MaxView

Timing Test Reference Updates
‹ 02_syntheticCase: 03_dropout04_int8 ›

Metrics: main (574ad3fb9) vs test/pipeline-scan-nnx (574ad3fb9)

Metricmain  574ad3fb9test/pipeline-scan-nnx  574ad3fb9Diff (test/pipeline-scan-nnx − main)
Parameters1.104 billion1.104 billion—
Final loss6.98106.98100
TFLOP/s11.33811.111-0.227
Tok/s1812.11775.9-36.202
Avg s/step1.3741.381+0.007
Memory %1.311.310
JAX0.9.20.8.1—

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

main  ·  574ad3fb9  ·  main_20260418_180002  ·  full log 
XPK Start: Sat Apr 18 18:08:48 UTC 2026
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
2026-04-18 18:09:12.777945: 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)
I0418 18:09:13.317142 135278194456384 max_utils.py:273] Attempting to initialize the jax distributed system...
I0418 18:09:22.359464 135278194456384 distributed.py:149] Starting JAX distributed service on [::]:8482
I0418 18:09:22.361876 135278194456384 distributed.py:172] Connecting to JAX distributed service on mt-03-dropout-e6fds-slice-job-0-0.mt-03-dropout-e6fds:8482
I0418 18:09:23.537461 135278194456384 max_utils.py:284] Jax distributed system initialized!
I0418 18:09:29.500558 135278194456384 max_utils.py:800] System Information: Jax Version: 0.9.2
I0418 18:09:29.500661 135278194456384 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0418 18:09:29.500701 135278194456384 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
I0418 18:09:29.500735 135278194456384 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0418 18:09:30.249334 135278194456384 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0418 18:09:30.249608 135278194456384 checkpointing.py:677] Setting up checkpoint logger...
I0418 18:09:30.249664 135278194456384 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0418 18:09:30.249708 135278194456384 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0418 18:09:30.250046 135278194456384 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 0x7b085431ce00>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0418 18:09:33.881184 135278194456384 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0418 18:09:33.881514 135278194456384 checkpoint_manager.py:708] [process=7][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7af4a03ea630>}, handler_registry=None
I0418 18:09:33.881750 135278194456384 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7af4a03ea630>` 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`.
I0418 18:09:33.881804 135278194456384 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7af4a00bf560>` 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`.
I0418 18:09:33.881839 135278194456384 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 0x7af4a03ea630>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7af4a03ea630>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7af4a00bf560>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7af4a00bf560>}).
I0418 18:09:33.882162 135278194456384 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0418 18:09:33.882233 135278194456384 async_checkpointer.py:192] [process=7][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7af4805d3920> timeout: 1200 secs and primary_host=0 for async checkpoint writes
I0418 18:09:35.278334 135278194456384 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_03_dropout/checkpoints
I0418 18:09:36.182208 135278194456384 checkpoint_manager.py:929] [process=7][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=10), 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_20260418_180002/linen_xpk_main_20260418_180002_03_dropout/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7af4a00bbfe0>
I0418 18:09:36.182367 135278194456384 checkpointing.py:301] Checkpoint manager created!
I0418 18:09:36.527550 135278194456384 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0418 18:09:36.527672 135278194456384 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None).
I0418 18:09:36.904780 135278194456384 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0418 18:09:36.904876 135278194456384 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None).
I0418 18:09:36.921031 135278194456384 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0418 18:09:36.921113 135278194456384 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None).
I0418 18:09:36.944603 135278194456384 attentions.py:1154] attentions/query Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0418 18:09:36.944672 135278194456384 attentions.py:1154] attentions/query Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None).
I0418 18:09:36.961120 135278194456384 attentions.py:1155] attentions/key Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0418 18:09:36.961184 135278194456384 attentions.py:1155] attentions/key Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None).
I0418 18:09:36.977460 135278194456384 attentions.py:1156] attentions/value Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0418 18:09:36.977526 135278194456384 attentions.py:1156] attentions/value Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None).
I0418 18:09:37.002854 135278194456384 attentions.py:1197] attentions/out Logical: bfloat16[32,128,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0418 18:09:37.002933 135278194456384 attentions.py:1197] attentions/out Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None).
I0418 18:09:37.028928 135278194456384 linears.py:525] linears/x Logical: bfloat16[32,128,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0418 18:09:37.028995 135278194456384 linears.py:525] linears/x Physical: bfloat16[32,128,7168]....................................... ('fsdp', None, None).
I0418 18:09:37.256135 135278194456384 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0418 18:09:37.256243 135278194456384 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0418 18:09:38.690651 135278194456384 maxtext_utils.py:1654]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0418 18:09:38.690775 135278194456384 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0418 18:09:38.690829 135278194456384 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0418 18:09:38.690893 135278194456384 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   P('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0418 18:09:38.690970 135278194456384 maxtext_utils.py:1654]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0418 18:09:38.691020 135278194456384 maxtext_utils.py:1654]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0418 18:09:38.691117 135278194456384 maxtext_utils.py:1654]  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)
I0418 18:09:38.691187 135278194456384 maxtext_utils.py:1654]  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')
I0418 18:09:38.691230 135278194456384 maxtext_utils.py:1654]  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)
I0418 18:09:38.691268 135278194456384 maxtext_utils.py:1654]  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)
I0418 18:09:38.691315 135278194456384 maxtext_utils.py:1654]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)

I0418 18:09:38.691363 135278194456384 maxtext_utils.py:1654]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')
I0418 18:09:39.213724 135278194456384 train.py:155] train/xent Logical: float32[32,128]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0418 18:09:39.213819 135278194456384 train.py:155] train/xent Physical: float32[32,128]............................................. ('fsdp', None).
I0418 18:09:39.229641 135278194456384 train.py:162] train/z_loss Logical: float32[32,128]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0418 18:09:39.229702 135278194456384 train.py:162] train/z_loss Physical: float32[32,128]............................................. ('fsdp', None).
I0418 18:09:42.679100 135278194456384 max_utils.py:791] Total memory size: 0.8 GB, Output size: 0.4 GB, Temp size: 0.4 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0418 18:09:42.679884 135278194456384 metric_logger.py:301] number parameters: 1.104 billion
I0418 18:09:46.956998 135278194456384 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0418 18:09:47.030601 135278194456384 checkpointing.py:776] Waited 0.07358765602111816 seconds for step 0 to finish before starting checkpointing.
I0418 18:09:47.033181 135278194456384 checkpoint_manager.py:2009] [process=7][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0418 18:09:47.035274 135278194456384 checkpoint_manager.py:1512] [process=7] Saving checkpoint at step 0
I0418 18:09:47.037438 135278194456384 event_tracking.py:70] [process=7] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_03_dropout/checkpoints/0.
I0418 18:09:47.755404 135278194456384 signaling_client.py:364] Using JaxDistributedSignalingClient
I0418 18:09:47.756417 135278194456384 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0418 18:09:47.756473 135278194456384 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
I0418 18:09:48.031235 135278194456384 base_pytree_checkpoint_handler.py:154] [process=7][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.275900s
I0418 18:09:48.031420 135278194456384 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/blocking_gbytes_per_sec: 5.486 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2812025547027588 s) (per-host)
I0418 18:09:48.031490 135278194456384 base_pytree_checkpoint_handler.py:768] [process=7][thread=MainThread] Initiated Pytree async_save. Time taken: 0.281285s (batch_requests_ready=0.002159s, total_serialization_initiated=0.279034s, others=0.000092s)
I0418 18:09:48.031605 135278194456384 composite_checkpoint_handler.py:715] [process=7][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.285461s (all_items=0.000018s, per_item={'items': '0.00001836'}, temp_paths=0.285443)
I0418 18:09:48.032428 135278194456384 event_tracking.py:125] [process=7] [async] Finished blocking save in 1.00 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_03_dropout/checkpoints/0.
I0418 18:09:48.032763 135149937227520 async_checkpointer.py:76] [process=7][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 18:29:48.032719
I0418 18:09:48.044974 135278194456384 checkpoint_manager.py:1560] [process=7][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0418 18:09:48.045357 135152604796672 async_checkpointer.py:280] [process=7][thread=save_finalize] Waiting for background save thread=async_save.
I0418 18:09:48.045511 135278194456384 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_03_dropout/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776535787.0331633, 'wait_for_prev_duration_secs': 6.127357482910156e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776535787.0353577, 'checkpointer_blocking_duration_secs': 0.9975700378417969, 'get_old_steps_start_time': 1776535788.0329514, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776535787.0311491, 'checkpoint_manager_blocking_duration_secs': 1.0143227577209473}
I0418 18:09:48.045617 135278194456384 checkpointing.py:408] Started an asynchronous checkpoint save for step 0
I0418 18:09:48.045666 135278194456384 max_utils.py:750] 
Memstats: After params initialized:
I0418 18:09:48.045714 135278194456384 max_utils.py:756] 	Using (GB) 0.41 / 31.25 (1.312000%) on TPU_26(process=7,(2,6,0,0))
I0418 18:09:48.045747 135278194456384 max_utils.py:756] 	Using (GB) 0.41 / 31.25 (1.312000%) on TPU_27(process=7,(3,6,0,0))
I0418 18:09:48.045774 135278194456384 max_utils.py:756] 	Using (GB) 0.41 / 31.25 (1.312000%) on TPU_30(process=7,(2,7,0,0))
I0418 18:09:48.045799 135278194456384 max_utils.py:756] 	Using (GB) 0.41 / 31.25 (1.312000%) on TPU_31(process=7,(3,7,0,0))
I0418 18:09:48.357237 135278194456384 metric_logger.py:196] completed step: 0, seconds: 4.277, TFLOP/s/device: 0.187, Tokens/s/device: 29.928, total_weights: 4096, loss: 10.889, lm_loss: 10.889, perplexity: 53560.699
I0418 18:09:48.461656 135278194456384 metric_logger.py:196] completed step: 1, seconds: 1.399, TFLOP/s/device: 0.573, Tokens/s/device: 91.517, total_weights: 4096, loss: 10.902, lm_loss: 10.902, perplexity: 54298.523
I0418 18:09:49.284912 135278194456384 metric_logger.py:196] completed step: 2, seconds: 0.035, TFLOP/s/device: 23.210, Tokens/s/device: 3709.607, total_weights: 4096, loss: 9.956, lm_loss: 9.956, perplexity: 21082.051
I0418 18:09:49.355606 135278194456384 metric_logger.py:196] completed step: 3, seconds: 0.823, TFLOP/s/device: 0.973, Tokens/s/device: 155.448, total_weights: 4096, loss: 9.139, lm_loss: 9.139, perplexity: 9314.691
I0418 18:09:49.497471 135278194456384 metric_logger.py:196] completed step: 4, seconds: 0.076, TFLOP/s/device: 10.559, Tokens/s/device: 1687.630, total_weights: 4096, loss: 8.459, lm_loss: 8.459, perplexity: 4715.493
I0418 18:09:49.503385 135278194456384 metric_logger.py:196] completed step: 5, seconds: 0.071, TFLOP/s/device: 11.348, Tokens/s/device: 1813.673, total_weights: 4096, loss: 7.924, lm_loss: 7.924, perplexity: 2762.373
I0418 18:09:50.863830    2771 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0418 18:09:53.872585 135151492839168 array_metadata_store.py:203] [process=7][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_03_dropout/checkpoints/0/items/array_metadatas/process_7
I0418 18:09:59.243610 135278194456384 metric_logger.py:196] completed step: 6, seconds: 0.142, TFLOP/s/device: 5.624, Tokens/s/device: 898.839, total_weights: 4096, loss: 7.553, lm_loss: 7.553, perplexity: 1906.617
I0418 18:09:59.314127 135278194456384 metric_logger.py:196] completed step: 7, seconds: 9.671, TFLOP/s/device: 0.083, Tokens/s/device: 13.236, total_weights: 4096, loss: 7.267, lm_loss: 7.267, perplexity: 1431.653
I0418 18:09:59.386180 135278194456384 metric_logger.py:196] completed step: 8, seconds: 0.075, TFLOP/s/device: 10.681, Tokens/s/device: 1707.145, total_weights: 4096, loss: 7.111, lm_loss: 7.111, perplexity: 1225.983
I0418 18:09:59.457588 135278194456384 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0418 18:09:59.458462 135278194456384 checkpointing.py:776] Waited 0.0008862018585205078 seconds for step 9 to finish before starting checkpointing.
I0418 18:09:59.460584 135278194456384 checkpoint_manager.py:2020] [process=7][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0418 18:10:20.024253 135149937227520 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/gbytes_per_sec: 48.943 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 32.27400207519531 s) (per-host)
I0418 18:10:20.024379 135149937227520 async_checkpointer.py:90] [process=7][thread=async_save] 3 Handler Commit operations completed. Time taken: 31.991492s.
I0418 18:10:32.291517 135149937227520 async_checkpointer.py:160] [process=7][thread=async_save] Background save thread done. Time taken: 44.258613s.
I0418 18:10:32.291756 135152604796672 async_checkpointer.py:288] [process=7][thread=save_finalize] Done with waiting for background save thread=async_save.
I0418 18:10:32.291820 135152604796672 async_checkpointer.py:298] [process=7][thread=save_finalize] No errors found in background save thread=async_save.
I0418 18:10:32.291871 135152604796672 checkpoint_manager.py:2137] [process=7][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0418 18:10:32.293576 135152604796672 checkpoint_manager.py:2146] [process=7][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0418 18:10:32.293772 135278194456384 checkpoint_manager.py:2032] [process=7][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0418 18:10:32.293942 135278194456384 checkpoint_manager.py:1452] Waiting for previous save to complete took 32.833347 seconds. If this number is high, consider checkpointing less frequently.
I0418 18:10:32.295870 135278194456384 checkpoint_manager.py:1512] [process=7] Saving checkpoint at step 9
I0418 18:10:32.297894 135278194456384 event_tracking.py:70] [process=7] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_03_dropout/checkpoints/9.
I0418 18:10:33.014365 135278194456384 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0418 18:10:33.014456 135278194456384 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
I0418 18:10:33.049299 135278194456384 base_pytree_checkpoint_handler.py:154] [process=7][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036150s
I0418 18:10:33.049465 135278194456384 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/blocking_gbytes_per_sec: 39.039 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.039513587951660156 s) (per-host)
I0418 18:10:33.049515 135278194456384 base_pytree_checkpoint_handler.py:768] [process=7][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039574s (batch_requests_ready=0.001720s, total_serialization_initiated=0.037786s, others=0.000068s)
I0418 18:10:33.049602 135278194456384 composite_checkpoint_handler.py:715] [process=7][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.043552s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.043536)
I0418 18:10:33.050279 135278194456384 event_tracking.py:125] [process=7] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_03_dropout/checkpoints/9.
I0418 18:10:33.050647 135151492839168 async_checkpointer.py:76] [process=7][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 18:30:33.050610
I0418 18:10:33.055156 135278194456384 checkpoint_manager.py:1560] [process=7][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0418 18:10:33.055509 135152604796672 async_checkpointer.py:280] [process=7][thread=save_finalize] Waiting for background save thread=async_save.
I0418 18:10:33.055669 135278194456384 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_03_dropout/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776535799.4605536, 'wait_for_prev_duration_secs': 32.8333466053009, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776535832.2959163, 'checkpointer_blocking_duration_secs': 0.7548732757568359, 'get_old_steps_start_time': 1776535833.0508122, 'get_old_steps_duration_secs': 2.86102294921875e-05, 'checkpoint_manager_blocking_start_time': 1776535799.4587278, 'checkpoint_manager_blocking_duration_secs': 33.59690737724304}
I0418 18:10:33.055786 135278194456384 checkpointing.py:408] Started an asynchronous checkpoint save for step 9
I0418 18:10:33.055831 135278194456384 checkpoint_manager.py:2020] [process=7][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0418 18:10:42.079926 135125735982848 array_metadata_store.py:203] [process=7][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_03_dropout/checkpoints/9/items/array_metadatas/process_7
I0418 18:11:16.778497 135151492839168 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/gbytes_per_sec: 36.090 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.768510580062866 s) (per-host)
I0418 18:11:16.778608 135151492839168 async_checkpointer.py:90] [process=7][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.727848s.
I0418 18:11:25.390923 135151492839168 async_checkpointer.py:160] [process=7][thread=async_save] Background save thread done. Time taken: 52.340147s.
