2026-04-23 02:11:54.656938: 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 00:00:1776910314.670293 2308838 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776910314.674151 2308838 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776910314.685682 2308838 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776910314.685697 2308838 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776910314.685698 2308838 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776910314.685700 2308838 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 02:11:58.350081: 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 02:11:58.485320 134992847800128 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0423 02:12:37.613826 134992847800128 max_utils.py:800] System Information: Jax Version: 0.8.1 I0423 02:12:37.614114 134992847800128 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0423 02:12:37.614167 134992847800128 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 02:12:37.614198 134992847800128 train_utils.py:348] WARNING: 'dataset_path' might be pointing your local file system I0423 02:12:37.614220 134992847800128 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0423 02:12:37.614392 134992847800128 train.py:703] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0423 02:12:38.275323 134992847800128 maxtext_utils.py:1604] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 02:12:38.275578 134992847800128 checkpointing.py:677] Setting up checkpoint logger... I0423 02:12:38.275621 134992847800128 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0423 02:12:38.275656 134992847800128 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0423 02:12:38.276103 134992847800128 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 0x7ac633bb1c10>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0423 02:12:39.910512 134992847800128 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0423 02:12:39.910891 134992847800128 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 0x7ac4d1d7cad0>}, handler_registry=None I0423 02:12:39.911139 134992847800128 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ac4d1d7cad0>` 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 02:12:39.911183 134992847800128 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ac4d1d7c710>` 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 02:12:39.911213 134992847800128 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 0x7ac4d1d7cad0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ac4d1d7cad0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ac4d1d7c710>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ac4d1d7c710>}). I0423 02:12:39.911503 134992847800128 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0423 02:12:40.329594 134963129222720 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776910162928643755, 'commit_timestamp_nsecs': 1776910201392794057, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA I0423 02:12:40.412035 134963129222720 checkpoint_manager.py:1847] Missing metrics for step 0 I0423 02:12:40.506679 134992847800128 checkpoint_manager.py:1788] Found 3 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints I0423 02:12:40.507270 134992847800128 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=False, 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_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7ac4d1d7dc70> I0423 02:12:40.507380 134992847800128 checkpointing.py:301] Checkpoint manager created! I0423 02:12:40.897916 134992847800128 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0423 02:12:40.898022 134992847800128 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 02:12:41.177842 134992847800128 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 02:12:41.177936 134992847800128 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 02:12:41.190165 134992847800128 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 02:12:41.190213 134992847800128 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 02:12:41.208851 134992847800128 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 02:12:41.208913 134992847800128 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 02:12:41.221354 134992847800128 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 02:12:41.221406 134992847800128 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 02:12:41.233678 134992847800128 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 02:12:41.233730 134992847800128 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 02:12:41.253963 134992847800128 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0423 02:12:41.254027 134992847800128 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 02:12:41.269573 134992847800128 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0423 02:12:41.269629 134992847800128 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0423 02:12:41.420153 134992847800128 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0423 02:12:41.420250 134992847800128 checkpointing.py:581] restoring from this run's directory step 9 I0423 02:12:41.608660 134992847800128 checkpointer.py:304] Restoring checkpoint from gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/9. I0423 02:12:42.168317 2310356 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0423 02:12:42.168389 2310356 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0423 02:12:51.457459 134992847800128 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/read/gbytes_per_sec: 1.297 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 9 seconds) (per-host) I0423 02:12:51.457939 134992847800128 checkpointer.py:318] Finished restoring checkpoint in 9.94 seconds from gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/9. I0423 02:12:51.458039 134992847800128 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints', 'checkpointer_start_time': 1776910361.5131783, 'checkpointer_duration_secs': 9.94482135772705, 'checkpoint_manager_start_time': 1776910361.4204535, 'checkpoint_manager_duration_secs': 10.03754997253418} [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 02:12:51.835025 134992847800128 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0423 02:12:51.835159 134992847800128 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 02:12:51.835203 134992847800128 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 02:12:51.835252 134992847800128 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 02:12:51.835293 134992847800128 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 02:12:51.835323 134992847800128 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 02:12:51.835367 134992847800128 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 02:12:51.835409 134992847800128 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 02:12:51.835440 134992847800128 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 02:12:51.835468 134992847800128 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 02:12:51.835505 134992847800128 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0423 02:12:51.835540 134992847800128 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0423 02:12:52.245517 134992847800128 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 02:12:52.245605 134992847800128 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0423 02:12:52.257185 134992847800128 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 02:12:52.257238 134992847800128 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0423 02:12:53.074851 134992847800128 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 02:12:53.078398 134992847800128 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0423 02:12:53.078565 134992847800128 metric_logger.py:301] number parameters: 1.104 billion I0423 02:12:54.962252 134992847800128 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0423 02:12:55.079821 134992847800128 checkpointing.py:776] Waited 0.11754298210144043 seconds for step 10 to finish before starting checkpointing. I0423 02:12:55.080477 134992847800128 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 02:12:55.080630 134992847800128 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 10 I0423 02:12:55.080691 134992847800128 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/10. I0423 02:12:55.231656 134992847800128 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/10 I0423 02:12:56.050101 134992847800128 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776910375815656109, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776910375815656109, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/10/_CHECKPOINT_METADATA I0423 02:12:56.051277 134992847800128 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient W0423 02:12:56.051590 134992847800128 future.py:328] [process=0][thread=MainThread][operation_id=1] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0423 02:12:56.134017 134960723789376 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/10/items I0423 02:12:56.681914 134992847800128 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 02:12:56.682034 134992847800128 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 02:12:58.852708 134992847800128 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 2.171228s I0423 02:12:58.856895 134992847800128 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.458 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 2 seconds) (per-host) I0423 02:12:58.856968 134992847800128 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 2.261069s (batch_requests_ready=0.082817s, total_serialization_initiated=2.174207s, others=0.004045s) I0423 02:12:58.857073 134992847800128 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.805861s (all_items=0.000032s, per_item={'items': '0.00003171'}, temp_paths=2.805830) W0423 02:12:58.857115 134992847800128 future.py:328] [process=0][thread=MainThread][operation_id=1] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0423 02:12:59.577560 134960713303616 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/10/items/array_metadatas/process_0 I0423 02:13:36.833449 134960734275136 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 37.975949s (commit=37.480115s, array_metadata_write=0.495834s) I0423 02:13:36.834581 134992847800128 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 314.046 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 40 seconds) (per-host) I0423 02:13:37.503787 134992847800128 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776910375815656109, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/10/_CHECKPOINT_METADATA I0423 02:13:37.676009 134992847800128 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 02:13:38.135592 134992847800128 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 02:13:38.135821 134992847800128 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.631596s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/10/items I0423 02:13:38.136693 134992847800128 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/10/items I0423 02:13:38.375392 134992847800128 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/10 I0423 02:13:39.051732 134992847800128 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/10`. I0423 02:13:39.052485 134992847800128 checkpointer.py:290] Finished synchronous save in 43.97 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/10 I0423 02:13:39.052669 134992847800128 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0423 02:13:39.052720 134992847800128 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=10] CheckpointManager Save Finalize is done on all hosts. I0423 02:13:39.052743 134992847800128 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=10] Finished synchronous save. I0423 02:13:39.052779 134992847800128 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776910375.0804646, 'wait_for_prev_duration_secs': 4.00543212890625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776910375.0806515, 'checkpointer_blocking_duration_secs': 43.97187614440918, 'get_old_steps_start_time': 1776910419.0525622, 'get_old_steps_duration_secs': 3.7670135498046875e-05, 'checkpoint_manager_blocking_start_time': 1776910375.0803792, 'checkpoint_manager_blocking_duration_secs': 43.97237849235535} I0423 02:13:39.052894 134992847800128 checkpointing.py:410] Saved a checkpoint at step 10. I0423 02:13:39.052944 134992847800128 max_utils.py:750] Memstats: After params initialized: I0423 02:13:39.053006 134992847800128 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0423 02:13:39.053032 134992847800128 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0423 02:13:39.053082 134992847800128 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0423 02:13:39.053104 134992847800128 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0423 02:13:39.053123 134992847800128 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0423 02:13:39.053141 134992847800128 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0423 02:13:39.053159 134992847800128 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0423 02:13:39.053178 134992847800128 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0423 02:13:40.001290 134992847800128 metric_logger.py:196] completed step: 10, seconds: 1.883, TFLOP/s/device: 7.215, Tokens/s/device: 1087.553, total_weights: 16384, loss: 3.602, lm_loss: 3.602, perplexity: 36.686 I0423 02:13:40.007851 134992847800128 metric_logger.py:196] completed step: 11, seconds: 44.332, TFLOP/s/device: 0.306, Tokens/s/device: 46.197, total_weights: 16384, loss: 3.101, lm_loss: 3.101, perplexity: 22.211 I0423 02:13:40.125876 134992847800128 metric_logger.py:196] completed step: 12, seconds: 0.713, TFLOP/s/device: 19.060, Tokens/s/device: 2872.898, total_weights: 16384, loss: 2.684, lm_loss: 2.684, perplexity: 14.644 I0423 02:13:40.251318 134992847800128 metric_logger.py:196] completed step: 13, seconds: 0.005, TFLOP/s/device: 2766.109, Tokens/s/device: 416938.111, total_weights: 16384, loss: 2.348, lm_loss: 2.348, perplexity: 10.461 I0423 02:13:40.256939 134992847800128 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0423 02:13:40.488734 134992847800128 checkpointing.py:776] Waited 0.23175883293151855 seconds for step 15 to finish before starting checkpointing. I0423 02:13:40.489672 134992847800128 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 02:13:40.489860 134992847800128 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 15 I0423 02:13:40.489928 134992847800128 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/15. I0423 02:13:40.659604 134992847800128 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/15 W0423 02:13:41.499378 134992847800128 future.py:328] [process=0][thread=MainThread][operation_id=2] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0423 02:13:41.584066 134960734275136 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/15/items I0423 02:13:42.124918 134992847800128 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 02:13:42.125033 134992847800128 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 02:13:42.688005 134992847800128 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.563492s I0423 02:13:42.693631 134992847800128 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 18.645 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 661 milliseconds) (per-host) I0423 02:13:42.693696 134992847800128 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.662003s (batch_requests_ready=0.089633s, total_serialization_initiated=0.566876s, others=0.005493s) I0423 02:13:42.693775 134992847800128 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.194808s (all_items=0.000031s, per_item={'items': '0.00003123'}, temp_paths=1.194777) W0423 02:13:42.693812 134992847800128 future.py:328] [process=0][thread=MainThread][operation_id=2] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0423 02:13:43.406697 134960713303616 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/15/items/array_metadatas/process_0 I0423 02:14:26.502794 134960518268480 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 43.808734s (commit=43.347523s, array_metadata_write=0.461211s) I0423 02:14:26.503960 134992847800128 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 284.150 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 44 seconds) (per-host) I0423 02:14:27.341056 134992847800128 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 02:14:27.746393 134992847800128 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 02:14:27.748087 134992847800128 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.578713s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/15/items I0423 02:14:27.748845 134992847800128 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/15/items I0423 02:14:27.973651 134992847800128 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/15 I0423 02:14:28.635855 134992847800128 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/15`. I0423 02:14:28.636643 134992847800128 checkpointer.py:290] Finished synchronous save in 48.15 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/15 I0423 02:14:28.636795 134992847800128 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0423 02:14:28.636834 134992847800128 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=15] CheckpointManager Save Finalize is done on all hosts. I0423 02:14:28.636859 134992847800128 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=15] Finished synchronous save. I0423 02:14:28.636892 134992847800128 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776910420.4896445, 'wait_for_prev_duration_secs': 5.650520324707031e-05, 'time_between_consecutive_saves_sec': 1.4369080066680908, 'checkpointer_blocking_start_time': 1776910420.4898822, 'checkpointer_blocking_duration_secs': 48.14680624008179, 'get_old_steps_start_time': 1776910468.636709, 'get_old_steps_duration_secs': 3.147125244140625e-05, 'checkpoint_manager_blocking_start_time': 1776910420.4895675, 'checkpoint_manager_blocking_duration_secs': 48.14730477333069} I0423 02:14:28.636967 134992847800128 checkpointing.py:410] Saved a checkpoint at step 15. I0423 02:14:28.637811 134992847800128 metric_logger.py:196] completed step: 14, seconds: 0.122, TFLOP/s/device: 111.545, Tokens/s/device: 16813.346, total_weights: 16384, loss: 2.083, lm_loss: 2.083, perplexity: 8.030 I0423 02:14:29.252018 134992847800128 metric_logger.py:196] completed step: 15, seconds: 0.123, TFLOP/s/device: 110.527, Tokens/s/device: 16659.888, total_weights: 16384, loss: 1.880, lm_loss: 1.880, perplexity: 6.551 I0423 02:14:29.259033 134992847800128 metric_logger.py:196] completed step: 16, seconds: 48.388, TFLOP/s/device: 0.281, Tokens/s/device: 42.325, total_weights: 16384, loss: 1.726, lm_loss: 1.726, perplexity: 5.616 I0423 02:14:29.377367 134992847800128 metric_logger.py:196] completed step: 17, seconds: 0.614, TFLOP/s/device: 22.146, Tokens/s/device: 3338.120, total_weights: 16384, loss: 1.610, lm_loss: 1.610, perplexity: 5.002 I0423 02:14:29.496621 134992847800128 metric_logger.py:196] completed step: 18, seconds: 0.005, TFLOP/s/device: 2708.758, Tokens/s/device: 408293.461, total_weights: 16384, loss: 1.521, lm_loss: 1.521, perplexity: 4.577 I0423 02:14:29.615140 134992847800128 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0423 02:14:29.616259 134992847800128 checkpointing.py:776] Waited 0.0011451244354248047 seconds for step 19 to finish before starting checkpointing. I0423 02:14:29.616508 134992847800128 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 02:14:29.616647 134992847800128 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 19 I0423 02:14:29.616710 134992847800128 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/19. I0423 02:14:29.787387 134992847800128 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/19 W0423 02:14:30.598785 134992847800128 future.py:328] [process=0][thread=MainThread][operation_id=3] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0423 02:14:30.684933 134960518268480 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/19/items I0423 02:14:31.234320 134992847800128 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 02:14:31.234432 134992847800128 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 02:14:37.027935 134992847800128 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 5.794025s I0423 02:14:37.032683 134992847800128 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.094 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 5 seconds) (per-host) I0423 02:14:37.032753 134992847800128 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 5.892349s (batch_requests_ready=0.090436s, total_serialization_initiated=5.797329s, others=0.004585s) I0423 02:14:37.032862 134992847800128 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 6.434443s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=6.434425) W0423 02:14:37.032901 134992847800128 future.py:328] [process=0][thread=MainThread][operation_id=3] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0423 02:14:37.744221 134960713303616 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/19/items/array_metadatas/process_0 I0423 02:15:18.951968 134962944673344 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.918656s (commit=41.377203s, array_metadata_write=0.541452s) I0423 02:15:18.953310 134992847800128 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 264.297 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 47 seconds) (per-host) I0423 02:15:19.802571 134992847800128 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 02:15:20.213289 134992847800128 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 02:15:20.213474 134992847800128 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.578105s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/19/items I0423 02:15:20.214251 134992847800128 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/19/items I0423 02:15:20.457624 134992847800128 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/19 I0423 02:15:21.202897 134992847800128 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/19`. I0423 02:15:21.203541 134992847800128 checkpointer.py:290] Finished synchronous save in 51.59 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints/19 I0423 02:15:21.203681 134992847800128 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0423 02:15:21.203718 134992847800128 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=19] CheckpointManager Save Finalize is done on all hosts. I0423 02:15:21.203749 134992847800128 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=19] Finished synchronous save. I0423 02:15:21.203785 134992847800128 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776910469.6164944, 'wait_for_prev_duration_secs': 3.790855407714844e-05, 'time_between_consecutive_saves_sec': 0.9796450138092041, 'checkpointer_blocking_start_time': 1776910469.6166675, 'checkpointer_blocking_duration_secs': 51.58692765235901, 'get_old_steps_start_time': 1776910521.203607, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1776910469.6164627, 'checkpoint_manager_blocking_duration_secs': 51.58730101585388} I0423 02:15:21.203915 134992847800128 checkpointing.py:410] Saved a checkpoint at step 19. I0423 02:15:21.203943 134992847800128 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 02:15:21.203964 134992847800128 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 02:15:21.204823 134992847800128 metric_logger.py:196] completed step: 19, seconds: 0.119, TFLOP/s/device: 113.786, Tokens/s/device: 17150.992, 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