2026-04-16 00:48:43.888726: 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) I0416 00:48:44.009399 137137970302080 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0416 00:49:14.980705 137137970302080 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 00:49:14.980842 137137970302080 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 00:49:14.980881 137137970302080 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 I0416 00:49:14.980919 137137970302080 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system I0416 00:49:14.981073 137137970302080 train.py:812] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 00:49:15.082491 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 00:49:15.489654 137137970302080 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 00:49:15.489842 137137970302080 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 00:49:15.490125 137137970302080 checkpointing.py:688] Setting up checkpoint logger... I0416 00:49:15.490175 137137970302080 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0416 00:49:15.490229 137137970302080 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 00:49:15.490877 137137970302080 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 0x7cb954728350>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 00:49:18.310851 137137970302080 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 00:49:18.311108 137137970302080 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 0x7cb2da74ccb0>}, handler_registry=None I0416 00:49:18.311421 137137970302080 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cb2da74ccb0>` 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`. I0416 00:49:18.311462 137137970302080 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cb2d8f5af60>` 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`. I0416 00:49:18.311491 137137970302080 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 0x7cb2da74ccb0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cb2da74ccb0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cb2d8f5af60>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cb2d8f5af60>}). I0416 00:49:18.311798 137137970302080 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 00:49:18.311851 137137970302080 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7cb2d8f5f4c0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 00:49:18.439485 137137970302080 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints I0416 00:49:18.439743 137137970302080 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_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7cb2da74f200> I0416 00:49:18.439837 137137970302080 checkpointing.py:302] Checkpoint manager created! I0416 00:49:18.520005 137137970302080 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0416 00:49:18.523660 137137970302080 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. I0416 00:49:18.578956 137137970302080 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 I0416 00:49:18.609826 137137970302080 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2 I0416 00:49:18.609898 137137970302080 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2 I0416 00:49:19.637076 137137970302080 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 00:49:19.637181 137137970302080 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 00:49:19.740695 137137970302080 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 00:49:19.740791 137137970302080 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 00:49:19.755892 137137970302080 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 00:49:19.755946 137137970302080 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 00:49:19.777427 137137970302080 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 00:49:19.777488 137137970302080 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 00:49:19.792656 137137970302080 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 00:49:19.792711 137137970302080 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 00:49:19.807838 137137970302080 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 00:49:19.807893 137137970302080 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 00:49:19.830404 137137970302080 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 00:49:19.830462 137137970302080 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 00:49:19.849384 137137970302080 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 00:49:19.849442 137137970302080 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0416 00:49:20.060770 137137970302080 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 00:49:20.060871 137137970302080 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. W0416 00:49:20.112439 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. [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 I0416 00:49:20.505542 137137970302080 maxtext_utils.py:1790] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0416 00:49:20.505654 137137970302080 maxtext_utils.py:1790] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0416 00:49:20.505696 137137970302080 maxtext_utils.py:1790] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0416 00:49:20.505740 137137970302080 maxtext_utils.py:1790] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0416 00:49:20.505780 137137970302080 maxtext_utils.py:1790] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0416 00:49:20.505807 137137970302080 maxtext_utils.py:1790] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0416 00:49:20.505845 137137970302080 maxtext_utils.py:1790] 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) I0416 00:49:20.505883 137137970302080 maxtext_utils.py:1790] 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') I0416 00:49:20.505911 137137970302080 maxtext_utils.py:1790] 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) I0416 00:49:20.505937 137137970302080 maxtext_utils.py:1790] 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) I0416 00:49:20.505973 137137970302080 maxtext_utils.py:1790] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0416 00:49:20.506007 137137970302080 maxtext_utils.py:1790] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0416 00:49:20.945375 137137970302080 train.py:157] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 00:49:20.945468 137137970302080 train.py:157] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0416 00:49:20.960447 137137970302080 train.py:164] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 00:49:20.960499 137137970302080 train.py:164] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). W0416 00:49:21.255092 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 00:49:21.372640 137137970302080 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. I0416 00:49:21.373219 137137970302080 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 00:49:21.373363 137137970302080 metric_logger.py:289] number parameters: 1.104 billion W0416 00:49:24.484689 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 00:49:24.616780 137137970302080 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0416 00:49:24.737552 137137970302080 checkpointing.py:798] Waited 0.12074875831604004 seconds for step 0 to finish before starting checkpointing. I0416 00:49:24.738060 137137970302080 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 00:49:24.738215 137137970302080 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0416 00:49:24.738608 137137970302080 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/0. I0416 00:49:24.826964 137137970302080 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 00:49:24.919840 137024726107712 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/0 I0416 00:49:24.930142 137137970302080 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 00:49:24.930243 137137970302080 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 W0416 00:49:25.157162 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 00:49:25.170132 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 00:49:25.179962 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 00:49:25.190147 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 00:49:25.203443 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 00:49:25.212234 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 00:49:25.217305 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 00:49:25.221254 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 00:49:25.673426 137137970302080 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.743762s I0416 00:49:25.673911 137137970302080 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 14.596 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 845 milliseconds) (per-host) I0416 00:49:25.673963 137137970302080 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.845563s (batch_requests_ready=0.097378s, total_serialization_initiated=0.747812s, others=0.000374s) I0416 00:49:25.674027 137137970302080 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.846481s (all_items=0.000021s, per_item={'items': '0.00002050'}, temp_paths=0.846460) I0416 00:49:25.675328 137024654804544 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 00:49:25.675433 137137970302080 async_checkpointer.py:561] Finished blocking save. Time taken: 0.937179s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/0. I0416 00:49:25.675614 137137970302080 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0416 00:49:25.675885 137137970302080 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776300564.7380326, 'wait_for_prev_duration_secs': 5.7697296142578125e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776300564.7382357, 'checkpointer_blocking_duration_secs': 0.9373066425323486, 'get_old_steps_start_time': 1776300565.6755607, 'get_old_steps_duration_secs': 2.002716064453125e-05, 'checkpoint_manager_blocking_start_time': 1776300564.7379372, 'checkpoint_manager_blocking_duration_secs': 0.937924861907959} I0416 00:49:25.675955 137137970302080 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0416 00:49:25.676000 137137970302080 max_utils.py:750] Memstats: After params initialized: I0416 00:49:25.676099 137137970302080 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0416 00:49:25.676126 137137970302080 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0416 00:49:25.676146 137137970302080 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0416 00:49:25.676164 137137970302080 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0416 00:49:25.676181 137137970302080 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0416 00:49:25.676198 137137970302080 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0416 00:49:25.676214 137137970302080 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0416 00:49:25.676230 137137970302080 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0416 00:49:25.682704 137024715621952 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/0/items W0416 00:49:25.682928 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 00:49:25.688729 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 00:49:25.693535 3443372 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 00:49:25.826701 137024675776064 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776300565575113535, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776300565575113535, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/0/_CHECKPOINT_METADATA I0416 00:49:25.827032 137024736593472 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 00:49:26.049117 137137970302080 metric_logger.py:185] completed step: 0, seconds: 3.243, TFLOP/s/device: 4.190, Tokens/s/device: 631.538, total_weights: 13328, loss: 10.828 I0416 00:49:26.050068 137137970302080 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/tensorboard/' I0416 00:49:26.165521 137137970302080 metric_logger.py:185] completed step: 1, seconds: 1.431, TFLOP/s/device: 9.498, Tokens/s/device: 1431.653, total_weights: 12332, loss: 10.828 I0416 00:49:26.283132 137137970302080 metric_logger.py:185] completed step: 2, seconds: 0.009, TFLOP/s/device: 1431.882, Tokens/s/device: 215828.854, total_weights: 15161, loss: 9.823 I0416 00:49:26.395329 3446100 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 00:49:26.395383 3446100 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 00:49:26.400403 137137970302080 metric_logger.py:185] completed step: 3, seconds: 0.117, TFLOP/s/device: 115.975, Tokens/s/device: 17480.966, total_weights: 13327, loss: 9.405 I0416 00:49:27.360358 137024705136192 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_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/0/items/array_metadatas/process_0 I0416 00:49:46.460515 137137970302080 metric_logger.py:185] completed step: 4, seconds: 0.118, TFLOP/s/device: 115.325, Tokens/s/device: 17383.038, total_weights: 11939, loss: 9.045 I0416 00:49:46.469681 137137970302080 metric_logger.py:185] completed step: 5, seconds: 0.136, TFLOP/s/device: 99.644, Tokens/s/device: 15019.398, total_weights: 15502, loss: 8.953 I0416 00:49:46.586502 137137970302080 metric_logger.py:185] completed step: 6, seconds: 20.041, TFLOP/s/device: 0.678, Tokens/s/device: 102.193, total_weights: 13864, loss: 8.775 I0416 00:49:46.703850 137137970302080 metric_logger.py:185] completed step: 7, seconds: 0.006, TFLOP/s/device: 2206.777, Tokens/s/device: 332629.527, total_weights: 12988, loss: 8.688 I0416 00:49:46.821549 137137970302080 metric_logger.py:185] completed step: 8, seconds: 0.118, TFLOP/s/device: 115.231, Tokens/s/device: 17368.885, total_weights: 13820, loss: 8.740 I0416 00:49:46.938842 137137970302080 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0416 00:49:46.939723 137137970302080 checkpointing.py:798] Waited 0.0009195804595947266 seconds for step 9 to finish before starting checkpointing. I0416 00:49:46.939999 137137970302080 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 00:49:52.659922 137024665290304 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 26.260588s (commit=25.814273s, array_metadata_write=0.446316s) I0416 00:49:52.661205 137024654804544 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 454.026 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 27 seconds) (per-host) I0416 00:49:52.661326 137024654804544 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 26.985847s. I0416 00:49:52.886731 137024654804544 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776300565575113535, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/0/_CHECKPOINT_METADATA I0416 00:49:53.073955 137024654804544 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0416 00:49:53.283863 137024675776064 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776300565575113535, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/0/_CHECKPOINT_METADATA I0416 00:49:53.558775 137024654804544 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 00:49:53.558976 137024654804544 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.623612s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/0/items I0416 00:49:53.559751 137024654804544 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/0/items I0416 00:49:53.813680 137024654804544 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/0 I0416 00:49:54.484264 137024654804544 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_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/0`. I0416 00:49:54.485029 137024654804544 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 29.746781s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/0 I0416 00:49:54.485123 137024654804544 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 28.809645s. I0416 00:49:54.485333 137024736593472 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 00:49:54.485456 137024736593472 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 00:49:54.485515 137024736593472 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0416 00:49:54.485561 137024736593472 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0416 00:49:54.485695 137137970302080 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. W0416 00:49:54.485840 137137970302080 checkpoint_manager.py:1441] Waiting for previous save to complete took 7.545842 seconds. If this number is high, consider checkpointing less frequently. I0416 00:49:54.486635 137137970302080 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0416 00:49:54.486939 137137970302080 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/9. I0416 00:49:54.681930 137024736593472 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/9 I0416 00:49:54.682122 137137970302080 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 00:49:54.682374 137137970302080 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 I0416 00:49:55.416857 137024512198208 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/9/items I0416 00:50:04.499867 137137970302080 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 9.818480s I0416 00:50:04.500312 137137970302080 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.244 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 9 seconds) (per-host) I0416 00:50:04.503806 137137970302080 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 9.922220s (batch_requests_ready=0.096577s, total_serialization_initiated=9.821888s, others=0.003755s) I0416 00:50:04.503875 137137970302080 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 9.922787s (all_items=0.000021s, per_item={'items': '0.00002074'}, temp_paths=9.922766) I0416 00:50:04.505110 137024736593472 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 00:50:04.505201 137137970302080 async_checkpointer.py:561] Finished blocking save. Time taken: 10.018524s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/9. I0416 00:50:04.505381 137137970302080 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0416 00:50:04.505542 137024614958656 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 00:50:04.505623 137137970302080 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776300586.9399636, 'wait_for_prev_duration_secs': 7.545842170715332, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776300594.486659, 'checkpointer_blocking_duration_secs': 10.018641233444214, 'get_old_steps_start_time': 1776300604.5053165, 'get_old_steps_duration_secs': 2.7418136596679688e-05, 'checkpoint_manager_blocking_start_time': 1776300586.9399195, 'checkpoint_manager_blocking_duration_secs': 17.565677165985107} I0416 00:50:04.505714 137137970302080 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0416 00:50:04.505747 137137970302080 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 00:50:05.284942 137024705136192 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_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/9/items/array_metadatas/process_0 I0416 00:50:39.273164 137024512198208 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 34.768877s (commit=34.266034s, array_metadata_write=0.502842s) I0416 00:50:39.274555 137024736593472 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 282.747 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 44 seconds) (per-host) I0416 00:50:39.274672 137024736593472 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 34.769427s. I0416 00:50:39.730005 137024736593472 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0416 00:50:40.159254 137024736593472 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 00:50:40.159866 137024736593472 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.571933s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/9/items I0416 00:50:40.160623 137024736593472 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/9/items I0416 00:50:40.410665 137024736593472 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/9 I0416 00:50:41.141254 137024736593472 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_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/9`. I0416 00:50:41.142008 137024736593472 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 46.655337s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_01_base/checkpoints/9 I0416 00:50:41.142108 137024736593472 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 36.636863s. I0416 00:50:41.142246 137024614958656 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 00:50:41.142295 137024614958656 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 00:50:41.142363 137024614958656 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0416 00:50:41.142404 137024614958656 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0416 00:50:41.142673 137137970302080 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. I0416 00:50:41.142837 137137970302080 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 00:50:41.143806 137137970302080 metric_logger.py:185] completed step: 9, seconds: 0.117, TFLOP/s/device: 115.829, Tokens/s/device: 17459.059, total_weights: 12300, loss: 8.710 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops