2026-04-23 03:25:14.438579: 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 00:00:1776914714.452438 2376516 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776914714.456413 2376516 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776914714.468905 2376516 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776914714.468920 2376516 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776914714.468922 2376516 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776914714.468923 2376516 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 03:25:18.137365: 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 03:25:18.274352 138855308793664 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 03:26:27.404103 138855308793664 max_utils.py:800] System Information: Jax Version: 0.8.1 I0423 03:26:27.404365 138855308793664 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0423 03:26:27.404419 138855308793664 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 03:26:27.404450 138855308793664 train_utils.py:348] WARNING: 'dataset_path' might be pointing your local file system I0423 03:26:27.404475 138855308793664 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0423 03:26:27.404636 138855308793664 train.py:672] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0423 03:26:28.061769 138855308793664 maxtext_utils.py:1604] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 03:26:28.062028 138855308793664 checkpointing.py:677] Setting up checkpoint logger... I0423 03:26:28.062087 138855308793664 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0423 03:26:28.062124 138855308793664 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0423 03:26:28.062546 138855308793664 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 0x7e497e5b1640>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0423 03:26:30.519395 138855308793664 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0423 03:26:30.519761 138855308793664 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 0x7e479df56390>}, handler_registry=None I0423 03:26:30.519982 138855308793664 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e479df56390>` 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 03:26:30.520022 138855308793664 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e479df56930>` 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 03:26:30.520066 138855308793664 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 0x7e479df56390>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e479df56390>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e479df56930>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e479df56930>}). I0423 03:26:30.520334 138855308793664 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0423 03:26:30.520384 138855308793664 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7e479df88540> timeout: 600 secs and primary_host=0 for async checkpoint writes I0423 03:26:30.640472 138855308793664 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_11_optimizer_offload_true/checkpoints I0423 03:26:30.640693 138855308793664 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=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_11_optimizer_offload_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e479df56870> I0423 03:26:30.640965 138855308793664 checkpointing.py:301] Checkpoint manager created! I0423 03:26:31.027542 138855308793664 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0423 03:26:31.027635 138855308793664 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 03:26:31.305562 138855308793664 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 03:26:31.305647 138855308793664 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 03:26:31.318014 138855308793664 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 03:26:31.318078 138855308793664 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 03:26:31.337367 138855308793664 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 03:26:31.337423 138855308793664 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 03:26:31.349988 138855308793664 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 03:26:31.350039 138855308793664 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 03:26:31.362461 138855308793664 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 03:26:31.362511 138855308793664 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 03:26:31.382243 138855308793664 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0423 03:26:31.382298 138855308793664 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 03:26:31.397862 138855308793664 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0423 03:26:31.397915 138855308793664 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0423 03:26:31.550602 138855308793664 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0423 03:26:31.550694 138855308793664 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 03:26:34.452845 138855308793664 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0423 03:26:34.452955 138855308793664 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 03:26:34.452998 138855308793664 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 03:26:34.453063 138855308793664 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 03:26:34.453110 138855308793664 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 03:26:34.453140 138855308793664 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 03:26:34.453180 138855308793664 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 03:26:34.453219 138855308793664 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 03:26:34.453249 138855308793664 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 03:26:34.453276 138855308793664 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 03:26:34.453313 138855308793664 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0423 03:26:34.453350 138855308793664 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0423 03:26:34.854393 138855308793664 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 03:26:34.854475 138855308793664 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0423 03:26:34.866128 138855308793664 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 03:26:34.866174 138855308793664 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0423 03:26:35.625273 138855308793664 max_utils.py:791] Total memory size: 2.4 GB, Output size: 0.5 GB, Temp size: 1.9 GB, Argument size: 0.5 GB, Host temp size: 0.0 GB. I0423 03:26:35.626060 138855308793664 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0423 03:26:35.626216 138855308793664 metric_logger.py:301] number parameters: 1.104 billion I0423 03:26:38.419274 138855308793664 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0423 03:26:38.614997 138855308793664 checkpointing.py:776] Waited 0.19579792022705078 seconds for step 0 to finish before starting checkpointing. I0423 03:26:38.616004 138855308793664 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 03:26:38.616291 138855308793664 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0423 03:26:38.617224 138855308793664 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_11_optimizer_offload_true/checkpoints/0. I0423 03:26:38.715635 138855308793664 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0423 03:26:38.808483 138817925285440 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_11_optimizer_offload_true/checkpoints/0 I0423 03:26:38.817951 138855308793664 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 03:26:38.818058 138855308793664 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 03:26:39.462934 138855308793664 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.645438s I0423 03:26:39.463323 138855308793664 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 16.530 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 746 milliseconds) (per-host) I0423 03:26:39.463379 138855308793664 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.746683s (batch_requests_ready=0.096817s, total_serialization_initiated=0.649586s, others=0.000280s) I0423 03:26:39.463449 138855308793664 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.747303s (all_items=0.000041s, per_item={'items': '0.00004148'}, temp_paths=0.747262) I0423 03:26:39.464212 138817321305664 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 03:26:39.464304 138855308793664 async_checkpointer.py:561] Finished blocking save. Time taken: 0.847965s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_11_optimizer_offload_true/checkpoints/0. I0423 03:26:39.464496 138855308793664 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0423 03:26:39.464650 138817935771200 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 03:26:39.464730 138855308793664 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_11_optimizer_offload_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776914798.6159806, 'wait_for_prev_duration_secs': 5.2928924560546875e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776914798.6163154, 'checkpointer_blocking_duration_secs': 0.8480880260467529, 'get_old_steps_start_time': 1776914799.4644217, 'get_old_steps_duration_secs': 3.5762786865234375e-05, 'checkpoint_manager_blocking_start_time': 1776914798.6157453, 'checkpoint_manager_blocking_duration_secs': 0.8489429950714111} I0423 03:26:39.464836 138855308793664 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0423 03:26:39.464901 138855308793664 max_utils.py:750] Memstats: After params initialized: I0423 03:26:39.464988 138855308793664 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_0(process=0,(0,0,0,0)) I0423 03:26:39.465018 138855308793664 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_1(process=0,(1,0,0,0)) I0423 03:26:39.465040 138855308793664 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_2(process=0,(0,1,0,0)) I0423 03:26:39.465072 138855308793664 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_3(process=0,(1,1,0,0)) I0423 03:26:39.465091 138855308793664 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_4(process=0,(0,2,0,0)) I0423 03:26:39.465110 138855308793664 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_5(process=0,(1,2,0,0)) I0423 03:26:39.465127 138855308793664 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_6(process=0,(0,3,0,0)) I0423 03:26:39.465144 138855308793664 max_utils.py:756] Using (GB) 0.56 / 31.25 (1.792000%) on TPU_7(process=0,(1,3,0,0)) I0423 03:26:39.569115 138817734444608 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_11_optimizer_offload_true/checkpoints/0/items I0423 03:26:39.814512 138817130464832 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776914799470612748, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776914799470612748, "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_11_optimizer_offload_true/checkpoints/0/_CHECKPOINT_METADATA I0423 03:26:39.822413 138855308793664 metric_logger.py:196] completed step: 0, seconds: 2.792, TFLOP/s/device: 4.866, Tokens/s/device: 733.420, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.840 I0423 03:26:39.823033 138855308793664 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_11_optimizer_offload_true/tensorboard/' I0423 03:26:40.011126 138855308793664 metric_logger.py:196] completed step: 1, seconds: 1.401, TFLOP/s/device: 9.700, Tokens/s/device: 1462.092, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.840 I0423 03:26:40.207951 138855308793664 metric_logger.py:196] completed step: 2, seconds: 0.007, TFLOP/s/device: 1922.888, Tokens/s/device: 289838.664, total_weights: 16384, loss: 9.775, lm_loss: 9.775, perplexity: 17589.299 I0423 03:26:40.292180 2378722 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0423 03:26:40.292248 2378722 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0423 03:26:40.405965 138855308793664 metric_logger.py:196] completed step: 3, seconds: 0.191, TFLOP/s/device: 71.309, Tokens/s/device: 10748.399, total_weights: 16384, loss: 8.838, lm_loss: 8.838, perplexity: 6892.454 I0423 03:26:57.452408 138817522632256 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_11_optimizer_offload_true/checkpoints/0/items/array_metadatas/process_0 I0423 03:26:57.454660 138855308793664 metric_logger.py:196] completed step: 4, seconds: 0.196, TFLOP/s/device: 69.281, Tokens/s/device: 10442.852, total_weights: 16384, loss: 7.965, lm_loss: 7.965, perplexity: 2879.114 I0423 03:26:57.479707 138855308793664 metric_logger.py:196] completed step: 5, seconds: 0.198, TFLOP/s/device: 68.678, Tokens/s/device: 10351.852, total_weights: 16384, loss: 7.233, lm_loss: 7.233, perplexity: 1383.691 I0423 03:26:57.671218 138855308793664 metric_logger.py:196] completed step: 6, seconds: 17.067, TFLOP/s/device: 0.796, Tokens/s/device: 120.001, total_weights: 16384, loss: 6.677, lm_loss: 6.677, perplexity: 793.605 I0423 03:26:57.868923 138855308793664 metric_logger.py:196] completed step: 7, seconds: 0.006, TFLOP/s/device: 2386.219, Tokens/s/device: 359676.853, total_weights: 16384, loss: 6.293, lm_loss: 6.293, perplexity: 540.987 I0423 03:26:58.069010 138855308793664 metric_logger.py:196] completed step: 8, seconds: 0.194, TFLOP/s/device: 70.162, Tokens/s/device: 10575.617, total_weights: 16384, loss: 6.055, lm_loss: 6.055, perplexity: 426.449 I0423 03:26:58.276951 138855308793664 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0423 03:26:58.284094 138855308793664 checkpointing.py:776] Waited 0.007378578186035156 seconds for step 9 to finish before starting checkpointing. I0423 03:26:58.289553 138855308793664 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 03:27:43.011038 138817331791424 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 62.724257s (commit=62.262294s, array_metadata_write=0.461963s) I0423 03:27:43.012130 138817321305664 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 196.543 MiB/s (total gbytes: 12.3 GiB) (time elapsed: a minute) (per-host) I0423 03:27:43.012241 138817321305664 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 63.547885s. I0423 03:27:43.324733 138817321305664 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776914799470612748, '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_11_optimizer_offload_true/checkpoints/0/_CHECKPOINT_METADATA I0423 03:27:43.556578 138817321305664 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 03:27:43.747013 138817130464832 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776914799470612748, '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_11_optimizer_offload_true/checkpoints/0/_CHECKPOINT_METADATA I0423 03:27:43.941243 138817321305664 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 03:27:43.941415 138817321305664 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.570049s. 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_11_optimizer_offload_true/checkpoints/0/items I0423 03:27:43.942171 138817321305664 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_11_optimizer_offload_true/checkpoints/0/items I0423 03:27:44.181950 138817321305664 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_11_optimizer_offload_true/checkpoints/0 I0423 03:27:44.891115 138817321305664 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_11_optimizer_offload_true/checkpoints/0`. I0423 03:27:44.891746 138817321305664 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 66.275410s. directory=gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_11_optimizer_offload_true/checkpoints/0 I0423 03:27:44.891830 138817321305664 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 65.427479s. I0423 03:27:44.892057 138817935771200 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 03:27:44.892167 138817935771200 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 03:27:44.892235 138817935771200 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0423 03:27:44.892289 138817935771200 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0423 03:27:44.892549 138855308793664 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 03:27:44.892672 138855308793664 checkpoint_manager.py:1441] Waiting for previous save to complete took 46.606097 seconds. If this number is high, consider checkpointing less frequently. I0423 03:27:44.893969 138855308793664 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0423 03:27:44.894294 138855308793664 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_11_optimizer_offload_true/checkpoints/9. I0423 03:27:45.070492 138855308793664 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 03:27:45.070585 138855308793664 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 03:27:45.073108 138817321305664 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_11_optimizer_offload_true/checkpoints/9 I0423 03:27:45.467651 138855308793664 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.397611s I0423 03:27:45.468221 138855308793664 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 25.390 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 486 milliseconds) (per-host) I0423 03:27:45.468280 138855308793664 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.486151s (batch_requests_ready=0.085186s, total_serialization_initiated=0.400511s, others=0.000454s) I0423 03:27:45.468357 138855308793664 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.486699s (all_items=0.000036s, per_item={'items': '0.00003552'}, temp_paths=0.486664) I0423 03:27:45.469464 138816719423040 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 03:27:45.469565 138855308793664 async_checkpointer.py:561] Finished blocking save. Time taken: 0.575549s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_11_optimizer_offload_true/checkpoints/9. I0423 03:27:45.469768 138855308793664 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0423 03:27:45.469976 138817935771200 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 03:27:45.470121 138855308793664 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_11_optimizer_offload_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776914818.286543, 'wait_for_prev_duration_secs': 46.60609722137451, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776914864.893994, 'checkpointer_blocking_duration_secs': 0.5756697654724121, 'get_old_steps_start_time': 1776914865.4696841, 'get_old_steps_duration_secs': 4.172325134277344e-05, 'checkpoint_manager_blocking_start_time': 1776914818.2845926, 'checkpoint_manager_blocking_duration_secs': 47.1854829788208} I0423 03:27:45.470247 138855308793664 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0423 03:27:45.470283 138855308793664 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 03:27:45.790706 138817522632256 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_11_optimizer_offload_true/checkpoints/9/items I0423 03:27:47.008422 138817533118016 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_11_optimizer_offload_true/checkpoints/9/items/array_metadatas/process_0 I0423 03:28:25.715974 138817119979072 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 39.471203s (commit=39.001491s, array_metadata_write=0.469711s) I0423 03:28:25.716954 138816719423040 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 310.221 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 40 seconds) (per-host) I0423 03:28:25.717077 138816719423040 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.247464s. I0423 03:28:26.288349 138816719423040 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 03:28:26.670162 138816719423040 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 03:28:26.670332 138816719423040 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.569860s. 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_11_optimizer_offload_true/checkpoints/9/items I0423 03:28:26.671032 138816719423040 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_11_optimizer_offload_true/checkpoints/9/items I0423 03:28:26.906011 138816719423040 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_11_optimizer_offload_true/checkpoints/9 I0423 03:28:27.580894 138816719423040 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_11_optimizer_offload_true/checkpoints/9`. I0423 03:28:27.581510 138816719423040 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 42.687501s. directory=gs://wanglance-maxtext/linen_ckpt_fix_gate_fp8_nan_sanitization_20260423_024158/linen_fix_gate_fp8_nan_sanitization_20260423_024158_11_optimizer_offload_true/checkpoints/9 I0423 03:28:27.581589 138816719423040 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 42.111979s. I0423 03:28:27.581774 138817935771200 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 03:28:27.581879 138817935771200 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 03:28:27.581931 138817935771200 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0423 03:28:27.581968 138817935771200 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0423 03:28:27.582108 138855308793664 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 03:28:27.582224 138855308793664 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 03:28:27.583105 138855308793664 metric_logger.py:196] completed step: 9, seconds: 0.198, TFLOP/s/device: 68.721, Tokens/s/device: 10358.344, total_weights: 16384, loss: 5.915, lm_loss: 5.915, perplexity: 370.686 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops