2026-04-20 14:23:52.012558: 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) I0420 14:23:52.130349 124388244503680 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0420 14:24:32.881696 124388244503680 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 14:24:32.881818 124388244503680 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 14:24:32.881852 124388244503680 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Apr 6 2026 20:48:10 (1775533690) cl/895581894 I0420 14:24:32.881879 124388244503680 train_utils.py:365] WARNING: 'dataset_path' might be pointing your local file system I0420 14:24:32.881969 124388244503680 train.py:806] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0420 14:24:33.404510 124388244503680 maxtext_utils.py:1718] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 14:24:33.404685 124388244503680 maxtext_utils.py:1718] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 14:24:33.404924 124388244503680 checkpointing.py:688] Setting up checkpoint logger... I0420 14:24:33.404967 124388244503680 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0420 14:24:33.405039 124388244503680 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0420 14:24:33.405631 124388244503680 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 0x7120ce3fa420>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 14:24:36.299952 124388244503680 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0420 14:24:36.300216 124388244503680 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 0x711a51d47560>}, handler_registry=None I0420 14:24:36.300548 124388244503680 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x711a51d47560>` 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`. I0420 14:24:36.300590 124388244503680 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x711a51d74350>` 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`. I0420 14:24:36.300619 124388244503680 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 0x711a51d47560>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x711a51d47560>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x711a51d74350>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x711a51d74350>}). I0420 14:24:36.300938 124388244503680 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0420 14:24:36.300998 124388244503680 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x711a51d5fb00> timeout: 600 secs and primary_host=0 for async checkpoint writes I0420 14:24:36.433115 124388244503680 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints I0420 14:24:36.433369 124388244503680 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_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x711a53311160> I0420 14:24:36.433459 124388244503680 checkpointing.py:302] Checkpoint manager created! I0420 14:24:36.893344 124388244503680 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0420 14:24:36.897002 124388244503680 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. I0420 14:24:36.952253 124388244503680 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 I0420 14:24:36.981151 124388244503680 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2 I0420 14:24:36.981219 124388244503680 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2 I0420 14:24:37.596606 124388244503680 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 14:24:37.596719 124388244503680 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0420 14:24:37.701105 124388244503680 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 14:24:37.701192 124388244503680 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0420 14:24:37.716060 124388244503680 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 14:24:37.716112 124388244503680 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0420 14:24:37.737713 124388244503680 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 14:24:37.737777 124388244503680 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0420 14:24:37.752646 124388244503680 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 14:24:37.752700 124388244503680 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0420 14:24:37.767620 124388244503680 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 14:24:37.767672 124388244503680 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0420 14:24:37.790842 124388244503680 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 14:24:37.790908 124388244503680 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0420 14:24:37.809957 124388244503680 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 14:24:37.810012 124388244503680 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0420 14:24:38.001664 124388244503680 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0420 14:24:38.001787 124388244503680 checkpointing.py:676] 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 I0420 14:24:39.524473 124388244503680 maxtext_utils.py:1821] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 14:24:39.524582 124388244503680 maxtext_utils.py:1821] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0420 14:24:39.524623 124388244503680 maxtext_utils.py:1821] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0420 14:24:39.524668 124388244503680 maxtext_utils.py:1821] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0420 14:24:39.524713 124388244503680 maxtext_utils.py:1821] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0420 14:24:39.524743 124388244503680 maxtext_utils.py:1821] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0420 14:24:39.524791 124388244503680 maxtext_utils.py:1821] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0420 14:24:39.524831 124388244503680 maxtext_utils.py:1821] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0420 14:24:39.524860 124388244503680 maxtext_utils.py:1821] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0420 14:24:39.524887 124388244503680 maxtext_utils.py:1821] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0420 14:24:39.524923 124388244503680 maxtext_utils.py:1821] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 14:24:39.524958 124388244503680 maxtext_utils.py:1821] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 14:24:39.959321 124388244503680 train.py:157] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0420 14:24:39.959413 124388244503680 train.py:157] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0420 14:24:39.973988 124388244503680 train.py:164] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0420 14:24:39.974059 124388244503680 train.py:164] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0420 14:24:47.388489 124388244503680 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. I0420 14:24:47.389147 124388244503680 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0420 14:24:47.389679 124388244503680 metric_logger.py:301] number parameters: 1.104 billion I0420 14:24:58.212757 124388244503680 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0420 14:24:58.343385 124388244503680 checkpointing.py:798] Waited 0.13062024116516113 seconds for step 0 to finish before starting checkpointing. I0420 14:24:58.344040 124388244503680 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 14:24:58.344256 124388244503680 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0420 14:24:58.344743 124388244503680 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/0. I0420 14:24:58.438508 124388244503680 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0420 14:24:58.526283 124273297458752 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/0 I0420 14:24:58.547573 124388244503680 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0420 14:24:58.547641 124388244503680 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 I0420 14:24:59.310748 124388244503680 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.763597s I0420 14:24:59.311113 124388244503680 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 14.170 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 870 milliseconds) (per-host) I0420 14:24:59.311165 124388244503680 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.871035s (batch_requests_ready=0.103346s, total_serialization_initiated=0.767442s, others=0.000248s) I0420 14:24:59.311230 124388244503680 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.872138s (all_items=0.000047s, per_item={'items': '0.00004697'}, temp_paths=0.872091) I0420 14:24:59.311977 124273217766976 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0420 14:24:59.312097 124388244503680 async_checkpointer.py:561] Finished blocking save. Time taken: 0.967799s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/0. I0420 14:24:59.312295 124388244503680 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0420 14:24:59.312500 124388244503680 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776695098.3440256, 'wait_for_prev_duration_secs': 6.413459777832031e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776695098.344278, 'checkpointer_blocking_duration_secs': 0.9679350852966309, 'get_old_steps_start_time': 1776695099.3122306, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776695098.3439116, 'checkpoint_manager_blocking_duration_secs': 0.9685649871826172} I0420 14:24:59.312577 124388244503680 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0420 14:24:59.312623 124388244503680 max_utils.py:750] Memstats: After params initialized: I0420 14:24:59.312711 124388244503680 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0420 14:24:59.312736 124388244503680 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0420 14:24:59.312757 124388244503680 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0420 14:24:59.312775 124388244503680 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0420 14:24:59.312793 124388244503680 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0420 14:24:59.312809 124388244503680 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0420 14:24:59.312827 124388244503680 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0420 14:24:59.312844 124388244503680 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0420 14:24:59.380377 124273286972992 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/0/items I0420 14:24:59.518630 124273238738496 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776695099249335081, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776695099249335081, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/0/_CHECKPOINT_METADATA I0420 14:24:59.518865 124273207281216 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 14:24:59.757290 124388244503680 metric_logger.py:196] completed step: 0, seconds: 10.822, TFLOP/s/device: 1.256, Tokens/s/device: 189.243, total_weights: 13328, loss: 10.828, lm_loss: 10.828, perplexity: 50404.086 I0420 14:24:59.758312 124388244503680 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/tensorboard/' I0420 14:24:59.873890 1035300 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0420 14:24:59.873969 1035300 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0420 14:24:59.877352 124388244503680 metric_logger.py:196] completed step: 1, seconds: 1.542, TFLOP/s/device: 8.813, Tokens/s/device: 1328.393, total_weights: 12332, loss: 10.828, lm_loss: 10.828, perplexity: 50391.758 I0420 14:24:59.989298 124388244503680 metric_logger.py:196] completed step: 2, seconds: 0.011, TFLOP/s/device: 1244.927, Tokens/s/device: 187648.891, total_weights: 15161, loss: 9.823, lm_loss: 9.823, perplexity: 18444.867 I0420 14:25:00.107020 124388244503680 metric_logger.py:196] completed step: 3, seconds: 0.139, TFLOP/s/device: 97.779, Tokens/s/device: 14738.266, total_weights: 13327, loss: 9.405, lm_loss: 9.405, perplexity: 12149.331 I0420 14:25:00.866266 124273276487232 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/0/items/array_metadatas/process_0 I0420 14:25:19.953068 124388244503680 metric_logger.py:196] completed step: 4, seconds: 0.094, TFLOP/s/device: 144.098, Tokens/s/device: 21719.994, total_weights: 11939, loss: 9.045, lm_loss: 9.045, perplexity: 8471.842 I0420 14:25:19.961721 124388244503680 metric_logger.py:196] completed step: 5, seconds: 0.117, TFLOP/s/device: 115.639, Tokens/s/device: 17430.381, total_weights: 15502, loss: 8.953, lm_loss: 8.953, perplexity: 7728.546 I0420 14:25:20.078545 124388244503680 metric_logger.py:196] completed step: 6, seconds: 19.844, TFLOP/s/device: 0.685, Tokens/s/device: 103.203, total_weights: 13864, loss: 8.775, lm_loss: 8.775, perplexity: 6468.536 I0420 14:25:20.195785 124388244503680 metric_logger.py:196] completed step: 7, seconds: 0.006, TFLOP/s/device: 2117.695, Tokens/s/device: 319201.995, total_weights: 12988, loss: 8.688, lm_loss: 8.688, perplexity: 5928.887 I0420 14:25:20.313331 124388244503680 metric_logger.py:196] completed step: 8, seconds: 0.118, TFLOP/s/device: 114.907, Tokens/s/device: 17319.971, total_weights: 13820, loss: 8.740, lm_loss: 8.740, perplexity: 6250.123 I0420 14:25:20.430271 124388244503680 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0420 14:25:20.431104 124388244503680 checkpointing.py:798] Waited 0.0008690357208251953 seconds for step 9 to finish before starting checkpointing. I0420 14:25:20.431386 124388244503680 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 14:25:26.644812 124273228252736 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 26.768290s (commit=26.323886s, array_metadata_write=0.444404s) I0420 14:25:26.645978 124273217766976 