MaxView

Timing Test Reference Updates
‹ 09_pdb_lt_1Case: 10_shardy_false10_shardy_true ›

Metrics: Linen vs NNX  ·  feat/nnx-set-defaults-true

MetricLinen  73213e044NNX  73213e044Diff (NNX − Linen)
Parameters1.104 billion1.104 billion—
Final loss8.17908.1750-0.004
TFLOP/s91.94292.830+0.888
Tok/s13858.513992.3+133.789
Avg s/step2.5362.631+0.095
Memory %1.381.380
JAX0.8.10.8.1—

Diff = NNX value − Linen value. Green = NNX improved. Red = NNX regressed.

Linen  ·  73213e044  ·  feat_nnx_set_defaults_true_20260422_154647  ·  full log 
XPK Start: Wed Apr 22 16:26:29 UTC 2026
2026-04-22 16:26:33.518591: 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 00:00:1776875193.531103      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1776875193.534751      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1776875193.545918      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776875193.545937      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776875193.545939      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776875193.545941      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-22 16:26:52.659073: 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)
I0422 16:26:53.184142 137983096088384 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-22 16:27:02,225:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0422 16:27:02.225820 137983096088384 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-22 16:27:02,228:jax._src.distributed:157: Connecting to JAX distributed service on mt-10-shardy-false-ajp01-slice-job-0-0.mt-10-shardy-false-ajp01:8482
I0422 16:27:02.228128 137983096088384 distributed.py:157] Connecting to JAX distributed service on mt-10-shardy-false-ajp01-slice-job-0-0.mt-10-shardy-false-ajp01:8482
I0422 16:27:03.132904 137983096088384 max_utils.py:284] Jax distributed system initialized!
I0422 16:27:10.040101 137983096088384 max_utils.py:800] System Information: Jax Version: 0.8.1
I0422 16:27:10.040206 137983096088384 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0422 16:27:10.040246 137983096088384 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
I0422 16:27:10.040282 137983096088384 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0422 16:27:10.660392 137983096088384 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0422 16:27:10.661007 137983096088384 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0422 16:27:10.661194 137983096088384 checkpointing.py:688] Setting up checkpoint logger...
I0422 16:27:10.661245 137983096088384 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0422 16:27:10.661289 137983096088384 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0422 16:27:10.661622 137983096088384 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 0x7d7e3716d100>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0422 16:27:13.721049 137983096088384 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0422 16:27:13.721289 137983096088384 checkpoint_manager.py:709] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d6bbb55db80>}, handler_registry=None
I0422 16:27:13.721523 137983096088384 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d6bbb55db80>` 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`.
I0422 16:27:13.721572 137983096088384 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d6bbb562a50>` 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`.
I0422 16:27:13.721609 137983096088384 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 0x7d6bbb55db80>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d6bbb55db80>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d6bbb562a50>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d6bbb562a50>}).
I0422 16:27:13.722132 137983096088384 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0422 16:27:13.722213 137983096088384 async_checkpointer.py:177] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7d6bbac64180> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0422 16:27:14.794959 137983096088384 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints
I0422 16:27:14.797120 137983096088384 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=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_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7d6bbb4c08f0>
I0422 16:27:14.797234 137983096088384 checkpointing.py:302] Checkpoint manager created!
I0422 16:27:15.683481 137983096088384 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0422 16:27:15.683584 137983096088384 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0422 16:27:16.050937 137983096088384 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0422 16:27:16.051027 137983096088384 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0422 16:27:16.067007 137983096088384 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0422 16:27:16.067064 137983096088384 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0422 16:27:16.089759 137983096088384 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0422 16:27:16.089826 137983096088384 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0422 16:27:16.105762 137983096088384 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0422 16:27:16.105825 137983096088384 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0422 16:27:16.121747 137983096088384 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0422 16:27:16.121805 137983096088384 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0422 16:27:16.146718 137983096088384 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0422 16:27:16.146783 137983096088384 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0422 16:27:16.166397 137983096088384 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0422 16:27:16.166463 137983096088384 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0422 16:27:16.359201 137983096088384 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0422 16:27:16.359311 137983096088384 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0422 16:27:17.759937 137983096088384 maxtext_utils.py:1740]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0422 16:27:17.760078 137983096088384 maxtext_utils.py:1740]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0422 16:27:17.760135 137983096088384 maxtext_utils.py:1740]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   PartitionSpec('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0422 16:27:17.760195 137983096088384 maxtext_utils.py:1740]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   PartitionSpec('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')
I0422 16:27:17.760247 137983096088384 maxtext_utils.py:1740]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  (None, None)
I0422 16:27:17.760292 137983096088384 maxtext_utils.py:1740]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   PartitionSpec('norm', 'layers')
    Physical:  (None, None)
I0422 16:27:17.760350 137983096088384 maxtext_utils.py:1740]  params/params/decoder/layers/self_attention/key/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0422 16:27:17.760407 137983096088384 maxtext_utils.py:1740]  params/params/decoder/layers/self_attention/out/kernel
    Shape:     float32[16,16,128,2048]
    Logical:   PartitionSpec('heads', 'layers', 'kv', 'embed')
    Physical:  (None, None, None, 'fsdp')
I0422 16:27:17.760450 137983096088384 maxtext_utils.py:1740]  params/params/decoder/layers/self_attention/query/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'q_heads', 'kv')
    Physical:  ('fsdp', None, None, None)
I0422 16:27:17.760488 137983096088384 maxtext_utils.py:1740]  params/params/decoder/layers/self_attention/value/kernel
    Shape:     float32[2048,16,16,128]
    Logical:   PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim')
    Physical:  ('fsdp', None, None, None)
I0422 16:27:17.760538 137983096088384 maxtext_utils.py:1740]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed', 'vocab')
    Physical:  ('fsdp', None)

I0422 16:27:17.760587 137983096088384 maxtext_utils.py:1740]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed')
    Physical:  (None, 'fsdp')
I0422 16:27:18.255928 137983096088384 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0422 16:27:18.256021 137983096088384 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0422 16:27:18.270984 137983096088384 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0422 16:27:18.271045 137983096088384 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0422 16:27:28.990761 137983096088384 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.
I0422 16:27:28.991609 137983096088384 metric_logger.py:289] number parameters: 1.104 billion
I0422 16:27:39.986252 137983096088384 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0422 16:27:40.254198 137983096088384 checkpointing.py:798] Waited 0.26793503761291504 seconds for step 0 to finish before starting checkpointing.
I0422 16:27:40.256580 137983096088384 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0422 16:27:40.258208 137983096088384 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 0
I0422 16:27:40.260917 137983096088384 async_checkpointer.py:452] [process=5] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints/0.
I0422 16:27:41.240067 137983096088384 signaling_client.py:364] Using JaxDistributedSignalingClient
I0422 16:27:41.241029 137983096088384 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0422 16:27:41.241089 137983096088384 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
I0422 16:27:41.485728 137983096088384 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.245739s
I0422 16:27:41.485899 137983096088384 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.941 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5244388580322266 s) (per-host)
I0422 16:27:41.485957 137983096088384 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.524506s (batch_requests_ready=0.275064s, total_serialization_initiated=0.249366s, others=0.000076s)
I0422 16:27:41.486047 137983096088384 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.528525s (all_items=0.000019s, per_item={'items': '0.00001931'}, temp_paths=0.528505)
I0422 16:27:41.487038 137859986089728 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started.
I0422 16:27:41.487212 137983096088384 async_checkpointer.py:561] Finished blocking save. Time taken: 1.228929s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints/0.
I0422 16:27:41.581784 137983096088384 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0422 16:27:41.582152 137855674337024 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0422 16:27:41.582317 137983096088384 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776875260.2565618, 'wait_for_prev_duration_secs': 6.222724914550781e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776875260.2582505, 'checkpointer_blocking_duration_secs': 1.229069471359253, 'get_old_steps_start_time': 1776875261.48734, 'get_old_steps_duration_secs': 2.4080276489257812e-05, 'checkpoint_manager_blocking_start_time': 1776875260.2547045, 'checkpoint_manager_blocking_duration_secs': 1.3275721073150635}
I0422 16:27:41.582425 137983096088384 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0422 16:27:41.582479 137983096088384 max_utils.py:750] 
Memstats: After params initialized:
I0422 16:27:41.582535 137983096088384 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0))
I0422 16:27:41.582570 137983096088384 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0))
I0422 16:27:41.582596 137983096088384 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0))
I0422 16:27:41.582622 137983096088384 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0))
I0422 16:27:41.847489 137983096088384 metric_logger.py:185] completed step: 0, seconds: 10.995, TFLOP/s/device: 1.236, Tokens/s/device: 186.275, total_weights: 65536, loss: 10.874
I0422 16:27:42.015884 137983096088384 metric_logger.py:185] completed step: 1, seconds: 1.860, TFLOP/s/device: 7.305, Tokens/s/device: 1101.151, total_weights: 65536, loss: 10.874
I0422 16:27:42.416666 137983096088384 metric_logger.py:185] completed step: 2, seconds: 0.021, TFLOP/s/device: 642.782, Tokens/s/device: 96887.123, total_weights: 65536, loss: 10.263
I0422 16:27:42.564398 137983096088384 metric_logger.py:185] completed step: 3, seconds: 0.401, TFLOP/s/device: 33.870, Tokens/s/device: 5105.309, total_weights: 65536, loss: 9.734
I0422 16:27:42.859880 137983096088384 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.847, Tokens/s/device: 13392.011, total_weights: 65536, loss: 9.278
I0422 16:27:42.865754 137983096088384 metric_logger.py:185] completed step: 5, seconds: 0.148, TFLOP/s/device: 92.106, Tokens/s/device: 13883.146, total_weights: 65536, loss: 8.892
I0422 16:28:02.658953 137983096088384 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.888, Tokens/s/device: 6916.746, total_weights: 65536, loss: 8.592
I0422 16:28:02.806688 137983096088384 metric_logger.py:185] completed step: 7, seconds: 19.646, TFLOP/s/device: 0.692, Tokens/s/device: 104.243, total_weights: 65536, loss: 8.384
I0422 16:28:02.954114 137983096088384 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.384, Tokens/s/device: 13472.886, total_weights: 65536, loss: 8.255
I0422 16:28:03.101175 137983096088384 checkpointing.py:794] Waiting for step 9 to finish before checkpoint...
I0422 16:28:03.101838 137983096088384 checkpointing.py:798] Waited 0.0006854534149169922 seconds for step 9 to finish before starting checkpointing.
I0422 16:28:03.103763 137983096088384 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0422 16:28:04.080481    2642 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0422 16:28:06.022377 137859994482432 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints/0/items/array_metadatas/process_5
I0422 16:28:38.082841 137859986089728 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 27.653 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 57.12134337425232 s) (per-host)
I0422 16:28:38.082941 137859986089728 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 56.595790s.
I0422 16:28:46.037692 137859986089728 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 64.550524s.
I0422 16:28:46.038008 137855674337024 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0422 16:28:46.038142 137855674337024 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0422 16:28:46.038200 137855674337024 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0422 16:28:46.039610 137855674337024 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0422 16:28:46.039814 137983096088384 checkpoint_manager.py:2036] [process=5][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0422 16:28:46.039960 137983096088384 checkpoint_manager.py:1458] Waiting for previous save to complete took 42.936200 seconds. If this number is high, consider checkpointing less frequently.
I0422 16:28:46.041670 137983096088384 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 9
I0422 16:28:46.045141 137983096088384 async_checkpointer.py:452] [process=5] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints/9.
I0422 16:28:46.619083 137983096088384 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0422 16:28:46.619182 137983096088384 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
I0422 16:28:46.663125 137983096088384 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.045233s
I0422 16:28:46.663295 137983096088384 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.074 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.30402565002441406 s) (per-host)
I0422 16:28:46.663344 137983096088384 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.304083s (batch_requests_ready=0.256910s, total_serialization_initiated=0.047107s, others=0.000066s)
I0422 16:28:46.663431 137983096088384 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.307893s (all_items=0.000015s, per_item={'items': '0.00001478'}, temp_paths=0.307878)
I0422 16:28:46.664283 137856236386048 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started.
I0422 16:28:46.664439 137983096088384 async_checkpointer.py:561] Finished blocking save. Time taken: 0.622697s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints/9.
I0422 16:28:46.694015 137983096088384 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0422 16:28:46.694310 137855674337024 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0422 16:28:46.694452 137983096088384 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776875283.103735, 'wait_for_prev_duration_secs': 42.93619966506958, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776875326.0417094, 'checkpointer_blocking_duration_secs': 0.622838020324707, 'get_old_steps_start_time': 1776875326.6645656, 'get_old_steps_duration_secs': 2.5510787963867188e-05, 'checkpoint_manager_blocking_start_time': 1776875283.1020617, 'checkpoint_manager_blocking_duration_secs': 43.5923535823822}
I0422 16:28:46.694557 137983096088384 checkpointing.py:409] Started an asynchronous checkpoint save for step 9
I0422 16:28:46.694604 137983096088384 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0422 16:28:51.376443 137859994482432 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints/9/items/array_metadatas/process_5
I0422 16:29:28.201958 137856236386048 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 37.751 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.8426456451416 s) (per-host)
I0422 16:29:28.202073 137856236386048 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.537681s.
I0422 16:29:35.105921 137856236386048 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 48.441514s.
I0422 16:29:35.106171 137855674337024 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0422 16:29:35.106235 137855674337024 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0422 16:29:35.106280 137855674337024 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0422 16:29:35.107958 137855674337024 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0422 16:29:35.108121 137983096088384 checkpoint_manager.py:2036] [process=5][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0422 16:29:35.108264 137983096088384 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0422 16:29:35.108889 137983096088384 metric_logger.py:185] completed step: 9, seconds: 0.148, TFLOP/s/device: 91.942, Tokens/s/device: 13858.532, total_weights: 65536, loss: 8.179
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Wed Apr 22 16:29:45 UTC 2026
EXIT_CODE=0
NNX  ·  73213e044  ·  feat_nnx_set_defaults_true_20260422_154647  ·  full log 
XPK Start: Wed Apr 22 18:04:05 UTC 2026
2026-04-22 18:04:09.413691: 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 00:00:1776881049.426492      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1776881049.430223      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1776881049.441736      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776881049.441753      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776881049.441756      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776881049.441758      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-22 18:04:28.665590: 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)
I0422 18:04:29.198515 137460067555136 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-22 18:04:38,240:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0422 18:04:38.240710 137460067555136 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-22 18:04:38,243:jax._src.distributed:157: Connecting to JAX distributed service on mt-10-shardy-false-4vm1u-slice-job-0-0.mt-10-shardy-false-4vm1u:8482
I0422 18:04:38.243032 137460067555136 distributed.py:157] Connecting to JAX distributed service on mt-10-shardy-false-4vm1u-slice-job-0-0.mt-10-shardy-false-4vm1u:8482
I0422 18:04:39.174724 137460067555136 max_utils.py:284] Jax distributed system initialized!
I0422 18:04:46.194042 137460067555136 max_utils.py:800] System Information: Jax Version: 0.8.1
I0422 18:04:46.194168 137460067555136 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0422 18:04:46.194211 137460067555136 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
I0422 18:04:46.194246 137460067555136 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0422 18:04:46.820435 137460067555136 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0422 18:04:46.937961 137460067555136 checkpointing.py:688] Setting up checkpoint logger...
I0422 18:04:46.938078 137460067555136 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0422 18:04:46.938136 137460067555136 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0422 18:04:46.938355 137460067555136 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 0x7d046e178170>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0422 18:04:49.972575 137460067555136 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0422 18:04:49.972812 137460067555136 checkpoint_manager.py:709] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cefec3a4350>}, handler_registry=None
I0422 18:04:49.973045 137460067555136 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cefec3a4350>` 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`.
I0422 18:04:49.973119 137460067555136 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cefc457dbb0>` 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`.
I0422 18:04:49.973158 137460067555136 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 0x7cefec3a4350>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cefec3a4350>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cefc457dbb0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cefc457dbb0>}).
I0422 18:04:49.973466 137460067555136 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0422 18:04:49.973549 137460067555136 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7cefc44fcf40> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0422 18:04:51.110426 137460067555136 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints
I0422 18:04:51.553947 137460067555136 checkpoint_manager.py:929] [process=6][thread=MainThread] CheckpointManager created,  primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=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/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7cefec3ee210>
I0422 18:04:51.554144 137460067555136 checkpointing.py:302] Checkpoint manager created!
I0422 18:04:52.502566 137460067555136 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0422 18:04:52.502680 137460067555136 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0422 18:04:53.884403 137460067555136 maxtext_utils.py:1749]  decoder/decoder_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0422 18:04:53.884521 137460067555136 maxtext_utils.py:1749]  decoder/layers/mlp/wi_0/kernel/value
    Shape:     float32[2048,16,7168]
    Physical:  ('fsdp', None, None)
I0422 18:04:53.884576 137460067555136 maxtext_utils.py:1749]  decoder/layers/mlp/wi_1/kernel/value
    Shape:     float32[2048,16,7168]
    Physical:  ('fsdp', None, None)
I0422 18:04:53.884623 137460067555136 maxtext_utils.py:1749]  decoder/layers/mlp/wo/kernel/value
    Shape:     float32[7168,16,2048]
    Physical:  (None, None, 'fsdp')
I0422 18:04:53.884662 137460067555136 maxtext_utils.py:1749]  decoder/layers/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048,16]
    Physical:  (None, None)
I0422 18:04:53.884699 137460067555136 maxtext_utils.py:1749]  decoder/layers/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048,16]
    Physical:  (None, None)
I0422 18:04:53.884737 137460067555136 maxtext_utils.py:1749]  decoder/layers/self_attention/key/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0422 18:04:53.884775 137460067555136 maxtext_utils.py:1749]  decoder/layers/self_attention/out/kernel/value
    Shape:     float32[16,16,128,2048]
    Physical:  (None, None, None, 'fsdp')
I0422 18:04:53.884810 137460067555136 maxtext_utils.py:1749]  decoder/layers/self_attention/query/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0422 18:04:53.884845 137460067555136 maxtext_utils.py:1749]  decoder/layers/self_attention/value/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0422 18:04:53.884878 137460067555136 maxtext_utils.py:1749]  decoder/logits_dense/kernel/value
    Shape:     float32[2048,32000]
    Physical:  ('fsdp', None)

I0422 18:04:53.884911 137460067555136 maxtext_utils.py:1749]  token_embedder/embedding/value
    Shape:     float32[32000,2048]
    Physical:  (None, 'fsdp')
I0422 18:04:54.281016 137460067555136 nnx_decoders.py:616] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0422 18:04:54.281188 137460067555136 nnx_decoders.py:616] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0422 18:04:54.286686 137460067555136 nnx_decoders.py:616] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0422 18:04:54.286783 137460067555136 nnx_decoders.py:616] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0422 18:04:54.446623 137460067555136 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0422 18:04:54.446760 137460067555136 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0422 18:04:54.608082 137460067555136 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0422 18:04:54.608240 137460067555136 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0422 18:04:54.779300 137460067555136 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0422 18:04:54.779442 137460067555136 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0422 18:04:54.929375 137460067555136 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0422 18:04:54.929513 137460067555136 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0422 18:04:55.075795 137460067555136 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0422 18:04:55.075932 137460067555136 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0422 18:04:55.236580 137460067555136 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0422 18:04:55.236720 137460067555136 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0422 18:04:55.259672 137460067555136 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0422 18:04:55.259769 137460067555136 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0422 18:05:06.371985 137460067555136 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.
I0422 18:05:06.374427 137460067555136 metric_logger.py:289] number parameters: 1.104 billion
I0422 18:05:17.184775 137460067555136 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0422 18:05:17.338980 137460067555136 checkpointing.py:798] Waited 0.1541900634765625 seconds for step 0 to finish before starting checkpointing.
I0422 18:05:17.341678 137460067555136 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0422 18:05:17.343359 137460067555136 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 0
I0422 18:05:17.346184 137460067555136 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints/0.
I0422 18:05:18.329746 137460067555136 signaling_client.py:364] Using JaxDistributedSignalingClient
I0422 18:05:18.330849 137460067555136 jax_array_handlers.py:358] Scheduling D2H of 69 prioritized jax.Array.
I0422 18:05:18.330911 137460067555136 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
I0422 18:05:18.573704 137460067555136 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.245146s
I0422 18:05:18.573873 137460067555136 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.950 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5229945182800293 s) (per-host)
I0422 18:05:18.573930 137460067555136 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.523062s (batch_requests_ready=0.270392s, total_serialization_initiated=0.252594s, others=0.000075s)
I0422 18:05:18.574028 137460067555136 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.527231s (all_items=0.000019s, per_item={'items': '0.00001931'}, temp_paths=0.527211)
I0422 18:05:18.574985 137337325471488 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0422 18:05:18.575187 137460067555136 async_checkpointer.py:561] Finished blocking save. Time taken: 1.231758s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints/0.
I0422 18:05:19.043124 137460067555136 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0422 18:05:19.043504 137333005326080 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0422 18:05:19.043685 137460067555136 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776881117.3416617, 'wait_for_prev_duration_secs': 6.29425048828125e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776881117.3433967, 'checkpointer_blocking_duration_secs': 1.2318956851959229, 'get_old_steps_start_time': 1776881118.575315, 'get_old_steps_duration_secs': 2.4318695068359375e-05, 'checkpoint_manager_blocking_start_time': 1776881117.3394773, 'checkpoint_manager_blocking_duration_secs': 1.7041640281677246}
I0422 18:05:19.043798 137460067555136 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0422 18:05:19.043874 137460067555136 max_utils.py:750] 
Memstats: After params initialized:
I0422 18:05:19.043930 137460067555136 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0422 18:05:19.043965 137460067555136 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0422 18:05:19.043993 137460067555136 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0422 18:05:19.044018 137460067555136 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0422 18:05:19.311650 137460067555136 metric_logger.py:185] completed step: 0, seconds: 10.809, TFLOP/s/device: 1.257, Tokens/s/device: 189.467, total_weights: 65536, loss: 10.874
I0422 18:05:19.476243 137460067555136 metric_logger.py:185] completed step: 1, seconds: 2.125, TFLOP/s/device: 6.394, Tokens/s/device: 963.731, total_weights: 65536, loss: 10.874
I0422 18:05:19.920889 137460067555136 metric_logger.py:185] completed step: 2, seconds: 0.019, TFLOP/s/device: 719.657, Tokens/s/device: 108474.576, total_weights: 65536, loss: 10.262
I0422 18:05:20.068348 137460067555136 metric_logger.py:185] completed step: 3, seconds: 0.436, TFLOP/s/device: 31.185, Tokens/s/device: 4700.579, total_weights: 65536, loss: 9.732
I0422 18:05:20.363834 137460067555136 metric_logger.py:185] completed step: 4, seconds: 0.163, TFLOP/s/device: 83.608, Tokens/s/device: 12602.379, total_weights: 65536, loss: 9.274
I0422 18:05:20.371090 137460067555136 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.241, Tokens/s/device: 13903.504, total_weights: 65536, loss: 8.888
I0422 18:05:40.711605 137460067555136 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.920, Tokens/s/device: 6921.608, total_weights: 65536, loss: 8.589
I0422 18:05:40.859116 137460067555136 metric_logger.py:185] completed step: 7, seconds: 20.194, TFLOP/s/device: 0.673, Tokens/s/device: 101.415, total_weights: 65536, loss: 8.380
I0422 18:05:41.006559 137460067555136 metric_logger.py:185] completed step: 8, seconds: 0.154, TFLOP/s/device: 88.231, Tokens/s/device: 13299.133, total_weights: 65536, loss: 8.251
I0422 18:05:41.154882 137460067555136 checkpointing.py:794] Waiting for step 9 to finish before checkpoint...
I0422 18:05:41.155475 137460067555136 checkpointing.py:798] Waited 0.0006034374237060547 seconds for step 9 to finish before starting checkpointing.
I0422 18:05:41.158241 137460067555136 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0422 18:05:41.575673    2600 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0422 18:05:43.971422 137337333864192 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints/0/items/array_metadatas/process_6
I0422 18:06:15.803615 137337325471488 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 27.351 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 57.75269627571106 s) (per-host)
I0422 18:06:15.803731 137337325471488 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 57.228606s.
I0422 18:06:23.112175 137337325471488 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 64.537036s.
I0422 18:06:23.112465 137333005326080 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0422 18:06:23.112605 137333005326080 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0422 18:06:23.112664 137333005326080 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0422 18:06:23.113955 137333005326080 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0422 18:06:23.114182 137460067555136 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0422 18:06:23.114316 137460067555136 checkpoint_manager.py:1458] Waiting for previous save to complete took 41.956079 seconds. If this number is high, consider checkpointing less frequently.
I0422 18:06:23.117702 137460067555136 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 9
I0422 18:06:23.121089 137460067555136 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints/9.
I0422 18:06:23.675139 137460067555136 jax_array_handlers.py:358] Scheduling D2H of 69 prioritized jax.Array.
I0422 18:06:23.675294 137460067555136 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
I0422 18:06:23.723170 137460067555136 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.050021s
I0422 18:06:23.723340 137460067555136 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.822 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.31987667083740234 s) (per-host)
I0422 18:06:23.723392 137460067555136 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.319937s (batch_requests_ready=0.265152s, total_serialization_initiated=0.054717s, others=0.000068s)
I0422 18:06:23.723489 137460067555136 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.323864s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.323848)
I0422 18:06:23.724419 137334087460608 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0422 18:06:23.724599 137460067555136 async_checkpointer.py:561] Finished blocking save. Time taken: 0.606823s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints/9.
I0422 18:06:23.726446 137460067555136 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0422 18:06:23.726706 137333005326080 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0422 18:06:23.726846 137460067555136 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776881141.1582124, 'wait_for_prev_duration_secs': 41.95607948303223, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776881183.1177425, 'checkpointer_blocking_duration_secs': 0.6069681644439697, 'get_old_steps_start_time': 1776881183.7247312, 'get_old_steps_duration_secs': 2.384185791015625e-05, 'checkpoint_manager_blocking_start_time': 1776881141.1557002, 'checkpoint_manager_blocking_duration_secs': 42.57111072540283}
I0422 18:06:23.727014 137460067555136 checkpointing.py:409] Started an asynchronous checkpoint save for step 9
I0422 18:06:23.727060 137460067555136 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0422 18:06:28.493867 137337333864192 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_10_shardy_false/checkpoints/9/items/array_metadatas/process_6
I0422 18:07:04.990595 137334087460608 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 37.983 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.58709120750427 s) (per-host)
I0422 18:07:04.990712 137334087460608 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.266175s.
I0422 18:07:13.627832 137334087460608 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 49.903278s.
I0422 18:07:13.628131 137333005326080 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0422 18:07:13.628249 137333005326080 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0422 18:07:13.628294 137333005326080 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0422 18:07:13.629625 137333005326080 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0422 18:07:13.629801 137460067555136 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0422 18:07:13.629971 137460067555136 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0422 18:07:13.630600 137460067555136 metric_logger.py:185] completed step: 9, seconds: 0.146, TFLOP/s/device: 92.830, Tokens/s/device: 13992.321, total_weights: 65536, loss: 8.175
Per train step:
 Total TFLOPs: 13.59 
 split as 93.93% learnable weight flops and 6.07% attention flops
XPK End: Wed Apr 22 18:07:25 UTC 2026
EXIT_CODE=0