MaxView

Timing Test Reference Updates
‹ 07_evalCase: 08_checkpoint_async_true_resume08_checkpoint_async_true_save ›

Metrics: main (586e69205) vs test/pipeline-scan-nnx (5b21b66fc)

Metricmain  586e69205test/pipeline-scan-nnx  5b21b66fcDiff (test/pipeline-scan-nnx − main)
Parameters1.104 billion1.104 billion—
Final loss5.39105.39100
TFLOP/s92.25092.224-0.026
Tok/s13904.913901.0-3.964
Avg s/step8.4658.450-0.015
Memory %1.381.380
JAX0.9.20.9.2—

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

main  ·  586e69205  ·  main_20260423_071538  ·  full log 
XPK Start: Thu Apr 23 07:46:13 UTC 2026
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
`rope_parameters`'s factor field must be a float >= 1, got 40
`rope_parameters`'s beta_fast field must be a float, got 32
`rope_parameters`'s beta_slow field must be a float, got 1
DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 
2026-04-23 07:46:38.179632: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303)
I0423 07:46:38.392082 134158149904192 max_utils.py:273] Attempting to initialize the jax distributed system...
I0423 07:46:47.433145 134158149904192 distributed.py:149] Starting JAX distributed service on [::]:8482
I0423 07:46:47.435507 134158149904192 distributed.py:172] Connecting to JAX distributed service on mt-08-checkpoint-async--snehc-slice-job-0-0.mt-08-checkpoint-async--snehc:8482
I0423 07:46:48.652136 134158149904192 max_utils.py:284] Jax distributed system initialized!
I0423 07:46:54.731050 134158149904192 max_utils.py:800] System Information: Jax Version: 0.9.2
I0423 07:46:54.731156 134158149904192 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0423 07:46:54.731196 134158149904192 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365
I0423 07:46:54.731232 134158149904192 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0423 07:46:55.427778 134158149904192 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0423 07:46:55.428062 134158149904192 checkpointing.py:677] Setting up checkpoint logger...
I0423 07:46:55.428118 134158149904192 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0423 07:46:55.428163 134158149904192 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0423 07:46:55.428502 134158149904192 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7a039cbe66c0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0423 07:46:57.124609 134158149904192 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0423 07:46:57.124857 134158149904192 checkpoint_manager.py:708] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79ee7c4a37d0>}, handler_registry=None
I0423 07:46:57.125095 134158149904192 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79ee7c4a37d0>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`.
I0423 07:46:57.125144 134158149904192 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79ee7c6ca600>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`.
I0423 07:46:57.125178 134158149904192 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 0x79ee7c4a37d0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79ee7c4a37d0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79ee7c6ca600>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79ee7c6ca600>}).
I0423 07:46:57.125511 134158149904192 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0423 07:46:57.125579 134158149904192 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x79ee587d9580> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0423 07:46:59.093984 134036861216512 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776930198718786951, 'commit_timestamp_nsecs': 1776930246313887161, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/5/_CHECKPOINT_METADATA
I0423 07:46:59.368760 134036869609216 checkpoint_manager.py:1871] Missing metrics for step 0
I0423 07:46:59.770248 134158149904192 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints
I0423 07:47:00.645240 134158149904192 checkpoint_manager.py:929] [process=5][thread=MainThread] CheckpointManager created,  primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x79ee7c4aaf90>
I0423 07:47:00.645412 134158149904192 checkpointing.py:301] Checkpoint manager created!
I0423 07:47:01.571860 134158149904192 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0423 07:47:01.571978 134158149904192 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 07:47:01.955543 134158149904192 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 07:47:01.955631 134158149904192 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 07:47:01.972343 134158149904192 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 07:47:01.972406 134158149904192 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 07:47:01.996778 134158149904192 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 07:47:01.996853 134158149904192 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 07:47:02.013299 134158149904192 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 07:47:02.013362 134158149904192 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 07:47:02.029924 134158149904192 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 07:47:02.029988 134158149904192 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 07:47:02.054886 134158149904192 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0423 07:47:02.054965 134158149904192 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 07:47:02.076147 134158149904192 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0423 07:47:02.076219 134158149904192 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0423 07:47:02.285558 134158149904192 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0423 07:47:02.285681 134158149904192 checkpointing.py:581] restoring from this run's directory step 9
I0423 07:47:02.531019 134158149904192 event_tracking.py:70] [process=5] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/9.
I0423 07:47:02.782293 134158149904192 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/9.
I0423 07:47:02.782727 134158149904192 event_tracking.py:125] [process=5] [sync] Finished blocking load in 0.25 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/9.
I0423 07:47:04.199686    1451 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0423 07:47:11.171474 134158149904192 jax_array_handlers.py:843] [process=5] /jax/orbax/read/worker/io/requested throughput: 226.476 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.977666139602661 s) (per-host)
I0423 07:47:11.176182 134158149904192 base_pytree_checkpoint_handler.py:130] [process=5] /jax/checkpoint/read/gbytes_per_sec: 213.846 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 7.400630950927734 s) (per-host)
I0423 07:47:12.757000 134158149904192 event_tracking.py:138] [process=5] [sync] Finished load in 10.23 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/9
I0423 07:47:12.757167 134158149904192 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints', 'checkpointer_start_time': 1776930422.5309172, 'checkpointer_duration_secs': 10.226218223571777, 'checkpoint_manager_start_time': 1776930422.2859213, 'checkpoint_manager_duration_secs': 10.471216678619385}
fsdp: 32
I0423 07:47:13.173856 134158149904192 maxtext_utils.py:1707]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0423 07:47:13.173983 134158149904192 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0423 07:47:13.174042 134158149904192 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0423 07:47:13.174101 134158149904192 maxtext_utils.py:1707]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   P('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0423 07:47:13.174150 134158149904192 maxtext_utils.py:1707]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0423 07:47:13.174189 134158149904192 maxtext_utils.py:1707]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0423 07:47:13.174242 134158149904192 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/key/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0423 07:47:13.174293 134158149904192 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/out/kernel
    Shape:     float32[16,16,128,2048]
    Logical:   P('heads', 'layers', 'kv', 'embed')
    Physical:  (None, None, None, 'fsdp')
I0423 07:47:13.174334 134158149904192 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/query/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None, None)
I0423 07:47:13.174367 134158149904192 maxtext_utils.py:1707]  params/params/decoder/layers/self_attention/value/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   P('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0423 07:47:13.174411 134158149904192 maxtext_utils.py:1707]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)

I0423 07:47:13.174456 134158149904192 maxtext_utils.py:1707]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')
I0423 07:47:13.656805 134158149904192 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 07:47:13.656898 134158149904192 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 07:47:13.672348 134158149904192 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 07:47:13.672409 134158149904192 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 07:47:24.555872 134158149904192 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0423 07:47:24.556757 134158149904192 metric_logger.py:301] number parameters: 1.104 billion
I0423 07:47:35.689789 134158149904192 checkpointing.py:772] Waiting for step 10 to finish before checkpoint...
I0423 07:47:36.388756 134158149904192 checkpointing.py:776] Waited 0.6989443302154541 seconds for step 10 to finish before starting checkpointing.
I0423 07:47:36.391450 134158149904192 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 07:47:36.393545 134158149904192 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 10
I0423 07:47:36.395130 134158149904192 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/10.
I0423 07:47:37.156774 134158149904192 signaling_client.py:364] Using JaxDistributedSignalingClient
I0423 07:47:37.157816 134158149904192 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 07:47:37.157870 134158149904192 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0423 07:47:37.440810 134158149904192 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.284111s
I0423 07:47:37.440985 134158149904192 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.355 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.288083553314209 s) (per-host)
I0423 07:47:37.441039 134158149904192 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.288149s (batch_requests_ready=0.002271s, total_serialization_initiated=0.285804s, others=0.000074s)
I0423 07:47:37.441133 134158149904192 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.292330s (all_items=0.000016s, per_item={'items': '0.00001574'}, temp_paths=0.292314)
I0423 07:47:37.441878 134158149904192 event_tracking.py:125] [process=5] [async] Finished blocking save in 1.05 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/10.
I0423 07:47:37.442164 134036332738304 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 08:07:37.442133
I0423 07:47:37.444200 134158149904192 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 07:47:37.444501 134034713732864 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0423 07:47:37.444666 134158149904192 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776930456.391431, 'wait_for_prev_duration_secs': 6.127357482910156e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776930456.3935847, 'checkpointer_blocking_duration_secs': 1.0487265586853027, 'get_old_steps_start_time': 1776930457.4423356, 'get_old_steps_duration_secs': 3.147125244140625e-05, 'checkpoint_manager_blocking_start_time': 1776930456.3892853, 'checkpoint_manager_blocking_duration_secs': 1.055332899093628}
I0423 07:47:37.444769 134158149904192 checkpointing.py:408] Started an asynchronous checkpoint save for step 10
I0423 07:47:37.444819 134158149904192 max_utils.py:750] 
Memstats: After params initialized:
I0423 07:47:37.444870 134158149904192 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0))
I0423 07:47:37.444903 134158149904192 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0))
I0423 07:47:37.444930 134158149904192 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0))
I0423 07:47:37.444953 134158149904192 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0))
I0423 07:47:37.764380 134158149904192 metric_logger.py:196] completed step: 10, seconds: 11.133, TFLOP/s/device: 1.220, Tokens/s/device: 183.959, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869
I0423 07:47:38.350332 134158149904192 metric_logger.py:196] completed step: 11, seconds: 2.073, TFLOP/s/device: 6.555, Tokens/s/device: 987.996, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659
I0423 07:47:38.498010 134158149904192 metric_logger.py:196] completed step: 12, seconds: 0.440, TFLOP/s/device: 30.897, Tokens/s/device: 4657.202, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068
I0423 07:47:38.645189 134158149904192 metric_logger.py:196] completed step: 13, seconds: 0.152, TFLOP/s/device: 89.107, Tokens/s/device: 13431.182, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001
I0423 07:47:38.649782 134158149904192 checkpointing.py:772] Waiting for step 15 to finish before checkpoint...
I0423 07:47:38.939350 134158149904192 checkpointing.py:776] Waited 0.28952932357788086 seconds for step 15 to finish before starting checkpointing.
I0423 07:47:38.942499 134158149904192 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 07:47:42.453522 134035636471552 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/10/items/array_metadatas/process_5
I0423 07:48:19.124465 134036332738304 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.635 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.97153091430664 s) (per-host)
I0423 07:48:19.124573 134036332738304 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.682300s.
I0423 07:48:26.874339 134036332738304 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 49.432049s.
I0423 07:48:26.874627 134034713732864 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 07:48:26.874753 134034713732864 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0423 07:48:26.874802 134034713732864 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0423 07:48:26.876379 134034713732864 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0423 07:48:26.876560 134158149904192 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10.
W0423 07:48:26.876707 134158149904192 checkpoint_manager.py:1452] Waiting for previous save to complete took 47.934209 seconds. If this number is high, consider checkpointing less frequently.
I0423 07:48:26.878777 134158149904192 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 15
I0423 07:48:26.881195 134158149904192 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/15.
I0423 07:48:27.243251 134158149904192 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 07:48:27.243347 134158149904192 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0423 07:48:27.269161 134158149904192 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026992s
I0423 07:48:27.269310 134158149904192 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 50.619 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030473947525024414 s) (per-host)
I0423 07:48:27.269360 134158149904192 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030533s (batch_requests_ready=0.001862s, total_serialization_initiated=0.028606s, others=0.000065s)
I0423 07:48:27.269442 134158149904192 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.035305s (all_items=0.000012s, per_item={'items': '0.00001240'}, temp_paths=0.035293)
I0423 07:48:27.270139 134158149904192 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.39 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/15.