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 448.021 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 28 seconds) (per-host) I0420 14:25:26.646107 124273217766976 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 27.333962s. I0420 14:25:26.880035 124273217766976 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776695099249335081, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/0/_CHECKPOINT_METADATA I0420 14:25:27.062664 124273217766976 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0420 14:25:27.263345 124273238738496 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776695099249335081, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/0/_CHECKPOINT_METADATA I0420 14:25:27.485026 124273217766976 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0420 14:25:27.485736 124273217766976 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.564552s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/0/items I0420 14:25:27.486488 124273217766976 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/0/items I0420 14:25:27.761339 124273217766976 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/0 I0420 14:25:28.435314 124273217766976 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/0`. I0420 14:25:28.436064 124273217766976 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 30.091756s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/0 I0420 14:25:28.436152 124273217766976 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 29.124009s. I0420 14:25:28.436268 124273207281216 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 14:25:28.436319 124273207281216 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 14:25:28.436375 124273207281216 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0420 14:25:28.436419 124273207281216 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0420 14:25:28.436574 124388244503680 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. W0420 14:25:28.436713 124388244503680 checkpoint_manager.py:1441] Waiting for previous save to complete took 8.005335 seconds. If this number is high, consider checkpointing less frequently. I0420 14:25:28.437345 124388244503680 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0420 14:25:28.437664 124388244503680 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/9. I0420 14:25:28.615616 124273207281216 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/9 I0420 14:25:28.627954 124388244503680 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0420 14:25:28.628074 124388244503680 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 I0420 14:25:29.333348 124273024828992 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/9/items I0420 14:25:32.501695 124388244503680 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 3.874220s I0420 14:25:32.506129 124388244503680 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.104 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 3 seconds) (per-host) I0420 14:25:32.506214 124388244503680 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 3.976446s (batch_requests_ready=0.094701s, total_serialization_initiated=3.877419s, others=0.004326s) I0420 14:25:32.506295 124388244503680 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 3.977047s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=3.977031) I0420 14:25:32.507874 124273207281216 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0420 14:25:32.507979 124388244503680 async_checkpointer.py:561] Finished blocking save. Time taken: 4.070586s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/9. I0420 14:25:32.508241 124388244503680 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0420 14:25:32.508391 124273217766976 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 14:25:32.508471 124388244503680 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776695120.4313533, 'wait_for_prev_duration_secs': 8.005335092544556, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776695128.4373693, 'checkpointer_blocking_duration_secs': 4.070793867111206, 'get_old_steps_start_time': 1776695132.5081823, 'get_old_steps_duration_secs': 2.2172927856445312e-05, 'checkpoint_manager_blocking_start_time': 1776695120.4313018, 'checkpoint_manager_blocking_duration_secs': 12.077144622802734} I0420 14:25:32.508563 124388244503680 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0420 14:25:32.508594 124388244503680 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 14:25:33.322972 124273307944512 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/9/items/array_metadatas/process_0 I0420 14:26:08.897475 124273024828992 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 36.390588s (commit=35.901019s, array_metadata_write=0.489568s) I0420 14:26:08.898767 124273207281216 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 313.033 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 40 seconds) (per-host) I0420 14:26:08.898900 124273207281216 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.390810s. I0420 14:26:09.308975 124273207281216 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0420 14:26:09.785839 124273207281216 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0420 14:26:09.786540 124273207281216 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.621317s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/9/items I0420 14:26:09.787307 124273207281216 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/9/items I0420 14:26:10.043115 124273207281216 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/9 I0420 14:26:10.746062 124273207281216 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/9`. I0420 14:26:10.746784 124273207281216 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 42.309402s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260420_142345/linen_feat_nnx_trainstate_and_training_loop_20260420_142345_01_base/checkpoints/9 I0420 14:26:10.746868 124273207281216 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 38.238780s. I0420 14:26:10.746986 124273217766976 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 14:26:10.747035 124273217766976 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 14:26:10.747120 124273217766976 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0420 14:26:10.747166 124273217766976 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0420 14:26:10.747295 124388244503680 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. I0420 14:26:10.747440 124388244503680 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 14:26:10.748723 124388244503680 metric_logger.py:196] completed step: 9, seconds: 0.117, TFLOP/s/device: 115.813, Tokens/s/device: 17456.529, 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