2026-04-23 03:56:25.480393: 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 00:00:1776916585.494085 2402236 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776916585.498129 2402236 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776916585.510619 2402236 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776916585.510631 2402236 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776916585.510633 2402236 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776916585.510634 2402236 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 03:56:29.195071: 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 03:56:29.330565 131615831762752 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. ~/maxtext/maxtext_venv/lib/python3.12/site-packages/jax/_src/xla_bridge.py:204: UserWarning: TPU backend initialization is taking more than 60.0 seconds. Did you run your code on all TPU hosts? See https://docs.jax.dev/en/latest/multi_process.html for more information. warnings.warn( I0423 03:57:34.435028 131615831762752 max_utils.py:800] System Information: Jax Version: 0.8.1 I0423 03:57:34.435290 131615831762752 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0423 03:57:34.435343 131615831762752 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 03:57:34.435372 131615831762752 train_utils.py:348] WARNING: 'dataset_path' might be pointing your local file system I0423 03:57:34.435397 131615831762752 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0423 03:57:34.435558 131615831762752 train.py:672] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0423 03:57:35.100258 131615831762752 maxtext_utils.py:1604] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 03:57:35.100519 131615831762752 checkpointing.py:677] Setting up checkpoint logger... I0423 03:57:35.100565 131615831762752 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0423 03:57:35.100601 131615831762752 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0423 03:57:35.101039 131615831762752 base_pytree_checkpoint_handler.py:411] 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 0x77b3e95aaba0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0423 03:57:37.491532 131615831762752 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0423 03:57:37.491894 131615831762752 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 0x77b28dde63f0>}, handler_registry=None I0423 03:57:37.492130 131615831762752 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x77b28dde63f0>` 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 03:57:37.492174 131615831762752 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x77b2187c7020>` 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 03:57:37.492204 131615831762752 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 0x77b28dde63f0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x77b28dde63f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x77b2187c7020>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x77b2187c7020>}). I0423 03:57:37.492469 131615831762752 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0423 03:57:37.492519 131615831762752 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x77b2187f8540> timeout: 600 secs and primary_host=0 for async checkpoint writes I0423 03:57:37.620479 131615831762752 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints I0423 03:57:37.620691 131615831762752 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_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x77b2187c6e40> I0423 03:57:37.620967 131615831762752 checkpointing.py:301] Checkpoint manager created! I0423 03:57:38.006972 131615831762752 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0423 03:57:38.007078 131615831762752 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 03:57:38.286628 131615831762752 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 03:57:38.286709 131615831762752 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 03:57:38.298911 131615831762752 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 03:57:38.298959 131615831762752 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 03:57:38.317788 131615831762752 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 03:57:38.317849 131615831762752 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 03:57:38.330070 131615831762752 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 03:57:38.330120 131615831762752 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 03:57:38.342246 131615831762752 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 03:57:38.342295 131615831762752 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 03:57:38.361490 131615831762752 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0423 03:57:38.361546 131615831762752 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 03:57:38.376862 131615831762752 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0423 03:57:38.376915 131615831762752 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0423 03:57:38.526602 131615831762752 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0423 03:57:38.526697 131615831762752 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. [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 03:57:39.155143 131615831762752 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0423 03:57:39.155257 131615831762752 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 03:57:39.155301 131615831762752 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 03:57:39.155347 131615831762752 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 03:57:39.155387 131615831762752 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 03:57:39.155420 131615831762752 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 03:57:39.155460 131615831762752 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 03:57:39.155502 131615831762752 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 03:57:39.155531 131615831762752 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 03:57:39.155557 131615831762752 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 03:57:39.155598 131615831762752 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0423 03:57:39.155637 131615831762752 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0423 03:57:39.560697 131615831762752 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 03:57:39.560779 131615831762752 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0423 03:57:39.572834 131615831762752 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 03:57:39.572882 131615831762752 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0423 03:57:40.330583 131615831762752 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 03:57:40.331132 131615831762752 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0423 03:57:40.331284 131615831762752 metric_logger.py:301] number parameters: 1.104 billion I0423 03:57:42.406959 131615831762752 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0423 03:57:42.524999 131615831762752 checkpointing.py:776] Waited 0.1180257797241211 seconds for step 0 to finish before starting checkpointing. I0423 03:57:42.525603 131615831762752 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 03:57:42.525754 131615831762752 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0423 03:57:42.526131 131615831762752 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_16_zarr3_true/checkpoints/0. I0423 03:57:42.605789 131615831762752 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0423 03:57:42.692011 131584497813056 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_16_zarr3_true/checkpoints/0 I0423 03:57:42.701496 131615831762752 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 03:57:42.701584 131615831762752 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 03:57:43.368291 131584487327296 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_16_zarr3_true/checkpoints/0/items I0423 03:57:43.571066 131584271320640 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776916663267789272, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776916663267789272, "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_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0423 03:57:43.833895 2404485 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0423 03:57:43.833945 2404485 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0423 03:57:45.165856 131615831762752 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 2.464740s I0423 03:57:45.170601 131615831762752 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.813 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 2 seconds) (per-host) I0423 03:57:45.170673 131615831762752 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 2.563966s (batch_requests_ready=0.090601s, total_serialization_initiated=2.468710s, others=0.004654s) I0423 03:57:45.170750 131615831762752 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.564497s (all_items=0.000018s, per_item={'items': '0.00001836'}, temp_paths=2.564478) I0423 03:57:45.172164 131584508298816 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 03:57:45.172244 131615831762752 async_checkpointer.py:561] Finished blocking save. Time taken: 2.646450s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints/0. I0423 03:57:45.172421 131615831762752 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0423 03:57:45.172596 131583667340864 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 03:57:45.172689 131615831762752 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776916662.525584, 'wait_for_prev_duration_secs': 4.935264587402344e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776916662.5257752, 'checkpointer_blocking_duration_secs': 2.6465747356414795, 'get_old_steps_start_time': 1776916665.172366, 'get_old_steps_duration_secs': 2.002716064453125e-05, 'checkpoint_manager_blocking_start_time': 1776916662.525505, 'checkpoint_manager_blocking_duration_secs': 2.6471590995788574} I0423 03:57:45.172798 131615831762752 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0423 03:57:45.172836 131615831762752 max_utils.py:750] Memstats: After params initialized: I0423 03:57:45.172890 131615831762752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0423 03:57:45.172915 131615831762752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0423 03:57:45.172936 131615831762752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0423 03:57:45.172954 131615831762752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0423 03:57:45.172972 131615831762752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0423 03:57:45.172989 131615831762752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0423 03:57:45.173006 131615831762752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0423 03:57:45.173023 131615831762752 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0423 03:57:45.432932 131615831762752 metric_logger.py:196] completed step: 0, seconds: 2.075, TFLOP/s/device: 6.547, Tokens/s/device: 986.906, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0423 03:57:45.433876 131615831762752 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/tensorboard/' I0423 03:57:45.549205 131615831762752 metric_logger.py:196] completed step: 1, seconds: 3.024, TFLOP/s/device: 4.493, Tokens/s/device: 677.257, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0423 03:57:45.668035 131615831762752 metric_logger.py:196] completed step: 2, seconds: 0.006, TFLOP/s/device: 2200.701, Tokens/s/device: 331713.638, total_weights: 16384, loss: 10.276, lm_loss: 10.276, perplexity: 29019.973 I0423 03:57:45.787272 131615831762752 metric_logger.py:196] completed step: 3, seconds: 0.117, TFLOP/s/device: 115.936, Tokens/s/device: 17475.148, total_weights: 16384, loss: 9.282, lm_loss: 9.282, perplexity: 10739.224 I0423 03:57:45.792405 131615831762752 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0423 03:57:45.943372 131584466355776 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_16_zarr3_true/checkpoints/0/items/array_metadatas/process_0 I0423 03:57:46.024926 131615831762752 checkpointing.py:776] Waited 0.23249077796936035 seconds for step 5 to finish before starting checkpointing. I0423 03:57:46.025884 131615831762752 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 03:58:20.414854 131583868667456 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 35.243376s (commit=34.770367s, array_metadata_write=0.473009s) I0423 03:58:20.415869 131584508298816 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 334.226 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 37 seconds) (per-host) I0423 03:58:20.415970 131584508298816 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 35.243681s. I0423 03:58:20.667309 131584508298816 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776916663267789272, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0423 03:58:20.865987 131584508298816 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 03:58:21.098575 131584271320640 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776916663267789272, '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_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0423 03:58:21.299934 131584508298816 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 03:58:21.300122 131584508298816 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.596863s. 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_16_zarr3_true/checkpoints/0/items I0423 03:58:21.300763 131584508298816 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints/0/items I0423 03:58:21.553631 131584508298816 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints/0 I0423 03:58:22.273903 131584508298816 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_16_zarr3_true/checkpoints/0`. I0423 03:58:22.274541 131584508298816 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 39.748755s. directory=gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints/0 I0423 03:58:22.274616 131584508298816 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 37.102329s. I0423 03:58:22.274810 131583667340864 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 03:58:22.274920 131583667340864 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 03:58:22.274987 131583667340864 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0423 03:58:22.275034 131583667340864 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0423 03:58:22.275189 131615831762752 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0423 03:58:22.275315 131615831762752 checkpoint_manager.py:1441] Waiting for previous save to complete took 36.249432 seconds. If this number is high, consider checkpointing less frequently. I0423 03:58:22.276670 131615831762752 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 5 I0423 03:58:22.276962 131615831762752 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_16_zarr3_true/checkpoints/5. I0423 03:58:22.456656 131584508298816 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_16_zarr3_true/checkpoints/5 I0423 03:58:22.457014 131615831762752 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 03:58:22.457180 131615831762752 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 03:58:23.171941 131584466355776 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_16_zarr3_true/checkpoints/5/items I0423 03:58:23.223780 131615831762752 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.767398s I0423 03:58:23.224197 131615831762752 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 14.437 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 854 milliseconds) (per-host) I0423 03:58:23.224253 131615831762752 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.854872s (batch_requests_ready=0.083614s, total_serialization_initiated=0.770937s, others=0.000321s) I0423 03:58:23.224316 131615831762752 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.855388s (all_items=0.000016s, per_item={'items': '0.00001621'}, temp_paths=0.855371) I0423 03:58:23.225082 131584286000704 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 03:58:23.225175 131615831762752 async_checkpointer.py:561] Finished blocking save. Time taken: 0.948463s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints/5. I0423 03:58:23.225359 131615831762752 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0423 03:58:23.225536 131615831762752 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776916666.0258598, 'wait_for_prev_duration_secs': 36.24943208694458, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776916702.2766943, 'checkpointer_blocking_duration_secs': 0.948584794998169, 'get_old_steps_start_time': 1776916703.225298, 'get_old_steps_duration_secs': 2.4080276489257812e-05, 'checkpoint_manager_blocking_start_time': 1776916666.025787, 'checkpoint_manager_blocking_duration_secs': 37.1997287273407} I0423 03:58:23.225641 131615831762752 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0423 03:58:23.226258 131615831762752 metric_logger.py:196] completed step: 4, seconds: 0.119, TFLOP/s/device: 114.455, Tokens/s/device: 17251.836, total_weights: 16384, loss: 8.327, lm_loss: 8.327, perplexity: 4134.812 I0423 03:58:23.234792 131615831762752 metric_logger.py:196] completed step: 5, seconds: 0.119, TFLOP/s/device: 113.980, Tokens/s/device: 17180.343, total_weights: 16384, loss: 7.367, lm_loss: 7.367, perplexity: 1583.522 I0423 03:58:23.326968 131583667340864 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 03:58:23.351033 131615831762752 metric_logger.py:196] completed step: 6, seconds: 37.441, TFLOP/s/device: 0.363, Tokens/s/device: 54.700, total_weights: 16384, loss: 6.459, lm_loss: 6.459, perplexity: 638.363 I0423 03:58:23.470361 131615831762752 metric_logger.py:196] completed step: 7, seconds: 0.006, TFLOP/s/device: 2355.197, Tokens/s/device: 355000.867, total_weights: 16384, loss: 5.622, lm_loss: 5.622, perplexity: 276.577 I0423 03:58:23.589433 131615831762752 metric_logger.py:196] completed step: 8, seconds: 0.118, TFLOP/s/device: 115.512, Tokens/s/device: 17411.265, total_weights: 16384, loss: 4.864, lm_loss: 4.864, perplexity: 129.523 I0423 03:58:23.707586 131615831762752 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0423 03:58:23.708672 131615831762752 checkpointing.py:776] Waited 0.001116037368774414 seconds for step 9 to finish before starting checkpointing. I0423 03:58:23.708937 131615831762752 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 03:58:24.373878 131584296486464 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_16_zarr3_true/checkpoints/5/items/array_metadatas/process_0 I0423 03:59:09.240955 131583868667456 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 45.614224s (commit=45.110739s, array_metadata_write=0.503485s) I0423 03:59:09.241926 131584286000704 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 269.599 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 46 seconds) (per-host) I0423 03:59:09.242030 131584286000704 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 46.016808s. I0423 03:59:09.723865 131584286000704 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 03:59:10.139188 131584286000704 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 03:59:10.139344 131584286000704 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.603550s. 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_16_zarr3_true/checkpoints/5/items I0423 03:59:10.140131 131584286000704 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints/5/items I0423 03:59:10.382139 131584286000704 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints/5 I0423 03:59:11.113436 131584286000704 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_16_zarr3_true/checkpoints/5`. I0423 03:59:11.114061 131584286000704 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 48.837343s. directory=gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints/5 I0423 03:59:11.114135 131584286000704 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 47.888916s. I0423 03:59:11.114321 131583667340864 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 03:59:11.114429 131583667340864 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 03:59:11.114481 131583667340864 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0423 03:59:11.114518 131583667340864 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0423 03:59:11.114665 131615831762752 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0423 03:59:11.114782 131615831762752 checkpoint_manager.py:1441] Waiting for previous save to complete took 47.405852 seconds. If this number is high, consider checkpointing less frequently. I0423 03:59:11.115069 131615831762752 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0423 03:59:11.115351 131615831762752 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_16_zarr3_true/checkpoints/9. I0423 03:59:11.278173 131615831762752 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 03:59:11.278340 131615831762752 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 03:59:11.279432 131584286000704 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_16_zarr3_true/checkpoints/9 I0423 03:59:11.988210 131584296486464 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_16_zarr3_true/checkpoints/9/items I0423 03:59:16.747714 131615831762752 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 5.470017s I0423 03:59:16.752299 131615831762752 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.221 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 5 seconds) (per-host) I0423 03:59:16.752361 131615831762752 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 5.556989s (batch_requests_ready=0.079195s, total_serialization_initiated=5.473372s, others=0.004422s) I0423 03:59:16.752443 131615831762752 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 5.557494s (all_items=0.000011s, per_item={'items': '0.00001121'}, temp_paths=5.557482) I0423 03:59:16.755122 131584286000704 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 03:59:16.755222 131615831762752 async_checkpointer.py:561] Finished blocking save. Time taken: 5.640108s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints/9. I0423 03:59:16.755425 131615831762752 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0423 03:59:16.755588 131583667340864 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 03:59:16.755697 131615831762752 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776916703.7089124, 'wait_for_prev_duration_secs': 47.40585160255432, 'time_between_consecutive_saves_sec': 1.4338507652282715, 'checkpointer_blocking_start_time': 1776916751.115096, 'checkpointer_blocking_duration_secs': 5.640244245529175, 'get_old_steps_start_time': 1776916756.7553577, 'get_old_steps_duration_secs': 2.6464462280273438e-05, 'checkpoint_manager_blocking_start_time': 1776916703.708883, 'checkpoint_manager_blocking_duration_secs': 53.04678726196289} I0423 03:59:16.755821 131615831762752 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0423 03:59:16.755862 131615831762752 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 03:59:17.552564 131584508298816 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_16_zarr3_true/checkpoints/9/items/array_metadatas/process_0 I0423 03:59:52.796628 131584296486464 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 36.043736s (commit=35.543324s, array_metadata_write=0.500412s) I0423 03:59:52.803464 131584286000704 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 303.710 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 41 seconds) (per-host) I0423 03:59:52.803570 131584286000704 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.048299s. I0423 03:59:53.283363 131584286000704 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 03:59:53.676087 131584286000704 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 03:59:53.676288 131584286000704 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.579756s. 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_16_zarr3_true/checkpoints/9/items I0423 03:59:53.676980 131584286000704 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints/9/items I0423 03:59:53.913180 131584286000704 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints/9 I0423 03:59:54.611621 131584286000704 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_16_zarr3_true/checkpoints/9`. I0423 03:59:54.612249 131584286000704 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 43.497143s. directory=gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_16_zarr3_true/checkpoints/9 I0423 03:59:54.612314 131584286000704 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 37.857045s. I0423 03:59:54.612481 131583667340864 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 03:59:54.612586 131583667340864 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 03:59:54.612639 131583667340864 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0423 03:59:54.612679 131583667340864 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0423 03:59:54.612797 131615831762752 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0423 03:59:54.612870 131615831762752 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 03:59:54.613720 131615831762752 metric_logger.py:196] completed step: 9, seconds: 0.120, TFLOP/s/device: 113.607, Tokens/s/device: 17124.032, total_weights: 16384, loss: 4.189, lm_loss: 4.189, perplexity: 65.979 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops