2026-04-22 17:45:10.314061: 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 00:00:1776879910.327352 2014639 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776879910.331156 2014639 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776879910.342684 2014639 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776879910.342702 2014639 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776879910.342704 2014639 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776879910.342705 2014639 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'} 2026-04-22 17:45:14.008018: 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) I0422 17:45:14.143130 123430403606336 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0422 17:45:44.614172 123430403606336 max_utils.py:800] System Information: Jax Version: 0.8.1 I0422 17:45:44.614443 123430403606336 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0422 17:45:44.614496 123430403606336 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 I0422 17:45:44.614528 123430403606336 train_utils.py:348] WARNING: 'dataset_path' might be pointing your local file system I0422 17:45:44.614710 123430403606336 train.py:706] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0422 17:45:45.273963 123430403606336 maxtext_utils.py:1604] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 17:45:45.274247 123430403606336 checkpointing.py:677] Setting up checkpoint logger... I0422 17:45:45.274292 123430403606336 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 17:45:45.274329 123430403606336 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0422 17:45:45.274753 123430403606336 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 0x70421a5b31d0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 17:45:48.203854 123430403606336 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0422 17:45:48.204246 123430403606336 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 0x70403f761c40>}, handler_registry=None I0422 17:45:48.204483 123430403606336 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x70403f761c40>` 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`. I0422 17:45:48.204526 123430403606336 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x70403f761e80>` 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`. I0422 17:45:48.204555 123430403606336 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 0x70403f761c40>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x70403f761c40>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x70403f761e80>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x70403f761e80>}). I0422 17:45:48.204854 123430403606336 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0422 17:45:48.204905 123430403606336 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x70403f774540> timeout: 600 secs and primary_host=0 for async checkpoint writes I0422 17:45:48.338485 123430403606336 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints I0422 17:45:48.338717 123430403606336 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_20260422_174509/linen_main_20260422_174509_01_base/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7040ba7424b0> I0422 17:45:48.338999 123430403606336 checkpointing.py:301] Checkpoint manager created! I0422 17:45:48.416877 123430403606336 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0422 17:45:48.419796 123430403606336 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. I0422 17:45:48.480828 123430403606336 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 I0422 17:45:48.503493 123430403606336 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2 I0422 17:45:48.503557 123430403606336 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2 I0422 17:45:49.117880 123430403606336 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 17:45:49.117991 123430403606336 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0422 17:45:49.420724 123430403606336 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 17:45:49.420813 123430403606336 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0422 17:45:49.433643 123430403606336 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 17:45:49.433695 123430403606336 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0422 17:45:49.452848 123430403606336 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 17:45:49.452909 123430403606336 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0422 17:45:49.465760 123430403606336 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 17:45:49.465813 123430403606336 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0422 17:45:49.478581 123430403606336 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 17:45:49.478632 123430403606336 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0422 17:45:49.498808 123430403606336 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 17:45:49.498869 123430403606336 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0422 17:45:49.514861 123430403606336 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 17:45:49.514917 123430403606336 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0422 17:45:49.669745 123430403606336 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0422 17:45:49.669849 123430403606336 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 I0422 17:45:50.295269 123430403606336 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0422 17:45:50.295384 123430403606336 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) I0422 17:45:50.295427 123430403606336 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) I0422 17:45:50.295473 123430403606336 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') I0422 17:45:50.295513 123430403606336 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) I0422 17:45:50.295543 123430403606336 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) I0422 17:45:50.295584 123430403606336 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) I0422 17:45:50.295624 123430403606336 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') I0422 17:45:50.295655 123430403606336 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) I0422 17:45:50.295682 123430403606336 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) I0422 17:45:50.295718 123430403606336 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 17:45:50.295763 123430403606336 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 17:45:50.983472 123430403606336 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0422 17:45:50.983562 123430403606336 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0422 17:45:50.995706 123430403606336 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0422 17:45:50.995764 123430403606336 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0422 17:45:58.342614 123430403606336 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. I0422 17:45:58.343384 123430403606336 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0422 17:45:58.343622 123430403606336 metric_logger.py:301] number parameters: 1.104 billion I0422 17:46:08.027452 123430403606336 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0422 17:46:08.155571 123430403606336 checkpointing.py:776] Waited 0.128096342086792 seconds for step 0 to finish before starting checkpointing. I0422 17:46:08.156506 123430403606336 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 17:46:08.156670 123430403606336 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0422 17:46:08.157142 123430403606336 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/0. I0422 17:46:08.259069 123430403606336 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0422 17:46:08.348609 123305168209472 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/0 I0422 17:46:08.373931 123430403606336 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0422 17:46:08.374034 123430403606336 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 I0422 17:46:09.204406 123304952202816 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/0/items I0422 17:46:09.365890 123304226588224 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776879969107124895, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776879969107124895, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/0/_CHECKPOINT_METADATA I0422 17:46:09.736606 2017793 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0422 17:46:09.736669 2017793 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0422 17:46:10.343320 123430403606336 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.969855s I0422 17:46:10.348206 123430403606336 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.911 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 2 seconds) (per-host) I0422 17:46:10.348344 123430403606336 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 2.087847s (batch_requests_ready=0.109290s, total_serialization_initiated=1.973730s, others=0.004827s) I0422 17:46:10.348434 123430403606336 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.088811s (all_items=0.000023s, per_item={'items': '0.00002337'}, temp_paths=2.088787) I0422 17:46:10.350094 123303563888192 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0422 17:46:10.350181 123430403606336 async_checkpointer.py:561] Finished blocking save. Time taken: 2.193468s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/0. I0422 17:46:10.350378 123430403606336 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 17:46:10.350590 123305178695232 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0422 17:46:10.350687 123430403606336 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776879968.1564915, 'wait_for_prev_duration_secs': 4.57763671875e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776879968.1566916, 'checkpointer_blocking_duration_secs': 2.193596601486206, 'get_old_steps_start_time': 1776879970.3503091, 'get_old_steps_duration_secs': 3.0994415283203125e-05, 'checkpoint_manager_blocking_start_time': 1776879968.1563873, 'checkpoint_manager_blocking_duration_secs': 2.194272994995117} I0422 17:46:10.350829 123430403606336 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0422 17:46:10.350894 123430403606336 max_utils.py:750] Memstats: After params initialized: I0422 17:46:10.350945 123430403606336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0422 17:46:10.350970 123430403606336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0422 17:46:10.350992 123430403606336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0422 17:46:10.351011 123430403606336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0422 17:46:10.351028 123430403606336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0422 17:46:10.351066 123430403606336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0422 17:46:10.351085 123430403606336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0422 17:46:10.351102 123430403606336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0422 17:46:10.730848 123430403606336 metric_logger.py:196] completed step: 0, seconds: 9.683, TFLOP/s/device: 1.403, Tokens/s/device: 211.500, total_weights: 13328, loss: 10.828, lm_loss: 10.828, perplexity: 50404.086 I0422 17:46:10.731862 123430403606336 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/tensorboard/' I0422 17:46:10.846887 123430403606336 metric_logger.py:196] completed step: 1, seconds: 2.701, TFLOP/s/device: 5.030, Tokens/s/device: 758.224, total_weights: 12332, loss: 10.828, lm_loss: 10.828, perplexity: 50391.758 I0422 17:46:10.964455 123430403606336 metric_logger.py:196] completed step: 2, seconds: 0.012, TFLOP/s/device: 1136.048, Tokens/s/device: 171237.458, total_weights: 15161, loss: 9.823, lm_loss: 9.823, perplexity: 18444.867 I0422 17:46:11.081855 123430403606336 metric_logger.py:196] completed step: 3, seconds: 0.116, TFLOP/s/device: 116.682, Tokens/s/device: 17587.551, total_weights: 13327, loss: 9.405, lm_loss: 9.405, perplexity: 12149.331 I0422 17:46:11.167327 123304931231296 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/0/items/array_metadatas/process_0 I0422 17:46:27.665935 123430403606336 metric_logger.py:196] completed step: 4, seconds: 0.117, TFLOP/s/device: 115.874, Tokens/s/device: 17465.759, total_weights: 11939, loss: 9.045, lm_loss: 9.045, perplexity: 8471.842 I0422 17:46:27.693098 123430403606336 metric_logger.py:196] completed step: 5, seconds: 0.117, TFLOP/s/device: 115.790, Tokens/s/device: 17453.108, total_weights: 15502, loss: 8.953, lm_loss: 8.953, perplexity: 7728.546 I0422 17:46:27.811119 123430403606336 metric_logger.py:196] completed step: 6, seconds: 16.600, TFLOP/s/device: 0.818, Tokens/s/device: 123.371, total_weights: 13864, loss: 8.775, lm_loss: 8.775, perplexity: 6468.536 I0422 17:46:27.928714 123430403606336 metric_logger.py:196] completed step: 7, seconds: 0.008, TFLOP/s/device: 1636.412, Tokens/s/device: 246657.835, total_weights: 12988, loss: 8.688, lm_loss: 8.688, perplexity: 5928.887 I0422 17:46:28.046241 123430403606336 metric_logger.py:196] completed step: 8, seconds: 0.120, TFLOP/s/device: 113.508, Tokens/s/device: 17109.154, total_weights: 13820, loss: 8.740, lm_loss: 8.740, perplexity: 6250.123 I0422 17:46:28.164726 123430403606336 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0422 17:46:28.166204 123430403606336 checkpointing.py:776] Waited 0.0015106201171875 seconds for step 9 to finish before starting checkpointing. I0422 17:46:28.166644 123430403606336 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 17:46:57.987448 123303765214784 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 47.638451s (commit=47.154181s, array_metadata_write=0.484271s) I0422 17:46:57.988608 123303563888192 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 254.118 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 49 seconds) (per-host) I0422 17:46:57.988668 123303563888192 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 47.638443s. I0422 17:46:58.277987 123303563888192 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776879969107124895, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/0/_CHECKPOINT_METADATA I0422 17:46:58.497689 123303563888192 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0422 17:46:58.694890 123304226588224 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776879969107124895, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/0/_CHECKPOINT_METADATA I0422 17:46:58.946726 123303563888192 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0422 17:46:58.946916 123303563888192 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.627062s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/0/items I0422 17:46:58.947718 123303563888192 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/0/items I0422 17:46:59.178606 123303563888192 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/0 I0422 17:46:59.936516 123303563888192 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/0`. I0422 17:46:59.937290 123303563888192 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 51.780584s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/0 I0422 17:46:59.937368 123303563888192 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 49.587144s. I0422 17:46:59.937587 123305178695232 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 17:46:59.937703 123305178695232 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0422 17:46:59.937773 123305178695232 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 17:46:59.937824 123305178695232 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 17:46:59.939129 123430403606336 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. W0422 17:46:59.939277 123430403606336 checkpoint_manager.py:1441] Waiting for previous save to complete took 31.772679 seconds. If this number is high, consider checkpointing less frequently. I0422 17:46:59.940694 123430403606336 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0422 17:46:59.941008 123430403606336 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/9. I0422 17:47:00.136088 123303563888192 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/9 I0422 17:47:00.141340 123430403606336 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0422 17:47:00.141442 123430403606336 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 I0422 17:47:00.857803 123430403606336 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.717007s I0422 17:47:00.858271 123430403606336 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 15.153 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 814 milliseconds) (per-host) I0422 17:47:00.858330 123430403606336 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.814520s (batch_requests_ready=0.092628s, total_serialization_initiated=0.721534s, others=0.000358s) I0422 17:47:00.858395 123430403606336 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.815114s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=0.815096) I0422 17:47:00.859495 123304899774016 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0422 17:47:00.859594 123430403606336 async_checkpointer.py:561] Finished blocking save. Time taken: 0.918858s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/9. I0422 17:47:00.859785 123430403606336 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 17:47:00.862928 123430403606336 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776879988.1665692, 'wait_for_prev_duration_secs': 31.77267861366272, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776880019.9407191, 'checkpointer_blocking_duration_secs': 0.9189789295196533, 'get_old_steps_start_time': 1776880020.8597176, 'get_old_steps_duration_secs': 2.5510787963867188e-05, 'checkpoint_manager_blocking_start_time': 1776879988.1665187, 'checkpoint_manager_blocking_duration_secs': 32.69638133049011} I0422 17:47:00.863032 123430403606336 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0422 17:47:00.863079 123430403606336 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 17:47:00.892548 123304931231296 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/9/items I0422 17:47:01.043071 123304698447424 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0422 17:47:02.149328 123304910259776 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/9/items/array_metadatas/process_0 I0422 17:47:34.974122 123303765214784 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 33.595548s (commit=33.139815s, array_metadata_write=0.455733s) I0422 17:47:34.975355 123304899774016 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 361.759 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 34 seconds) (per-host) I0422 17:47:34.975494 123304899774016 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 34.115854s. I0422 17:47:35.437117 123304899774016 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0422 17:47:35.851134 123304899774016 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0422 17:47:35.851320 123304899774016 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.580629s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/9/items I0422 17:47:35.852100 123304899774016 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/9/items I0422 17:47:36.103268 123304899774016 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/9 I0422 17:47:36.829711 123304899774016 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/9`. I0422 17:47:36.830452 123304899774016 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 36.889721s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_01_base/checkpoints/9 I0422 17:47:36.830525 123304899774016 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 35.970887s. I0422 17:47:36.830716 123304698447424 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 17:47:36.830851 123304698447424 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0422 17:47:36.830922 123304698447424 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 17:47:36.830968 123304698447424 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 17:47:36.831185 123430403606336 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. I0422 17:47:36.831340 123430403606336 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 17:47:36.832467 123430403606336 metric_logger.py:196] completed step: 9, seconds: 0.120, TFLOP/s/device: 112.923, Tokens/s/device: 17020.994, 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