MaxView

‹ —Case: 01_base02_synthetic ›

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/s92.15892.354+0.196
Tok/s13891.113920.6+29.553
Avg s/step2.3472.443+0.096
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 14:58:21 UTC 2026
2026-04-24 14:58:25.432427: 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 00:00:1777042705.445241      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1777042705.449009      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1777042705.460278      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777042705.460296      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777042705.460298      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777042705.460300      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-24 14:58:44.599068: 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 14:58:45.122118 139543169230656 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-24 14:58:54,162:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0424 14:58:54.162940 139543169230656 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-24 14:58:54,165:jax._src.distributed:157: Connecting to JAX distributed service on mt-01-base-8ah2w-slice-job-0-0.mt-01-base-8ah2w:8482
I0424 14:58:54.165254 139543169230656 distributed.py:157] Connecting to JAX distributed service on mt-01-base-8ah2w-slice-job-0-0.mt-01-base-8ah2w:8482
I0424 14:58:55.603101 139543169230656 max_utils.py:284] Jax distributed system initialized!
I0424 14:59:02.645855 139543169230656 max_utils.py:800] System Information: Jax Version: 0.8.1
I0424 14:59:02.645963 139543169230656 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0424 14:59:02.646005 139543169230656 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 14:59:02.646039 139543169230656 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 14:59:03.263400 139543169230656 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 14:59:03.264006 139543169230656 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 14:59:03.264197 139543169230656 checkpointing.py:688] Setting up checkpoint logger...
I0424 14:59:03.264248 139543169230656 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0424 14:59:03.264291 139543169230656 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0424 14:59:03.264660 139543169230656 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 0x7ee963b4dc10>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 14:59:07.048093 139543169230656 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0424 14:59:07.048440 139543169230656 checkpoint_manager.py:709] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ed4d0349580>}, handler_registry=None
I0424 14:59:07.048688 139543169230656 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ed4d0349580>` 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 14:59:07.048738 139543169230656 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ed4d034e300>` 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 14:59:07.048774 139543169230656 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 0x7ed4d0349580>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ed4d0349580>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ed4d034e300>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ed4d034e300>}).
I0424 14:59:07.049295 139543169230656 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0424 14:59:07.049376 139543169230656 async_checkpointer.py:177] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7ed4b82b8180> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0424 14:59:07.929190 139543169230656 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_01_base/checkpoints
I0424 14:59:08.189613 139543169230656 checkpoint_manager.py:929] [process=5][thread=MainThread] CheckpointManager created,  primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7ed4d034e5a0>
I0424 14:59:08.189815 139543169230656 checkpointing.py:302] Checkpoint manager created!
I0424 14:59:09.083861 139543169230656 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 14:59:09.083965 139543169230656 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 14:59:09.455369 139543169230656 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 14:59:09.455464 139543169230656 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 14:59:09.471327 139543169230656 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 14:59:09.471391 139543169230656 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 14:59:09.494197 139543169230656 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 14:59:09.494269 139543169230656 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 14:59:09.510301 139543169230656 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 14:59:09.510364 139543169230656 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 14:59:09.526245 139543169230656 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 14:59:09.526306 139543169230656 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 14:59:09.551008 139543169230656 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0424 14:59:09.551076 139543169230656 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 14:59:09.570612 139543169230656 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 14:59:09.570692 139543169230656 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 14:59:09.762631 139543169230656 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 14:59:09.762749 139543169230656 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0424 14:59:11.209847 139543169230656 maxtext_utils.py:1740]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   PartitionSpec('norm',)
    Physical:  (None,)
I0424 14:59:11.209974 139543169230656 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 14:59:11.210035 139543169230656 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 14:59:11.210096 139543169230656 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 14:59:11.210152 139543169230656 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 14:59:11.210191 139543169230656 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 14:59:11.210244 139543169230656 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 14:59:11.210297 139543169230656 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 14:59:11.210336 139543169230656 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 14:59:11.210373 139543169230656 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 14:59:11.210422 139543169230656 maxtext_utils.py:1740]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   PartitionSpec('embed', 'vocab')
    Physical:  ('fsdp', None)

I0424 14:59:11.210470 139543169230656 maxtext_utils.py:1740]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   PartitionSpec('vocab', 'embed')
    Physical:  (None, 'fsdp')
I0424 14:59:11.709036 139543169230656 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 14:59:11.709129 139543169230656 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 14:59:11.724051 139543169230656 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0424 14:59:11.724105 139543169230656 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0424 14:59:22.452657 139543169230656 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 14:59:22.453560 139543169230656 metric_logger.py:289] number parameters: 1.104 billion
I0424 14:59:33.569235 139543169230656 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0424 14:59:33.728039 139543169230656 checkpointing.py:798] Waited 0.1587843894958496 seconds for step 0 to finish before starting checkpointing.
I0424 14:59:33.730541 139543169230656 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 14:59:33.732290 139543169230656 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 0
I0424 14:59:33.735033 139543169230656 async_checkpointer.py:452] [process=5] 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_01_base/checkpoints/0.
I0424 14:59:34.503138 139543169230656 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 14:59:34.504190 139543169230656 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0424 14:59:34.504252 139543169230656 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 14:59:34.752350 139543169230656 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.249294s
I0424 14:59:34.752523 139543169230656 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.030 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5090720653533936 s) (per-host)
I0424 14:59:34.752576 139543169230656 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.509137s (batch_requests_ready=0.256173s, total_serialization_initiated=0.252890s, others=0.000074s)
I0424 14:59:34.752680 139543169230656 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.513377s (all_items=0.000020s, per_item={'items': '0.00002003'}, temp_paths=0.513357)
I0424 14:59:34.753600 139415435339520 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started.
I0424 14:59:34.753777 139543169230656 async_checkpointer.py:561] Finished blocking save. Time taken: 1.021419s. 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_01_base/checkpoints/0.
I0424 14:59:34.808055 139543169230656 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 14:59:34.808365 139414797788928 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0424 14:59:34.808512 139543169230656 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_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777042773.7305224, 'wait_for_prev_duration_secs': 6.580352783203125e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777042773.732329, 'checkpointer_blocking_duration_secs': 1.0215535163879395, 'get_old_steps_start_time': 1777042774.7539046, 'get_old_steps_duration_secs': 2.3603439331054688e-05, 'checkpoint_manager_blocking_start_time': 1777042773.7285812, 'checkpoint_manager_blocking_duration_secs': 1.079890251159668}
I0424 14:59:34.808619 139543169230656 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0424 14:59:34.808685 139543169230656 max_utils.py:750] 
Memstats: After params initialized:
I0424 14:59:34.808740 139543169230656 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0))
I0424 14:59:34.808777 139543169230656 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0))
I0424 14:59:34.808805 139543169230656 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0))
I0424 14:59:34.808830 139543169230656 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0))
I0424 14:59:35.073723 139543169230656 metric_logger.py:185] completed step: 0, seconds: 11.116, TFLOP/s/device: 1.222, Tokens/s/device: 184.246, total_weights: 65536, loss: 10.874
I0424 14:59:35.246284 139543169230656 metric_logger.py:185] completed step: 1, seconds: 1.503, TFLOP/s/device: 9.039, Tokens/s/device: 1362.476, total_weights: 65536, loss: 10.874
I0424 14:59:35.633995 139543169230656 metric_logger.py:185] completed step: 2, seconds: 0.025, TFLOP/s/device: 534.905, Tokens/s/device: 80626.747, total_weights: 65536, loss: 10.263
I0424 14:59:35.781360 139543169230656 metric_logger.py:185] completed step: 3, seconds: 0.388, TFLOP/s/device: 35.013, Tokens/s/device: 5277.480, total_weights: 65536, loss: 9.734
I0424 14:59:36.076925 139543169230656 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 89.005, Tokens/s/device: 13415.785, total_weights: 65536, loss: 9.278
I0424 14:59:36.082576 139543169230656 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.273, Tokens/s/device: 13908.414, total_weights: 65536, loss: 8.892
I0424 14:59:54.545460 139543169230656 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.897, Tokens/s/device: 6918.124, total_weights: 65536, loss: 8.592
I0424 14:59:54.692885 139543169230656 metric_logger.py:185] completed step: 7, seconds: 18.316, TFLOP/s/device: 0.742, Tokens/s/device: 111.815, total_weights: 65536, loss: 8.384
I0424 14:59:54.840260 139543169230656 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.276, Tokens/s/device: 13456.598, total_weights: 65536, loss: 8.255
I0424 14:59:54.987380 139543169230656 checkpointing.py:794] Waiting for step 9 to finish before checkpoint...