I0423 07:48:27.270400 134034713732864 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 08:08:27.270377
I0423 07:48:27.283287 134158149904192 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 07:48:27.283508 134030418765568 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0423 07:48:27.283628 134158149904192 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776930458.942469, 'wait_for_prev_duration_secs': 47.9342086315155, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776930506.8788152, 'checkpointer_blocking_duration_secs': 0.39168286323547363, 'get_old_steps_start_time': 1776930507.2705224, 'get_old_steps_duration_secs': 3.0040740966796875e-05, 'checkpoint_manager_blocking_start_time': 1776930458.9399946, 'checkpoint_manager_blocking_duration_secs': 48.34359884262085}
I0423 07:48:27.283743 134158149904192 checkpointing.py:408] Started an asynchronous checkpoint save for step 15
I0423 07:48:27.284719 134158149904192 metric_logger.py:196] completed step: 14, seconds: 0.148, TFLOP/s/device: 92.067, Tokens/s/device: 13877.407, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193
I0423 07:48:27.290805 134158149904192 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.318, Tokens/s/device: 13915.123, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661
I0423 07:48:33.322978 134035636471552 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/15/items/array_metadatas/process_5
I0423 07:48:49.053203 134158149904192 metric_logger.py:196] completed step: 16, seconds: 48.640, TFLOP/s/device: 0.279, Tokens/s/device: 42.106, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872
I0423 07:48:49.200552 134158149904192 metric_logger.py:196] completed step: 17, seconds: 21.616, TFLOP/s/device: 0.629, Tokens/s/device: 94.743, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975
I0423 07:48:49.347944 134158149904192 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.207, Tokens/s/device: 13446.261, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333
I0423 07:48:49.494909 134158149904192 checkpointing.py:772] Waiting for step 19 to finish before checkpoint...
I0423 07:48:49.495615 134158149904192 checkpointing.py:776] Waited 0.0007269382476806641 seconds for step 19 to finish before starting checkpointing.
I0423 07:48:49.497611 134158149904192 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 07:49:08.485059 134034713732864 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 38.297 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.2461838722229 s) (per-host)
I0423 07:49:08.485186 134034713732864 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.214710s.
I0423 07:49:18.711140 134034713732864 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 51.440649s.
I0423 07:49:18.711362 134030418765568 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 07:49:18.711432 134030418765568 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0423 07:49:18.711479 134030418765568 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0423 07:49:18.712950 134030418765568 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0423 07:49:18.713135 134158149904192 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15.
W0423 07:49:18.713263 134158149904192 checkpoint_manager.py:1452] Waiting for previous save to complete took 29.215660 seconds. If this number is high, consider checkpointing less frequently.
I0423 07:49:18.714834 134158149904192 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 19
I0423 07:49:18.717007 134158149904192 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/19.
I0423 07:49:19.428848 134158149904192 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 07:49:19.428948 134158149904192 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0423 07:49:19.454299 134158149904192 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026298s
I0423 07:49:19.454431 134158149904192 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 51.708 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.02983260154724121 s) (per-host)
I0423 07:49:19.454492 134158149904192 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.029899s (batch_requests_ready=0.001760s, total_serialization_initiated=0.028068s, others=0.000071s)
I0423 07:49:19.454568 134158149904192 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034168s (all_items=0.000012s, per_item={'items': '0.00001240'}, temp_paths=0.034155)
I0423 07:49:19.455290 134158149904192 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.74 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/19.
I0423 07:49:19.455642 134035636471552 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 08:09:19.455601
I0423 07:49:19.458974 134158149904192 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 07:49:19.459202 134030418765568 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0423 07:49:19.459326 134158149904192 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776930529.4975805, 'wait_for_prev_duration_secs': 29.215660333633423, 'time_between_consecutive_saves_sec': 22.621154308319092, 'checkpointer_blocking_start_time': 1776930558.7148738, 'checkpointer_blocking_duration_secs': 0.7409257888793945, 'get_old_steps_start_time': 1776930559.4558163, 'get_old_steps_duration_secs': 2.6226043701171875e-05, 'checkpoint_manager_blocking_start_time': 1776930529.495855, 'checkpoint_manager_blocking_duration_secs': 29.96344017982483}
I0423 07:49:19.459434 134158149904192 checkpointing.py:408] Started an asynchronous checkpoint save for step 19
I0423 07:49:19.459476 134158149904192 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 07:49:24.438692 134030427158272 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_08_checkpoint_async_true/checkpoints/19/items/array_metadatas/process_5
I0423 07:50:01.196527 134035636471552 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.815 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.77189040184021 s) (per-host)
I0423 07:50:01.196643 134035636471552 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.740872s.
I0423 07:50:09.609428 134035636471552 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 50.153642s.
I0423 07:50:09.609692 134030418765568 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 07:50:09.609805 134030418765568 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0423 07:50:09.609846 134030418765568 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0423 07:50:09.611560 134030418765568 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0423 07:50:09.611749 134158149904192 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19.
I0423 07:50:09.611894 134158149904192 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 07:50:09.612846 134158149904192 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.250, Tokens/s/device: 13904.920, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Thu Apr 23 07:50:18 UTC 2026
EXIT_CODE=0
test/pipeline-scan-nnx  ·  5b21b66fc  ·  test_pipeline_scan_nnx_20260423_211617  ·  full log 
XPK Start: Thu Apr 23 21:42:44 UTC 2026
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
`rope_parameters`'s factor field must be a float >= 1, got 40
`rope_parameters`'s beta_fast field must be a float, got 32
`rope_parameters`'s beta_slow field must be a float, got 1
DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 
2026-04-23 21:43:09.456160: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303)
I0423 21:43:09.667482 136420393178944 max_utils.py:273] Attempting to initialize the jax distributed system...
I0423 21:43:18.709647 136420393178944 distributed.py:149] Starting JAX distributed service on [::]:8482
I0423 21:43:18.712023 136420393178944 distributed.py:172] Connecting to JAX distributed service on mt-08-checkpoint-async--uzvlk-slice-job-0-0.mt-08-checkpoint-async--uzvlk:8482
I0423 21:43:19.733192 136420393178944 max_utils.py:284] Jax distributed system initialized!
I0423 21:43:24.815222 136420393178944 max_utils.py:800] System Information: Jax Version: 0.9.2
I0423 21:43:24.815326 136420393178944 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0423 21:43:24.815367 136420393178944 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365
I0423 21:43:24.815402 136420393178944 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0423 21:43:25.512062 136420393178944 maxtext_utils.py:1565] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0423 21:43:25.512342 136420393178944 checkpointing.py:677] Setting up checkpoint logger...
I0423 21:43:25.512398 136420393178944 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0423 21:43:25.512444 136420393178944 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0423 21:43:25.512789 136420393178944 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7c12451184a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0423 21:43:27.060161 136420393178944 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0423 21:43:27.060397 136420393178944 checkpoint_manager.py:708] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bfeac5e7770>}, handler_registry=None
I0423 21:43:27.060659 136420393178944 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bfeac5e7770>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`.
I0423 21:43:27.060722 136420393178944 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bfeac5e8f80>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`.
I0423 21:43:27.060765 136420393178944 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 0x7bfeac5e7770>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bfeac5e7770>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bfeac5e8f80>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bfeac5e8f80>}).
I0423 21:43:27.061119 136420393178944 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0423 21:43:27.061188 136420393178944 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7bfeac41b560> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0423 21:43:28.683079 136301617587968 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776980343988706270, 'commit_timestamp_nsecs': 1776980386346458565, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/0/_CHECKPOINT_METADATA
I0423 21:43:28.943583 136301617587968 checkpoint_manager.py:1871] Missing metrics for step 0
I0423 21:43:29.238742 136420393178944 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints
I0423 21:43:29.746267 136420393178944 checkpoint_manager.py:929] [process=4][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_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7bfeac5e86e0>
I0423 21:43:29.746456 136420393178944 checkpointing.py:301] Checkpoint manager created!
I0423 21:43:30.673127 136420393178944 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0423 21:43:30.673246 136420393178944 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 21:43:31.056157 136420393178944 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 21:43:31.056250 136420393178944 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 21:43:31.072748 136420393178944 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0423 21:43:31.072818 136420393178944 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0423 21:43:31.096572 136420393178944 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 21:43:31.096641 136420393178944 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 21:43:31.113135 136420393178944 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 21:43:31.113193 136420393178944 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 21:43:31.129535 136420393178944 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0423 21:43:31.129600 136420393178944 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 21:43:31.154656 136420393178944 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0423 21:43:31.154742 136420393178944 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0423 21:43:31.177040 136420393178944 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0423 21:43:31.177115 136420393178944 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0423 21:43:31.388319 136420393178944 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0423 21:43:31.388422 136420393178944 checkpointing.py:581] restoring from this run's directory step 9
I0423 21:43:31.637459 136420393178944 event_tracking.py:70] [process=4] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/9.