I0418 18:11:25.391216 135152604796672 async_checkpointer.py:288] [process=7][thread=save_finalize] Done with waiting for background save thread=async_save.
I0418 18:11:25.391334 135152604796672 async_checkpointer.py:298] [process=7][thread=save_finalize] No errors found in background save thread=async_save.
I0418 18:11:25.391383 135152604796672 checkpoint_manager.py:2137] [process=7][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0418 18:11:25.393105 135152604796672 checkpoint_manager.py:2146] [process=7][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0418 18:11:25.393279 135278194456384 checkpoint_manager.py:2032] [process=7][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0418 18:11:25.393428 135278194456384 checkpoint_manager.py:2009] [process=7][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0418 18:11:25.394446 135278194456384 metric_logger.py:196] completed step: 9, seconds: 0.071, TFLOP/s/device: 11.338, Tokens/s/device: 1812.056, total_weights: 4096, loss: 6.981, lm_loss: 6.981, perplexity: 1075.867
Per train step:
 Total TFLOPs: 0.80 
 split as 99.60% learnable weight flops and 0.40% attention flops
XPK End: Sat Apr 18 18:11:34 UTC 2026
EXIT_CODE=0
test/pipeline-scan-nnx  ·  574ad3fb9  ·  test_pipeline_scan_nnx_20260419_085454  ·  full log 
XPK Start: Sun Apr 19 09:01:40 UTC 2026
2026-04-19 09:01:44.186848: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
E0000 00:00:1776589304.199472      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1776589304.203184      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1776589304.214290      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776589304.214308      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776589304.214310      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776589304.214312      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-19 09:02:03.288191: 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)
I0419 09:02:03.800300 137291389007680 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-19 09:02:12,840:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0419 09:02:12.840843 137291389007680 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-19 09:02:12,843:jax._src.distributed:157: Connecting to JAX distributed service on mt-03-dropout-99g6h-slice-job-0-0.mt-03-dropout-99g6h:8482
I0419 09:02:12.843200 137291389007680 distributed.py:157] Connecting to JAX distributed service on mt-03-dropout-99g6h-slice-job-0-0.mt-03-dropout-99g6h:8482
I0419 09:02:14.642414 137291389007680 max_utils.py:284] Jax distributed system initialized!
I0419 09:02:21.522285 137291389007680 max_utils.py:800] System Information: Jax Version: 0.8.1
I0419 09:02:21.522390 137291389007680 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0419 09:02:21.522430 137291389007680 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709
I0419 09:02:21.522464 137291389007680 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0419 09:02:22.143685 137291389007680 maxtext_utils.py:1398] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0419 09:02:22.143986 137291389007680 checkpointing.py:677] Setting up checkpoint logger...
I0419 09:02:22.144043 137291389007680 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0419 09:02:22.144087 137291389007680 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0419 09:02:22.144449 137291389007680 base_pytree_checkpoint_handler.py:415] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7cdd14154c80>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0419 09:02:25.473391 137291389007680 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0419 09:02:25.473633 137291389007680 checkpoint_manager.py:709] [process=3][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cc9eee855e0>}, handler_registry=None
I0419 09:02:25.473871 137291389007680 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cc9eee855e0>` 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`.
I0419 09:02:25.473921 137291389007680 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cc9eee8a930>` 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`.
I0419 09:02:25.473958 137291389007680 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 0x7cc9eee855e0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cc9eee855e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cc9eee8a930>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cc9eee8a930>}).
I0419 09:02:25.474474 137291389007680 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0419 09:02:25.474554 137291389007680 async_checkpointer.py:177] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7cc9ec674040> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0419 09:02:26.212734 137291389007680 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_03_dropout/checkpoints
I0419 09:02:26.651614 137291389007680 checkpoint_manager.py:929] [process=3][thread=MainThread] CheckpointManager created,  primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), 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_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_03_dropout/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7cc9eee85bb0>
I0419 09:02:26.651795 137291389007680 checkpointing.py:301] Checkpoint manager created!
I0419 09:02:26.968576 137291389007680 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0419 09:02:26.968680 137291389007680 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None).
I0419 09:02:27.341688 137291389007680 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0419 09:02:27.341781 137291389007680 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None).
I0419 09:02:27.357657 137291389007680 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,128,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0419 09:02:27.357718 137291389007680 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,128,2048]....................................... ('fsdp', None, None).
I0419 09:02:27.380873 137291389007680 attentions.py:1150] attentions/query Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0419 09:02:27.380942 137291389007680 attentions.py:1150] attentions/query Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None).
I0419 09:02:27.396908 137291389007680 attentions.py:1151] attentions/key Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0419 09:02:27.396971 137291389007680 attentions.py:1151] attentions/key Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None).
I0419 09:02:27.412960 137291389007680 attentions.py:1152] attentions/value Logical: bfloat16[32,128,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0419 09:02:27.413024 137291389007680 attentions.py:1152] attentions/value Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None).
I0419 09:02:27.438266 137291389007680 attentions.py:1193] attentions/out Logical: bfloat16[32,128,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0419 09:02:27.438338 137291389007680 attentions.py:1193] attentions/out Physical: bfloat16[32,128,16,128]..................................... ('fsdp', None, None, None).
I0419 09:02:27.464205 137291389007680 linears.py:525] linears/x Logical: bfloat16[32,128,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0419 09:02:27.464287 137291389007680 linears.py:525] linears/x Physical: bfloat16[32,128,7168]....................................... ('fsdp', None, None).
I0419 09:02:27.671814 137291389007680 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0419 09:02:27.671932 137291389007680 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0419 09:02:29.113749 137291389007680 maxtext_utils.py:1501]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0419 09:02:29.113881 137291389007680 maxtext_utils.py:1501]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0419 09:02:29.113936 137291389007680 maxtext_utils.py:1501]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0419 09:02:29.113995 137291389007680 maxtext_utils.py:1501]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   PartitionSpec('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0419 09:02:29.114050 137291389007680 maxtext_utils.py:1501]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  (None, None)
I0419 09:02:29.114089 137291389007680 maxtext_utils.py:1501]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  (None, None)
I0419 09:02:29.114149 137291389007680 maxtext_utils.py:1501]  params/params/decoder/layers/self_attention/key/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0419 09:02:29.114204 137291389007680 maxtext_utils.py:1501]  params/params/decoder/layers/self_attention/out/kernel
    Shape:     float32[16,16,128,2048]
    Logical:   PartitionSpec('heads', 'layers', 'kv', 'embed')
    Physical:  (None, None, None, 'fsdp')
I0419 09:02:29.114274 137291389007680 maxtext_utils.py:1501]  params/params/decoder/layers/self_attention/query/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None, None)
I0419 09:02:29.114324 137291389007680 maxtext_utils.py:1501]  params/params/decoder/layers/self_attention/value/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0419 09:02:29.114377 137291389007680 maxtext_utils.py:1501]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed', 'vocab')
    Physical:  ('fsdp', None)
I0419 09:02:29.114426 137291389007680 maxtext_utils.py:1501]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed')
    Physical:  (None, 'fsdp')

I0419 09:02:29.645651 137291389007680 train.py:156] train/xent Logical: float32[32,128]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0419 09:02:29.645754 137291389007680 train.py:156] train/xent Physical: float32[32,128]............................................. ('fsdp', None).
I0419 09:02:29.660943 137291389007680 train.py:163] train/z_loss Logical: float32[32,128]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0419 09:02:29.661008 137291389007680 train.py:163] train/z_loss Physical: float32[32,128]............................................. ('fsdp', None).
I0419 09:02:33.077452 137291389007680 max_utils.py:791] Total memory size: 0.8 GB, Output size: 0.4 GB, Temp size: 0.4 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0419 09:02:33.078337 137291389007680 metric_logger.py:289] number parameters: 1.104 billion
I0419 09:02:37.024916 137291389007680 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0419 09:02:37.253314 137291389007680 checkpointing.py:776] Waited 0.22838449478149414 seconds for step 0 to finish before starting checkpointing.
I0419 09:02:37.256110 137291389007680 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0419 09:02:37.258006 137291389007680 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 0
I0419 09:02:37.260777 137291389007680 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_03_dropout/checkpoints/0.
I0419 09:02:37.863741 137291389007680 signaling_client.py:364] Using JaxDistributedSignalingClient
I0419 09:02:37.864768 137291389007680 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0419 09:02:37.864827 137291389007680 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
I0419 09:02:38.109198 137291389007680 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.245536s
I0419 09:02:38.109384 137291389007680 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.944 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5239486694335938 s) (per-host)
I0419 09:02:38.109439 137291389007680 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.524014s (batch_requests_ready=0.274882s, total_serialization_initiated=0.249059s, others=0.000073s)
I0419 09:02:38.109532 137291389007680 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.528003s (all_items=0.000019s, per_item={'items': '0.00001931'}, temp_paths=0.527984)
I0419 09:02:38.110473 137163463190272 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started.
I0419 09:02:38.110640 137291389007680 async_checkpointer.py:561] Finished blocking save. Time taken: 0.852561s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_03_dropout/checkpoints/0.
I0419 09:02:38.527562 137291389007680 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0419 09:02:38.527961 137163438012160 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save.
I0419 09:02:38.528124 137291389007680 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_03_dropout/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776589357.2560918, 'wait_for_prev_duration_secs': 6.341934204101562e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776589357.2580462, 'checkpointer_blocking_duration_secs': 0.8526999950408936, 'get_old_steps_start_time': 1776589358.1107676, 'get_old_steps_duration_secs': 2.6702880859375e-05, 'checkpoint_manager_blocking_start_time': 1776589357.2539039, 'checkpoint_manager_blocking_duration_secs': 1.2741801738739014}
I0419 09:02:38.528246 137291389007680 checkpointing.py:408] Started an asynchronous checkpoint save for step 0
I0419 09:02:38.528301 137291389007680 max_utils.py:750] 
Memstats: After params initialized:
I0419 09:02:38.528360 137291389007680 max_utils.py:756] 	Using (GB) 0.41 / 31.25 (1.312000%) on TPU_10(process=3,(2,2,0,0))
I0419 09:02:38.528395 137291389007680 max_utils.py:756] 	Using (GB) 0.41 / 31.25 (1.312000%) on TPU_11(process=3,(3,2,0,0))
I0419 09:02:38.528424 137291389007680 max_utils.py:756] 	Using (GB) 0.41 / 31.25 (1.312000%) on TPU_14(process=3,(2,3,0,0))
I0419 09:02:38.528450 137291389007680 max_utils.py:756] 	Using (GB) 0.41 / 31.25 (1.312000%) on TPU_15(process=3,(3,3,0,0))
I0419 09:02:38.792181 137291389007680 metric_logger.py:185] completed step: 0, seconds: 3.946, TFLOP/s/device: 0.203, Tokens/s/device: 32.434, total_weights: 4096, loss: 10.889
I0419 09:02:38.897509 137291389007680 metric_logger.py:185] completed step: 1, seconds: 1.766, TFLOP/s/device: 0.454, Tokens/s/device: 72.487, total_weights: 4096, loss: 10.902
I0419 09:02:39.301612 137291389007680 metric_logger.py:185] completed step: 2, seconds: 0.035, TFLOP/s/device: 22.719, Tokens/s/device: 3631.000, total_weights: 4096, loss: 9.956
I0419 09:02:39.372351 137291389007680 metric_logger.py:185] completed step: 3, seconds: 0.405, TFLOP/s/device: 1.979, Tokens/s/device: 316.223, total_weights: 4096, loss: 9.139
I0419 09:02:39.514444 137291389007680 metric_logger.py:185] completed step: 4, seconds: 0.075, TFLOP/s/device: 10.622, Tokens/s/device: 1697.680, total_weights: 4096, loss: 8.458
I0419 09:02:39.520284 137291389007680 metric_logger.py:185] completed step: 5, seconds: 0.071, TFLOP/s/device: 11.332, Tokens/s/device: 1811.082, total_weights: 4096, loss: 7.924
I0419 09:02:49.372355 137291389007680 metric_logger.py:185] completed step: 6, seconds: 0.143, TFLOP/s/device: 5.615, Tokens/s/device: 897.358, total_weights: 4096, loss: 7.553
I0419 09:02:49.444517 137291389007680 metric_logger.py:185] completed step: 7, seconds: 9.783, TFLOP/s/device: 0.082, Tokens/s/device: 13.084, total_weights: 4096, loss: 7.267
I0419 09:02:49.516996 137291389007680 metric_logger.py:185] completed step: 8, seconds: 0.075, TFLOP/s/device: 10.711, Tokens/s/device: 1711.962, total_weights: 4096, loss: 7.111
I0419 09:02:49.589073 137291389007680 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0419 09:02:49.589725 137291389007680 checkpointing.py:776] Waited 0.0006687641143798828 seconds for step 9 to finish before starting checkpointing.
I0419 09:02:49.591827 137291389007680 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0419 09:02:50.546663    2632 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0419 09:02:53.321293 137163471582976 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_03_dropout/checkpoints/0/items/array_metadatas/process_3
I0419 09:03:19.692799 137163463190272 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 37.514 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.107320070266724 s) (per-host)
I0419 09:03:19.692920 137163463190272 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.582337s.
I0419 09:03:28.910514 137163463190272 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 50.799915s.
I0419 09:03:28.910739 137163438012160 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save.
I0419 09:03:28.910800 137163438012160 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save.
I0419 09:03:28.910847 137163438012160 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0419 09:03:28.913549 137163438012160 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0419 09:03:28.913721 137291389007680 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0419 09:03:28.913857 137291389007680 checkpoint_manager.py:1458] Waiting for previous save to complete took 39.322037 seconds. If this number is high, consider checkpointing less frequently.
I0419 09:03:28.915732 137291389007680 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 9
I0419 09:03:28.919156 137291389007680 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_03_dropout/checkpoints/9.
I0419 09:03:29.452023 137291389007680 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0419 09:03:29.452127 137291389007680 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
I0419 09:03:29.497467 137291389007680 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.046420s
I0419 09:03:29.497635 137291389007680 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.137 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3002626895904541 s) (per-host)
I0419 09:03:29.497685 137291389007680 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.300323s (batch_requests_ready=0.251972s, total_serialization_initiated=0.048282s, others=0.000068s)
I0419 09:03:29.497774 137291389007680 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.304254s (all_items=0.000014s, per_item={'items': '0.00001431'}, temp_paths=0.304240)
I0419 09:03:29.498681 137163471582976 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started.
I0419 09:03:29.498837 137291389007680 async_checkpointer.py:561] Finished blocking save. Time taken: 0.583033s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_03_dropout/checkpoints/9.
I0419 09:03:29.924955 137291389007680 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0419 09:03:29.925347 137163438012160 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save.
I0419 09:03:29.925519 137291389007680 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_03_dropout/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776589369.5917957, 'wait_for_prev_duration_secs': 39.32203698158264, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776589408.9157715, 'checkpointer_blocking_duration_secs': 0.5831789970397949, 'get_old_steps_start_time': 1776589409.4989696, 'get_old_steps_duration_secs': 2.574920654296875e-05, 'checkpoint_manager_blocking_start_time': 1776589369.5899546, 'checkpoint_manager_blocking_duration_secs': 40.33552622795105}
I0419 09:03:29.925626 137291389007680 checkpointing.py:408] Started an asynchronous checkpoint save for step 9
I0419 09:03:29.925672 137291389007680 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0419 09:03:34.671882 137139048048384 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_03_dropout/checkpoints/9/items/array_metadatas/process_3
I0419 09:04:09.925243 137163471582976 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 38.784 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 40.72781586647034 s) (per-host)
I0419 09:04:09.925359 137163471582976 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.426566s.
I0419 09:04:19.325308 137163471582976 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 49.826501s.
I0419 09:04:19.325630 137163438012160 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save.
I0419 09:04:19.325749 137163438012160 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save.
I0419 09:04:19.325797 137163438012160 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0419 09:04:19.327578 137163438012160 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0419 09:04:19.327757 137291389007680 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0419 09:04:19.327897 137291389007680 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0419 09:04:19.328549 137291389007680 metric_logger.py:185] completed step: 9, seconds: 0.072, TFLOP/s/device: 11.111, Tokens/s/device: 1775.854, total_weights: 4096, loss: 6.981
Per train step:
 Total TFLOPs: 0.80 
 split as 99.60% learnable weight flops and 0.40% attention flops
XPK End: Sun Apr 19 09:04:31 UTC 2026
EXIT_CODE=0