2026-04-16 19:35:28.022934: 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 19:35:28.138286 129740583242880 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0416 19:36:25.213494 129740583242880 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 19:36:25.213606 129740583242880 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 19:36:25.213641 129740583242880 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 19:36:25.213671 129740583242880 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system I0416 19:36:25.213693 129740583242880 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0416 19:36:25.213770 129740583242880 train.py:811] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 19:36:25.308404 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:36:25.710386 129740583242880 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 19:36:25.710553 129740583242880 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 19:36:25.710774 129740583242880 checkpointing.py:688] Setting up checkpoint logger... I0416 19:36:25.710813 129740583242880 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=False I0416 19:36:25.710847 129740583242880 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 19:36:25.711491 129740583242880 base_pytree_checkpoint_handler.py:411] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x75fefe7fca10>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 19:36:27.345992 129740583242880 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 19:36:27.346189 129740583242880 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 0x75f9529551c0>}, handler_registry=None I0416 19:36:27.346423 129740583242880 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x75f9529551c0>` 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 19:36:27.346461 129740583242880 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x75f882134440>` 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 19:36:27.346491 129740583242880 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 0x75f9529551c0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x75f9529551c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x75f882134440>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x75f882134440>}). I0416 19:36:27.346782 129740583242880 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 19:36:27.346831 129740583242880 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x75f882131ee0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 19:36:27.760991 129634473608768 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776367992525514067, 'commit_timestamp_nsecs': 1776368041868254653, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/5/_CHECKPOINT_METADATA I0416 19:36:27.856414 129634473608768 checkpoint_manager.py:1847] Missing metrics for step 5 I0416 19:36:27.870886 129740583242880 checkpoint_manager.py:1788] Found 3 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints I0416 19:36:27.871323 129740583242880 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=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x75fa005148f0> I0416 19:36:27.871407 129740583242880 checkpointing.py:302] Checkpoint manager created! W0416 19:36:27.887982 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:36:28.074679 129740583242880 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 19:36:28.074868 129740583242880 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 19:36:28.163866 129740583242880 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 19:36:28.163945 129740583242880 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 19:36:28.178068 129740583242880 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 19:36:28.178116 129740583242880 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 19:36:28.199272 129740583242880 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 19:36:28.199327 129740583242880 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 19:36:28.213504 129740583242880 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 19:36:28.213551 129740583242880 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 19:36:28.227705 129740583242880 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 19:36:28.227753 129740583242880 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 19:36:28.249258 129740583242880 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 19:36:28.249314 129740583242880 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 19:36:28.267199 129740583242880 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 19:36:28.267250 129740583242880 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0416 19:36:28.454619 129740583242880 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 19:36:28.454710 129740583242880 checkpointing.py:582] restoring from this run's directory step 9 I0416 19:36:28.608216 129740583242880 checkpointer.py:304] Restoring checkpoint from gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/9. I0416 19:36:29.161136 4044021 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 19:36:29.161202 4044021 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 19:36:34.716173 129740583242880 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/read/gbytes_per_sec: 2.143 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 5 seconds) (per-host) I0416 19:36:34.716503 129740583242880 checkpointer.py:318] Finished restoring checkpoint in 6.18 seconds from gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/9. I0416 19:36:34.716607 129740583242880 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints', 'checkpointer_start_time': 1776368188.5350246, 'checkpointer_duration_secs': 6.181530952453613, 'checkpoint_manager_start_time': 1776368188.4549367, 'checkpoint_manager_duration_secs': 6.261622428894043} [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 19:36:35.228989 129740583242880 maxtext_utils.py:1796] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0416 19:36:35.229113 129740583242880 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0416 19:36:35.229156 129740583242880 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0416 19:36:35.229200 129740583242880 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0416 19:36:35.229240 129740583242880 maxtext_utils.py:1796] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0416 19:36:35.229270 129740583242880 maxtext_utils.py:1796] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0416 19:36:35.229310 129740583242880 maxtext_utils.py:1796] 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 19:36:35.229352 129740583242880 maxtext_utils.py:1796] 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 19:36:35.229379 129740583242880 maxtext_utils.py:1796] 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 19:36:35.229404 129740583242880 maxtext_utils.py:1796] 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 19:36:35.229439 129740583242880 maxtext_utils.py:1796] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0416 19:36:35.229475 129740583242880 maxtext_utils.py:1796] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0416 19:36:35.647729 129740583242880 train.py:157] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 19:36:35.647808 129740583242880 train.py:157] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0416 19:36:35.661557 129740583242880 train.py:164] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 19:36:35.661615 129740583242880 train.py:164] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). W0416 19:36:36.324299 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:36:36.387284 129740583242880 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 19:36:36.387865 129740583242880 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 19:36:36.388007 129740583242880 metric_logger.py:289] number parameters: 1.104 billion W0416 19:36:37.231463 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:36:37.299536 129740583242880 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0416 19:36:37.416954 129740583242880 checkpointing.py:798] Waited 0.11740851402282715 seconds for step 10 to finish before starting checkpointing. I0416 19:36:37.417363 129740583242880 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 19:36:37.417538 129740583242880 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 10 I0416 19:36:37.417837 129740583242880 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10. I0416 19:36:37.499394 129740583242880 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 19:36:37.590219 129632777012800 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10 I0416 19:36:37.594096 129740583242880 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 19:36:37.594188 129740583242880 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 19:36:37.814567 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:36:37.830038 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:36:37.837678 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:36:37.844829 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:36:37.849091 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:36:37.853232 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:36:37.857175 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:36:37.861178 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:36:38.283195 129632766527040 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10/items I0416 19:36:38.444740 129632735069760 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776368198206459696, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776368198206459696, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10/_CHECKPOINT_METADATA I0416 19:36:42.742693 129740583242880 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 5.149015s I0416 19:36:42.746679 129740583242880 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.352 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 5 seconds) (per-host) I0416 19:36:42.746743 129740583242880 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 5.246169s (batch_requests_ready=0.090253s, total_serialization_initiated=5.152027s, others=0.003888s) I0416 19:36:42.746824 129740583242880 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 5.246981s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=5.246963) I0416 19:36:42.748181 129632787498560 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 19:36:42.748276 129740583242880 async_checkpointer.py:561] Finished blocking save. Time taken: 5.330687s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10. I0416 19:36:42.748511 129740583242880 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0416 19:36:42.748678 129632756041280 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 19:36:42.748771 129740583242880 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776368197.4173508, 'wait_for_prev_duration_secs': 4.029273986816406e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776368197.4175587, 'checkpointer_blocking_duration_secs': 5.330859422683716, 'get_old_steps_start_time': 1776368202.74844, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1776368197.417266, 'checkpoint_manager_blocking_duration_secs': 5.331476211547852} I0416 19:36:42.748886 129740583242880 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0416 19:36:42.748926 129740583242880 max_utils.py:750] Memstats: After params initialized: I0416 19:36:42.748983 129740583242880 max_utils.py:756] Using (GB) 1.6 / 31.25 (5.120000%) on TPU_0(process=0,(0,0,0,0)) I0416 19:36:42.749010 129740583242880 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0416 19:36:42.749030 129740583242880 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0416 19:36:42.749064 129740583242880 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0416 19:36:42.749084 129740583242880 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0416 19:36:42.749102 129740583242880 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0416 19:36:42.749120 129740583242880 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0416 19:36:42.749137 129740583242880 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0416 19:36:42.755285 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:36:42.761582 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:36:42.765839 4042427 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:36:42.958106 129740583242880 metric_logger.py:185] completed step: 10, seconds: 0.911, TFLOP/s/device: 14.914, Tokens/s/device: 2247.926, total_weights: 16384, loss: 3.602 I0416 19:36:43.074215 129740583242880 metric_logger.py:185] completed step: 11, seconds: 5.657, TFLOP/s/device: 2.402, Tokens/s/device: 362.034, total_weights: 16384, loss: 3.101 I0416 19:36:43.192778 129740583242880 metric_logger.py:185] completed step: 12, seconds: 0.005, TFLOP/s/device: 2577.714, Tokens/s/device: 388541.074, total_weights: 16384, loss: 2.684 I0416 19:36:43.310998 129740583242880 metric_logger.py:185] completed step: 13, seconds: 0.117, TFLOP/s/device: 116.418, Tokens/s/device: 17547.768, total_weights: 16384, loss: 2.348 I0416 19:36:43.315308 129740583242880 checkpointing.py:794] Waiting for step 15 to finish before checkpoint... I0416 19:36:43.463903 129632745555520 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10/items/array_metadatas/process_0 I0416 19:36:43.546859 129740583242880 checkpointing.py:798] Waited 0.2315235137939453 seconds for step 15 to finish before starting checkpointing. I0416 19:36:43.547253 129740583242880 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 19:37:18.332271 129632766527040 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 35.584862s (commit=35.117964s, array_metadata_write=0.466898s) I0416 19:37:18.333346 129632787498560 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 309.477 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 40 seconds) (per-host) I0416 19:37:18.333404 129632787498560 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 35.585071s. I0416 19:37:18.793282 129632787498560 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 19:37:18.991287 129632735069760 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776368198206459696, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10/_CHECKPOINT_METADATA I0416 19:37:19.311307 129632787498560 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.667042s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10/items I0416 19:37:19.312063 129632787498560 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10/items I0416 19:37:19.574292 129632787498560 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10 I0416 19:37:20.300784 129632787498560 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10`. I0416 19:37:20.301425 129632787498560 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 42.883854s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10 I0416 19:37:20.301503 129632787498560 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 37.553171s. I0416 19:37:20.301715 129632756041280 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 19:37:20.301822 129632756041280 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 19:37:20.301895 129632756041280 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0416 19:37:20.301943 129632756041280 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0416 19:37:20.302160 129740583242880 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0416 19:37:20.303052 129740583242880 checkpoint_manager.py:1441] Waiting for previous save to complete took 36.755788 seconds. If this number is high, consider checkpointing less frequently. I0416 19:37:20.304325 129740583242880 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 15 I0416 19:37:20.304613 129740583242880 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15. I0416 19:37:20.468162 129632756041280 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15 I0416 19:37:20.485634 129740583242880 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 19:37:20.485723 129740583242880 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 19:37:21.174699 129632724584000 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15/items I0416 19:37:25.182901 129740583242880 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 4.697670s I0416 19:37:25.186827 129740583242880 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.570 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 4 seconds) (per-host) I0416 19:37:25.186885 129740583242880 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 4.801025s (batch_requests_ready=0.096185s, total_serialization_initiated=4.701026s, others=0.003813s) I0416 19:37:25.186959 129740583242880 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 4.801540s (all_items=0.000016s, per_item={'items': '0.00001574'}, temp_paths=4.801524) I0416 19:37:25.188194 129632756041280 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 19:37:25.188298 129740583242880 async_checkpointer.py:561] Finished blocking save. Time taken: 4.883927s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15. I0416 19:37:25.188548 129740583242880 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0416 19:37:25.188692 129632787498560 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 19:37:25.188783 129740583242880 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776368203.5472288, 'wait_for_prev_duration_secs': 36.75578784942627, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776368240.3043528, 'checkpointer_blocking_duration_secs': 4.884101867675781, 'get_old_steps_start_time': 1776368245.1884751, 'get_old_steps_duration_secs': 3.0279159545898438e-05, 'checkpoint_manager_blocking_start_time': 1776368203.547167, 'checkpoint_manager_blocking_duration_secs': 41.64159035682678} I0416 19:37:25.188902 129740583242880 checkpointing.py:409] Started an asynchronous checkpoint save for step 15 I0416 19:37:25.189352 129740583242880 metric_logger.py:185] completed step: 14, seconds: 0.119, TFLOP/s/device: 114.293, Tokens/s/device: 17227.456, total_weights: 16384, loss: 2.083 I0416 19:37:25.196495 129740583242880 metric_logger.py:185] completed step: 15, seconds: 0.118, TFLOP/s/device: 115.282, Tokens/s/device: 17376.548, total_weights: 16384, loss: 1.880 I0416 19:37:25.312465 129740583242880 metric_logger.py:185] completed step: 16, seconds: 41.880, TFLOP/s/device: 0.324, Tokens/s/device: 48.902, total_weights: 16384, loss: 1.726 I0416 19:37:25.430565 129740583242880 metric_logger.py:185] completed step: 17, seconds: 0.005, TFLOP/s/device: 2500.852, Tokens/s/device: 376955.641, total_weights: 16384, loss: 1.610 I0416 19:37:25.549371 129740583242880 metric_logger.py:185] completed step: 18, seconds: 0.117, TFLOP/s/device: 116.181, Tokens/s/device: 17512.057, total_weights: 16384, loss: 1.521 I0416 19:37:25.667666 129740583242880 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0416 19:37:25.668509 129740583242880 checkpointing.py:798] Waited 0.0008699893951416016 seconds for step 19 to finish before starting checkpointing. I0416 19:37:25.668753 129740583242880 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 19:37:25.945832 129634421179968 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15/items/array_metadatas/process_0 I0416 19:37:57.457672 129632724584000 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 32.270132s (commit=31.818405s, array_metadata_write=0.451727s) I0416 19:37:57.458713 129632756041280 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 340.864 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 37 seconds) (per-host) I0416 19:37:57.458824 129632756041280 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 32.270479s. I0416 19:37:57.869791 129632756041280 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 19:37:58.349802 129632756041280 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.624149s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15/items I0416 19:37:58.350545 129632756041280 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15/items I0416 19:37:58.584954 129632756041280 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15 I0416 19:37:59.270218 129632756041280 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15`. I0416 19:37:59.270845 129632756041280 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 38.966479s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15 I0416 19:37:59.270938 129632756041280 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 34.082597s. I0416 19:37:59.271067 129632787498560 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 19:37:59.271115 129632787498560 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 19:37:59.271165 129632787498560 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0416 19:37:59.271206 129632787498560 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0416 19:37:59.271450 129740583242880 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0416 19:37:59.271574 129740583242880 checkpoint_manager.py:1441] Waiting for previous save to complete took 33.602821 seconds. If this number is high, consider checkpointing less frequently. I0416 19:37:59.271794 129740583242880 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 19 I0416 19:37:59.272284 129740583242880 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19. I0416 19:37:59.431078 129632787498560 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19 I0416 19:37:59.440746 129740583242880 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 19:37:59.440836 129740583242880 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 19:38:00.122384 129634410694208 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19/items I0416 19:38:06.183905 129740583242880 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 6.743538s I0416 19:38:06.187768 129740583242880 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.806 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host) I0416 19:38:06.187833 129740583242880 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 6.832352s (batch_requests_ready=0.081890s, total_serialization_initiated=6.746712s, others=0.003750s) I0416 19:38:06.187903 129740583242880 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 6.832818s (all_items=0.000013s, per_item={'items': '0.00001335'}, temp_paths=6.832804) I0416 19:38:06.189132 129632724584000 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 19:38:06.189215 129740583242880 async_checkpointer.py:561] Finished blocking save. Time taken: 6.917380s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19. I0416 19:38:06.189390 129740583242880 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0416 19:38:06.189592 129634368751168 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 19:38:06.189701 129740583242880 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776368245.6687284, 'wait_for_prev_duration_secs': 33.6028208732605, 'time_between_consecutive_saves_sec': 5.366770505905151, 'checkpointer_blocking_start_time': 1776368279.2718167, 'checkpointer_blocking_duration_secs': 6.917504549026489, 'get_old_steps_start_time': 1776368286.1893334, 'get_old_steps_duration_secs': 2.288818359375e-05, 'checkpoint_manager_blocking_start_time': 1776368245.668695, 'checkpoint_manager_blocking_duration_secs': 40.520981311798096} I0416 19:38:06.189821 129740583242880 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0416 19:38:06.189852 129740583242880 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 19:38:06.931303 129634421179968 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19/items/array_metadatas/process_0 I0416 19:38:49.561891 129634410694208 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 43.373511s (commit=42.928482s, array_metadata_write=0.445029s) I0416 19:38:49.562834 129632724584000 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 251.692 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 50 seconds) (per-host) I0416 19:38:49.562886 129632724584000 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.373627s. I0416 19:38:50.008772 129632724584000 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 19:38:50.527335 129632724584000 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.658978s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19/items I0416 19:38:50.528030 129632724584000 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19/items I0416 19:38:50.808376 129632724584000 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19 I0416 19:38:51.468495 129632724584000 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19`. I0416 19:38:51.469137 129632724584000 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 52.197310s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19 I0416 19:38:51.469208 129632724584000 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 45.279950s. I0416 19:38:51.469356 129634368751168 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 19:38:51.469457 129634368751168 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 19:38:51.469500 129634368751168 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0416 19:38:51.469544 129634368751168 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0416 19:38:51.469672 129740583242880 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0416 19:38:51.469801 129740583242880 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 19:38:51.470476 129740583242880 metric_logger.py:185] completed step: 19, seconds: 0.118, TFLOP/s/device: 115.225, Tokens/s/device: 17368.001, total_weights: 16384, loss: 1.449 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops