MaxView

‹ 01_baseCase: 02_synthetic03_dropout ›

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

MetricLinen  574ad3fb9NNX  574ad3fb9Diff (NNX − Linen)
Parameters1.104 billion1.104 billion
Final loss8.17908.1750-0.004
TFLOP/s92.20492.318+0.114
Tok/s13897.913915.2+17.281
Avg s/step2.5502.590+0.04
Memory %1.381.380
JAX0.8.10.8.1

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

Linen  ·  574ad3fb9  ·  feat_nnx_set_defaults_true_20260419_025828  ·  full log
XPK Start: Sun Apr 19 03:20:10 UTC 2026
2026-04-19 03:20:14.541171: 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 00:00:1776568814.553381      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1776568814.557023      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1776568814.568198      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776568814.568218      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776568814.568220      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776568814.568222      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-19 03:20:34.069184: 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 03:20:34.595479 134789606065984 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-19 03:20:43,635:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0419 03:20:43.635188 134789606065984 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-19 03:20:43,637:jax._src.distributed:157: Connecting to JAX distributed service on mt-02-synthetic-601k9-slice-job-0-0.mt-02-synthetic-601k9:8482
I0419 03:20:43.637505 134789606065984 distributed.py:157] Connecting to JAX distributed service on mt-02-synthetic-601k9-slice-job-0-0.mt-02-synthetic-601k9:8482
I0419 03:20:45.702836 134789606065984 max_utils.py:284] Jax distributed system initialized!
I0419 03:20:52.728485 134789606065984 max_utils.py:800] System Information: Jax Version: 0.8.1
I0419 03:20:52.728593 134789606065984 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0419 03:20:52.728635 134789606065984 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 03:20:52.728671 134789606065984 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0419 03:20:53.346567 134789606065984 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0419 03:20:53.347158 134789606065984 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0419 03:20:53.347357 134789606065984 checkpointing.py:688] Setting up checkpoint logger...
I0419 03:20:53.347410 134789606065984 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0419 03:20:53.347451 134789606065984 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0419 03:20:53.347829 134789606065984 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 0x7a969ab75250>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0419 03:20:57.420400 134789606065984 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0419 03:20:57.420636 134789606065984 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 0x7a823845d3a0>}, handler_registry=None
I0419 03:20:57.420883 134789606065984 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a823845d3a0>` 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 03:20:57.420934 134789606065984 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a823845a750>` 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 03:20:57.420970 134789606065984 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 0x7a823845d3a0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a823845d3a0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a823845a750>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a823845a750>}).
I0419 03:20:57.421480 134789606065984 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0419 03:20:57.421559 134789606065984 async_checkpointer.py:177] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7a8238208180> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0419 03:20:59.017167 134789606065984 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints
I0419 03:20:59.442145 134789606065984 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_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7a8238685670>
I0419 03:20:59.442350 134789606065984 checkpointing.py:302] Checkpoint manager created!
I0419 03:21:00.331946 134789606065984 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0419 03:21:00.332052 134789606065984 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0419 03:21:00.697630 134789606065984 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0419 03:21:00.697725 134789606065984 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0419 03:21:00.713572 134789606065984 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0419 03:21:00.713641 134789606065984 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0419 03:21:00.736272 134789606065984 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0419 03:21:00.736338 134789606065984 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0419 03:21:00.752228 134789606065984 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0419 03:21:00.752300 134789606065984 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0419 03:21:00.768185 134789606065984 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0419 03:21:00.768258 134789606065984 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0419 03:21:00.792850 134789606065984 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0419 03:21:00.792920 134789606065984 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0419 03:21:00.812480 134789606065984 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0419 03:21:00.812550 134789606065984 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0419 03:21:01.000218 134789606065984 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0419 03:21:01.000343 134789606065984 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0419 03:21:02.434762 134789606065984 maxtext_utils.py:1740]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0419 03:21:02.434911 134789606065984 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)
I0419 03:21:02.434967 134789606065984 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)
I0419 03:21:02.435026 134789606065984 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')
I0419 03:21:02.435078 134789606065984 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)
I0419 03:21:02.435116 134789606065984 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)
I0419 03:21:02.435167 134789606065984 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)
I0419 03:21:02.435220 134789606065984 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')
I0419 03:21:02.435275 134789606065984 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)
I0419 03:21:02.435316 134789606065984 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)
I0419 03:21:02.435366 134789606065984 maxtext_utils.py:1740]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed', 'vocab')
    Physical:  ('fsdp', None)
I0419 03:21:02.435415 134789606065984 maxtext_utils.py:1740]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed')
    Physical:  (None, 'fsdp')

I0419 03:21:02.932815 134789606065984 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0419 03:21:02.932913 134789606065984 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0419 03:21:02.947887 134789606065984 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0419 03:21:02.947946 134789606065984 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0419 03:21:13.622918 134789606065984 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.
I0419 03:21:13.623817 134789606065984 metric_logger.py:289] number parameters: 1.104 billion
I0419 03:21:24.767302 134789606065984 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0419 03:21:24.922299 134789606065984 checkpointing.py:798] Waited 0.15494871139526367 seconds for step 0 to finish before starting checkpointing.
I0419 03:21:24.924805 134789606065984 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0419 03:21:24.926606 134789606065984 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 0
I0419 03:21:24.929439 134789606065984 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints/0.
I0419 03:21:26.346125 134789606065984 signaling_client.py:364] Using JaxDistributedSignalingClient
I0419 03:21:26.347121 134789606065984 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0419 03:21:26.347177 134789606065984 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 03:21:26.599310 134789606065984 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.253260s
I0419 03:21:26.599484 134789606065984 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.650 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.9347474575042725 s) (per-host)
I0419 03:21:26.599537 134789606065984 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.934810s (batch_requests_ready=0.677885s, total_serialization_initiated=0.256855s, others=0.000070s)
I0419 03:21:26.599628 134789606065984 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.939213s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=0.939194)
I0419 03:21:26.600572 134666778552064 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started.
I0419 03:21:26.600761 134789606065984 async_checkpointer.py:561] Finished blocking save. Time taken: 1.674087s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints/0.
I0419 03:21:26.643737 134789606065984 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0419 03:21:26.644079 134661955106560 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save.
I0419 03:21:26.644251 134789606065984 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776568884.9247873, 'wait_for_prev_duration_secs': 6.413459777832031e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776568884.9266446, 'checkpointer_blocking_duration_secs': 1.6742265224456787, 'get_old_steps_start_time': 1776568886.6008897, 'get_old_steps_duration_secs': 2.4080276489257812e-05, 'checkpoint_manager_blocking_start_time': 1776568884.922904, 'checkpoint_manager_blocking_duration_secs': 1.7212984561920166}
I0419 03:21:26.644357 134789606065984 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0419 03:21:26.644408 134789606065984 max_utils.py:750] 
Memstats: After params initialized:
I0419 03:21:26.644459 134789606065984 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_10(process=3,(2,2,0,0))
I0419 03:21:26.644493 134789606065984 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_11(process=3,(3,2,0,0))
I0419 03:21:26.644520 134789606065984 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_14(process=3,(2,3,0,0))
I0419 03:21:26.644545 134789606065984 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_15(process=3,(3,3,0,0))
I0419 03:21:26.913165 134789606065984 metric_logger.py:185] completed step: 0, seconds: 11.143, TFLOP/s/device: 1.219, Tokens/s/device: 183.786, total_weights: 65536, loss: 10.874
I0419 03:21:27.089809 134789606065984 metric_logger.py:185] completed step: 1, seconds: 2.145, TFLOP/s/device: 6.336, Tokens/s/device: 954.962, total_weights: 65536, loss: 10.874
I0419 03:21:27.466427 134789606065984 metric_logger.py:185] completed step: 2, seconds: 0.029, TFLOP/s/device: 466.575, Tokens/s/device: 70327.255, total_weights: 65536, loss: 10.263
I0419 03:21:27.614052 134789606065984 metric_logger.py:185] completed step: 3, seconds: 0.377, TFLOP/s/device: 36.051, Tokens/s/device: 5433.975, total_weights: 65536, loss: 9.734
I0419 03:21:27.909728 134789606065984 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.698, Tokens/s/device: 13369.543, total_weights: 65536, loss: 9.278
I0419 03:21:27.915311 134789606065984 metric_logger.py:185] completed step: 5, seconds: 0.148, TFLOP/s/device: 91.855, Tokens/s/device: 13845.415, total_weights: 65536, loss: 8.892
I0419 03:21:47.567597 134789606065984 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.971, Tokens/s/device: 6929.266, total_weights: 65536, loss: 8.592
I0419 03:21:47.715138 134789606065984 metric_logger.py:185] completed step: 7, seconds: 19.505, TFLOP/s/device: 0.697, Tokens/s/device: 104.997, total_weights: 65536, loss: 8.384
I0419 03:21:47.862983 134789606065984 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.159, Tokens/s/device: 13439.026, total_weights: 65536, loss: 8.255
I0419 03:21:48.010313 134789606065984 checkpointing.py:794] Waiting for step 9 to finish before checkpoint...
I0419 03:21:48.010964 134789606065984 checkpointing.py:798] Waited 0.0006723403930664062 seconds for step 9 to finish before starting checkpointing.
I0419 03:21:48.013509 134789606065984 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0419 03:21:49.512957    2619 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0419 03:21:51.269009 134666786944768 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints/0/items/array_metadatas/process_3
I0419 03:22:23.815577 134666778552064 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 27.164 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 58.15079998970032 s) (per-host)
I0419 03:22:23.815695 134666778552064 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 57.214995s.
I0419 03:22:34.943951 134666778552064 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 68.343226s.
I0419 03:22:34.944212 134661955106560 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save.
I0419 03:22:34.944380 134661955106560 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save.
I0419 03:22:34.944431 134661955106560 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0419 03:22:34.946129 134661955106560 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0419 03:22:34.946346 134789606065984 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 03:22:34.946498 134789606065984 checkpoint_manager.py:1458] Waiting for previous save to complete took 46.932993 seconds. If this number is high, consider checkpointing less frequently.
I0419 03:22:34.948254 134789606065984 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 9
I0419 03:22:34.951735 134789606065984 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints/9.
I0419 03:22:36.786077 134789606065984 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0419 03:22:36.786181 134789606065984 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 03:22:36.830171 134789606065984 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.045336s
I0419 03:22:36.830341 134789606065984 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.072 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.7446529865264893 s) (per-host)
I0419 03:22:36.830392 134789606065984 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.744714s (batch_requests_ready=0.697464s, total_serialization_initiated=0.047183s, others=0.000067s)
I0419 03:22:36.830472 134789606065984 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.748774s (all_items=0.000015s, per_item={'items': '0.00001454'}, temp_paths=0.748760)
I0419 03:22:36.831393 134662525548288 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started.
I0419 03:22:36.831552 134789606065984 async_checkpointer.py:561] Finished blocking save. Time taken: 1.883228s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints/9.
I0419 03:22:36.833451 134789606065984 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0419 03:22:36.833724 134661955106560 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save.
I0419 03:22:36.833863 134789606065984 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776568908.0134807, 'wait_for_prev_duration_secs': 46.932992696762085, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776568954.9482937, 'checkpointer_blocking_duration_secs': 1.8833637237548828, 'get_old_steps_start_time': 1776568956.8316743, 'get_old_steps_duration_secs': 2.5272369384765625e-05, 'checkpoint_manager_blocking_start_time': 1776568908.0111928, 'checkpoint_manager_blocking_duration_secs': 48.822635650634766}
I0419 03:22:36.834021 134789606065984 checkpointing.py:409] Started an asynchronous checkpoint save for step 9
I0419 03:22:36.834067 134789606065984 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0419 03:22:42.377150 134666786944768 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints/9/items/array_metadatas/process_3
I0419 03:23:18.131028 134662525548288 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 37.569 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.045305013656616 s) (per-host)
I0419 03:23:18.131130 134662525548288 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.299625s.
I0419 03:23:27.278446 134662525548288 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 50.446924s.
I0419 03:23:27.278714 134661955106560 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save.
I0419 03:23:27.278833 134661955106560 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save.
I0419 03:23:27.278892 134661955106560 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0419 03:23:27.280648 134661955106560 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0419 03:23:27.280834 134789606065984 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 03:23:27.280981 134789606065984 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0419 03:23:27.281625 134789606065984 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.204, Tokens/s/device: 13897.937, 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: Sun Apr 19 03:23:40 UTC 2026
EXIT_CODE=0
NNX  ·  574ad3fb9  ·  feat_nnx_set_defaults_true_20260419_025828  ·  full log
XPK Start: Sun Apr 19 04:59:14 UTC 2026
2026-04-19 04:59:18.285046: 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 00:00:1776574758.297419      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1776574758.301203      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1776574758.312343      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776574758.312363      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776574758.312365      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1776574758.312367      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-19 04:59:37.458749: 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 04:59:37.990556 132540212778816 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-19 04:59:47,030:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0419 04:59:47.030530 132540212778816 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-19 04:59:47,032:jax._src.distributed:157: Connecting to JAX distributed service on mt-02-synthetic-tj2y8-slice-job-0-0.mt-02-synthetic-tj2y8:8482
I0419 04:59:47.032905 132540212778816 distributed.py:157] Connecting to JAX distributed service on mt-02-synthetic-tj2y8-slice-job-0-0.mt-02-synthetic-tj2y8:8482
I0419 04:59:48.914491 132540212778816 max_utils.py:284] Jax distributed system initialized!
I0419 04:59:55.747631 132540212778816 max_utils.py:800] System Information: Jax Version: 0.8.1
I0419 04:59:55.747739 132540212778816 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0419 04:59:55.747780 132540212778816 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 04:59:55.747815 132540212778816 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0419 04:59:56.364531 132540212778816 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0419 04:59:56.479824 132540212778816 checkpointing.py:688] Setting up checkpoint logger...
I0419 04:59:56.479951 132540212778816 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0419 04:59:56.479997 132540212778816 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0419 04:59:56.480203 132540212778816 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 0x788ae5755e50>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0419 05:00:00.733654 132540212778816 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0419 05:00:00.734008 132540212778816 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 0x7877d32d4fe0>}, handler_registry=None
I0419 05:00:00.734264 132540212778816 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7877d32d4fe0>` 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 05:00:00.734314 132540212778816 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7877d32d4da0>` 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 05:00:00.734351 132540212778816 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 0x7877d32d4fe0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7877d32d4fe0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7877d32d4da0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7877d32d4da0>}).
I0419 05:00:00.734678 132540212778816 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0419 05:00:00.734754 132540212778816 async_checkpointer.py:177] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7877d0ad0f40> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0419 05:00:03.421878 132540212778816 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints
I0419 05:00:03.890930 132540212778816 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/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7877d13c8d70>
I0419 05:00:03.891110 132540212778816 checkpointing.py:302] Checkpoint manager created!
I0419 05:00:04.820084 132540212778816 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0419 05:00:04.820207 132540212778816 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0419 05:00:06.220580 132540212778816 maxtext_utils.py:1749]  decoder/decoder_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0419 05:00:06.220711 132540212778816 maxtext_utils.py:1749]  decoder/layers/mlp/wi_0/kernel/value
    Shape:     float32[2048,16,7168]
    Physical:  ('fsdp', None, None)
I0419 05:00:06.220767 132540212778816 maxtext_utils.py:1749]  decoder/layers/mlp/wi_1/kernel/value
    Shape:     float32[2048,16,7168]
    Physical:  ('fsdp', None, None)
I0419 05:00:06.220812 132540212778816 maxtext_utils.py:1749]  decoder/layers/mlp/wo/kernel/value
    Shape:     float32[7168,16,2048]
    Physical:  (None, None, 'fsdp')
I0419 05:00:06.220851 132540212778816 maxtext_utils.py:1749]  decoder/layers/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048,16]
    Physical:  (None, None)
I0419 05:00:06.220886 132540212778816 maxtext_utils.py:1749]  decoder/layers/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048,16]
    Physical:  (None, None)
I0419 05:00:06.220924 132540212778816 maxtext_utils.py:1749]  decoder/layers/self_attention/key/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0419 05:00:06.220962 132540212778816 maxtext_utils.py:1749]  decoder/layers/self_attention/out/kernel/value
    Shape:     float32[16,16,128,2048]
    Physical:  (None, None, None, 'fsdp')
I0419 05:00:06.220999 132540212778816 maxtext_utils.py:1749]  decoder/layers/self_attention/query/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0419 05:00:06.221034 132540212778816 maxtext_utils.py:1749]  decoder/layers/self_attention/value/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0419 05:00:06.221068 132540212778816 maxtext_utils.py:1749]  decoder/logits_dense/kernel/value
    Shape:     float32[2048,32000]
    Physical:  ('fsdp', None)

I0419 05:00:06.221101 132540212778816 maxtext_utils.py:1749]  token_embedder/embedding/value
    Shape:     float32[32000,2048]
    Physical:  (None, 'fsdp')
I0419 05:00:06.621770 132540212778816 nnx_decoders.py:616] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0419 05:00:06.621936 132540212778816 nnx_decoders.py:616] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0419 05:00:06.627362 132540212778816 nnx_decoders.py:616] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0419 05:00:06.627463 132540212778816 nnx_decoders.py:616] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0419 05:00:06.783345 132540212778816 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0419 05:00:06.783480 132540212778816 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0419 05:00:06.943272 132540212778816 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0419 05:00:06.943409 132540212778816 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0419 05:00:07.113028 132540212778816 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0419 05:00:07.113176 132540212778816 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0419 05:00:07.269972 132540212778816 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0419 05:00:07.270116 132540212778816 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0419 05:00:07.427519 132540212778816 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0419 05:00:07.427660 132540212778816 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0419 05:00:07.594815 132540212778816 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0419 05:00:07.594962 132540212778816 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0419 05:00:07.617806 132540212778816 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0419 05:00:07.617906 132540212778816 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0419 05:00:18.729788 132540212778816 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.
I0419 05:00:18.732134 132540212778816 metric_logger.py:289] number parameters: 1.104 billion
I0419 05:00:29.858728 132540212778816 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0419 05:00:30.009023 132540212778816 checkpointing.py:798] Waited 0.1502833366394043 seconds for step 0 to finish before starting checkpointing.
I0419 05:00:30.011529 132540212778816 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0419 05:00:30.013437 132540212778816 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 0
I0419 05:00:30.016403 132540212778816 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints/0.
I0419 05:00:31.447193 132540212778816 signaling_client.py:364] Using JaxDistributedSignalingClient
I0419 05:00:31.448273 132540212778816 jax_array_handlers.py:358] Scheduling D2H of 69 prioritized jax.Array.
I0419 05:00:31.448330 132540212778816 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 05:00:31.694811 132540212778816 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.248749s
I0419 05:00:31.694990 132540212778816 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.057 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.504608154296875 s) (per-host)
I0419 05:00:31.695043 132540212778816 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.504671s (batch_requests_ready=0.249282s, total_serialization_initiated=0.255317s, others=0.000072s)
I0419 05:00:31.695144 132540212778816 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.508784s (all_items=0.000020s, per_item={'items': '0.00001955'}, temp_paths=0.508765)
I0419 05:00:31.696137 132418287671040 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started.
I0419 05:00:31.696261 132540212778816 async_checkpointer.py:561] Finished blocking save. Time taken: 1.682739s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints/0.
I0419 05:00:32.140397 132540212778816 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0419 05:00:32.140789 132414579930880 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save.
I0419 05:00:32.140951 132540212778816 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776574830.011497, 'wait_for_prev_duration_secs': 0.00012731552124023438, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776574830.0134766, 'checkpointer_blocking_duration_secs': 1.6828949451446533, 'get_old_steps_start_time': 1776574831.6963937, 'get_old_steps_duration_secs': 2.6702880859375e-05, 'checkpoint_manager_blocking_start_time': 1776574830.009518, 'checkpoint_manager_blocking_duration_secs': 2.1313905715942383}
I0419 05:00:32.141062 132540212778816 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0419 05:00:32.141135 132540212778816 max_utils.py:750] 
Memstats: After params initialized:
I0419 05:00:32.141192 132540212778816 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_10(process=3,(2,2,0,0))
I0419 05:00:32.141225 132540212778816 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_11(process=3,(3,2,0,0))
I0419 05:00:32.141266 132540212778816 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_14(process=3,(2,3,0,0))
I0419 05:00:32.141292 132540212778816 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_15(process=3,(3,3,0,0))
I0419 05:00:32.405157 132540212778816 metric_logger.py:185] completed step: 0, seconds: 11.126, TFLOP/s/device: 1.221, Tokens/s/device: 184.081, total_weights: 65536, loss: 10.874
I0419 05:00:32.580580 132540212778816 metric_logger.py:185] completed step: 1, seconds: 2.545, TFLOP/s/device: 5.339, Tokens/s/device: 804.815, total_weights: 65536, loss: 10.874
I0419 05:00:32.990377 132540212778816 metric_logger.py:185] completed step: 2, seconds: 0.029, TFLOP/s/device: 461.269, Tokens/s/device: 69527.431, total_weights: 65536, loss: 10.262
I0419 05:00:33.137994 132540212778816 metric_logger.py:185] completed step: 3, seconds: 0.401, TFLOP/s/device: 33.906, Tokens/s/device: 5110.660, total_weights: 65536, loss: 9.732
I0419 05:00:33.433907 132540212778816 metric_logger.py:185] completed step: 4, seconds: 0.163, TFLOP/s/device: 83.338, Tokens/s/device: 12561.566, total_weights: 65536, loss: 9.274
I0419 05:00:33.441374 132540212778816 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.157, Tokens/s/device: 13890.961, total_weights: 65536, loss: 8.888
I0419 05:00:53.011213 132540212778816 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.857, Tokens/s/device: 6912.054, total_weights: 65536, loss: 8.589
I0419 05:00:53.158769 132540212778816 metric_logger.py:185] completed step: 7, seconds: 19.424, TFLOP/s/device: 0.700, Tokens/s/device: 105.439, total_weights: 65536, loss: 8.380
I0419 05:00:53.306481 132540212778816 metric_logger.py:185] completed step: 8, seconds: 0.154, TFLOP/s/device: 88.480, Tokens/s/device: 13336.633, total_weights: 65536, loss: 8.251
I0419 05:00:53.455291 132540212778816 checkpointing.py:794] Waiting for step 9 to finish before checkpoint...
I0419 05:00:53.455911 132540212778816 checkpointing.py:798] Waited 0.00063323974609375 seconds for step 9 to finish before starting checkpointing.
I0419 05:00:53.457828 132540212778816 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0419 05:00:55.330015    2756 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0419 05:00:58.954433 132418296063744 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints/0/items/array_metadatas/process_3
I0419 05:01:29.145472 132418287671040 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 27.256 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 57.955047607421875 s) (per-host)
I0419 05:01:29.145594 132418287671040 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 57.449394s.
I0419 05:01:40.318725 132418287671040 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 68.622511s.
I0419 05:01:40.319049 132414579930880 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save.
I0419 05:01:40.319167 132414579930880 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save.
I0419 05:01:40.319218 132414579930880 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0419 05:01:40.333133 132414579930880 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0419 05:01:40.333343 132540212778816 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 05:01:40.333492 132540212778816 checkpoint_manager.py:1458] Waiting for previous save to complete took 46.875668 seconds. If this number is high, consider checkpointing less frequently.
I0419 05:01:40.335438 132540212778816 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 9
I0419 05:01:40.338804 132540212778816 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints/9.
I0419 05:01:41.355080 132540212778816 jax_array_handlers.py:358] Scheduling D2H of 69 prioritized jax.Array.
I0419 05:01:41.355253 132540212778816 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 05:01:41.401386 132540212778816 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.048377s
I0419 05:01:41.401558 132540212778816 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.786 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3222792148590088 s) (per-host)
I0419 05:01:41.401611 132540212778816 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.322340s (batch_requests_ready=0.269252s, total_serialization_initiated=0.053019s, others=0.000069s)
I0419 05:01:41.401704 132540212778816 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.326463s (all_items=0.000015s, per_item={'items': '0.00001502'}, temp_paths=0.326447)
I0419 05:01:41.402558 132414535137024 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started.
I0419 05:01:41.402720 132540212778816 async_checkpointer.py:561] Finished blocking save. Time taken: 1.067211s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints/9.
I0419 05:01:41.826541 132540212778816 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0419 05:01:41.826913 132414579930880 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save.
I0419 05:01:41.827081 132540212778816 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776574853.4577987, 'wait_for_prev_duration_secs': 46.87566781044006, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776574900.335476, 'checkpointer_blocking_duration_secs': 1.067354440689087, 'get_old_steps_start_time': 1776574901.4028492, 'get_old_steps_duration_secs': 2.6226043701171875e-05, 'checkpoint_manager_blocking_start_time': 1776574853.4561303, 'checkpoint_manager_blocking_duration_secs': 48.37091279029846}
I0419 05:01:41.827204 132540212778816 checkpointing.py:409] Started an asynchronous checkpoint save for step 9
I0419 05:01:41.827265 132540212778816 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0419 05:01:50.056725 132418296063744 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_02_synthetic/checkpoints/9/items/array_metadatas/process_3
I0419 05:02:25.406258 132414535137024 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 35.635 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 44.32692503929138 s) (per-host)
I0419 05:02:25.406377 132414535137024 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 44.003706s.
I0419 05:02:34.761058 132414535137024 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 53.358371s.
I0419 05:02:34.761273 132414579930880 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save.
I0419 05:02:34.761336 132414579930880 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save.
I0419 05:02:34.761381 132414579930880 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0419 05:02:34.763900 132414579930880 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0419 05:02:34.764060 132540212778816 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 05:02:34.764257 132540212778816 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0419 05:02:34.764930 132540212778816 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.318, Tokens/s/device: 13915.218, 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: Sun Apr 19 05:02:44 UTC 2026
EXIT_CODE=0