I0424 14:59:54.988053 139543169230656 checkpointing.py:798] Waited 0.0006933212280273438 seconds for step 9 to finish before starting checkpointing.
I0424 14:59:54.990210 139543169230656 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 14:59:55.663399    2689 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 14:59:57.625800 139415443732224 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/0/items/array_metadatas/process_5
I0424 15:00:28.885317 139415435339520 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 28.908 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 54.64182448387146 s) (per-host)
I0424 15:00:28.885431 139415435339520 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 54.131709s.
I0424 15:00:37.566341 139415435339520 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 62.812602s.
I0424 15:00:37.566560 139414797788928 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 15:00:37.566640 139414797788928 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0424 15:00:37.566744 139414797788928 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0424 15:00:37.568264 139414797788928 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0424 15:00:37.568458 139543169230656 checkpoint_manager.py:2036] [process=5][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0424 15:00:37.568603 139543169230656 checkpoint_manager.py:1458] Waiting for previous save to complete took 42.578396 seconds. If this number is high, consider checkpointing less frequently.
I0424 15:00:37.570404 139543169230656 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 9
I0424 15:00:37.573742 139543169230656 async_checkpointer.py:452] [process=5] 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_01_base/checkpoints/9.
I0424 15:00:38.125104 139543169230656 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array.
I0424 15:00:38.125203 139543169230656 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 15:00:38.168919 139543169230656 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.045240s
I0424 15:00:38.169098 139543169230656 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.271 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.29263806343078613 s) (per-host)
I0424 15:00:38.169151 139543169230656 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.292701s (batch_requests_ready=0.245536s, total_serialization_initiated=0.047093s, others=0.000072s)
I0424 15:00:38.169243 139543169230656 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.296667s (all_items=0.000016s, per_item={'items': '0.00001574'}, temp_paths=0.296651)
I0424 15:00:38.170091 139415393376000 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started.
I0424 15:00:38.170255 139543169230656 async_checkpointer.py:561] Finished blocking save. Time taken: 0.599780s. 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_01_base/checkpoints/9.
I0424 15:00:38.196958 139543169230656 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 15:00:38.197286 139414797788928 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save.
I0424 15:00:38.197455 139543169230656 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_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777042794.9901814, 'wait_for_prev_duration_secs': 42.57839584350586, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777042837.570442, 'checkpointer_blocking_duration_secs': 0.5999181270599365, 'get_old_steps_start_time': 1777042838.1703784, 'get_old_steps_duration_secs': 2.5033950805664062e-05, 'checkpoint_manager_blocking_start_time': 1777042794.9882803, 'checkpoint_manager_blocking_duration_secs': 43.20913910865784}
I0424 15:00:38.197560 139543169230656 checkpointing.py:409] Started an asynchronous checkpoint save for step 9
I0424 15:00:38.197605 139543169230656 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 15:00:43.111798 139415443732224 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/linen_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/9/items/array_metadatas/process_5
I0424 15:01:19.751045 139415393376000 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 37.722 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.87454319000244 s) (per-host)
I0424 15:01:19.751165 139415393376000 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.580957s.
I0424 15:01:27.183747 139415393376000 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 49.013523s.
I0424 15:01:27.184070 139414797788928 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 15:01:27.184196 139414797788928 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save.
I0424 15:01:27.184244 139414797788928 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0424 15:01:27.186443 139414797788928 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0424 15:01:27.186633 139543169230656 checkpoint_manager.py:2036] [process=5][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0424 15:01:27.186787 139543169230656 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 15:01:27.187444 139543169230656 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.158, Tokens/s/device: 13891.056, 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 15:01:37 UTC 2026
EXIT_CODE=0
NNX  ·  73213e044  ·  feat_nnx_set_defaults_true_20260424_145744  ·  full log
XPK Start: Fri Apr 24 16:32:01 UTC 2026
2026-04-24 16:32:05.452051: 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 00:00:1777048325.465380      10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1777048325.469134      10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
W0000 00:00:1777048325.480758      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777048325.480777      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777048325.480780      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
W0000 00:00:1777048325.480782      10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
2026-04-24 16:32:24.664733: 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 16:32:25.191114 135970079123264 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-24 16:32:34,234:jax._src.distributed:140: Starting JAX distributed service on [::]:8482
I0424 16:32:34.234088 135970079123264 distributed.py:140] Starting JAX distributed service on [::]:8482
INFO:2026-04-24 16:32:34,238:jax._src.distributed:157: Connecting to JAX distributed service on mt-01-base-02xnr-slice-job-0-0.mt-01-base-02xnr:8482
I0424 16:32:34.238742 135970079123264 distributed.py:157] Connecting to JAX distributed service on mt-01-base-02xnr-slice-job-0-0.mt-01-base-02xnr:8482
I0424 16:32:35.181607 135970079123264 max_utils.py:284] Jax distributed system initialized!
I0424 16:32:41.998013 135970079123264 max_utils.py:800] System Information: Jax Version: 0.8.1
I0424 16:32:41.998122 135970079123264 max_utils.py:801] System Information: Jaxlib Version: 0.8.1
I0424 16:32:41.998164 135970079123264 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 16:32:41.998200 135970079123264 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0424 16:32:42.609136 135970079123264 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0424 16:32:42.724076 135970079123264 checkpointing.py:688] Setting up checkpoint logger...
I0424 16:32:42.724193 135970079123264 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0424 16:32:42.724238 135970079123264 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None
I0424 16:32:42.724445 135970079123264 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 0x7ba971f6d1f0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0424 16:32:46.433434 135970079123264 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0424 16:32:46.434180 135970079123264 checkpoint_manager.py:709] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b96642fbbf0>}, handler_registry=None
I0424 16:32:46.434506 135970079123264 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b96642fbbf0>` 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 16:32:46.434559 135970079123264 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b9534194170>` 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 16:32:46.434612 135970079123264 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 0x7b96642fbbf0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b96642fbbf0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b9534194170>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b9534194170>}).
I0424 16:32:46.434952 135970079123264 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33
I0424 16:32:46.435031 135970079123264 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7b96642a0f40> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0424 16:32:47.220637 135970079123264 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_01_base/checkpoints
I0424 16:32:47.315219 135970079123264 checkpoint_manager.py:929] [process=0][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_01_base/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7b9534194c20>
I0424 16:32:47.315393 135970079123264 checkpointing.py:302] Checkpoint manager created!
I0424 16:32:48.243367 135970079123264 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0424 16:32:48.243490 135970079123264 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0424 16:32:49.661172 135970079123264 maxtext_utils.py:1749]  decoder/decoder_norm/scale/value
    Shape:     float32[2048]
    Physical:  (None,)
I0424 16:32:49.661290 135970079123264 maxtext_utils.py:1749]  decoder/layers/mlp/wi_0/kernel/value
    Shape:     float32[2048,16,7168]
    Physical:  ('fsdp', None, None)
I0424 16:32:49.661345 135970079123264 maxtext_utils.py:1749]  decoder/layers/mlp/wi_1/kernel/value
    Shape:     float32[2048,16,7168]
    Physical:  ('fsdp', None, None)
I0424 16:32:49.661390 135970079123264 maxtext_utils.py:1749]  decoder/layers/mlp/wo/kernel/value
    Shape:     float32[7168,16,2048]
    Physical:  (None, None, 'fsdp')
I0424 16:32:49.661427 135970079123264 maxtext_utils.py:1749]  decoder/layers/post_self_attention_layer_norm/scale/value
    Shape:     float32[2048,16]
    Physical:  (None, None)
I0424 16:32:49.661464 135970079123264 maxtext_utils.py:1749]  decoder/layers/pre_self_attention_layer_norm/scale/value
    Shape:     float32[2048,16]
    Physical:  (None, None)
I0424 16:32:49.661501 135970079123264 maxtext_utils.py:1749]  decoder/layers/self_attention/key/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0424 16:32:49.661538 135970079123264 maxtext_utils.py:1749]  decoder/layers/self_attention/out/kernel/value
    Shape:     float32[16,16,128,2048]
    Physical:  (None, None, None, 'fsdp')
I0424 16:32:49.661585 135970079123264 maxtext_utils.py:1749]  decoder/layers/self_attention/query/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0424 16:32:49.661624 135970079123264 maxtext_utils.py:1749]  decoder/layers/self_attention/value/kernel/value
    Shape:     float32[2048,16,16,128]
    Physical:  ('fsdp', None, None, None)
I0424 16:32:49.661663 135970079123264 maxtext_utils.py:1749]  decoder/logits_dense/kernel/value
    Shape:     float32[2048,32000]
    Physical:  ('fsdp', None)

I0424 16:32:49.661725 135970079123264 maxtext_utils.py:1749]  token_embedder/embedding/value
    Shape:     float32[32000,2048]
    Physical:  (None, 'fsdp')