I0423 21:43:31.879034 136420393178944 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/9.
I0423 21:43:31.879469 136420393178944 event_tracking.py:125] [process=4] [sync] Finished blocking load in 0.24 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/9.
I0423 21:43:33.303149    1409 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0423 21:43:39.141689 136420393178944 jax_array_handlers.py:843] [process=4] /jax/orbax/read/worker/io/requested throughput: 270.380 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 5.844638109207153 s) (per-host)
I0423 21:43:39.146182 136420393178944 base_pytree_checkpoint_handler.py:130] [process=4] /jax/checkpoint/read/gbytes_per_sec: 252.165 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.2760186195373535 s) (per-host)
I0423 21:43:40.784192 136420393178944 event_tracking.py:138] [process=4] [sync] Finished load in 9.15 seconds @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/9
I0423 21:43:40.784351 136420393178944 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints', 'checkpointer_start_time': 1776980611.637361, 'checkpointer_duration_secs': 9.146957635879517, 'checkpoint_manager_start_time': 1776980611.3886652, 'checkpoint_manager_duration_secs': 9.395655393600464}
fsdp: 32
I0423 21:43:41.206809 136420393178944 maxtext_utils.py:1668]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0423 21:43:41.206951 136420393178944 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)
I0423 21:43:41.207007 136420393178944 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)
I0423 21:43:41.207065 136420393178944 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')
I0423 21:43:41.207117 136420393178944 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)
I0423 21:43:41.207264 136420393178944 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)
I0423 21:43:41.207353 136420393178944 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)
I0423 21:43:41.207412 136420393178944 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')
I0423 21:43:41.207452 136420393178944 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)
I0423 21:43:41.207489 136420393178944 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)
I0423 21:43:41.207534 136420393178944 maxtext_utils.py:1668]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)

I0423 21:43:41.207580 136420393178944 maxtext_utils.py:1668]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')
I0423 21:43:41.696731 136420393178944 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 21:43:41.696820 136420393178944 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 21:43:41.712426 136420393178944 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0423 21:43:41.712486 136420393178944 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0423 21:43:52.798577 136420393178944 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0423 21:43:52.799387 136420393178944 metric_logger.py:301] number parameters: 1.104 billion
I0423 21:44:04.440773 136420393178944 checkpointing.py:772] Waiting for step 10 to finish before checkpoint...
I0423 21:44:04.593586 136420393178944 checkpointing.py:776] Waited 0.15279769897460938 seconds for step 10 to finish before starting checkpointing.
I0423 21:44:04.596086 136420393178944 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 21:44:04.598214 136420393178944 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 10
I0423 21:44:04.599699 136420393178944 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/10.
I0423 21:44:04.919905 136420393178944 signaling_client.py:364] Using JaxDistributedSignalingClient
I0423 21:44:04.920907 136420393178944 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 21:44:04.920961 136420393178944 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0423 21:44:05.194182 136420393178944 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.274362s
I0423 21:44:05.194361 136420393178944 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 5.542 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2783544063568115 s) (per-host)
I0423 21:44:05.194416 136420393178944 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.278420s (batch_requests_ready=0.002276s, total_serialization_initiated=0.276068s, others=0.000075s)
I0423 21:44:05.194507 136420393178944 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.282832s (all_items=0.000018s, per_item={'items': '0.00001836'}, temp_paths=0.282814)
I0423 21:44:05.195251 136420393178944 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.60 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/10.
I0423 21:44:05.195558 136301256898304 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 22:04:05.195523
I0423 21:44:05.212564 136420393178944 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 21:44:05.212936 136297944119040 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save.
I0423 21:44:05.213099 136420393178944 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776980644.5960615, 'wait_for_prev_duration_secs': 8.726119995117188e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776980644.5982728, 'checkpointer_blocking_duration_secs': 0.5974423885345459, 'get_old_steps_start_time': 1776980645.195737, 'get_old_steps_duration_secs': 3.147125244140625e-05, 'checkpoint_manager_blocking_start_time': 1776980644.5941238, 'checkpoint_manager_blocking_duration_secs': 0.6189382076263428}
I0423 21:44:05.213208 136420393178944 checkpointing.py:408] Started an asynchronous checkpoint save for step 10
I0423 21:44:05.213259 136420393178944 max_utils.py:750] 
Memstats: After params initialized:
I0423 21:44:05.213310 136420393178944 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0))
I0423 21:44:05.213342 136420393178944 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0))
I0423 21:44:05.213368 136420393178944 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0))
I0423 21:44:05.213392 136420393178944 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0))
I0423 21:44:05.523177 136420393178944 metric_logger.py:196] completed step: 10, seconds: 11.641, TFLOP/s/device: 1.167, Tokens/s/device: 175.926, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869
I0423 21:44:05.696981 136420393178944 metric_logger.py:196] completed step: 11, seconds: 1.081, TFLOP/s/device: 12.570, Tokens/s/device: 1894.733, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659
I0423 21:44:05.847059 136420393178944 metric_logger.py:196] completed step: 12, seconds: 0.027, TFLOP/s/device: 512.219, Tokens/s/device: 77207.268, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068
I0423 21:44:05.994663 136420393178944 metric_logger.py:196] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.692, Tokens/s/device: 13368.583, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001
I0423 21:44:05.999079 136420393178944 checkpointing.py:772] Waiting for step 15 to finish before checkpoint...
I0423 21:44:06.289177 136420393178944 checkpointing.py:776] Waited 0.29006338119506836 seconds for step 15 to finish before starting checkpointing.
I0423 21:44:06.292230 136420393178944 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 21:44:10.163744 136297968563968 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/10/items/array_metadatas/process_4
I0423 21:44:46.048341 136301256898304 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 38.403 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.13229846954346 s) (per-host)
I0423 21:44:46.048467 136301256898304 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.852795s.
I0423 21:44:54.916657 136301256898304 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 49.720968s.
I0423 21:44:54.916936 136297944119040 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 21:44:54.917064 136297944119040 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save.
I0423 21:44:54.917125 136297944119040 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0423 21:44:54.918803 136297944119040 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0423 21:44:54.918992 136420393178944 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10.
W0423 21:44:54.919124 136420393178944 checkpoint_manager.py:1452] Waiting for previous save to complete took 48.626894 seconds. If this number is high, consider checkpointing less frequently.
I0423 21:44:54.921199 136420393178944 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 15
I0423 21:44:54.923581 136420393178944 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/15.
I0423 21:44:55.210743 136420393178944 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 21:44:55.210838 136420393178944 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0423 21:44:55.236186 136420393178944 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027244s
I0423 21:44:55.236367 136420393178944 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 50.006 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030848026275634766 s) (per-host)
I0423 21:44:55.236425 136420393178944 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030917s (batch_requests_ready=0.001924s, total_serialization_initiated=0.028918s, others=0.000075s)
I0423 21:44:55.236546 136420393178944 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.035766s (all_items=0.000013s, per_item={'items': '0.00001287'}, temp_paths=0.035753)
I0423 21:44:55.237282 136420393178944 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.32 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/15.
I0423 21:44:55.237607 136297968563968 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 22:04:55.237571
I0423 21:44:55.248190 136420393178944 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 21:44:55.248487 136297944119040 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save.
I0423 21:44:55.248651 136420393178944 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776980646.2922008, 'wait_for_prev_duration_secs': 48.62689399719238, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776980694.921238, 'checkpointer_blocking_duration_secs': 0.31650853157043457, 'get_old_steps_start_time': 1776980695.237768, 'get_old_steps_duration_secs': 3.0040740966796875e-05, 'checkpoint_manager_blocking_start_time': 1776980646.2897995, 'checkpoint_manager_blocking_duration_secs': 48.958818435668945}
I0423 21:44:55.248769 136420393178944 checkpointing.py:408] Started an asynchronous checkpoint save for step 15
I0423 21:44:55.249530 136420393178944 metric_logger.py:196] completed step: 14, seconds: 0.150, TFLOP/s/device: 90.460, Tokens/s/device: 13635.062, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193
I0423 21:44:55.255496 136420393178944 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.184, Tokens/s/device: 13894.920, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661
I0423 21:45:00.775727 136296869644032 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/15/items/array_metadatas/process_4
I0423 21:45:17.146805 136420393178944 metric_logger.py:196] completed step: 16, seconds: 49.255, TFLOP/s/device: 0.276, Tokens/s/device: 41.580, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872
I0423 21:45:17.294301 136420393178944 metric_logger.py:196] completed step: 17, seconds: 21.745, TFLOP/s/device: 0.625, Tokens/s/device: 94.181, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975
I0423 21:45:17.441518 136420393178944 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.376, Tokens/s/device: 13471.734, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333
I0423 21:45:17.588404 136420393178944 checkpointing.py:772] Waiting for step 19 to finish before checkpoint...
I0423 21:45:17.589151 136420393178944 checkpointing.py:776] Waited 0.000766754150390625 seconds for step 19 to finish before starting checkpointing.
I0423 21:45:17.591213 136420393178944 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 21:45:37.861824 136297968563968 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 37.031 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.656264781951904 s) (per-host)
I0423 21:45:37.861959 136297968563968 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.624240s.
I0423 21:45:45.695439 136297968563968 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.457704s.
I0423 21:45:45.695734 136297944119040 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 21:45:45.695852 136297944119040 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save.
I0423 21:45:45.695918 136297944119040 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts...
I0423 21:45:45.697741 136297944119040 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts.
I0423 21:45:45.697951 136420393178944 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15.
W0423 21:45:45.698077 136420393178944 checkpoint_manager.py:1452] Waiting for previous save to complete took 28.106873 seconds. If this number is high, consider checkpointing less frequently.
I0423 21:45:45.699518 136420393178944 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 19
I0423 21:45:45.701587 136420393178944 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/19.
I0423 21:45:45.996231 136420393178944 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0423 21:45:45.996324 136420393178944 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0423 21:45:46.021992 136420393178944 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026641s
I0423 21:45:46.022118 136420393178944 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 51.056 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030213117599487305 s) (per-host)
I0423 21:45:46.022166 136420393178944 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030270s (batch_requests_ready=0.001759s, total_serialization_initiated=0.028449s, others=0.000062s)
I0423 21:45:46.022239 136420393178944 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034286s (all_items=0.000012s, per_item={'items': '0.00001192'}, temp_paths=0.034274)
I0423 21:45:46.022919 136420393178944 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.32 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/19.
I0423 21:45:46.023182 136297944119040 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 22:05:46.023159
I0423 21:45:46.027142 136420393178944 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0423 21:45:46.027398 136296309061376 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save.
I0423 21:45:46.027554 136420393178944 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776980717.591181, 'wait_for_prev_duration_secs': 28.106873035430908, 'time_between_consecutive_saves_sec': 22.67233157157898, 'checkpointer_blocking_start_time': 1776980745.6995578, 'checkpointer_blocking_duration_secs': 0.3237333297729492, 'get_old_steps_start_time': 1776980746.0233114, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1776980717.5893989, 'checkpoint_manager_blocking_duration_secs': 28.438122749328613}
I0423 21:45:46.027675 136420393178944 checkpointing.py:408] Started an asynchronous checkpoint save for step 19
I0423 21:45:46.027717 136420393178944 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0423 21:45:51.551764 136296869644032 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260423_211617/linen_xpk_test_pipeline_scan_nnx_20260423_211617_08_checkpoint_async_true/checkpoints/19/items/array_metadatas/process_4
I0423 21:46:28.276372 136297944119040 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 37.356 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.284422874450684 s) (per-host)
I0423 21:46:28.276479 136297944119040 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.253216s.
I0423 21:46:36.274205 136297944119040 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.250925s.
I0423 21:46:36.274486 136296309061376 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save.
I0423 21:46:36.274621 136296309061376 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save.
I0423 21:46:36.274663 136296309061376 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0423 21:46:36.276854 136296309061376 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0423 21:46:36.277040 136420393178944 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19.
I0423 21:46:36.277186 136420393178944 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0423 21:46:36.278197 136420393178944 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.224, Tokens/s/device: 13900.956, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Thu Apr 23 21:46:44 UTC 2026
EXIT_CODE=0