: 0.148, TFLOP/s/device: 92.041, Tokens/s/device: 13873.459, total_weights: 65536, loss: 5.864
I0423 17:20:07.742813 138370872182592 metric_logger.py:185] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.150, Tokens/s/device: 13889.831, total_weights: 65536, loss: 5.719
I0423 17:20:26.212639 138370872182592 metric_logger.py:185] completed step: 16, seconds: 47.898, TFLOP/s/device: 0.284, Tokens/s/device: 42.757, total_weights: 65536, loss: 5.605
I0423 17:20:26.360221 138370872182592 metric_logger.py:185] completed step: 17, seconds: 18.324, TFLOP/s/device: 0.741, Tokens/s/device: 111.767, total_weights: 65536, loss: 5.517
I0423 17:20:26.507471 138370872182592 metric_logger.py:185] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.361, Tokens/s/device: 13469.431, total_weights: 65536, loss: 5.448
I0423 17:20:26.654477 138370872182592 checkpointing.py:794] Waiting for step 19 to finish before checkpoint...
I0423 17:20:26.655170 138370872182592 checkpointing.py:798] Waited 0.0007123947143554688 seconds for step 19 to finish before starting checkpointing.
I0423 17:20:26.657111 138370872182592 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 17:20:29.648294 138245694269184 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_true/checkpoints/15/items/array_metadatas/process_6
I0423 17:21:07.926517 138243563570944 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 25.994 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 60.76886963844299 s) (per-host)
I0423 17:21:07.926633 138243563570944 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 60.458353s.
I0423 17:21:13.600569 138243563570944 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 66.132273s.
I0423 17:21:13.600815 138245199361792 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 17:21:13.600883 138245199361792 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0423 17:21:13.600933 138245199361792 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0423 17:21:13.603457 138245199361792 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0423 17:21:13.603643 138370872182592 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 17:21:13.603796 138370872182592 checkpoint_manager.py:1458] Waiting for previous save to complete took 46.946709 seconds. If this number is high, consider checkpointing less frequently.
I0423 17:21:13.605387 138370872182592 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 19
I0423 17:21:13.608888 138370872182592 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_true/checkpoints/19.
I0423 17:21:14.174787 138370872182592 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0423 17:21:14.174885 138370872182592 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 17:21:14.218260 138370872182592 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044381s
I0423 17:21:14.218424 138370872182592 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.743 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3252098560333252 s) (per-host)
I0423 17:21:14.218471 138370872182592 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.325268s (batch_requests_ready=0.278768s, total_serialization_initiated=0.046438s, others=0.000062s)
I0423 17:21:14.218563 138370872182592 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.329278s (all_items=0.000012s, per_item={'items': '0.00001168'}, temp_paths=0.329266)
I0423 17:21:14.219425 138249477543680 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0423 17:21:14.219599 138370872182592 async_checkpointer.py:561] Finished blocking save. Time taken: 0.614141s. 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_true/checkpoints/19.
I0423 17:21:14.224990 138370872182592 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 17:21:14.225229 138245199361792 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0423 17:21:14.225366 138370872182592 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_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776964826.6570704, 'wait_for_prev_duration_secs': 46.9467089176178, 'time_between_consecutive_saves_sec': 19.789621114730835, 'checkpointer_blocking_start_time': 1776964873.6054254, 'checkpointer_blocking_duration_secs': 0.614285945892334, 'get_old_steps_start_time': 1776964874.2197242, 'get_old_steps_duration_secs': 2.2649765014648438e-05, 'checkpoint_manager_blocking_start_time': 1776964826.6554055, 'checkpoint_manager_blocking_duration_secs': 47.569923400878906}
I0423 17:21:14.225471 138370872182592 checkpointing.py:409] Started an asynchronous checkpoint save for step 19
I0423 17:21:14.225514 138370872182592 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 17:21:19.913733 138249485936384 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_true/checkpoints/19/items/array_metadatas/process_6
I0423 17:21:56.182330 138249477543680 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 37.352 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.28908729553223 s) (per-host)
I0423 17:21:56.182431 138249477543680 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.962887s.
I0423 17:22:03.331635 138249477543680 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 49.112076s.
I0423 17:22:03.331855 138245199361792 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 17:22:03.331922 138245199361792 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0423 17:22:03.331965 138245199361792 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0423 17:22:03.334690 138245199361792 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0423 17:22:03.334853 138370872182592 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 17:22:03.335003 138370872182592 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 17:22:03.335718 138370872182592 metric_logger.py:185] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.214, Tokens/s/device: 13899.541, 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 17:22:15 UTC 2026
EXIT_CODE=0