MaxView

Timing Test Reference Updates
‹ 14_async_ckpt_false_saveCase: 15_ocdbt_false_resume15_ocdbt_false_save ›

Metrics: main (b117f50cf) vs test/pipeline-scan-nnx (5b21b66fc)

Metricmain  b117f50cftest/pipeline-scan-nnx  5b21b66fcDiff (test/pipeline-scan-nnx − main)
Parameters1.104 billion1.104 billion—
Final loss5.39105.39100
TFLOP/s92.29892.165-0.133
Tok/s13912.213892.2-20.007
Avg s/step4.9735.097+0.124
Memory %1.381.380
JAX0.9.20.9.2—

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

main  ·  b117f50cf  ·  main_20260424_070227  ·  full log 
XPK Start: Fri Apr 24 08:12:50 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 08:13:15.223302: 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 08:13:15.427043 132554245596992 max_utils.py:273] Attempting to initialize the jax distributed system...
I0424 08:13:24.468811 132554245596992 distributed.py:149] Starting JAX distributed service on [::]:8482
I0424 08:13:24.471210 132554245596992 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-false-rs-kl43v-slice-job-0-0.mt-15-ocdbt-false-rs-kl43v:8482
I0424 08:13:25.748331 132554245596992 max_utils.py:284] Jax distributed system initialized!
I0424 08:13:31.849847 132554245596992 max_utils.py:800] System Information: Jax Version: 0.9.2
I0424 08:13:31.849952 132554245596992 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0424 08:13:31.849992 132554245596992 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 08:13:31.850027 132554245596992 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 08:13:32.551555 132554245596992 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 08:13:32.551838 132554245596992 checkpointing.py:677] Setting up checkpoint logger...
I0424 08:13:32.551895 132554245596992 checkpointing.py:233] Creating checkpoint manager with ocdbt=False and zarr3=True
I0424 08:13:32.551940 132554245596992 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0424 08:13:32.552306 132554245596992 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 0x788e145143b0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 08:13:33.800186 132554245596992 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0424 08:13:33.800425 132554245596992 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 0x7879dc408470>}, handler_registry=None
I0424 08:13:33.800671 132554245596992 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7879dc408470>` 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 08:13:33.800721 132554245596992 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7879dc40c170>` 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 08:13:33.800757 132554245596992 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 0x7879dc408470>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7879dc408470>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7879dc40c170>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7879dc40c170>}).
I0424 08:13:33.801086 132554245596992 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0424 08:13:33.801169 132554245596992 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7878d0771800> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0424 08:13:35.704386 132433143908096 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1777018260516454388, 'commit_timestamp_nsecs': 1777018275133681194, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/5/_CHECKPOINT_METADATA
I0424 08:13:35.951899 132433169086208 checkpoint_manager.py:1871] Missing metrics for step 0
I0424 08:13:36.377387 132554245596992 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints
I0424 08:13:36.379996 132554245596992 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 0x7879dc409ca0>
I0424 08:13:36.380131 132554245596992 checkpointing.py:301] Checkpoint manager created!
I0424 08:13:37.311951 132554245596992 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 08:13:37.312073 132554245596992 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 08:13:37.695410 132554245596992 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn').
I0424 08:13:37.695502 132554245596992 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 08:13:37.711998 132554245596992 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn').
I0424 08:13:37.712057 132554245596992 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 08:13:37.735916 132554245596992 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 08:13:37.735987 132554245596992 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 08:13:37.752635 132554245596992 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 08:13:37.752696 132554245596992 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 08:13:37.769123 132554245596992 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 08:13:37.769181 132554245596992 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 08:13:37.793993 132554245596992 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch_attn', 'activation_length_attn', 'activation_heads', 'activation_kv').
I0424 08:13:37.794061 132554245596992 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 08:13:37.815841 132554245596992 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 08:13:37.815909 132554245596992 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 08:13:38.029474 132554245596992 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 08:13:38.029585 132554245596992 checkpointing.py:581] restoring from this run's directory step 9
I0424 08:13:38.274970 132554245596992 event_tracking.py:70] [process=6] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/9.
I0424 08:13:38.528152 132554245596992 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/9.
I0424 08:13:38.528595 132554245596992 event_tracking.py:125] [process=6] [sync] Finished blocking load in 0.25 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/9.
I0424 08:13:41.305197    1480 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 08:13:50.347204 132554245596992 jax_array_handlers.py:843] [process=6] /jax/orbax/read/worker/io/requested throughput: 153.686 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.28249979019165 s) (per-host)
I0424 08:13:50.355884 132554245596992 base_pytree_checkpoint_handler.py:130] [process=6] /jax/checkpoint/read/gbytes_per_sec: 148.152 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.682231664657593 s) (per-host)
I0424 08:13:50.358079 132554245596992 event_tracking.py:138] [process=6] [sync] Finished load in 12.08 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/9
I0424 08:13:50.358190 132554245596992 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints', 'checkpointer_start_time': 1777018418.274869, 'checkpointer_duration_secs': 12.08329176902771, 'checkpoint_manager_start_time': 1777018418.0298252, 'checkpoint_manager_duration_secs': 12.328337907791138}
fsdp: 32
I0424 08:13:50.777605 132554245596992 maxtext_utils.py:1707]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0424 08:13:50.777733 132554245596992 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 08:13:50.777787 132554245596992 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 08:13:50.777843 132554245596992 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 08:13:50.777895 132554245596992 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 08:13:50.777934 132554245596992 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 08:13:50.777986 132554245596992 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 08:13:50.778037 132554245596992 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 08:13:50.778078 132554245596992 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 08:13:50.778128 132554245596992 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 08:13:50.778178 132554245596992 maxtext_utils.py:1707]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)

I0424 08:13:50.778225 132554245596992 maxtext_utils.py:1707]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')
I0424 08:13:51.279529 132554245596992 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 08:13:51.279634 132554245596992 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 08:13:51.295418 132554245596992 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 08:13:51.295479 132554245596992 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 08:14:02.249771 132554245596992 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 08:14:02.250674 132554245596992 metric_logger.py:301] number parameters: 1.104 billion
I0424 08:14:13.660865 132554245596992 checkpointing.py:772] Waiting for step 10 to finish before checkpoint...
I0424 08:14:14.110754 132554245596992 checkpointing.py:776] Waited 0.44986748695373535 seconds for step 10 to finish before starting checkpointing.
I0424 08:14:14.113121 132554245596992 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 08:14:14.115116 132554245596992 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 10
I0424 08:14:14.117917 132554245596992 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/10.
I0424 08:14:14.717549 132554245596992 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 08:14:14.718564 132554245596992 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 08:14:14.718633 132554245596992 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 08:14:15.003401 132554245596992 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.285928s
I0424 08:14:15.003573 132554245596992 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.321 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2898893356323242 s) (per-host)
I0424 08:14:15.003644 132554245596992 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.289972s (batch_requests_ready=0.002281s, total_serialization_initiated=0.287601s, others=0.000090s)
I0424 08:14:15.003752 132554245596992 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.294075s (all_items=0.000024s, per_item={'items': '0.00002384'}, temp_paths=0.294051)
I0424 08:14:15.004535 132554245596992 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.89 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/10.
I0424 08:14:15.004831 132430434375424 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 08:34:15.004797
I0424 08:14:15.007806 132554245596992 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 08:14:15.008033 132429378926336 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 08:14:15.008163 132554245596992 standard_logger.py:34] {'step': 10, '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': 1777018454.113085, 'wait_for_prev_duration_secs': 7.605552673339844e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777018454.115154, 'checkpointer_blocking_duration_secs': 0.8898262977600098, 'get_old_steps_start_time': 1777018455.005003, 'get_old_steps_duration_secs': 3.1948089599609375e-05, 'checkpoint_manager_blocking_start_time': 1777018454.1112185, 'checkpoint_manager_blocking_duration_secs': 0.8969078063964844}
I0424 08:14:15.008270 132554245596992 checkpointing.py:408] Started an asynchronous checkpoint save for step 10
I0424 08:14:15.008321 132554245596992 max_utils.py:750] 
Memstats: After params initialized:
I0424 08:14:15.008368 132554245596992 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0424 08:14:15.008400 132554245596992 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0424 08:14:15.008427 132554245596992 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0424 08:14:15.008450 132554245596992 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0424 08:14:15.321252 132554245596992 metric_logger.py:196] completed step: 10, seconds: 11.410, TFLOP/s/device: 1.191, Tokens/s/device: 179.490, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869
I0424 08:14:15.500161 132554245596992 metric_logger.py:196] completed step: 11, seconds: 1.659, TFLOP/s/device: 8.191, Tokens/s/device: 1234.616, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659
I0424 08:14:15.647858 132554245596992 metric_logger.py:196] completed step: 12, seconds: 0.032, TFLOP/s/device: 427.537, Tokens/s/device: 64443.046, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068
I0424 08:14:15.795200 132554245596992 metric_logger.py:196] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.790, Tokens/s/device: 13383.434, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001
I0424 08:14:15.799686 132554245596992 checkpointing.py:772] Waiting for step 15 to finish before checkpoint...
I0424 08:14:16.089279 132554245596992 checkpointing.py:776] Waited 0.2895548343658447 seconds for step 15 to finish before starting checkpointing.
I0424 08:14:16.092383 132554245596992 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 08:14:20.443988 132429387319040 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/10/items/array_metadatas/process_6
I0424 08:14:24.961184 132430434375424 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 154.145 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.247466325759888 s) (per-host)
I0424 08:14:24.961309 132430434375424 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 9.956369s.
I0424 08:14:30.687379 132430434375424 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 15.682421s.
I0424 08:14:30.687615 132429378926336 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 08:14:30.687672 132429378926336 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 08:14:30.687716 132429378926336 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0424 08:14:30.690411 132429378926336 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0424 08:14:30.690576 132554245596992 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10.
W0424 08:14:30.690707 132554245596992 checkpoint_manager.py:1452] Waiting for previous save to complete took 14.598322 seconds. If this number is high, consider checkpointing less frequently.
I0424 08:14:30.692813 132554245596992 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 15
I0424 08:14:30.695323 132554245596992 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/15.
I0424 08:14:30.982076 132554245596992 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 08:14:30.982183 132554245596992 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 08:14:31.008629 132554245596992 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027738s
I0424 08:14:31.008761 132554245596992 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 49.393 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.031230688095092773 s) (per-host)
I0424 08:14:31.008809 132554245596992 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.031287s (batch_requests_ready=0.001888s, total_serialization_initiated=0.029336s, others=0.000063s)
I0424 08:14:31.008887 132554245596992 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.036214s (all_items=0.000012s, per_item={'items': '0.00001240'}, temp_paths=0.036202)
I0424 08:14:31.009619 132554245596992 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/15.
I0424 08:14:31.009942 132429378926336 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 08:34:31.009905
I0424 08:14:31.022026 132554245596992 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 08:14:31.022341 132428874118912 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 08:14:31.022503 132554245596992 standard_logger.py:34] {'step': 15, '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': 1777018456.0923533, 'wait_for_prev_duration_secs': 14.598321914672852, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777018470.6928525, 'checkpointer_blocking_duration_secs': 0.3172283172607422, 'get_old_steps_start_time': 1777018471.0101168, 'get_old_steps_duration_secs': 3.1948089599609375e-05, 'checkpoint_manager_blocking_start_time': 1777018456.0899284, 'checkpoint_manager_blocking_duration_secs': 14.932541608810425}
I0424 08:14:31.022612 132554245596992 checkpointing.py:408] Started an asynchronous checkpoint save for step 15
I0424 08:14:31.023477 132554245596992 metric_logger.py:196] completed step: 14, seconds: 0.148, TFLOP/s/device: 91.994, Tokens/s/device: 13866.320, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193
I0424 08:14:31.029720 132554245596992 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.172, Tokens/s/device: 13893.223, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661
I0424 08:14:36.632544 132429387319040 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/15/items/array_metadatas/process_6
I0424 08:14:41.285461 132429378926336 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 153.242 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.307890892028809 s) (per-host)
I0424 08:14:41.285588 132429378926336 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.275537s.
I0424 08:14:46.440348 132429378926336 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 15.430282s.
I0424 08:14:46.440638 132428874118912 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 08:14:46.440768 132428874118912 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 08:14:46.440833 132428874118912 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0424 08:14:46.443281 132428874118912 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0424 08:14:51.830430 132554245596992 metric_logger.py:196] completed step: 16, seconds: 15.228, TFLOP/s/device: 0.892, Tokens/s/device: 134.485, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872
I0424 08:14:51.977844 132554245596992 metric_logger.py:196] completed step: 17, seconds: 20.655, TFLOP/s/device: 0.658, Tokens/s/device: 99.153, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975
I0424 08:14:52.125060 132554245596992 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.260, Tokens/s/device: 13454.299, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333
I0424 08:14:52.271696 132554245596992 checkpointing.py:772] Waiting for step 19 to finish before checkpoint...
I0424 08:14:52.272418 132554245596992 checkpointing.py:776] Waited 0.0007429122924804688 seconds for step 19 to finish before starting checkpointing.
I0424 08:14:52.274433 132554245596992 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 08:14:52.276169 132554245596992 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 19
I0424 08:14:52.277598 132554245596992 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/19.
I0424 08:14:52.578297 132554245596992 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 08:14:52.578392 132554245596992 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 08:14:52.604580 132554245596992 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027073s
I0424 08:14:52.604750 132554245596992 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 50.380 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030618667602539062 s) (per-host)
I0424 08:14:52.604801 132554245596992 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030680s (batch_requests_ready=0.001744s, total_serialization_initiated=0.028870s, others=0.000066s)
I0424 08:14:52.604895 132554245596992 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034935s (all_items=0.000010s, per_item={'items': '0.00000978'}, temp_paths=0.034925)
I0424 08:14:52.605585 132554245596992 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_15_ocdbt_false/checkpoints/19.
I0424 08:14:52.605941 132426172978944 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 08:34:52.605896
I0424 08:14:52.608130 132554245596992 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 08:14:52.608347 132428874118912 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 08:14:52.608444 132554245596992 standard_logger.py:34] {'step': 19, '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': 1777018492.2744036, 'wait_for_prev_duration_secs': 6.961822509765625e-05, 'time_between_consecutive_saves_sec': 5.831081867218018, 'checkpointer_blocking_start_time': 1777018492.2762065, 'checkpointer_blocking_duration_secs': 0.3298642635345459, 'get_old_steps_start_time': 1777018492.606089, 'get_old_steps_duration_secs': 4.315376281738281e-05, 'checkpoint_manager_blocking_start_time': 1777018492.2726564, 'checkpoint_manager_blocking_duration_secs': 0.3357553482055664}
I0424 08:14:52.608610 132554245596992 checkpointing.py:408] Started an asynchronous checkpoint save for step 19
I0424 08:14:52.608655 132554245596992 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 08:14:57.749529 132429387319040 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/19/items/array_metadatas/process_6
I0424 08:15:02.951599 132426172978944 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 152.215 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.377416849136353 s) (per-host)
I0424 08:15:02.951723 132426172978944 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.345680s.
I0424 08:15:08.285403 132426172978944 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 15.679343s.
I0424 08:15:08.285710 132428874118912 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 08:15:08.285766 132428874118912 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 08:15:08.285806 132428874118912 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0424 08:15:08.287160 132428874118912 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0424 08:15:08.287367 132554245596992 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19.
I0424 08:15:08.287532 132554245596992 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 08:15:08.288463 132554245596992 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.298, Tokens/s/device: 13912.193, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Fri Apr 24 08:15:20 UTC 2026
EXIT_CODE=0
test/pipeline-scan-nnx  ·  5b21b66fc  ·  test_pipeline_scan_nnx_20260424_200834  ·  full log 
XPK Start: Fri Apr 24 21:16:27 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 21:16:51.888349: 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 21:16:52.098695 133532101318464 max_utils.py:273] Attempting to initialize the jax distributed system...
I0424 21:17:01.139137 133532101318464 distributed.py:149] Starting JAX distributed service on [::]:8482
I0424 21:17:01.141405 133532101318464 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-false-rs-6zwv9-slice-job-0-0.mt-15-ocdbt-false-rs-6zwv9:8482
I0424 21:17:02.146010 133532101318464 max_utils.py:284] Jax distributed system initialized!
I0424 21:17:07.455844 133532101318464 max_utils.py:800] System Information: Jax Version: 0.9.2
I0424 21:17:07.455950 133532101318464 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0424 21:17:07.455990 133532101318464 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 21:17:07.456026 133532101318464 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 21:17:08.155268 133532101318464 maxtext_utils.py:1565] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 21:17:08.155536 133532101318464 checkpointing.py:677] Setting up checkpoint logger...
I0424 21:17:08.155589 133532101318464 checkpointing.py:233] Creating checkpoint manager with ocdbt=False and zarr3=True
I0424 21:17:08.155632 133532101318464 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0424 21:17:08.155970 133532101318464 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 0x7971a5e3cef0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 21:17:10.214585 133532101318464 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0424 21:17:10.214813 133532101318464 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 0x795e0046e7b0>}, handler_registry=None
I0424 21:17:10.215050 133532101318464 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x795e0046e7b0>` 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 21:17:10.215115 133532101318464 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x795e002771d0>` 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 21:17:10.215152 133532101318464 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 0x795e0046e7b0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x795e0046e7b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x795e002771d0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x795e002771d0>}).
I0424 21:17:10.215493 133532101318464 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0424 21:17:10.215562 133532101318464 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x795cfc7e3560> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0424 21:17:11.677171 133413126256384 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1777065258915381712, 'commit_timestamp_nsecs': 1777065271945999806, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints/0/_CHECKPOINT_METADATA
I0424 21:17:11.930181 133413126256384 checkpoint_manager.py:1871] Missing metrics for step 0
I0424 21:17:11.949311 133532101318464 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints
I0424 21:17:12.393962 133532101318464 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_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x795e00275220>
I0424 21:17:12.394183 133532101318464 checkpointing.py:301] Checkpoint manager created!
I0424 21:17:13.327252 133532101318464 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 21:17:13.327370 133532101318464 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 21:17:13.717616 133532101318464 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 21:17:13.717710 133532101318464 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 21:17:13.734135 133532101318464 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 21:17:13.734197 133532101318464 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 21:17:13.758022 133532101318464 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 21:17:13.758104 133532101318464 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 21:17:13.774692 133532101318464 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 21:17:13.774755 133532101318464 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 21:17:13.791231 133532101318464 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 21:17:13.791295 133532101318464 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 21:17:13.816258 133532101318464 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0424 21:17:13.816330 133532101318464 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 21:17:13.838133 133532101318464 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 21:17:13.838206 133532101318464 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 21:17:14.051167 133532101318464 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 21:17:14.051288 133532101318464 checkpointing.py:581] restoring from this run's directory step 9
I0424 21:17:14.287178 133532101318464 event_tracking.py:70] [process=6] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints/9.
I0424 21:17:14.555875 133532101318464 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints/9.
I0424 21:17:14.556342 133532101318464 event_tracking.py:125] [process=6] [sync] Finished blocking load in 0.27 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints/9.
I0424 21:17:17.284495    1446 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 21:17:27.363858 133532101318464 jax_array_handlers.py:843] [process=6] /jax/orbax/read/worker/io/requested throughput: 142.118 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 11.119435787200928 s) (per-host)
I0424 21:17:27.371854 133532101318464 base_pytree_checkpoint_handler.py:130] [process=6] /jax/checkpoint/read/gbytes_per_sec: 135.767 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 11.656717777252197 s) (per-host)
I0424 21:17:27.374044 133532101318464 event_tracking.py:138] [process=6] [sync] Finished load in 13.09 seconds @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints/9
I0424 21:17:27.374157 133532101318464 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints', 'checkpointer_start_time': 1777065434.287056, 'checkpointer_duration_secs': 13.087074041366577, 'checkpoint_manager_start_time': 1777065434.0515409, 'checkpoint_manager_duration_secs': 13.322590827941895}
fsdp: 32
I0424 21:17:27.795141 133532101318464 maxtext_utils.py:1668]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0424 21:17:27.795270 133532101318464 maxtext_utils.py:1668]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0424 21:17:27.795325 133532101318464 maxtext_utils.py:1668]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0424 21:17:27.795394 133532101318464 maxtext_utils.py:1668]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   P('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0424 21:17:27.795446 133532101318464 maxtext_utils.py:1668]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0424 21:17:27.795483 133532101318464 maxtext_utils.py:1668]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0424 21:17:27.795533 133532101318464 maxtext_utils.py:1668]  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 21:17:27.795585 133532101318464 maxtext_utils.py:1668]  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 21:17:27.795625 133532101318464 maxtext_utils.py:1668]  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 21:17:27.795659 133532101318464 maxtext_utils.py:1668]  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 21:17:27.795703 133532101318464 maxtext_utils.py:1668]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)
I0424 21:17:27.795748 133532101318464 maxtext_utils.py:1668]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')

I0424 21:17:28.285731 133532101318464 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 21:17:28.285825 133532101318464 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 21:17:28.301483 133532101318464 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 21:17:28.301547 133532101318464 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 21:17:39.338298 133532101318464 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 21:17:39.339109 133532101318464 metric_logger.py:301] number parameters: 1.104 billion
I0424 21:17:51.123813 133532101318464 checkpointing.py:772] Waiting for step 10 to finish before checkpoint...
I0424 21:17:51.278259 133532101318464 checkpointing.py:776] Waited 0.15442442893981934 seconds for step 10 to finish before starting checkpointing.
I0424 21:17:51.280521 133532101318464 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 21:17:51.282253 133532101318464 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 10
I0424 21:17:51.285170 133532101318464 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints/10.
I0424 21:17:51.601727 133532101318464 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 21:17:51.602580 133532101318464 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 21:17:51.602637 133532101318464 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 21:17:51.878826 133532101318464 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.277175s
I0424 21:17:51.879001 133532101318464 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.483 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2813301086425781 s) (per-host)
I0424 21:17:51.879055 133532101318464 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.281394s (batch_requests_ready=0.002410s, total_serialization_initiated=0.278912s, others=0.000072s)
I0424 21:17:51.879167 133532101318464 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.285389s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.285372)
I0424 21:17:51.879913 133532101318464 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.60 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints/10.
I0424 21:17:51.880267 133409830225664 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 21:37:51.880228
I0424 21:17:51.896799 133532101318464 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 21:17:51.897074 133409326229248 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 21:17:51.897223 133532101318464 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777065471.2805028, 'wait_for_prev_duration_secs': 6.628036499023438e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777065471.2822917, 'checkpointer_blocking_duration_secs': 0.5980827808380127, 'get_old_steps_start_time': 1777065471.8804002, 'get_old_steps_duration_secs': 3.1948089599609375e-05, 'checkpoint_manager_blocking_start_time': 1777065471.2786453, 'checkpoint_manager_blocking_duration_secs': 0.6185414791107178}
I0424 21:17:51.897330 133532101318464 checkpointing.py:408] Started an asynchronous checkpoint save for step 10
I0424 21:17:51.897381 133532101318464 max_utils.py:750] 
Memstats: After params initialized:
I0424 21:17:51.897433 133532101318464 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0424 21:17:51.897465 133532101318464 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0424 21:17:51.897492 133532101318464 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0424 21:17:51.897517 133532101318464 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0424 21:17:52.208039 133532101318464 metric_logger.py:196] completed step: 10, seconds: 11.785, TFLOP/s/device: 1.153, Tokens/s/device: 173.786, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869
I0424 21:17:52.376012 133532101318464 metric_logger.py:196] completed step: 11, seconds: 1.083, TFLOP/s/device: 12.550, Tokens/s/device: 1891.686, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659
I0424 21:17:52.523433 133532101318464 metric_logger.py:196] completed step: 12, seconds: 0.021, TFLOP/s/device: 660.885, Tokens/s/device: 99615.740, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068
I0424 21:17:52.670806 133532101318464 metric_logger.py:196] completed step: 13, seconds: 0.154, TFLOP/s/device: 88.405, Tokens/s/device: 13325.439, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001
I0424 21:17:52.675498 133532101318464 checkpointing.py:772] Waiting for step 15 to finish before checkpoint...
I0424 21:17:52.965133 133532101318464 checkpointing.py:776] Waited 0.28957104682922363 seconds for step 15 to finish before starting checkpointing.
I0424 21:17:52.968264 133532101318464 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 21:17:56.957486 133410358007552 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints/10/items/array_metadatas/process_6
I0424 21:18:02.170466 133409830225664 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 149.403 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.572755813598633 s) (per-host)
I0424 21:18:02.170594 133409830225664 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.289663s.
I0424 21:18:08.429599 133409830225664 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 16.548639s.
I0424 21:18:08.429876 133409326229248 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 21:18:08.429995 133409326229248 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 21:18:08.430052 133409326229248 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0424 21:18:08.431653 133409326229248 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0424 21:18:08.431832 133532101318464 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10.
W0424 21:18:08.431965 133532101318464 checkpoint_manager.py:1452] Waiting for previous save to complete took 15.463701 seconds. If this number is high, consider checkpointing less frequently.
I0424 21:18:08.433918 133532101318464 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 15
I0424 21:18:08.436348 133532101318464 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints/15.
I0424 21:18:08.974876 133532101318464 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 21:18:08.974970 133532101318464 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 21:18:09.001787 133532101318464 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.028043s
I0424 21:18:09.001916 133532101318464 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 47.835 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03224802017211914 s) (per-host)
I0424 21:18:09.001965 133532101318464 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.032306s (batch_requests_ready=0.002573s, total_serialization_initiated=0.029670s, others=0.000062s)
I0424 21:18:09.002040 133532101318464 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.037359s (all_items=0.000013s, per_item={'items': '0.00001311'}, temp_paths=0.037346)
I0424 21:18:09.002762 133532101318464 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.57 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints/15.
I0424 21:18:09.003068 133409326229248 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 21:38:09.003033
I0424 21:18:09.012578 133532101318464 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 21:18:09.012834 133406079825664 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 21:18:09.012965 133532101318464 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777065472.968233, 'wait_for_prev_duration_secs': 15.463700532913208, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777065488.4339592, 'checkpointer_blocking_duration_secs': 0.569232702255249, 'get_old_steps_start_time': 1777065489.003216, 'get_old_steps_duration_secs': 3.2901763916015625e-05, 'checkpoint_manager_blocking_start_time': 1777065472.9657738, 'checkpoint_manager_blocking_duration_secs': 16.04715657234192}
I0424 21:18:09.013074 133532101318464 checkpointing.py:408] Started an asynchronous checkpoint save for step 15
I0424 21:18:09.013937 133532101318464 metric_logger.py:196] completed step: 14, seconds: 0.147, TFLOP/s/device: 92.169, Tokens/s/device: 13892.658, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193
I0424 21:18:09.019991 133532101318464 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.254, Tokens/s/device: 13905.486, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661
I0424 21:18:14.198715 133410358007552 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints/15/items/array_metadatas/process_6
I0424 21:18:19.469631 133409326229248 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 150.439 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.499922752380371 s) (per-host)
I0424 21:18:19.469754 133409326229248 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.466018s.
I0424 21:18:25.113648 133409326229248 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 16.109895s.
I0424 21:18:25.113878 133406079825664 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 21:18:25.113940 133406079825664 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 21:18:25.113986 133406079825664 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0424 21:18:25.115482 133406079825664 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0424 21:18:30.160313 133532101318464 metric_logger.py:196] completed step: 16, seconds: 16.343, TFLOP/s/device: 0.831, Tokens/s/device: 125.314, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872
I0424 21:18:30.307714 133532101318464 metric_logger.py:196] completed step: 17, seconds: 20.994, TFLOP/s/device: 0.647, Tokens/s/device: 97.550, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975
I0424 21:18:30.455308 133532101318464 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.181, Tokens/s/device: 13442.378, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333
I0424 21:18:30.602140 133532101318464 checkpointing.py:772] Waiting for step 19 to finish before checkpoint...
I0424 21:18:30.602845 133532101318464 checkpointing.py:776] Waited 0.0007646083831787109 seconds for step 19 to finish before starting checkpointing.
I0424 21:18:30.604899 133532101318464 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 21:18:30.606507 133532101318464 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 19
I0424 21:18:30.607901 133532101318464 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints/19.
I0424 21:18:30.907026 133532101318464 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0424 21:18:30.907133 133532101318464 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 21:18:30.931957 133532101318464 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.025748s
I0424 21:18:30.932138 133532101318464 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 52.475 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.02939629554748535 s) (per-host)
I0424 21:18:30.932191 133532101318464 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.029460s (batch_requests_ready=0.001770s, total_serialization_initiated=0.027622s, others=0.000069s)
I0424 21:18:30.932285 133532101318464 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.033345s (all_items=0.000011s, per_item={'items': '0.00001121'}, temp_paths=0.033334)
I0424 21:18:30.932892 133532101318464 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints/19.
I0424 21:18:30.933223 133405039654656 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 21:38:30.933182
I0424 21:18:30.936123 133532101318464 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 21:18:30.936362 133406079825664 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 21:18:30.936489 133532101318464 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777065510.604868, 'wait_for_prev_duration_secs': 7.081031799316406e-05, 'time_between_consecutive_saves_sec': 5.489342927932739, 'checkpointer_blocking_start_time': 1777065510.606544, 'checkpointer_blocking_duration_secs': 0.32683563232421875, 'get_old_steps_start_time': 1777065510.933395, 'get_old_steps_duration_secs': 2.2649765014648438e-05, 'checkpoint_manager_blocking_start_time': 1777065510.6030786, 'checkpoint_manager_blocking_duration_secs': 0.33337855339050293}
I0424 21:18:30.936593 133532101318464 checkpointing.py:408] Started an asynchronous checkpoint save for step 19
I0424 21:18:30.936637 133532101318464 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 21:18:36.024659 133410358007552 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260424_200834/linen_xpk_test_pipeline_scan_nnx_20260424_200834_15_ocdbt_false/checkpoints/19/items/array_metadatas/process_6
I0424 21:18:41.393769 133405039654656 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 150.567 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.490998029708862 s) (per-host)
I0424 21:18:41.393881 133405039654656 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.460532s.
I0424 21:18:46.882679 133405039654656 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 15.949311s.
I0424 21:18:46.882892 133406079825664 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 21:18:46.882949 133406079825664 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 21:18:46.882990 133406079825664 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0424 21:18:46.884666 133406079825664 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0424 21:18:46.884848 133532101318464 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19.
I0424 21:18:46.885007 133532101318464 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 21:18:46.886118 133532101318464 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.165, Tokens/s/device: 13892.186, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Fri Apr 24 21:18:57 UTC 2026
EXIT_CODE=0