I0424 16:32:50.064435 135970079123264 nnx_decoders.py:616] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 16:32:50.064600 135970079123264 nnx_decoders.py:616] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 16:32:50.069936 135970079123264 nnx_decoders.py:616] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0424 16:32:50.070029 135970079123264 nnx_decoders.py:616] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 16:32:50.223621 135970079123264 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 16:32:50.223811 135970079123264 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 16:32:50.378989 135970079123264 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0424 16:32:50.379127 135970079123264 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0424 16:32:50.544947 135970079123264 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 16:32:50.545089 135970079123264 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:32:50.699171 135970079123264 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 16:32:50.699307 135970079123264 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:32:50.852551 135970079123264 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 16:32:50.852693 135970079123264 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:32:51.018725 135970079123264 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0424 16:32:51.018870 135970079123264 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0424 16:32:51.042350 135970079123264 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0424 16:32:51.042454 135970079123264 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0424 16:33:01.798755 135970079123264 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 16:33:01.806765 135970079123264 metric_logger.py:289] number parameters: 1.104 billion
I0424 16:33:13.303252 135970079123264 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0424 16:33:13.455528 135970079123264 checkpointing.py:798] Waited 0.15225863456726074 seconds for step 0 to finish before starting checkpointing.
I0424 16:33:13.458197 135970079123264 checkpoint_manager.py:2013] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 16:33:13.459973 135970079123264 checkpoint_manager.py:1518] [process=0] Saving checkpoint at step 0
I0424 16:33:13.462635 135970079123264 async_checkpointer.py:452] [process=0] 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_01_base/checkpoints/0.
I0424 16:33:14.064286 135970079123264 signaling_client.py:364] Using JaxDistributedSignalingClient
I0424 16:33:14.322655 135843599591168 atomicity.py:137] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/0
I0424 16:33:14.340214 135970079123264 jax_array_handlers.py:358] Scheduling D2H of 69 prioritized jax.Array.
I0424 16:33:14.340281 135970079123264 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 16:33:14.583018 135970079123264 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.244794s
I0424 16:33:14.583800 135970079123264 base_pytree_checkpoint_handler.py:129] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.015 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5116941928863525 s) (per-host)
I0424 16:33:14.583877 135970079123264 base_pytree_checkpoint_handler.py:737] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.511789s (batch_requests_ready=0.258824s, total_serialization_initiated=0.252268s, others=0.000697s)
I0424 16:33:14.583964 135970079123264 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.515724s (all_items=0.000020s, per_item={'items': '0.00001979'}, temp_paths=0.515704)
I0424 16:33:14.584941 135844664940288 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0424 16:33:14.585110 135970079123264 async_checkpointer.py:561] Finished blocking save. Time taken: 1.125062s. 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_01_base/checkpoints/0.
I0424 16:33:14.869871 135970079123264 checkpoint_manager.py:1566] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 16:33:14.870249 135843591198464 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0424 16:33:14.870419 135970079123264 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_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777048393.4581769, 'wait_for_prev_duration_secs': 6.4849853515625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777048393.4600143, 'checkpointer_blocking_duration_secs': 1.1252026557922363, 'get_old_steps_start_time': 1777048394.5852363, 'get_old_steps_duration_secs': 2.5510787963867188e-05, 'checkpoint_manager_blocking_start_time': 1777048393.4560478, 'checkpoint_manager_blocking_duration_secs': 1.41432785987854}
I0424 16:33:14.870532 135970079123264 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0424 16:33:14.870624 135970079123264 max_utils.py:750] 
Memstats: After params initialized:
I0424 16:33:14.870683 135970079123264 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0))
I0424 16:33:14.870718 135970079123264 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0))
I0424 16:33:14.870746 135970079123264 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0))
I0424 16:33:14.870772 135970079123264 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0))
I0424 16:33:15.179188 135970079123264 metric_logger.py:185] completed step: 0, seconds: 11.142, TFLOP/s/device: 1.219, Tokens/s/device: 183.807, total_weights: 65536, loss: 10.874
I0424 16:33:15.181929 135970079123264 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/tensorboard/'
I0424 16:33:15.601428 135970079123264 metric_logger.py:185] completed step: 1, seconds: 1.872, TFLOP/s/device: 7.256, Tokens/s/device: 1093.749, total_weights: 65536, loss: 10.874
I0424 16:33:15.743253 135970079123264 metric_logger.py:185] completed step: 2, seconds: 0.423, TFLOP/s/device: 32.093, Tokens/s/device: 4837.479, total_weights: 65536, loss: 10.262
I0424 16:33:15.890743 135970079123264 metric_logger.py:185] completed step: 3, seconds: 0.012, TFLOP/s/device: 1176.375, Tokens/s/device: 177316.017, total_weights: 65536, loss: 9.732
I0424 16:33:16.136202 135844681725696 atomicity.py:137] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/0/items
I0424 16:33:34.975746 135844153247488 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1777048395873680673, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1777048395873680673, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/0/_CHECKPOINT_METADATA
I0424 16:33:34.977081 135970079123264 metric_logger.py:185] completed step: 4, seconds: 0.143, TFLOP/s/device: 94.772, Tokens/s/device: 14285.116, total_weights: 65536, loss: 9.274
I0424 16:33:34.986873 135970079123264 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.450, Tokens/s/device: 13935.006, total_weights: 65536, loss: 8.888
I0424 16:33:35.131622 135970079123264 metric_logger.py:185] completed step: 6, seconds: 19.085, TFLOP/s/device: 0.712, Tokens/s/device: 107.310, total_weights: 65536, loss: 8.589
I0424 16:33:35.279061 135970079123264 metric_logger.py:185] completed step: 7, seconds: 0.008, TFLOP/s/device: 1803.920, Tokens/s/device: 271906.532, total_weights: 65536, loss: 8.380
I0424 16:33:35.426610 135970079123264 metric_logger.py:185] completed step: 8, seconds: 0.146, TFLOP/s/device: 92.772, Tokens/s/device: 13983.627, total_weights: 65536, loss: 8.251
I0424 16:33:35.574908 135970079123264 checkpointing.py:794] Waiting for step 9 to finish before checkpoint...
I0424 16:33:35.575564 135970079123264 checkpointing.py:798] Waited 0.000667572021484375 seconds for step 9 to finish before starting checkpointing.
I0424 16:33:35.577626 135970079123264 checkpoint_manager.py:2024] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 16:33:35.986164    2619 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0424 16:33:37.656588 135847894558464 array_metadata_store.py:203] [process=0][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_01_base/checkpoints/0/items/array_metadatas/process_0
I0424 16:34:11.695147 135847886165760 base_pytree_checkpoint_handler.py:1232] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 35.713574s (commit=34.738817s, array_metadata_write=0.974756s)
I0424 16:34:11.696305 135844664940288 base_pytree_checkpoint_handler.py:129] [process=0] /jax/checkpoint/write/gbytes_per_sec: 27.412 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 57.624167680740356 s) (per-host)
I0424 16:34:11.696428 135844664940288 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 57.111373s.
I0424 16:34:12.346709 135844664940288 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1777048395873680673, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/0/_CHECKPOINT_METADATA
I0424 16:34:13.101327 135844153247488 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1777048395873680673, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/0/_CHECKPOINT_METADATA
I0424 16:34:13.444124 135844664940288 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000421s.
I0424 16:34:13.904419 135844664940288 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389).
I0424 16:34:14.860780 135844664940288 base_pytree_checkpoint_handler.py:1356] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.389126s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/0/items
I0424 16:34:14.862536 135844664940288 atomicity.py:608] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/0/items
I0424 16:34:15.291466 135844664940288 atomicity.py:608] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/0
I0424 16:34:16.565647 135844664940288 atomicity.py:790] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/0`.
I0424 16:34:16.566682 135844664940288 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 63.106642s. directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/0
I0424 16:34:16.568200 135844664940288 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 61.983144s.
I0424 16:34:16.568414 135843591198464 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 16:34:16.568550 135843591198464 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0424 16:34:16.568638 135843591198464 checkpoint_manager.py:2133] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0424 16:34:16.570505 135843591198464 checkpoint_manager.py:2142] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0424 16:34:16.570694 135970079123264 checkpoint_manager.py:2036] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0.
W0424 16:34:16.570829 135970079123264 checkpoint_manager.py:1458] Waiting for previous save to complete took 40.993234 seconds. If this number is high, consider checkpointing less frequently.
I0424 16:34:16.574162 135970079123264 checkpoint_manager.py:1518] [process=0] Saving checkpoint at step 9
I0424 16:34:16.577667 135970079123264 async_checkpointer.py:452] [process=0] 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_01_base/checkpoints/9.
I0424 16:34:17.104737 135844664940288 atomicity.py:137] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/9
I0424 16:34:17.116933 135970079123264 jax_array_handlers.py:358] Scheduling D2H of 69 prioritized jax.Array.
I0424 16:34:17.117034 135970079123264 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 16:34:17.143927 135970079123264 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.029064s
I0424 16:34:17.144718 135970079123264 base_pytree_checkpoint_handler.py:129] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.358 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2878763675689697 s) (per-host)
I0424 16:34:17.144792 135970079123264 base_pytree_checkpoint_handler.py:737] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.287968s (batch_requests_ready=0.252470s, total_serialization_initiated=0.034790s, others=0.000708s)
I0424 16:34:17.144876 135970079123264 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.291810s (all_items=0.000016s, per_item={'items': '0.00001574'}, temp_paths=0.291794)
I0424 16:34:17.145662 135843641554688 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0424 16:34:17.145772 135970079123264 async_checkpointer.py:561] Finished blocking save. Time taken: 0.571537s. 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_01_base/checkpoints/9.
I0424 16:34:17.454830 135970079123264 checkpoint_manager.py:1566] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize
I0424 16:34:17.455192 135843591198464 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0424 16:34:17.455373 135970079123264 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_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777048415.5775685, 'wait_for_prev_duration_secs': 40.993234395980835, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777048456.5742025, 'checkpointer_blocking_duration_secs': 0.5716714859008789, 'get_old_steps_start_time': 1777048457.1458962, 'get_old_steps_duration_secs': 2.8848648071289062e-05, 'checkpoint_manager_blocking_start_time': 1777048415.575798, 'checkpoint_manager_blocking_duration_secs': 41.879539251327515}
I0424 16:34:17.455485 135970079123264 checkpointing.py:409] Started an asynchronous checkpoint save for step 9
I0424 16:34:17.455530 135970079123264 checkpoint_manager.py:2024] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0424 16:34:18.882972 135844673332992 atomicity.py:137] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/9/items
I0424 16:34:22.124593 135844681725696 array_metadata_store.py:203] [process=0][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_01_base/checkpoints/9/items/array_metadatas/process_0
I0424 16:35:01.651631 135844128069376 base_pytree_checkpoint_handler.py:1232] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.289135s (commit=40.318086s, array_metadata_write=0.971049s)
I0424 16:35:01.652701 135843641554688 base_pytree_checkpoint_handler.py:129] [process=0] /jax/checkpoint/write/gbytes_per_sec: 35.262 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 44.79585242271423 s) (per-host)
I0424 16:35:01.652766 135843641554688 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 44.507037s.
I0424 16:35:03.019855 135843641554688 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000378s.
I0424 16:35:03.253727 135843641554688 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389).
I0424 16:35:03.892770 135843641554688 base_pytree_checkpoint_handler.py:1356] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.474945s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/9/items
I0424 16:35:03.895123 135843641554688 atomicity.py:608] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/9/items
I0424 16:35:04.336121 135843641554688 atomicity.py:608] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/9
I0424 16:35:05.659288 135843641554688 atomicity.py:790] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/9`.
I0424 16:35:05.660125 135843641554688 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 49.085895s. directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260424_145744/nnx_xpk_feat_nnx_set_defaults_true_20260424_145744_01_base/checkpoints/9
I0424 16:35:05.661751 135843641554688 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 48.516020s.
I0424 16:35:05.661955 135843591198464 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0424 16:35:05.662089 135843591198464 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0424 16:35:05.662167 135843591198464 checkpoint_manager.py:2133] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0424 16:35:05.663675 135843591198464 checkpoint_manager.py:2142] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts.
I0424 16:35:05.663851 135970079123264 checkpoint_manager.py:2036] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9.
I0424 16:35:05.664016 135970079123264 checkpoint_manager.py:2013] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0424 16:35:05.664688 135970079123264 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.354, Tokens/s/device: 13920.609, 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 16:35:17 UTC 2026
EXIT_CODE=0