2026-04-23 03:54:26.956752: 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 00:00:1776916466.969775 2399682 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776916466.973664 2399682 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776916466.985712 2399682 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776916466.985726 2399682 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776916466.985728 2399682 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776916466.985730 2399682 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} `rope_scaling`'s factor field must be a float >= 1, got 40 `rope_scaling`'s beta_fast field must be a float, got 32 `rope_scaling`'s beta_slow field must be a float, got 1 Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} 2026-04-23 03:54:30.671368: 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) I0423 03:54:30.806473 127994456946496 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0423 03:55:09.388257 127994456946496 max_utils.py:800] System Information: Jax Version: 0.8.1 I0423 03:55:09.388497 127994456946496 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0423 03:55:09.388546 127994456946496 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 I0423 03:55:09.388577 127994456946496 train_utils.py:348] WARNING: 'dataset_path' might be pointing your local file system I0423 03:55:09.388603 127994456946496 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0423 03:55:09.388768 127994456946496 train.py:672] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0423 03:55:10.049587 127994456946496 maxtext_utils.py:1604] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 03:55:10.049847 127994456946496 checkpointing.py:677] Setting up checkpoint logger... I0423 03:55:10.049897 127994456946496 checkpointing.py:233] Creating checkpoint manager with ocdbt=False and zarr3=True I0423 03:55:10.049933 127994456946496 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0423 03:55:10.050373 127994456946496 base_pytree_checkpoint_handler.py:411] Created BasePyTreeCheckpointHandler: use_ocdbt=False, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7468be9aaf00>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0423 03:55:11.718663 127994456946496 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0423 03:55:11.719023 127994456946496 checkpoint_manager.py:702] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x746761d4db20>}, handler_registry=None I0423 03:55:11.719251 127994456946496 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x746761d4db20>` 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`. I0423 03:55:11.719292 127994456946496 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7466e7305880>` 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`. I0423 03:55:11.719322 127994456946496 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 0x746761d4db20>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x746761d4db20>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7466e7305880>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7466e7305880>}). I0423 03:55:11.719594 127994456946496 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0423 03:55:11.719644 127994456946496 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7466e73244a0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0423 03:55:12.249391 127964511471168 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776916423850688054, 'commit_timestamp_nsecs': 1776916437188208528, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/9/_CHECKPOINT_METADATA I0423 03:55:12.329169 127964511471168 checkpoint_manager.py:1847] Missing metrics for step 9 I0423 03:55:12.391304 127994456946496 checkpoint_manager.py:1788] Found 3 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints I0423 03:55:12.391822 127994456946496 checkpoint_manager.py:921] [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_hns=False, 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=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x74676314faa0> I0423 03:55:12.391918 127994456946496 checkpointing.py:301] Checkpoint manager created! I0423 03:55:12.790968 127994456946496 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0423 03:55:12.791087 127994456946496 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 03:55:13.070605 127994456946496 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 03:55:13.070685 127994456946496 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 03:55:13.082891 127994456946496 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 03:55:13.082939 127994456946496 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 03:55:13.101766 127994456946496 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 03:55:13.101823 127994456946496 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 03:55:13.114024 127994456946496 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 03:55:13.114090 127994456946496 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 03:55:13.126158 127994456946496 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 03:55:13.126206 127994456946496 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 03:55:13.145520 127994456946496 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0423 03:55:13.145574 127994456946496 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 03:55:13.160858 127994456946496 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0423 03:55:13.160912 127994456946496 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0423 03:55:13.312351 127994456946496 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0423 03:55:13.312443 127994456946496 checkpointing.py:581] restoring from this run's directory step 9 I0423 03:55:13.494854 127994456946496 checkpointer.py:304] Restoring checkpoint from gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/9. I0423 03:55:14.379539 2401244 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0423 03:55:14.379599 2401244 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0423 03:55:19.879741 127994456946496 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/read/gbytes_per_sec: 2.053 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host) I0423 03:55:19.879993 127994456946496 checkpointer.py:318] Finished restoring checkpoint in 6.47 seconds from gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/9. I0423 03:55:19.880071 127994456946496 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints', 'checkpointer_start_time': 1776916513.409995, 'checkpointer_duration_secs': 6.470040798187256, 'checkpoint_manager_start_time': 1776916513.3126452, 'checkpoint_manager_duration_secs': 6.567391633987427} [DECOUPLED NO-OP] gcs_storage: using stubs. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] workload_monitor: using stub. [DECOUPLED NO-OP] vertex_tensorboard: using stub. fsdp: 8 I0423 03:55:20.226250 127994456946496 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0423 03:55:20.226427 127994456946496 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0423 03:55:20.226471 127994456946496 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0423 03:55:20.226516 127994456946496 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0423 03:55:20.226557 127994456946496 maxtext_utils.py:1707] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0423 03:55:20.226588 127994456946496 maxtext_utils.py:1707] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0423 03:55:20.226630 127994456946496 maxtext_utils.py:1707] 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) I0423 03:55:20.226680 127994456946496 maxtext_utils.py:1707] 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') I0423 03:55:20.226710 127994456946496 maxtext_utils.py:1707] 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) I0423 03:55:20.226736 127994456946496 maxtext_utils.py:1707] 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) I0423 03:55:20.226772 127994456946496 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0423 03:55:20.226810 127994456946496 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0423 03:55:20.636449 127994456946496 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 03:55:20.636567 127994456946496 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0423 03:55:20.648187 127994456946496 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 03:55:20.648236 127994456946496 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0423 03:55:21.430387 127994456946496 max_utils.py:791] Total memory size: 3.6 GB, Output size: 1.5 GB, Temp size: 2.0 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0423 03:55:21.433965 127994456946496 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0423 03:55:21.434157 127994456946496 metric_logger.py:301] number parameters: 1.104 billion I0423 03:55:23.585439 127994456946496 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0423 03:55:23.702635 127994456946496 checkpointing.py:776] Waited 0.11718583106994629 seconds for step 10 to finish before starting checkpointing. I0423 03:55:23.703277 127994456946496 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 03:55:23.703444 127994456946496 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 10 I0423 03:55:23.703751 127994456946496 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/10. I0423 03:55:23.793298 127994456946496 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0423 03:55:23.872230 127994456946496 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 03:55:23.872369 127963674707520 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/10 I0423 03:55:23.872445 127994456946496 replica_slices.py:410] 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 I0423 03:55:24.605962 127963664221760 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/10/items I0423 03:55:24.782509 127962861012544 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776916524518170756, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776916524518170756, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/10/_CHECKPOINT_METADATA I0423 03:55:31.056236 127994456946496 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 7.184580s I0423 03:55:31.056816 127994456946496 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.699 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 7 seconds) (per-host) I0423 03:55:31.056871 127994456946496 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 7.262392s (batch_requests_ready=0.074515s, total_serialization_initiated=7.187394s, others=0.000483s) I0423 03:55:31.056937 127994456946496 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 7.263186s (all_items=0.000020s, per_item={'items': '0.00001955'}, temp_paths=7.263166) I0423 03:55:31.062099 127963685193280 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 03:55:31.062203 127994456946496 async_checkpointer.py:561] Finished blocking save. Time taken: 7.358717s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/10. I0423 03:55:31.062388 127994456946496 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0423 03:55:31.062574 127962663880256 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 03:55:31.062681 127994456946496 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776916523.7032638, 'wait_for_prev_duration_secs': 4.172325134277344e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776916523.703466, 'checkpointer_blocking_duration_secs': 7.358847618103027, 'get_old_steps_start_time': 1776916531.0623274, 'get_old_steps_duration_secs': 2.2649765014648438e-05, 'checkpoint_manager_blocking_start_time': 1776916523.7031846, 'checkpoint_manager_blocking_duration_secs': 7.359469175338745} I0423 03:55:31.062791 127994456946496 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0423 03:55:31.062829 127994456946496 max_utils.py:750] Memstats: After params initialized: I0423 03:55:31.062876 127994456946496 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0423 03:55:31.062903 127994456946496 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0423 03:55:31.062922 127994456946496 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0423 03:55:31.062941 127994456946496 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0423 03:55:31.062958 127994456946496 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0423 03:55:31.062975 127994456946496 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0423 03:55:31.062991 127994456946496 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0423 03:55:31.063007 127994456946496 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0423 03:55:31.334110 127994456946496 metric_logger.py:196] completed step: 10, seconds: 2.151, TFLOP/s/device: 6.318, Tokens/s/device: 952.258, total_weights: 16384, loss: 3.602, lm_loss: 3.602, perplexity: 36.686 I0423 03:55:31.450589 127994456946496 metric_logger.py:196] completed step: 11, seconds: 7.746, TFLOP/s/device: 1.754, Tokens/s/device: 264.378, total_weights: 16384, loss: 3.101, lm_loss: 3.101, perplexity: 22.211 I0423 03:55:31.570791 127994456946496 metric_logger.py:196] completed step: 12, seconds: 0.006, TFLOP/s/device: 2157.372, Tokens/s/device: 325182.598, total_weights: 16384, loss: 2.684, lm_loss: 2.684, perplexity: 14.644 I0423 03:55:31.719381 127994456946496 metric_logger.py:196] completed step: 13, seconds: 0.120, TFLOP/s/device: 113.505, Tokens/s/device: 17108.726, total_weights: 16384, loss: 2.348, lm_loss: 2.348, perplexity: 10.461 I0423 03:55:31.813428 127994456946496 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0423 03:55:31.838792 127963504838208 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/10/items/array_metadatas/process_0 I0423 03:55:31.930907 127994456946496 checkpointing.py:776] Waited 0.11746621131896973 seconds for step 15 to finish before starting checkpointing. I0423 03:55:31.932122 127994456946496 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 03:55:38.259312 127963664221760 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 7.198291s (commit=6.765560s, array_metadata_write=0.432730s) I0423 03:55:38.260269 127963685193280 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 873.567 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 14 seconds) (per-host) I0423 03:55:38.260374 127963685193280 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 7.198124s. W0423 03:55:38.597853 127963685193280 ocdbt_utils.py:175] [process=0][thread=async_save] Skipping merge of OCDBT checkpoints: No per-process OCDBT checkpoint subdirs found in gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/10/items, I0423 03:55:38.743536 127963685193280 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 03:55:38.743649 127963685193280 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.203984s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/10/items I0423 03:55:38.744347 127963685193280 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/10/items I0423 03:55:38.917402 127962861012544 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776916524518170756, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/10/_CHECKPOINT_METADATA I0423 03:55:38.994116 127963685193280 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/10 I0423 03:55:39.705807 127963685193280 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/10`. I0423 03:55:39.706410 127963685193280 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 16.002932s. directory=gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/10 I0423 03:55:39.706487 127963685193280 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 8.644240s. I0423 03:55:39.706677 127962663880256 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 03:55:39.706783 127962663880256 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 03:55:39.706839 127962663880256 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0423 03:55:39.706885 127962663880256 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0423 03:55:39.707084 127994456946496 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0423 03:55:39.707192 127994456946496 checkpoint_manager.py:1441] Waiting for previous save to complete took 7.775090 seconds. If this number is high, consider checkpointing less frequently. I0423 