MaxView

‹ 06_grad_accumCase: 07_eval08_checkpoint_async_true_resume ›

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/s92.02992.283+0.254
Tok/s13871.613909.8+38.251
Avg s/step2.4552.408-0.047
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_20260424_145744  ·  full log
XPK Start: Fri Apr 24 15:16:02 UTC 2026
2026-04-24 15:16:06.974500: 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 00:00:1777043766.987419      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1777043766.991171      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1777043767.002826      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777043767.002844      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777043767.002847      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777043767.002849      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-24 15:16:26.288192: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303)
I0424 15:16:26.819310 137185916966720 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-24 15:16:35,860:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0424 15:16:35.860323 137185916966720 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-24 15:16:35,862:jax._src.distributed:157: Connecting to JAX distributed service on mt-07-eval-ce7hg-slice-job-0-0.mt-07-eval-ce7hg:8482
I0424 15:16:35.862510 137185916966720 distributed.py:157] Connecting to JAX distributed service on mt-07-eval-ce7hg-slice-job-0-0.mt-07-eval-ce7hg:8482
I0424 15:16:37.078683 137185916966720 max_utils.py:284] Jax distributed system initialized!
I0424 15:16:42.947254 137185916966720 max_utils.py:800] System Information: Jax Version: 0.8.1
I0424 15:16:42.947369 137185916966720 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0424 15:16:42.947420 137185916966720 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
I0424 15:16:42.947467 137185916966720 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 15:16:43.571681 137185916966720 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 15:16:43.572318 137185916966720 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 15:16:43.572522 137185916966720 checkpointing.py:688] Setting up checkpoint logger...
I0424 15:16:43.572587 137185916966720 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0424 15:16:43.572652 137185916966720 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0424 15:16:43.573033 137185916966720 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 0x7cc49b779370>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 15:16:46.539395 137185916966720 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0424 15:16:46.539905 137185916966720 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 0x7cb203d3dcd0>}, handler_registry=None
I0424 15:16:46.540186 137185916966720 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cb203d3dcd0>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`.
I0424 15:16:46.540250 137185916966720 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cb203d401a0>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`.
I0424 15:16:46.540309 137185916966720 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 0x7cb203d3dcd0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cb203d3dcd0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cb203d401a0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cb203d401a0>}).
I0424 15:16:46.540854 137185916966720 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0424 15:16:46.540956 137185916966720 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7cb130190180> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0424 15:16:47.420965 137185916966720 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints
I0424 15:16:47.430704 137185916966720 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/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7cb27c55d760>
I0424 15:16:47.430824 137185916966720 checkpointing.py:302] Checkpoint manager created!
I0424 15:16:48.329058 137185916966720 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 15:16:48.329180 137185916966720 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 15:16:48.700543 137185916966720 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 15:16:48.700636 137185916966720 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 15:16:48.716627 137185916966720 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 15:16:48.716687 137185916966720 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 15:16:48.739682 137185916966720 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 15:16:48.739752 137185916966720 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 15:16:48.755777 137185916966720 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 15:16:48.755841 137185916966720 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 15:16:48.771790 137185916966720 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 15:16:48.771852 137185916966720 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 15:16:48.796829 137185916966720 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0424 15:16:48.796900 137185916966720 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 15:16:48.816705 137185916966720 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 15:16:48.816777 137185916966720 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 15:16:49.007162 137185916966720 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 15:16:49.007277 137185916966720 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0424 15:16:50.446359 137185916966720 maxtext_utils.py:1740]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0424 15:16:50.446493 137185916966720 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)
I0424 15:16:50.446549 137185916966720 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)
I0424 15:16:50.446610 137185916966720 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')
I0424 15:16:50.446663 137185916966720 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)
I0424 15:16:50.446702 137185916966720 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)
I0424 15:16:50.446756 137185916966720 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)
I0424 15:16:50.446811 137185916966720 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')
I0424 15:16:50.446853 137185916966720 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)
I0424 15:16:50.446892 137185916966720 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)
I0424 15:16:50.446943 137185916966720 maxtext_utils.py:1740]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed', 'vocab')
    Physical:  ('fsdp', None)
I0424 15:16:50.446992 137185916966720 maxtext_utils.py:1740]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed')
    Physical:  (None, 'fsdp')

I0424 15:16:50.951266 137185916966720 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 15:16:50.951360 137185916966720 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 15:16:50.966451 137185916966720 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 15:16:50.966516 137185916966720 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 15:17:01.914675 137185916966720 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0424 15:17:01.915637 137185916966720 metric_logger.py:289] number parameters: 1.104 billion
I0424 15:17:13.124420 137185916966720 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0424 15:17:13.278065 137185916966720 checkpointing.py:798] Waited 0.15362930297851562 seconds for step 0 to finish before starting checkpointing.
I0424 15:17:13.280713 137185916966720 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 15:17:13.282471 137185916966720 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 0
I0424 15:17:13.285195 137185916966720 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints/0.
I0424 15:17:13.830778 137185916966720 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 15:17:13.831837 137185916966720 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0424 15:17:13.831898 137185916966720 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0424 15:17:14.073423 137185916966720 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.242728s
I0424 15:17:14.073596 137185916966720 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.060 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5040979385375977 s) (per-host)
I0424 15:17:14.073649 137185916966720 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.504161s (batch_requests_ready=0.257749s, total_serialization_initiated=0.246341s, others=0.000072s)
I0424 15:17:14.073741 137185916966720 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.508170s (all_items=0.000019s, per_item={'items': '0.00001907'}, temp_paths=0.508151)
I0424 15:17:14.074712 137062315837184 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0424 15:17:14.074884 137185916966720 async_checkpointer.py:561] Finished blocking save. Time taken: 0.792341s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints/0.
I0424 15:17:14.522898 137185916966720 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 15:17:14.523276 137061801654016 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 15:17:14.523418 137185916966720 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777043833.2806933, 'wait_for_prev_duration_secs': 6.771087646484375e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777043833.2825081, 'checkpointer_blocking_duration_secs': 0.7924861907958984, 'get_old_steps_start_time': 1777043834.0750139, 'get_old_steps_duration_secs': 2.6464462280273438e-05, 'checkpoint_manager_blocking_start_time': 1777043833.2786803, 'checkpoint_manager_blocking_duration_secs': 1.2446980476379395}
I0424 15:17:14.523572 137185916966720 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0424 15:17:14.523627 137185916966720 max_utils.py:750] 
Memstats: After params initialized:
I0424 15:17:14.523684 137185916966720 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0424 15:17:14.523720 137185916966720 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0424 15:17:14.523746 137185916966720 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0424 15:17:14.523774 137185916966720 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0424 15:17:14.786248 137185916966720 metric_logger.py:185] completed step: 0, seconds: 11.209, TFLOP/s/device: 1.212, Tokens/s/device: 182.716, total_weights: 65536, loss: 10.874
I0424 15:17:14.966756 137185916966720 metric_logger.py:185] completed step: 1, seconds: 1.661, TFLOP/s/device: 8.182, Tokens/s/device: 1233.349, total_weights: 65536, loss: 10.874
I0424 15:17:15.359159 137185916966720 metric_logger.py:185] completed step: 2, seconds: 0.033, TFLOP/s/device: 408.451, Tokens/s/device: 61566.211, total_weights: 65536, loss: 10.263
I0424 15:17:15.506415 137185916966720 metric_logger.py:185] completed step: 3, seconds: 0.393, TFLOP/s/device: 34.583, Tokens/s/device: 5212.721, total_weights: 65536, loss: 9.734
I0424 15:17:15.801959 137185916966720 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 89.073, Tokens/s/device: 13425.987, total_weights: 65536, loss: 9.278
I0424 15:17:15.807773 137185916966720 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.298, Tokens/s/device: 13912.193, total_weights: 65536, loss: 8.892
I0424 15:17:35.062164 137185916966720 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.879, Tokens/s/device: 6915.321, total_weights: 65536, loss: 8.592
I0424 15:17:35.209710 137185916966720 metric_logger.py:185] completed step: 7, seconds: 19.108, TFLOP/s/device: 0.711, Tokens/s/device: 107.181, total_weights: 65536, loss: 8.384
I0424 15:17:35.357026 137185916966720 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.533, Tokens/s/device: 13495.437, total_weights: 65536, loss: 8.255
I0424 15:17:35.504173 137185916966720 checkpointing.py:794] Waiting for step 9 to finish before checkpoint...
I0424 15:17:35.504835 137185916966720 checkpointing.py:798] Waited 0.0006818771362304688 seconds for step 9 to finish before starting checkpointing.
I0424 15:17:35.506842 137185916966720 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 15:17:35.893007    2616 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 15:17:37.893246 137062324229888 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints/0/items/array_metadatas/process_6
I0424 15:18:09.995524 137062315837184 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 27.994 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 56.42598509788513 s) (per-host)
I0424 15:18:09.995644 137062315837184 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 55.920822s.
I0424 15:18:16.506334 137062315837184 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 62.431496s.
I0424 15:18:16.506625 137061801654016 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 15:18:16.506747 137061801654016 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 15:18:16.506800 137061801654016 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0424 15:18:16.508084 137061801654016 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0424 15:18:16.508304 137185916966720 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.
W0424 15:18:16.508453 137185916966720 checkpoint_manager.py:1458] Waiting for previous save to complete took 41.001617 seconds. If this number is high, consider checkpointing less frequently.
I0424 15:18:16.510338 137185916966720 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 9
I0424 15:18:16.513814 137185916966720 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints/9.
I0424 15:18:17.393599 137185916966720 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0424 15:18:17.393699 137185916966720 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0424 15:18:17.437394 137185916966720 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.045011s
I0424 15:18:17.437555 137185916966720 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.027 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.30687618255615234 s) (per-host)
I0424 15:18:17.437607 137185916966720 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.306937s (batch_requests_ready=0.259931s, total_serialization_initiated=0.046932s, others=0.000074s)
I0424 15:18:17.437693 137185916966720 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.310907s (all_items=0.000016s, per_item={'items': '0.00001574'}, temp_paths=0.310891)
I0424 15:18:17.438561 137062293120768 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0424 15:18:17.438690 137185916966720 async_checkpointer.py:561] Finished blocking save. Time taken: 0.928283s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints/9.
I0424 15:18:17.469385 137185916966720 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 15:18:17.469704 137061801654016 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 15:18:17.469870 137185916966720 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777043855.5068128, 'wait_for_prev_duration_secs': 41.00161671638489, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777043896.5103786, 'checkpointer_blocking_duration_secs': 0.9284822940826416, 'get_old_steps_start_time': 1777043897.4388783, 'get_old_steps_duration_secs': 2.6702880859375e-05, 'checkpoint_manager_blocking_start_time': 1777043855.505058, 'checkpoint_manager_blocking_duration_secs': 41.96477675437927}
I0424 15:18:17.469983 137185916966720 checkpointing.py:409] Started an asynchronous checkpoint save for step 9
I0424 15:18:17.470029 137185916966720 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 15:18:22.304352 137062324229888 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints/9/items/array_metadatas/process_6
I0424 15:19:00.259195 137062293120768 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 36.625 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.12847352027893 s) (per-host)
I0424 15:19:00.259314 137062293120768 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.820543s.
I0424 15:19:05.264843 137062293120768 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 47.826057s.
I0424 15:19:05.265153 137061801654016 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 15:19:05.265269 137061801654016 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 15:19:05.265318 137061801654016 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0424 15:19:05.266839 137061801654016 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0424 15:19:05.267019 137185916966720 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.
I0424 15:19:05.267171 137185916966720 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 15:19:05.267857 137185916966720 metric_logger.py:185] completed step: 9, seconds: 0.148, TFLOP/s/device: 92.029, Tokens/s/device: 13871.580, 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: Fri Apr 24 15:19:18 UTC 2026
EXIT_CODE=0
NNX  ·  73213e044  ·  feat_nnx_set_defaults_true_20260424_145744  ·  full log
XPK Start: Fri Apr 24 16:44:26 UTC 2026
2026-04-24 16:44:30.517711: 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 00:00:1777049070.532283      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1777049070.536160      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1777049070.547844      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777049070.547863      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777049070.547865      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777049070.547867      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-24 16:44:49.937261: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303)
I0424 16:44:50.469315 137589768787776 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-24 16:44:59,509:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0424 16:44:59.509200 137589768787776 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-24 16:44:59,511:jax._src.distributed:157: Connecting to JAX distributed service on mt-07-eval-kar1g-slice-job-0-0.mt-07-eval-kar1g:8482
I0424 16:44:59.511453 137589768787776 distributed.py:157] Connecting to JAX distributed service on mt-07-eval-kar1g-slice-job-0-0.mt-07-eval-kar1g:8482
I0424 16:45:00.509658 137589768787776 max_utils.py:284] Jax distributed system initialized!
I0424 16:45:07.500022 137589768787776 max_utils.py:800] System Information: Jax Version: 0.8.1
I0424 16:45:07.500173 137589768787776 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0424 16:45:07.500218 137589768787776 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
I0424 16:45:07.500251 137589768787776 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 16:45:08.121119 137589768787776 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 16:45:08.238845 137589768787776 checkpointing.py:688] Setting up checkpoint logger...
I0424 16:45:08.238966 137589768787776 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0424 16:45:08.239010 137589768787776 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0424 16:45:08.239234 137589768787776 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 0x7d2293f550a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 16:45:11.144126 137589768787776 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0424 16:45:11.144469 137589768787776 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 0x7d0e385803b0>}, handler_registry=None
I0424 16:45:11.144712 137589768787776 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d0e385803b0>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`.
I0424 16:45:11.144761 137589768787776 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d0e783bf560>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`.
I0424 16:45:11.144798 137589768787776 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 0x7d0e385803b0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d0e385803b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d0e783bf560>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d0e783bf560>}).
I0424 16:45:11.145154 137589768787776 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0424 16:45:11.145236 137589768787776 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7d0e38500f40> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0424 16:45:11.862469 137589768787776 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints
I0424 16:45:12.153469 137589768787776 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_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7d0e385cbfb0>
I0424 16:45:12.153655 137589768787776 checkpointing.py:302] Checkpoint manager created!
I0424 16:45:13.088128 137589768787776 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 16:45:13.088252 137589768787776 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0424 16:45:14.497681 137589768787776 maxtext_utils.py:1749]  decoder/decoder_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0424 16:45:14.497803 137589768787776 maxtext_utils.py:1749]  decoder/layers/mlp/wi_0/kernel/value
    Shape:     float32[2048,16,7168]
    Physical:  ('fsdp', None, None)
I0424 16:45:14.497858 137589768787776 maxtext_utils.py:1749]  decoder/layers/mlp/wi_1/kernel/value
    Shape:     float32[2048,16,7168]
    Physical:  ('fsdp', None, None)
I0424 16:45:14.497902 137589768787776 maxtext_utils.py:1749]  decoder/layers/mlp/wo/kernel/value
    Shape:     float32[7168,16,2048]
    Physical:  (None, None, 'fsdp')
I0424 16:45:14.497940 137589768787776 maxtext_utils.py:1749]  decoder/layers/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048,16]
    Physical:  (None, None)
I0424 16:45:14.497977 137589768787776 maxtext_utils.py:1749]  decoder/layers/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048,16]
    Physical:  (None, None)
I0424 16:45:14.498014 137589768787776 maxtext_utils.py:1749]  decoder/layers/self_attention/key/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0424 16:45:14.498053 137589768787776 maxtext_utils.py:1749]  decoder/layers/self_attention/out/kernel/value
    Shape:     float32[16,16,128,2048]
    Physical:  (None, None, None, 'fsdp')
I0424 16:45:14.498089 137589768787776 maxtext_utils.py:1749]  decoder/layers/self_attention/query/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0424 16:45:14.498141 137589768787776 maxtext_utils.py:1749]  decoder/layers/self_attention/value/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0424 16:45:14.498178 137589768787776 maxtext_utils.py:1749]  decoder/logits_dense/kernel/value
    Shape:     float32[2048,32000]
    Physical:  ('fsdp', None)

I0424 16:45:14.498213 137589768787776 maxtext_utils.py:1749]  token_embedder/embedding/value
    Shape:     float32[32000,2048]
    Physical:  (None, 'fsdp')
I0424 16:45:14.901979 137589768787776 nnx_decoders.py:616] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 16:45:14.902144 137589768787776 nnx_decoders.py:616] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 16:45:14.907563 137589768787776 nnx_decoders.py:616] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 16:45:14.907655 137589768787776 nnx_decoders.py:616] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 16:45:15.052803 137589768787776 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 16:45:15.052944 137589768787776 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 16:45:15.196398 137589768787776 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 16:45:15.196533 137589768787776 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 16:45:15.353903 137589768787776 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 16:45:15.354038 137589768787776 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:45:15.503600 137589768787776 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 16:45:15.503732 137589768787776 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:45:15.649867 137589768787776 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0424 16:45:15.650007 137589768787776 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:45:15.806736 137589768787776 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0424 16:45:15.806869 137589768787776 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:45:15.829716 137589768787776 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 16:45:15.829813 137589768787776 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 16:45:27.040102 137589768787776 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB.
I0424 16:45:27.042599 137589768787776 metric_logger.py:289] number parameters: 1.104 billion
I0424 16:45:38.182414 137589768787776 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0424 16:45:38.335667 137589768787776 checkpointing.py:798] Waited 0.15323996543884277 seconds for step 0 to finish before starting checkpointing.
I0424 16:45:38.338022 137589768787776 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 16:45:38.339842 137589768787776 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 0
I0424 16:45:38.342797 137589768787776 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints/0.
I0424 16:45:38.997956 137589768787776 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 16:45:38.999036 137589768787776 jax_array_handlers.py:358] Scheduling D2H of 69 prioritized jax.Array.
I0424 16:45:38.999105 137589768787776 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0424 16:45:39.251986 137589768787776 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.255227s
I0424 16:45:39.252172 137589768787776 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.635 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5853595733642578 s) (per-host)
I0424 16:45:39.252226 137589768787776 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.585423s (batch_requests_ready=0.323339s, total_serialization_initiated=0.262012s, others=0.000071s)
I0424 16:45:39.252336 137589768787776 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.589801s (all_items=0.000019s, per_item={'items': '0.00001884'}, temp_paths=0.589782)
I0424 16:45:39.253283 137464052184832 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0424 16:45:39.253486 137589768787776 async_checkpointer.py:561] Finished blocking save. Time taken: 0.913574s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints/0.
I0424 16:45:39.255417 137589768787776 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 16:45:39.255649 137463473350400 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 16:45:39.255744 137589768787776 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777049138.3380032, 'wait_for_prev_duration_secs': 6.771087646484375e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777049138.3398821, 'checkpointer_blocking_duration_secs': 0.9137136936187744, 'get_old_steps_start_time': 1777049139.2536156, 'get_old_steps_duration_secs': 2.6941299438476562e-05, 'checkpoint_manager_blocking_start_time': 1777049138.3360617, 'checkpoint_manager_blocking_duration_secs': 0.9196450710296631}
I0424 16:45:39.255902 137589768787776 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0424 16:45:39.255977 137589768787776 max_utils.py:750] 
Memstats: After params initialized:
I0424 16:45:39.256027 137589768787776 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0))
I0424 16:45:39.256062 137589768787776 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0))
I0424 16:45:39.256090 137589768787776 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0))
I0424 16:45:39.256132 137589768787776 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0))
I0424 16:45:39.526632 137589768787776 metric_logger.py:185] completed step: 0, seconds: 11.139, TFLOP/s/device: 1.220, Tokens/s/device: 183.864, total_weights: 65536, loss: 10.874
I0424 16:45:39.701855 137589768787776 metric_logger.py:185] completed step: 1, seconds: 1.343, TFLOP/s/device: 10.121, Tokens/s/device: 1525.500, total_weights: 65536, loss: 10.874
I0424 16:45:40.111790 137589768787776 metric_logger.py:185] completed step: 2, seconds: 0.030, TFLOP/s/device: 457.541, Tokens/s/device: 68965.517, total_weights: 65536, loss: 10.262
I0424 16:45:40.259474 137589768787776 metric_logger.py:185] completed step: 3, seconds: 0.401, TFLOP/s/device: 33.917, Tokens/s/device: 5112.306, total_weights: 65536, loss: 9.732
I0424 16:45:40.554771 137589768787776 metric_logger.py:185] completed step: 4, seconds: 0.163, TFLOP/s/device: 83.365, Tokens/s/device: 12565.728, total_weights: 65536, loss: 9.274
I0424 16:45:40.561873 137589768787776 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.489, Tokens/s/device: 13940.982, total_weights: 65536, loss: 8.888
I0424 16:45:59.700410 137589768787776 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.930, Tokens/s/device: 6923.035, total_weights: 65536, loss: 8.589
I0424 16:45:59.847885 137589768787776 metric_logger.py:185] completed step: 7, seconds: 18.993, TFLOP/s/device: 0.715, Tokens/s/device: 107.831, total_weights: 65536, loss: 8.380
I0424 16:45:59.995167 137589768787776 metric_logger.py:185] completed step: 8, seconds: 0.153, TFLOP/s/device: 88.793, Tokens/s/device: 13383.871, total_weights: 65536, loss: 8.251
I0424 16:46:00.143727 137589768787776 checkpointing.py:794] Waiting for step 9 to finish before checkpoint...
I0424 16:46:00.144347 137589768787776 checkpointing.py:798] Waited 0.0006325244903564453 seconds for step 9 to finish before starting checkpointing.
I0424 16:46:00.146258 137589768787776 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 16:46:01.151869    2643 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 16:46:03.092281 137464531113728 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_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints/0/items/array_metadatas/process_6
I0424 16:46:34.648691 137464052184832 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 28.216 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 55.98182439804077 s) (per-host)
I0424 16:46:34.648815 137464052184832 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 55.395403s.
I0424 16:46:43.227476 137464052184832 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 63.974046s.
I0424 16:46:43.227776 137463473350400 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 16:46:43.227906 137463473350400 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 16:46:43.227955 137463473350400 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0424 16:46:43.229535 137463473350400 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0424 16:46:43.229718 137589768787776 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.
W0424 16:46:43.229875 137589768787776 checkpoint_manager.py:1458] Waiting for previous save to complete took 43.083622 seconds. If this number is high, consider checkpointing less frequently.
I0424 16:46:43.231873 137589768787776 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 9
I0424 16:46:43.235294 137589768787776 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints/9.
I0424 16:46:44.208774 137589768787776 jax_array_handlers.py:358] Scheduling D2H of 69 prioritized jax.Array.
I0424 16:46:44.208931 137589768787776 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False
I0424 16:46:44.258501 137589768787776 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.051996s
I0424 16:46:44.258674 137589768787776 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.943 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.31207990646362305 s) (per-host)
I0424 16:46:44.258727 137589768787776 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.312143s (batch_requests_ready=0.255253s, total_serialization_initiated=0.056818s, others=0.000071s)
I0424 16:46:44.258826 137589768787776 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.316320s (all_items=0.000016s, per_item={'items': '0.00001574'}, temp_paths=0.316304)
I0424 16:46:44.259729 137464017426176 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started.
I0424 16:46:44.259851 137589768787776 async_checkpointer.py:561] Finished blocking save. Time taken: 1.027911s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints/9.
I0424 16:46:44.459174 137589768787776 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 16:46:44.459565 137463473350400 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save.
I0424 16:46:44.459740 137589768787776 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777049160.1462288, 'wait_for_prev_duration_secs': 43.08362150192261, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777049203.2319133, 'checkpointer_blocking_duration_secs': 1.028106689453125, 'get_old_steps_start_time': 1777049204.2600405, 'get_old_steps_duration_secs': 2.8133392333984375e-05, 'checkpoint_manager_blocking_start_time': 1777049160.1446073, 'checkpoint_manager_blocking_duration_secs': 44.31509256362915}
I0424 16:46:44.459855 137589768787776 checkpointing.py:409] Started an asynchronous checkpoint save for step 9
I0424 16:46:44.459902 137589768787776 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 16:46:49.410424 137464531113728 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_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_07_eval/checkpoints/9/items/array_metadatas/process_6
I0424 16:47:25.733246 137464017426176 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 37.802 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.78661346435547 s) (per-host)
I0424 16:47:25.733378 137464017426176 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.473449s.
I0424 16:47:34.038913 137464017426176 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 49.778967s.
I0424 16:47:34.039224 137463473350400 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 16:47:34.039354 137463473350400 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save.
I0424 16:47:34.039402 137463473350400 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0424 16:47:34.040997 137463473350400 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0424 16:47:34.041185 137589768787776 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.
I0424 16:47:34.041371 137589768787776 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 16:47:34.042014 137589768787776 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.283, Tokens/s/device: 13909.831, 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: Fri Apr 24 16:47:46 UTC 2026
EXIT_CODE=0