MaxView

Timing Test Reference Updates
‹ 15_ocdbt_false_resumeCase: 15_ocdbt_false_save15_ocdbt_true_resume ›

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

Metricmain  b117f50cffeat/nnx-set-defaults-true  73213e044Diff (feat/nnx-set-defaults-true − main)
Parameters1.104 billion1.104 billion—
Final loss7.18107.18100
TFLOP/s92.18992.187-0.002
Tok/s13895.713895.4-0.283
Avg s/step4.6504.225-0.425
Memory %1.381.380
JAX0.9.20.8.1—

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

main  ·  b117f50cf  ·  main_20260424_070227  ·  full log 
XPK Start: Fri Apr 24 08:09:28 UTC 2026
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
`rope_parameters`'s factor field must be a float >= 1, got 40
`rope_parameters`'s beta_fast field must be a float, got 32
`rope_parameters`'s beta_slow field must be a float, got 1
DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 
2026-04-24 08:09:53.043284: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303)
I0424 08:09:53.250511 136717846619968 max_utils.py:273] Attempting to initialize the jax distributed system...
I0424 08:10:02.291826 136717846619968 distributed.py:149] Starting JAX distributed service on [::]:8482
I0424 08:10:02.294214 136717846619968 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-false-sv-cf2hk-slice-job-0-0.mt-15-ocdbt-false-sv-cf2hk:8482
I0424 08:10:03.431330 136717846619968 max_utils.py:284] Jax distributed system initialized!
I0424 08:10:09.656880 136717846619968 max_utils.py:800] System Information: Jax Version: 0.9.2
I0424 08:10:09.656987 136717846619968 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0424 08:10:09.657026 136717846619968 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365
I0424 08:10:09.657061 136717846619968 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 08:10:10.356773 136717846619968 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 08:10:10.357053 136717846619968 checkpointing.py:677] Setting up checkpoint logger...
I0424 08:10:10.357121 136717846619968 checkpointing.py:233] Creating checkpoint manager with ocdbt=False and zarr3=True
I0424 08:10:10.357168 136717846619968 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0424 08:10:10.357503 136717846619968 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=False, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7c579e1ca6c0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 08:10:13.204048 136717846619968 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0424 08:10:13.204312 136717846619968 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 0x7c4370304140>}, handler_registry=None
I0424 08:10:13.204557 136717846619968 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c4370304140>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`.
I0424 08:10:13.204606 136717846619968 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c4370307e30>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`.
I0424 08:10:13.204642 136717846619968 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 0x7c4370304140>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c4370304140>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c4370307e30>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c4370307e30>}).
I0424 08:10:13.204967 136717846619968 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0424 08:10:13.205035 136717846619968 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7c43387d1a80> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0424 08:10:13.997034 136717846619968 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints
I0424 08:10:13.999293 136717846619968 checkpoint_manager.py:929] [process=6][thread=MainThread] CheckpointManager created,  primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7c4370465b50>
I0424 08:10:13.999407 136717846619968 checkpointing.py:301] Checkpoint manager created!
I0424 08:10:14.943631 136717846619968 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 08:10:14.943755 136717846619968 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 08:10:15.320410 136717846619968 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn').
I0424 08:10:15.320503 136717846619968 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 08:10:15.337074 136717846619968 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn').
I0424 08:10:15.337146 136717846619968 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 08:10:15.361024 136717846619968 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 08:10:15.361107 136717846619968 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 08:10:15.377774 136717846619968 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 08:10:15.377835 136717846619968 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 08:10:15.394383 136717846619968 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 08:10:15.394443 136717846619968 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 08:10:15.419265 136717846619968 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch_attn', 'activation_length_attn', 'activation_heads', 'activation_kv').
I0424 08:10:15.419342 136717846619968 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 08:10:15.440567 136717846619968 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 08:10:15.440632 136717846619968 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 08:10:15.652059 136717846619968 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 08:10:15.652191 136717846619968 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0424 08:10:17.090396 136717846619968 maxtext_utils.py:1707]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0424 08:10:17.090521 136717846619968 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0424 08:10:17.090573 136717846619968 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0424 08:10:17.090627 136717846619968 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   P('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0424 08:10:17.090679 136717846619968 maxtext_utils.py:1707]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0424 08:10:17.090722 136717846619968 maxtext_utils.py:1707]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0424 08:10:17.090776 136717846619968 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/key/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0424 08:10:17.090827 136717846619968 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/out/kernel
    Shape:     float32[16,16,128,2048]
    Logical:   P('heads', 'layers', 'kv', 'embed')
    Physical:  (None, None, None, 'fsdp')
I0424 08:10:17.090866 136717846619968 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/query/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None, None)
I0424 08:10:17.090901 136717846619968 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/value/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0424 08:10:17.090947 136717846619968 maxtext_utils.py:1707]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)

I0424 08:10:17.090994 136717846619968 maxtext_utils.py:1707]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')
I0424 08:10:17.591812 136717846619968 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 08:10:17.591908 136717846619968 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 08:10:17.607535 136717846619968 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 08:10:17.607597 136717846619968 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 08:10:28.352028 136717846619968 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0424 08:10:28.352849 136717846619968 metric_logger.py:301] number parameters: 1.104 billion
I0424 08:10:40.062044 136717846619968 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0424 08:10:40.230339 136717846619968 checkpointing.py:776] Waited 0.16826248168945312 seconds for step 0 to finish before starting checkpointing.
I0424 08:10:40.233064 136717846619968 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 08:10:40.236868 136717846619968 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0
I0424 08:10:40.238279 136717846619968 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/0.
I0424 08:10:40.520429 136717846619968 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 08:10:40.521486 136717846619968 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 08:10:40.521600 136717846619968 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0424 08:10:40.806139 136717846619968 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.285782s
I0424 08:10:40.806313 136717846619968 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.296 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.29126930236816406 s) (per-host)
I0424 08:10:40.806365 136717846619968 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.291330s (batch_requests_ready=0.002227s, total_serialization_initiated=0.289035s, others=0.000069s)
I0424 08:10:40.806456 136717846619968 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.295394s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.295377)
I0424 08:10:40.807373 136717846619968 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.57 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/0.
I0424 08:10:40.807678 136589982017280 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 08:30:40.807647
I0424 08:10:40.813739 136717846619968 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 08:10:40.814051 136588365371136 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 08:10:40.814223 136717846619968 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777018240.2330458, 'wait_for_prev_duration_secs': 7.510185241699219e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777018240.2369084, 'checkpointer_blocking_duration_secs': 0.5709133148193359, 'get_old_steps_start_time': 1777018240.8078492, 'get_old_steps_duration_secs': 3.0040740966796875e-05, 'checkpoint_manager_blocking_start_time': 1777018240.2308917, 'checkpoint_manager_blocking_duration_secs': 0.5832929611206055}
I0424 08:10:40.814327 136717846619968 checkpointing.py:408] Started an asynchronous checkpoint save for step 0
I0424 08:10:40.814378 136717846619968 max_utils.py:750] 
Memstats: After params initialized:
I0424 08:10:40.814426 136717846619968 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0424 08:10:40.814458 136717846619968 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0424 08:10:40.814486 136717846619968 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0424 08:10:40.814517 136717846619968 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0424 08:10:41.126300 136717846619968 metric_logger.py:196] completed step: 0, seconds: 11.709, TFLOP/s/device: 1.160, Tokens/s/device: 174.907, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875
I0424 08:10:41.301966 136717846619968 metric_logger.py:196] completed step: 1, seconds: 1.063, TFLOP/s/device: 12.786, Tokens/s/device: 1927.194, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875
I0424 08:10:41.743802 136717846619968 metric_logger.py:196] completed step: 2, seconds: 0.028, TFLOP/s/device: 476.842, Tokens/s/device: 71874.781, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719
I0424 08:10:41.891113 136717846619968 metric_logger.py:196] completed step: 3, seconds: 0.442, TFLOP/s/device: 30.726, Tokens/s/device: 4631.336, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506
I0424 08:10:41.895524 136717846619968 checkpointing.py:772] Waiting for step 5 to finish before checkpoint...
I0424 08:10:42.185058 136717846619968 checkpointing.py:776] Waited 0.28949832916259766 seconds for step 5 to finish before starting checkpointing.
I0424 08:10:42.187989 136717846619968 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 08:10:44.798974    2798 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 08:10:46.882181 136588373763840 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/0/items/array_metadatas/process_6
I0424 08:10:52.316746 136589982017280 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 133.845 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 11.80166220664978 s) (per-host)
I0424 08:10:52.316871 136589982017280 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 11.509086s.
I0424 08:10:58.316721 136589982017280 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 17.508919s.
I0424 08:10:58.317002 136588365371136 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 08:10:58.317156 136588365371136 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 08:10:58.317207 136588365371136 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0424 08:10:58.319393 136588365371136 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0424 08:10:58.319567 136717846619968 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0424 08:10:58.319704 136717846619968 checkpoint_manager.py:1452] Waiting for previous save to complete took 16.131716 seconds. If this number is high, consider checkpointing less frequently.
I0424 08:10:58.321832 136717846619968 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 5
I0424 08:10:58.324674 136717846619968 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/5.
I0424 08:10:58.774058 136717846619968 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 08:10:58.774167 136717846619968 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0424 08:10:58.811713 136717846619968 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.038859s
I0424 08:10:58.811887 136717846619968 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 36.470 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.0422971248626709 s) (per-host)
I0424 08:10:58.811939 136717846619968 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.042358s (batch_requests_ready=0.001786s, total_serialization_initiated=0.040503s, others=0.000069s)
I0424 08:10:58.812033 136717846619968 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.047261s (all_items=0.000015s, per_item={'items': '0.00001502'}, temp_paths=0.047246)
I0424 08:10:58.812753 136717846619968 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.49 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/5.
I0424 08:10:58.813084 136588365371136 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 08:30:58.813047
I0424 08:10:58.819482 136717846619968 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 08:10:58.819759 136589489469184 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 08:10:58.819917 136717846619968 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777018242.1879592, 'wait_for_prev_duration_secs': 16.13171648979187, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777018258.321873, 'checkpointer_blocking_duration_secs': 0.49132585525512695, 'get_old_steps_start_time': 1777018258.8132231, 'get_old_steps_duration_secs': 3.0994415283203125e-05, 'checkpoint_manager_blocking_start_time': 1777018242.185777, 'checkpoint_manager_blocking_duration_secs': 16.634105443954468}
I0424 08:10:58.820030 136717846619968 checkpointing.py:408] Started an asynchronous checkpoint save for step 5
I0424 08:10:58.820823 136717846619968 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.902, Tokens/s/device: 13400.335, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355
I0424 08:10:58.826952 136717846619968 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.331, Tokens/s/device: 13917.109, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467
I0424 08:11:03.875831 136588373763840 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/5/items/array_metadatas/process_6
I0424 08:11:09.419562 136588365371136 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 148.320 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.649940252304077 s) (per-host)
I0424 08:11:09.419682 136588365371136 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.605931s.
I0424 08:11:16.570963 136588365371136 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 17.757196s.
I0424 08:11:16.571222 136589489469184 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 08:11:16.571283 136589489469184 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 08:11:16.571331 136589489469184 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts...
I0424 08:11:16.572708 136589489469184 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts.
I0424 08:11:21.763310 136717846619968 metric_logger.py:196] completed step: 6, seconds: 16.930, TFLOP/s/device: 0.803, Tokens/s/device: 120.969, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740
I0424 08:11:21.910925 136717846619968 metric_logger.py:196] completed step: 7, seconds: 22.791, TFLOP/s/device: 0.596, Tokens/s/device: 89.862, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151
I0424 08:11:22.058425 136717846619968 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.282, Tokens/s/device: 13457.571, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098
I0424 08:11:22.205164 136717846619968 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0424 08:11:22.205847 136717846619968 checkpointing.py:776] Waited 0.0007479190826416016 seconds for step 9 to finish before starting checkpointing.
I0424 08:11:22.207845 136717846619968 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 08:11:22.209467 136717846619968 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 9
I0424 08:11:22.210908 136717846619968 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/9.
I0424 08:11:22.490381 136717846619968 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 08:11:22.490477 136717846619968 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0424 08:11:22.525655 136717846619968 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036169s
I0424 08:11:22.525844 136717846619968 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 38.933 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03962135314941406 s) (per-host)
I0424 08:11:22.525910 136717846619968 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039699s (batch_requests_ready=0.001630s, total_serialization_initiated=0.037986s, others=0.000083s)
I0424 08:11:22.526025 136717846619968 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.043703s (all_items=0.000011s, per_item={'items': '0.00001121'}, temp_paths=0.043691)
I0424 08:11:22.526786 136717846619968 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.32 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/9.
I0424 08:11:22.527146 136564449437440 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 08:31:22.527081
I0424 08:11:22.532660 136717846619968 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 08:11:22.532935 136564466222848 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 08:11:22.533102 136717846619968 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777018282.2078125, 'wait_for_prev_duration_secs': 7.43865966796875e-05, 'time_between_consecutive_saves_sec': 5.635066032409668, 'checkpointer_blocking_start_time': 1777018282.2095063, 'checkpointer_blocking_duration_secs': 0.3177816867828369, 'get_old_steps_start_time': 1777018282.5273068, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1777018282.206073, 'checkpoint_manager_blocking_duration_secs': 0.32698631286621094}
I0424 08:11:22.533212 136717846619968 checkpointing.py:408] Started an asynchronous checkpoint save for step 9
I0424 08:11:22.533260 136717846619968 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 08:11:27.989973 136583529346816 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/9/items/array_metadatas/process_6
I0424 08:11:32.996049 136564449437440 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 150.298 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.509793519973755 s) (per-host)
I0424 08:11:32.996188 136564449437440 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.468930s.
I0424 08:11:37.905251 136564449437440 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 15.377978s.
I0424 08:11:37.905581 136564466222848 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 08:11:37.905697 136564466222848 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 08:11:37.905752 136564466222848 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0424 08:11:37.906920 136564466222848 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0424 08:11:37.907118 136717846619968 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0424 08:11:37.907274 136717846619968 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 08:11:37.908231 136717846619968 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.189, Tokens/s/device: 13895.674, total_weights: 65536, loss: 7.181, lm_loss: 7.181, perplexity: 1314.495
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Fri Apr 24 08:11:48 UTC 2026
EXIT_CODE=0
feat/nnx-set-defaults-true  ·  73213e044  ·  feat_nnx_set_defaults_true_20260424_145744  ·  full log 
XPK Start: Fri Apr 24 16:03:27 UTC 2026
2026-04-24 16:03:31.363537: 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 00:00:1777046611.376746      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1777046611.380534      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1777046611.392064      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777046611.392083      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777046611.392085      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777046611.392087      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-24 16:03:50.834326: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303)
I0424 16:03:51.369877 132891047843648 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-24 16:04:00,411:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0424 16:04:00.411249 132891047843648 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-24 16:04:00,413:jax._src.distributed:157: Connecting to JAX distributed service on mt-15-ocdbt-false-sv-3lopc-slice-job-0-0.mt-15-ocdbt-false-sv-3lopc:8482
I0424 16:04:00.413623 132891047843648 distributed.py:157] Connecting to JAX distributed service on mt-15-ocdbt-false-sv-3lopc-slice-job-0-0.mt-15-ocdbt-false-sv-3lopc:8482
I0424 16:04:01.102343 132891047843648 max_utils.py:284] Jax distributed system initialized!
I0424 16:04:08.277094 132891047843648 max_utils.py:800] System Information: Jax Version: 0.8.1
I0424 16:04:08.277198 132891047843648 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0424 16:04:08.277238 132891047843648 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709
I0424 16:04:08.277272 132891047843648 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 16:04:08.903088 132891047843648 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 16:04:08.903673 132891047843648 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 16:04:08.903858 132891047843648 checkpointing.py:688] Setting up checkpoint logger...
I0424 16:04:08.903907 132891047843648 checkpointing.py:234] Creating checkpoint manager with ocdbt=False and zarr3=True
I0424 16:04:08.903963 132891047843648 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0424 16:04:08.904307 132891047843648 base_pytree_checkpoint_handler.py:415] Created BasePyTreeCheckpointHandler: use_ocdbt=False, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x78dc8ad79010>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 16:04:11.905342 132891047843648 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0424 16:04:11.905691 132891047843648 checkpoint_manager.py:709] [process=3][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78c8243f71a0>}, handler_registry=None
I0424 16:04:11.905938 132891047843648 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78c8243f71a0>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`.
I0424 16:04:11.905989 132891047843648 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78c824488e30>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`.
I0424 16:04:11.906025 132891047843648 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 0x78c8243f71a0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78c8243f71a0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78c824488e30>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78c824488e30>}).
I0424 16:04:11.906524 132891047843648 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0424 16:04:11.906600 132891047843648 async_checkpointer.py:177] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x78c80c79c180> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0424 16:04:12.837092 132891047843648 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_15_ocdbt_false/checkpoints
I0424 16:04:13.257691 132891047843648 checkpoint_manager.py:929] [process=3][thread=MainThread] CheckpointManager created,  primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x78c80c533ef0>
I0424 16:04:13.257865 132891047843648 checkpointing.py:302] Checkpoint manager created!
I0424 16:04:14.198916 132891047843648 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 16:04:14.199032 132891047843648 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 16:04:14.568553 132891047843648 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 16:04:14.568648 132891047843648 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 16:04:14.584509 132891047843648 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 16:04:14.584571 132891047843648 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 16:04:14.607457 132891047843648 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 16:04:14.607535 132891047843648 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:04:14.623524 132891047843648 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 16:04:14.623589 132891047843648 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:04:14.639440 132891047843648 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 16:04:14.639505 132891047843648 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:04:14.664357 132891047843648 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0424 16:04:14.664427 132891047843648 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:04:14.684029 132891047843648 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 16:04:14.684094 132891047843648 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 16:04:14.873828 132891047843648 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 16:04:14.873954 132891047843648 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0424 16:04:16.305114 132891047843648 maxtext_utils.py:1740]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0424 16:04:16.305254 132891047843648 maxtext_utils.py:1740]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0424 16:04:16.305311 132891047843648 maxtext_utils.py:1740]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0424 16:04:16.305371 132891047843648 maxtext_utils.py:1740]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   PartitionSpec('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0424 16:04:16.305425 132891047843648 maxtext_utils.py:1740]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  (None, None)
I0424 16:04:16.305472 132891047843648 maxtext_utils.py:1740]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  (None, None)
I0424 16:04:16.305527 132891047843648 maxtext_utils.py:1740]  params/params/decoder/layers/self_attention/key/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0424 16:04:16.305580 132891047843648 maxtext_utils.py:1740]  params/params/decoder/layers/self_attention/out/kernel
    Shape:     float32[16,16,128,2048]
    Logical:   PartitionSpec('heads', 'layers', 'kv', 'embed')
    Physical:  (None, None, None, 'fsdp')
I0424 16:04:16.305621 132891047843648 maxtext_utils.py:1740]  params/params/decoder/layers/self_attention/query/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None, None)
I0424 16:04:16.305660 132891047843648 maxtext_utils.py:1740]  params/params/decoder/layers/self_attention/value/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0424 16:04:16.305709 132891047843648 maxtext_utils.py:1740]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed', 'vocab')
    Physical:  ('fsdp', None)
I0424 16:04:16.305759 132891047843648 maxtext_utils.py:1740]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed')
    Physical:  (None, 'fsdp')

I0424 16:04:16.816078 132891047843648 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 16:04:16.816176 132891047843648 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 16:04:16.831166 132891047843648 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 16:04:16.831228 132891047843648 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 16:04:27.846232 132891047843648 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0424 16:04:27.847050 132891047843648 metric_logger.py:289] number parameters: 1.104 billion
I0424 16:04:39.276501 132891047843648 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0424 16:04:39.429283 132891047843648 checkpointing.py:798] Waited 0.15276408195495605 seconds for step 0 to finish before starting checkpointing.
I0424 16:04:39.431765 132891047843648 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 16:04:39.433381 132891047843648 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 0
I0424 16:04:39.436037 132891047843648 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_15_ocdbt_false/checkpoints/0.
I0424 16:04:40.066544 132891047843648 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 16:04:40.067603 132891047843648 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0424 16:04:40.067661 132891047843648 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0424 16:04:40.318796 132891047843648 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.252332s
I0424 16:04:40.318985 132891047843648 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.726 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5658533573150635 s) (per-host)
I0424 16:04:40.319044 132891047843648 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.565921s (batch_requests_ready=0.309896s, total_serialization_initiated=0.255948s, others=0.000077s)
I0424 16:04:40.319136 132891047843648 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.570237s (all_items=0.000019s, per_item={'items': '0.00001931'}, temp_paths=0.570218)
I0424 16:04:40.320096 132765223745280 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started.
I0424 16:04:40.320269 132891047843648 async_checkpointer.py:561] Finished blocking save. Time taken: 0.886817s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_15_ocdbt_false/checkpoints/0.
I0424 16:04:40.322096 132891047843648 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 16:04:40.322361 132764124825344 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save.
I0424 16:04:40.322488 132891047843648 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777046679.4317474, 'wait_for_prev_duration_secs': 6.198883056640625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777046679.4334188, 'checkpointer_blocking_duration_secs': 0.8869562149047852, 'get_old_steps_start_time': 1777046680.3203933, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1777046679.4298291, 'checkpoint_manager_blocking_duration_secs': 0.8926243782043457}
I0424 16:04:40.322644 132891047843648 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0424 16:04:40.322693 132891047843648 max_utils.py:750] 
Memstats: After params initialized:
I0424 16:04:40.322744 132891047843648 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_10(process=3,(2,2,0,0))
I0424 16:04:40.322777 132891047843648 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_11(process=3,(3,2,0,0))
I0424 16:04:40.322803 132891047843648 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_14(process=3,(2,3,0,0))
I0424 16:04:40.322828 132891047843648 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_15(process=3,(3,3,0,0))
I0424 16:04:40.592156 132891047843648 metric_logger.py:185] completed step: 0, seconds: 11.429, TFLOP/s/device: 1.189, Tokens/s/device: 179.188, total_weights: 65536, loss: 10.874
I0424 16:04:40.763133 132891047843648 metric_logger.py:185] completed step: 1, seconds: 1.314, TFLOP/s/device: 10.338, Tokens/s/device: 1558.183, total_weights: 65536, loss: 10.874
I0424 16:04:41.226662 132891047843648 metric_logger.py:185] completed step: 2, seconds: 0.024, TFLOP/s/device: 572.282, Tokens/s/device: 86260.635, total_weights: 65536, loss: 10.560
I0424 16:04:41.374178 132891047843648 metric_logger.py:185] completed step: 3, seconds: 0.464, TFLOP/s/device: 29.295, Tokens/s/device: 4415.725, total_weights: 65536, loss: 9.980
I0424 16:04:41.378766 132891047843648 checkpointing.py:794] Waiting for step 5 to finish before checkpoint...
I0424 16:04:41.668462 132891047843648 checkpointing.py:798] Waited 0.28966379165649414 seconds for step 5 to finish before starting checkpointing.
I0424 16:04:41.671442 132891047843648 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 16:04:42.945389    2595 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 16:04:44.858799 132765712779008 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_15_ocdbt_false/checkpoints/0/items/array_metadatas/process_3
I0424 16:04:50.499063 132765223745280 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 146.996 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.745883703231812 s) (per-host)
I0424 16:04:50.499172 132765223745280 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.178962s.
I0424 16:04:56.739610 132765223745280 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 16.419385s.
I0424 16:04:56.739982 132764124825344 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 16:04:56.740101 132764124825344 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save.
I0424 16:04:56.740158 132764124825344 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0424 16:04:56.741550 132764124825344 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0424 16:04:56.741741 132891047843648 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0424 16:04:56.741875 132891047843648 checkpoint_manager.py:1458] Waiting for previous save to complete took 15.070438 seconds. If this number is high, consider checkpointing less frequently.
I0424 16:04:56.744070 132891047843648 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 5
I0424 16:04:56.748488 132891047843648 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_15_ocdbt_false/checkpoints/5.
I0424 16:04:57.465686 132891047843648 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0424 16:04:57.465788 132891047843648 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0424 16:04:57.509685 132891047843648 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.045088s
I0424 16:04:57.509851 132891047843648 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.985 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.30944275856018066 s) (per-host)
I0424 16:04:57.509903 132891047843648 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.309505s (batch_requests_ready=0.262511s, total_serialization_initiated=0.046924s, others=0.000071s)
I0424 16:04:57.509999 132891047843648 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.314605s (all_items=0.000017s, per_item={'items': '0.00001669'}, temp_paths=0.314589)
I0424 16:04:57.510833 132759838250752 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started.
I0424 16:04:57.511004 132891047843648 async_checkpointer.py:561] Finished blocking save. Time taken: 0.766866s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_15_ocdbt_false/checkpoints/5.
I0424 16:04:57.619033 132891047843648 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 16:04:57.619397 132764124825344 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save.
I0424 16:04:57.619563 132891047843648 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777046681.6714134, 'wait_for_prev_duration_secs': 15.070437669754028, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777046696.7441084, 'checkpointer_blocking_duration_secs': 0.7670025825500488, 'get_old_steps_start_time': 1777046697.5111296, 'get_old_steps_duration_secs': 2.7894973754882812e-05, 'checkpoint_manager_blocking_start_time': 1777046681.6691365, 'checkpoint_manager_blocking_duration_secs': 15.950388669967651}
I0424 16:04:57.619681 132891047843648 checkpointing.py:409] Started an asynchronous checkpoint save for step 5
I0424 16:04:57.620301 132891047843648 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.898, Tokens/s/device: 13399.634, total_weights: 65536, loss: 9.460
I0424 16:04:57.626333 132891047843648 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.159, Tokens/s/device: 13891.150, total_weights: 65536, loss: 8.959
I0424 16:05:17.149136 132891047843648 metric_logger.py:185] completed step: 6, seconds: 16.246, TFLOP/s/device: 0.836, Tokens/s/device: 126.059, total_weights: 65536, loss: 8.469
I0424 16:05:17.296548 132891047843648 metric_logger.py:185] completed step: 7, seconds: 19.377, TFLOP/s/device: 0.701, Tokens/s/device: 105.693, total_weights: 65536, loss: 8.003
I0424 16:05:17.443898 132891047843648 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.337, Tokens/s/device: 13465.888, total_weights: 65536, loss: 7.572
I0424 16:05:17.591284 132891047843648 checkpointing.py:794] Waiting for step 9 to finish before checkpoint...
I0424 16:05:17.591967 132891047843648 checkpointing.py:798] Waited 0.0007004737854003906 seconds for step 9 to finish before starting checkpointing.
I0424 16:05:17.593983 132891047843648 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 16:05:21.712404 132765712779008 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_15_ocdbt_false/checkpoints/5/items/array_metadatas/process_3
I0424 16:05:27.928653 132759838250752 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 51.405 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 30.7281813621521 s) (per-host)
I0424 16:05:27.928780 132759838250752 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 30.417826s.
I0424 16:05:33.071619 132759838250752 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 35.560643s.
I0424 16:05:33.071907 132764124825344 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 16:05:33.072063 132764124825344 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save.
I0424 16:05:33.072118 132764124825344 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts...
I0424 16:05:33.074470 132764124825344 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts.
I0424 16:05:33.074682 132891047843648 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5.
W0424 16:05:33.074814 132891047843648 checkpoint_manager.py:1458] Waiting for previous save to complete took 15.480841 seconds. If this number is high, consider checkpointing less frequently.
I0424 16:05:33.076429 132891047843648 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 9
I0424 16:05:33.079843 132891047843648 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_15_ocdbt_false/checkpoints/9.
I0424 16:05:33.614356 132891047843648 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0424 16:05:33.614453 132891047843648 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0424 16:05:33.657802 132891047843648 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044367s
I0424 16:05:33.657979 132891047843648 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.062 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3047525882720947 s) (per-host)
I0424 16:05:33.658029 132891047843648 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.304813s (batch_requests_ready=0.258361s, total_serialization_initiated=0.046387s, others=0.000066s)
I0424 16:05:33.658117 132891047843648 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.309496s (all_items=0.000010s, per_item={'items': '0.00001049'}, temp_paths=0.309485)
I0424 16:05:33.659069 132760904333056 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started.
I0424 16:05:33.659231 132891047843648 async_checkpointer.py:561] Finished blocking save. Time taken: 0.582728s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_15_ocdbt_false/checkpoints/9.
I0424 16:05:33.661156 132891047843648 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 16:05:33.661409 132764124825344 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save.
I0424 16:05:33.661563 132891047843648 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777046717.5939567, 'wait_for_prev_duration_secs': 15.480840682983398, 'time_between_consecutive_saves_sec': 20.85236668586731, 'checkpointer_blocking_start_time': 1777046733.0764709, 'checkpointer_blocking_duration_secs': 0.5828666687011719, 'get_old_steps_start_time': 1777046733.6593506, 'get_old_steps_duration_secs': 1.9550323486328125e-05, 'checkpoint_manager_blocking_start_time': 1777046717.5921936, 'checkpoint_manager_blocking_duration_secs': 16.069338083267212}
I0424 16:05:33.661663 132891047843648 checkpointing.py:409] Started an asynchronous checkpoint save for step 9
I0424 16:05:33.661705 132891047843648 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 16:05:38.419622 132765712779008 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_15_ocdbt_false/checkpoints/9/items/array_metadatas/process_3
I0424 16:05:44.153481 132760904333056 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 146.256 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.80021357536316 s) (per-host)
I0424 16:05:44.153610 132760904333056 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.494429s.
I0424 16:05:49.602363 132760904333056 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 15.943166s.
I0424 16:05:49.602694 132764124825344 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 16:05:49.602827 132764124825344 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save.
I0424 16:05:49.602870 132764124825344 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0424 16:05:49.604302 132764124825344 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0424 16:05:49.604470 132891047843648 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0424 16:05:49.604623 132891047843648 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 16:05:49.605243 132891047843648 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.187, Tokens/s/device: 13895.391, total_weights: 65536, loss: 7.181
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Fri Apr 24 16:06:02 UTC 2026
EXIT_CODE=0