03:55:39.707805 127994456946496 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 15 I0423 03:55:39.708097 127994456946496 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/15. I0423 03:55:39.871441 127994456946496 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 03:55:39.871535 127994456946496 replica_slices.py:410] 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 I0423 03:55:39.882129 127963685193280 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/15 I0423 03:55:40.596170 127963504838208 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/15/items I0423 03:55:42.116026 127994456946496 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 2.244979s I0423 03:55:42.120390 127994456946496 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.307 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 2 seconds) (per-host) I0423 03:55:42.120459 127994456946496 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 2.325251s (batch_requests_ready=0.073004s, total_serialization_initiated=2.247989s, others=0.004258s) I0423 03:55:42.120540 127994456946496 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.325789s (all_items=0.000012s, per_item={'items': '0.00001240'}, temp_paths=2.325776) I0423 03:55:42.122169 127963685193280 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 03:55:42.122254 127994456946496 async_checkpointer.py:561] Finished blocking save. Time taken: 2.414404s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/15. I0423 03:55:42.122439 127994456946496 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0423 03:55:42.122595 127962663880256 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 03:55:42.122699 127994456946496 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776916531.9320827, 'wait_for_prev_duration_secs': 7.775090456008911, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776916539.7078302, 'checkpointer_blocking_duration_secs': 2.4145331382751465, 'get_old_steps_start_time': 1776916542.1223793, 'get_old_steps_duration_secs': 2.1219253540039062e-05, 'checkpoint_manager_blocking_start_time': 1776916531.931958, 'checkpoint_manager_blocking_duration_secs': 10.190710544586182} I0423 03:55:42.122811 127994456946496 checkpointing.py:408] Started an asynchronous checkpoint save for step 15 I0423 03:55:42.123358 127994456946496 metric_logger.py:196] completed step: 14, seconds: 0.119, TFLOP/s/device: 113.768, Tokens/s/device: 17148.407, total_weights: 16384, loss: 2.083, lm_loss: 2.083, perplexity: 8.030 I0423 03:55:42.131577 127994456946496 metric_logger.py:196] completed step: 15, seconds: 0.236, TFLOP/s/device: 57.661, Tokens/s/device: 8691.298, total_weights: 16384, loss: 1.880, lm_loss: 1.880, perplexity: 6.551 I0423 03:55:42.247539 127994456946496 metric_logger.py:196] completed step: 16, seconds: 10.316, TFLOP/s/device: 1.317, Tokens/s/device: 198.521, total_weights: 16384, loss: 1.726, lm_loss: 1.726, perplexity: 5.616 I0423 03:55:42.414181 127994456946496 metric_logger.py:196] completed step: 17, seconds: 0.007, TFLOP/s/device: 2078.814, Tokens/s/device: 313341.493, total_weights: 16384, loss: 1.610, lm_loss: 1.610, perplexity: 5.002 I0423 03:55:42.487272 127994456946496 metric_logger.py:196] completed step: 18, seconds: 0.117, TFLOP/s/device: 115.730, Tokens/s/device: 17444.039, total_weights: 16384, loss: 1.521, lm_loss: 1.521, perplexity: 4.577 I0423 03:55:42.610748 127994456946496 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0423 03:55:42.642467 127994456946496 checkpointing.py:776] Waited 0.03171849250793457 seconds for step 19 to finish before starting checkpointing. I0423 03:55:42.642958 127994456946496 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 03:55:42.924860 127963674707520 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/15/items/array_metadatas/process_0 I0423 03:55:50.300760 127963504838208 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 8.179504s (commit=7.749604s, array_metadata_write=0.429900s) I0423 03:55:50.301728 127963685193280 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 1.175 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 10 seconds) (per-host) I0423 03:55:50.301832 127963685193280 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 8.179531s. W0423 03:55:50.667234 127963685193280 ocdbt_utils.py:175] [process=0][thread=async_save] Skipping merge of OCDBT checkpoints: No per-process OCDBT checkpoint subdirs found in gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/15/items, I0423 03:55:50.779038 127963685193280 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 03:55:50.779162 127963685193280 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.171420s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/15/items I0423 03:55:50.779786 127963685193280 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/15/items I0423 03:55:51.015294 127963685193280 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/15 I0423 03:55:51.741398 127963685193280 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/15`. I0423 03:55:51.742022 127963685193280 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 12.034180s. directory=gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/15 I0423 03:55:51.742113 127963685193280 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 9.619813s. I0423 03:55:51.742308 127962663880256 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 03:55:51.742413 127962663880256 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 03:55:51.742466 127962663880256 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0423 03:55:51.742503 127962663880256 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0423 03:55:51.742604 127994456946496 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0423 03:55:51.742658 127994456946496 checkpoint_manager.py:1441] Waiting for previous save to complete took 9.099719 seconds. If this number is high, consider checkpointing less frequently. I0423 03:55:51.742922 127994456946496 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 19 I0423 03:55:51.743208 127994456946496 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/19. I0423 03:55:51.889718 127963685193280 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/19 I0423 03:55:51.902805 127994456946496 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 03:55:51.902893 127994456946496 replica_slices.py:410] 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 I0423 03:55:52.542857 127963674707520 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/19/items I0423 03:55:54.618999 127994456946496 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 2.716563s I0423 03:55:54.623175 127994456946496 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.400 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 2 seconds) (per-host) I0423 03:55:54.623231 127994456946496 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 2.804952s (batch_requests_ready=0.080741s, total_serialization_initiated=2.720150s, others=0.004061s) I0423 03:55:54.623308 127994456946496 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.805676s (all_items=0.000015s, per_item={'items': '0.00001454'}, temp_paths=2.805662) I0423 03:55:54.624834 127963685193280 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 03:55:54.624954 127994456946496 async_checkpointer.py:561] Finished blocking save. Time taken: 2.881985s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/19. I0423 03:55:54.625135 127994456946496 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0423 03:55:54.625299 127962663880256 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 03:55:54.625393 127994456946496 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776916542.642928, 'wait_for_prev_duration_secs': 9.099718809127808, 'time_between_consecutive_saves_sec': 2.9360227584838867, 'checkpointer_blocking_start_time': 1776916551.742947, 'checkpointer_blocking_duration_secs': 2.8821263313293457, 'get_old_steps_start_time': 1776916554.6250834, 'get_old_steps_duration_secs': 1.8835067749023438e-05, 'checkpoint_manager_blocking_start_time': 1776916542.6428838, 'checkpoint_manager_blocking_duration_secs': 11.982485055923462} I0423 03:55:54.625517 127994456946496 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0423 03:55:54.625549 127994456946496 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 03:55:55.399400 127963664221760 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/19/items/array_metadatas/process_0 I0423 03:56:01.302119 127963674707520 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 6.678245s (commit=6.224186s, array_metadata_write=0.454059s) I0423 03:56:01.303025 127963685193280 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 1.301 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 9 seconds) (per-host) I0423 03:56:01.303137 127963685193280 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 6.678140s. W0423 03:56:01.641923 127963685193280 ocdbt_utils.py:175] [process=0][thread=async_save] Skipping merge of OCDBT checkpoints: No per-process OCDBT checkpoint subdirs found in gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/19/items, I0423 03:56:01.762616 127963685193280 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 03:56:01.762728 127963685193280 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.164844s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/19/items I0423 03:56:01.763368 127963685193280 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/19/items I0423 03:56:01.992171 127963685193280 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/19 I0423 03:56:02.692705 127963685193280 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/19`. I0423 03:56:02.693338 127963685193280 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 10.950380s. directory=gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_15_ocdbt_false/checkpoints/19 I0423 03:56:02.693399 127963685193280 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 8.068404s. I0423 03:56:02.693534 127962663880256 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 03:56:02.693641 127962663880256 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 03:56:02.693690 127962663880256 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0423 03:56:02.693729 127962663880256 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0423 03:56:02.693925 127994456946496 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0423 03:56:02.694066 127994456946496 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 03:56:02.694766 127994456946496 metric_logger.py:196] completed step: 19, seconds: 0.228, TFLOP/s/device: 59.613, Tokens/s/device: 8985.530, total_weights: 16384, loss: 1.449, lm_loss: 1.449, perplexity: 4.260 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops