2026-04-21 14:46:58.577938: 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 00:00:1776782818.588701 1471922 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776782818.591792 1471922 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776782818.601518 1471922 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776782818.601530 1471922 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776782818.601532 1471922 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776782818.601534 1471922 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-21 14:47:02.343997: 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) I0421 14:47:02.458868 139640787012736 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0421 14:47:55.494300 139640787012736 max_utils.py:800] System Information: Jax Version: 0.8.1 I0421 14:47:55.494580 139640787012736 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0421 14:47:55.494632 139640787012736 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 I0421 14:47:55.494664 139640787012736 train_utils.py:335] WARNING: 'dataset_path' might be pointing your local file system I0421 14:47:55.494873 139640787012736 train.py:703] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0421 14:47:55.865215 139640787012736 maxtext_utils.py:1551] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 14:47:55.865601 139640787012736 checkpointing.py:677] Setting up checkpoint logger... I0421 14:47:55.865659 139640787012736 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0421 14:47:55.865755 139640787012736 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0421 14:47:55.866617 139640787012736 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 0x7f005b3a1f70>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 14:47:58.739166 139640787012736 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0421 14:47:58.739446 139640787012736 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 0x7efeb3b49550>}, handler_registry=None I0421 14:47:58.739859 139640787012736 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7efeb3b49550>` 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`. I0421 14:47:58.739905 139640787012736 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7efeb3fdffb0>` 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`. I0421 14:47:58.739934 139640787012736 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 0x7efeb3b49550>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7efeb3b49550>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7efeb3fdffb0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7efeb3fdffb0>}). I0421 14:47:58.740264 139640787012736 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0421 14:47:58.740324 139640787012736 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7efeb3b27380> timeout: 600 secs and primary_host=0 for async checkpoint writes I0421 14:47:58.890960 139640787012736 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints I0421 14:47:58.891262 139640787012736 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_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7efeb3b49100> I0421 14:47:58.891357 139640787012736 checkpointing.py:301] Checkpoint manager created! I0421 14:47:59.870454 139640787012736 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0421 14:47:59.879249 139640787012736 reader.py:262] Creating a tf.data.Dataset reading 8 files located in folders: tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0. I0421 14:47:59.960460 139640787012736 logging_logger.py:49] Constructing tf.data.Dataset __local_c4_builder for split train, from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0421 14:47:59.983269 139640787012736 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2 I0421 14:47:59.983333 139640787012736 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2 I0421 14:48:00.635003 139640787012736 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 14:48:00.635103 139640787012736 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 14:48:00.931725 139640787012736 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 14:48:00.931805 139640787012736 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 14:48:00.944326 139640787012736 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 14:48:00.944375 139640787012736 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 14:48:00.963214 139640787012736 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 14:48:00.963270 139640787012736 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 14:48:00.975871 139640787012736 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 14:48:00.975921 139640787012736 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 14:48:00.988405 139640787012736 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 14:48:00.988453 139640787012736 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 14:48:01.008270 139640787012736 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 14:48:01.008324 139640787012736 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 14:48:01.024116 139640787012736 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 14:48:01.024167 139640787012736 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0421 14:48:01.191423 139640787012736 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0421 14:48:01.191515 139640787012736 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 I0421 14:48:02.786726 139640787012736 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0421 14:48:02.786833 139640787012736 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0421 14:48:02.786874 139640787012736 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0421 14:48:02.786919 139640787012736 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0421 14:48:02.786960 139640787012736 maxtext_utils.py:1654] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0421 14:48:02.786989 139640787012736 maxtext_utils.py:1654] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0421 14:48:02.787029 139640787012736 maxtext_utils.py:1654] 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) I0421 14:48:02.787087 139640787012736 maxtext_utils.py:1654] 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') I0421 14:48:02.787118 139640787012736 maxtext_utils.py:1654] 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) I0421 14:48:02.787146 139640787012736 maxtext_utils.py:1654] 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) I0421 14:48:02.787188 139640787012736 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0421 14:48:02.787223 139640787012736 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0421 14:48:03.194704 139640787012736 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 14:48:03.194783 139640787012736 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0421 14:48:03.206507 139640787012736 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 14:48:03.206556 139640787012736 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0421 14:48:11.391195 139640787012736 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. I0421 14:48:11.391863 139640787012736 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0421 14:48:11.392064 139640787012736 metric_logger.py:301] number parameters: 1.104 billion I0421 14:48:22.585259 139640787012736 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0421 14:48:22.716688 139640787012736 checkpointing.py:776] Waited 0.1314377784729004 seconds for step 0 to finish before starting checkpointing. I0421 14:48:22.717283 139640787012736 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 14:48:22.717477 139640787012736 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0421 14:48:22.717997 139640787012736 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/0. I0421 14:48:22.807418 139640787012736 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0421 14:48:22.890770 139549204284992 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/0 I0421 14:48:22.925627 139640787012736 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 14:48:22.925716 139640787012736 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 I0421 14:48:23.663149 139640787012736 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.737897s I0421 14:48:23.663599 139640787012736 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 14.436 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 854 milliseconds) (per-host) I0421 14:48:23.663754 139640787012736 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.854977s (batch_requests_ready=0.112735s, total_serialization_initiated=0.741868s, others=0.000373s) I0421 14:48:23.663833 139640787012736 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.855929s (all_items=0.000035s, per_item={'items': '0.00003505'}, temp_paths=0.855894) I0421 14:48:23.664511 139549141370432 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 14:48:23.664610 139640787012736 async_checkpointer.py:561] Finished blocking save. Time taken: 0.947089s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/0. I0421 14:48:23.664790 139640787012736 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0421 14:48:23.664983 139640787012736 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776782902.7172658, 'wait_for_prev_duration_secs': 4.744529724121094e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776782902.717499, 'checkpointer_blocking_duration_secs': 0.9472134113311768, 'get_old_steps_start_time': 1776782903.6647286, 'get_old_steps_duration_secs': 2.4557113647460938e-05, 'checkpoint_manager_blocking_start_time': 1776782902.7171614, 'checkpoint_manager_blocking_duration_secs': 0.9477975368499756} I0421 14:48:23.665065 139640787012736 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0421 14:48:23.665120 139640787012736 max_utils.py:750] Memstats: After params initialized: I0421 14:48:23.665192 139640787012736 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0421 14:48:23.665218 139640787012736 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0421 14:48:23.665237 139640787012736 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0421 14:48:23.665254 139640787012736 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0421 14:48:23.665271 139640787012736 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0421 14:48:23.665287 139640787012736 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0421 14:48:23.665303 139640787012736 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0421 14:48:23.665319 139640787012736 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0421 14:48:23.711082 139549193799232 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/0/items I0421 14:48:23.857420 139549162341952 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776782903613123091, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776782903613123091, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/0/_CHECKPOINT_METADATA I0421 14:48:23.857722 139549130884672 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 14:48:24.117468 139640787012736 metric_logger.py:196] completed step: 0, seconds: 11.192, TFLOP/s/device: 1.214, Tokens/s/device: 182.980, total_weights: 13328, loss: 10.828, lm_loss: 10.828, perplexity: 50404.086 I0421 14:48:24.118537 139640787012736 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/tensorboard/' I0421 14:48:24.233081 139640787012736 metric_logger.py:196] completed step: 1, seconds: 1.530, TFLOP/s/device: 8.882, Tokens/s/device: 1338.840, total_weights: 12332, loss: 10.828, lm_loss: 10.828, perplexity: 50391.758 I0421 14:48:24.350564 139640787012736 metric_logger.py:196] completed step: 2, seconds: 0.011, TFLOP/s/device: 1282.773, Tokens/s/device: 193353.474, total_weights: 15161, loss: 9.823, lm_loss: 9.823, perplexity: 18444.867 I0421 14:48:24.438811 1475751 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0421 14:48:24.438900 1475751 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0421 14:48:24.468337 139640787012736 metric_logger.py:196] completed step: 3, seconds: 0.117, TFLOP/s/device: 116.495, Tokens/s/device: 17559.353, total_weights: 13327, loss: 9.405, lm_loss: 9.405, perplexity: 12149.331 I0421 14:48:39.460573 139640787012736 metric_logger.py:196] completed step: 4, seconds: 0.117, TFLOP/s/device: 115.896, Tokens/s/device: 17469.037, total_weights: 11939, loss: 9.045, lm_loss: 9.045, perplexity: 8471.842 I0421 14:48:39.486795 139640787012736 metric_logger.py:196] completed step: 5, seconds: 0.118, TFLOP/s/device: 115.029, Tokens/s/device: 17338.447, total_weights: 15502, loss: 8.953, lm_loss: 8.953, perplexity: 7728.546 I0421 14:48:39.603839 139640787012736 metric_logger.py:196] completed step: 6, seconds: 15.008, TFLOP/s/device: 0.905, Tokens/s/device: 136.461, total_weights: 13864, loss: 8.775, lm_loss: 8.775, perplexity: 6468.536 I0421 14:48:39.721942 139640787012736 metric_logger.py:196] completed step: 7, seconds: 0.007, TFLOP/s/device: 1911.795, Tokens/s/device: 288166.596, total_weights: 12988, loss: 8.688, lm_loss: 8.688, perplexity: 5928.887 I0421 14:48:39.788995 139549204284992 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/0/items/array_metadatas/process_0 I0421 14:48:39.839299 139640787012736 metric_logger.py:196] completed step: 8, seconds: 0.119, TFLOP/s/device: 114.534, Tokens/s/device: 17263.761, total_weights: 13820, loss: 8.740, lm_loss: 8.740, perplexity: 6250.123 I0421 14:48:39.956465 139640787012736 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0421 14:48:39.957303 139640787012736 checkpointing.py:776] Waited 0.0008721351623535156 seconds for step 9 to finish before starting checkpointing. I0421 14:48:39.957624 139640787012736 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 14:49:09.733274 139549151856192 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 45.293485s (commit=44.865844s, array_metadata_write=0.427641s) I0421 14:49:09.734349 139549141370432 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 269.294 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 46 seconds) (per-host) I0421 14:49:09.734465 139549141370432 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 46.069810s. I0421 14:49:10.005895 139549141370432 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776782903613123091, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/0/_CHECKPOINT_METADATA I0421 14:49:10.194126 139549141370432 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0421 14:49:10.412311 139549162341952 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776782903613123091, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/0/_CHECKPOINT_METADATA I0421 14:49:10.631789 139549141370432 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 14:49:10.632434 139549141370432 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.586081s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/0/items I0421 14:49:10.633128 139549141370432 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/0/items I0421 14:49:10.884600 139549141370432 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/0 I0421 14:49:11.563558 139549141370432 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/0`. I0421 14:49:11.564214 139549141370432 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 48.846699s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/0 I0421 14:49:11.564297 139549141370432 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 47.899644s. I0421 14:49:11.564425 139549130884672 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 14:49:11.564476 139549130884672 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 14:49:11.564528 139549130884672 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0421 14:49:11.564570 139549130884672 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0421 14:49:11.564760 139640787012736 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. W0421 14:49:11.564884 139640787012736 checkpoint_manager.py:1441] Waiting for previous save to complete took 31.607298 seconds. If this number is high, consider checkpointing less frequently. I0421 14:49:11.565969 139640787012736 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0421 14:49:11.566265 139640787012736 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/9. I0421 14:49:11.740090 139640787012736 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 14:49:11.740187 139640787012736 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 I0421 14:49:11.744990 139549130884672 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/9 I0421 14:49:12.505627 139549046998592 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/9/items I0421 14:49:16.810337 139640787012736 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 5.070682s I0421 14:49:16.814900 139640787012736 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.390 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 5 seconds) (per-host) I0421 14:49:16.814970 139640787012736 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 5.163078s (batch_requests_ready=0.084656s, total_serialization_initiated=5.073977s, others=0.004445s) I0421 14:49:16.815085 139640787012736 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 5.163953s (all_items=0.000049s, per_item={'items': '0.00004935'}, temp_paths=5.163904) I0421 14:49:16.816623 139549130884672 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 14:49:16.816707 139640787012736 async_checkpointer.py:561] Finished blocking save. Time taken: 5.250693s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/9. I0421 14:49:16.816896 139640787012736 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0421 14:49:16.817082 139549141370432 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 14:49:16.817214 139640787012736 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776782919.9575603, 'wait_for_prev_duration_secs': 31.607298135757446, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776782951.565993, 'checkpointer_blocking_duration_secs': 5.250819444656372, 'get_old_steps_start_time': 1776782956.8168323, 'get_old_steps_duration_secs': 2.47955322265625e-05, 'checkpoint_manager_blocking_start_time': 1776782919.9575083, 'checkpoint_manager_blocking_duration_secs': 36.85967659950256} I0421 14:49:16.817339 139640787012736 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0421 14:49:16.817377 139640787012736 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 14:49:17.558920 139549204284992 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/9/items/array_metadatas/process_0 I0421 14:49:53.186732 139549046998592 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 36.370989s (commit=35.924752s, array_metadata_write=0.446238s) I0421 14:49:53.187793 139549130884672 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 304.238 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 41 seconds) (per-host) I0421 14:49:53.187856 139549130884672 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.371105s. I0421 14:49:53.614203 139549130884672 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0421 14:49:54.041783 139549130884672 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 14:49:54.042505 139549130884672 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.570986s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/9/items I0421 14:49:54.043185 139549130884672 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/9/items I0421 14:49:54.305096 139549130884672 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/9 I0421 14:49:54.982197 139549130884672 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/9`. I0421 14:49:54.982874 139549130884672 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 43.416866s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_01_base/checkpoints/9 I0421 14:49:54.982954 139549130884672 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 38.166205s. I0421 14:49:54.983146 139549141370432 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 14:49:54.983256 139549141370432 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 14:49:54.983325 139549141370432 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0421 14:49:54.983369 139549141370432 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0421 14:49:54.983836 139640787012736 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. I0421 14:49:54.984135 139640787012736 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 14:49:54.984974 139640787012736 metric_logger.py:196] completed step: 9, seconds: 0.118, TFLOP/s/device: 115.158, Tokens/s/device: 17357.845, total_weights: 12300, loss: 8.710, lm_loss: 8.710, perplexity: 6065.008 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops