2026-04-16 21:01:16.866504: 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 21:01:16.983738 126092468206720 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0416 21:02:13.384231 126092468206720 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 21:02:13.384344 126092468206720 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 21:02:13.384378 126092468206720 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 21:02:13.384407 126092468206720 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system I0416 21:02:13.384431 126092468206720 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0416 21:02:13.384538 126092468206720 train.py:811] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 21:02:13.480200 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 21:02:13.884477 126092468206720 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 21:02:14.360792 126092468206720 checkpointing.py:688] Setting up checkpoint logger... I0416 21:02:14.360914 126092468206720 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=False I0416 21:02:14.360955 126092468206720 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 21:02:14.361147 126092468206720 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 0x72ad9a7fcd70>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 21:02:16.025317 126092468206720 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 21:02:16.025490 126092468206720 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 0x72a89b9262d0>}, handler_registry=None I0416 21:02:16.025708 126092468206720 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x72a89b9262d0>` 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 21:02:16.025746 126092468206720 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x72a89b9d1400>` 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 21:02:16.025775 126092468206720 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 0x72a89b9262d0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x72a89b9262d0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x72a89b9d1400>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x72a89b9d1400>}). I0416 21:02:16.026103 126092468206720 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 21:02:16.026153 126092468206720 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x72a71eda3240> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 21:02:16.446533 125986347681344 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776373205457704146, 'commit_timestamp_nsecs': 1776373247755319983, 'custom_metadata': {}} from gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/9/_CHECKPOINT_METADATA I0416 21:02:16.525984 125986347681344 checkpoint_manager.py:1847] Missing metrics for step 9 I0416 21:02:16.619169 126092468206720 checkpoint_manager.py:1788] Found 3 checkpoint steps in gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints I0416 21:02:16.619660 126092468206720 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/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x72a89b9d3b90> I0416 21:02:16.619744 126092468206720 checkpointing.py:302] Checkpoint manager created! W0416 21:02:16.636771 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 21:02:16.850768 126092468206720 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 21:02:16.850871 126092468206720 checkpointing.py:582] restoring from this run's directory step 9 I0416 21:02:17.018324 126092468206720 checkpointer.py:304] Restoring checkpoint from gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/9. I0416 21:02:17.533148 4144453 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 21:02:17.533214 4144453 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 21:02:23.721316 126092468206720 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/read/gbytes_per_sec: 1.936 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host) I0416 21:02:23.722152 126092468206720 checkpointer.py:318] Finished restoring checkpoint in 6.79 seconds from gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/9. I0416 21:02:23.722253 126092468206720 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints', 'checkpointer_start_time': 1776373336.9289453, 'checkpointer_duration_secs': 6.793268442153931, 'checkpoint_manager_start_time': 1776373336.8513844, 'checkpoint_manager_duration_secs': 6.870831489562988} [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 21:02:23.795938 126092468206720 maxtext_utils.py:1805] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0416 21:02:23.796088 126092468206720 maxtext_utils.py:1805] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0416 21:02:23.796132 126092468206720 maxtext_utils.py:1805] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0416 21:02:23.796165 126092468206720 maxtext_utils.py:1805] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0416 21:02:23.796195 126092468206720 maxtext_utils.py:1805] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0416 21:02:23.796221 126092468206720 maxtext_utils.py:1805] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0416 21:02:23.796248 126092468206720 maxtext_utils.py:1805] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 21:02:23.796275 126092468206720 maxtext_utils.py:1805] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0416 21:02:23.796301 126092468206720 maxtext_utils.py:1805] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 21:02:23.796327 126092468206720 maxtext_utils.py:1805] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 21:02:23.796351 126092468206720 maxtext_utils.py:1805] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0416 21:02:23.796375 126092468206720 maxtext_utils.py:1805] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0416 21:02:23.915269 126092468206720 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 21:02:23.915382 126092468206720 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 21:02:23.928647 126092468206720 nnx_decoders.py:465] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 21:02:23.928704 126092468206720 nnx_decoders.py:465] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 21:02:23.944905 126092468206720 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 21:02:23.944962 126092468206720 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 21:02:23.959499 126092468206720 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 21:02:23.959553 126092468206720 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 21:02:23.992918 126092468206720 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 21:02:23.993008 126092468206720 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 21:02:24.007696 126092468206720 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 21:02:24.007756 126092468206720 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 21:02:24.022383 126092468206720 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 21:02:24.022441 126092468206720 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 21:02:24.083443 126092468206720 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 21:02:24.083548 126092468206720 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 21:02:24.103722 126092468206720 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 21:02:24.103781 126092468206720 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). W0416 21:02:24.671466 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 21:02:24.741455 126092468206720 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 21:02:24.742194 126092468206720 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 21:02:24.744800 126092468206720 metric_logger.py:289] number parameters: 1.104 billion W0416 21:02:26.026195 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 21:02:26.100499 126092468206720 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0416 21:02:26.216819 126092468206720 checkpointing.py:798] Waited 0.11630392074584961 seconds for step 10 to finish before starting checkpointing. I0416 21:02:26.217253 126092468206720 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 21:02:26.217399 126092468206720 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 10 I0416 21:02:26.217701 126092468206720 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10. I0416 21:02:26.310616 126092468206720 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 21:02:26.395214 126092468206720 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 21:02:26.395299 126092468206720 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 21:02:26.401602 125984848217664 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10 W0416 21:02:27.053401 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 21:02:27.068718 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 21:02:27.082936 125984837731904 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10/items W0416 21:02:27.083347 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 21:02:27.088569 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 21:02:27.092817 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 21:02:27.096898 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 21:02:27.100898 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 21:02:27.105029 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 21:02:27.234966 125984797886016 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776373346989589242, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776373346989589242, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10/_CHECKPOINT_METADATA I0416 21:02:31.761474 126092468206720 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 5.367250s I0416 21:02:31.766866 126092468206720 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.262 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 5 seconds) (per-host) I0416 21:02:31.767267 126092468206720 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 5.455706s (batch_requests_ready=0.075655s, total_serialization_initiated=5.374427s, others=0.005624s) I0416 21:02:31.767354 126092468206720 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 5.456259s (all_items=0.000020s, per_item={'items': '0.00001979'}, temp_paths=5.456239) I0416 21:02:31.768473 125984858703424 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 21:02:31.768557 126092468206720 async_checkpointer.py:561] Finished blocking save. Time taken: 5.551119s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10. I0416 21:02:31.768761 126092468206720 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0416 21:02:31.768931 125984787400256 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 21:02:31.769023 126092468206720 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_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': 1776373346.2172408, 'wait_for_prev_duration_secs': 4.076957702636719e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776373346.2174191, 'checkpointer_blocking_duration_secs': 5.5512611865997314, 'get_old_steps_start_time': 1776373351.7687027, 'get_old_steps_duration_secs': 2.2172927856445312e-05, 'checkpoint_manager_blocking_start_time': 1776373346.2171586, 'checkpoint_manager_blocking_duration_secs': 5.551839828491211} I0416 21:02:31.769159 126092468206720 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0416 21:02:31.769217 126092468206720 max_utils.py:750] Memstats: After params initialized: I0416 21:02:31.769261 126092468206720 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0416 21:02:31.769285 126092468206720 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0416 21:02:31.769306 126092468206720 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0416 21:02:31.769324 126092468206720 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0416 21:02:31.769342 126092468206720 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0416 21:02:31.769359 126092468206720 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0416 21:02:31.769377 126092468206720 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0416 21:02:31.769393 126092468206720 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0416 21:02:31.774981 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 21:02:31.781117 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 21:02:31.785929 4142496 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 21:02:31.978270 126092468206720 metric_logger.py:185] completed step: 10, seconds: 1.354, TFLOP/s/device: 10.033, Tokens/s/device: 1512.219, total_weights: 16384, loss: 3.587 I0416 21:02:32.091902 126092468206720 metric_logger.py:185] completed step: 11, seconds: 5.875, TFLOP/s/device: 2.313, Tokens/s/device: 348.575, total_weights: 16384, loss: 3.086 I0416 21:02:32.209970 126092468206720 metric_logger.py:185] completed step: 12, seconds: 0.007, TFLOP/s/device: 1877.713, Tokens/s/device: 283029.298, total_weights: 16384, loss: 2.671 I0416 21:02:32.327995 126092468206720 metric_logger.py:185] completed step: 13, seconds: 0.115, TFLOP/s/device: 118.314, Tokens/s/device: 17833.507, total_weights: 16384, loss: 2.335 I0416 21:02:32.334408 126092468206720 checkpointing.py:794] Waiting for step 15 to finish before checkpoint... I0416 21:02:32.538067 125984816760384 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10/items/array_metadatas/process_0 I0416 21:02:32.565037 126092468206720 checkpointing.py:798] Waited 0.23060393333435059 seconds for step 15 to finish before starting checkpointing. I0416 21:02:32.565449 126092468206720 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 21:03:08.523310 125984837731904 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 36.756102s (commit=36.305133s, array_metadata_write=0.450969s) I0416 21:03:08.524374 125984858703424 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 299.359 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 42 seconds) (per-host) I0416 21:03:08.524429 125984858703424 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.755828s. I0416 21:03:08.922783 125984858703424 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 21:03:09.167377 125984797886016 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776373346989589242, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10/_CHECKPOINT_METADATA I0416 21:03:09.430328 125984858703424 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.641083s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10/items I0416 21:03:09.431038 125984858703424 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10/items I0416 21:03:09.678996 125984858703424 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10 I0416 21:03:10.399825 125984858703424 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10`. I0416 21:03:10.400480 125984858703424 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 44.183050s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/10 I0416 21:03:10.400561 125984858703424 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 38.631960s. I0416 21:03:10.400724 125984787400256 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 21:03:10.400828 125984787400256 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 21:03:10.400899 125984787400256 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0416 21:03:10.400957 125984787400256 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0416 21:03:10.401263 126092468206720 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 21:03:10.401760 126092468206720 checkpoint_manager.py:1441] Waiting for previous save to complete took 37.836308 seconds. If this number is high, consider checkpointing less frequently. I0416 21:03:10.402744 126092468206720 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 15 I0416 21:03:10.403058 126092468206720 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15. I0416 21:03:10.584843 125984787400256 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15 I0416 21:03:10.585023 126092468206720 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 21:03:10.585520 126092468206720 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 21:03:11.301760 125984848217664 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15/items I0416 21:03:19.954237 126092468206720 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 9.370514s I0416 21:03:19.959792 126092468206720 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.303 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 9 seconds) (per-host) I0416 21:03:19.959855 126092468206720 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 9.472589s (batch_requests_ready=0.088530s, total_serialization_initiated=9.378616s, others=0.005443s) I0416 21:03:19.959939 126092468206720 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 9.473363s (all_items=0.000014s, per_item={'items': '0.00001431'}, temp_paths=9.473349) I0416 21:03:19.961144 125986316224064 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 21:03:19.961228 126092468206720 async_checkpointer.py:561] Finished blocking save. Time taken: 9.558438s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15. I0416 21:03:19.961438 126092468206720 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0416 21:03:19.961676 125984858703424 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 21:03:19.961804 126092468206720 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_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': 1776373352.5654254, 'wait_for_prev_duration_secs': 37.836308002471924, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776373390.4027708, 'checkpointer_blocking_duration_secs': 9.558570384979248, 'get_old_steps_start_time': 1776373399.961366, 'get_old_steps_duration_secs': 2.6941299438476562e-05, 'checkpoint_manager_blocking_start_time': 1776373352.5653625, 'checkpoint_manager_blocking_duration_secs': 47.39640212059021} I0416 21:03:19.961939 126092468206720 checkpointing.py:409] Started an asynchronous checkpoint save for step 15 I0416 21:03:19.962395 126092468206720 metric_logger.py:185] completed step: 14, seconds: 0.117, TFLOP/s/device: 115.685, Tokens/s/device: 17437.356, total_weights: 16384, loss: 2.072 I0416 21:03:19.971663 126092468206720 metric_logger.py:185] completed step: 15, seconds: 0.119, TFLOP/s/device: 114.440, Tokens/s/device: 17249.657, total_weights: 16384, loss: 1.869 I0416 21:03:20.085402 126092468206720 metric_logger.py:185] completed step: 16, seconds: 47.635, TFLOP/s/device: 0.285, Tokens/s/device: 42.994, total_weights: 16384, loss: 1.716 I0416 21:03:20.203869 126092468206720 metric_logger.py:185] completed step: 17, seconds: 0.008, TFLOP/s/device: 1801.768, Tokens/s/device: 271582.018, total_weights: 16384, loss: 1.601 I0416 21:03:20.322252 126092468206720 metric_logger.py:185] completed step: 18, seconds: 0.116, TFLOP/s/device: 117.386, Tokens/s/device: 17693.763, total_weights: 16384, loss: 1.513 I0416 21:03:20.441069 126092468206720 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0416 21:03:20.442250 126092468206720 checkpointing.py:798] Waited 0.0012135505676269531 seconds for step 19 to finish before starting checkpointing. I0416 21:03:20.443197 126092468206720 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 21:03:20.759690 125986326709824 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15/items/array_metadatas/process_0 I0416 21:03:56.820670 125984837731904 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 36.860140s (commit=36.403296s, array_metadata_write=0.456844s) I0416 21:03:56.821897 125986316224064 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 272.729 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 46 seconds) (per-host) I0416 21:03:56.822015 125986316224064 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.860742s. I0416 21:03:57.244383 125986316224064 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 21:03:57.774809 125986316224064 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.666127s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15/items I0416 21:03:57.775568 125986316224064 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15/items I0416 21:03:58.012161 125986316224064 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15 I0416 21:03:58.725524 125986316224064 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15`. I0416 21:03:58.726166 125986316224064 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 48.323382s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/15 I0416 21:03:58.726243 125986316224064 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 38.764971s. I0416 21:03:58.726423 125984858703424 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 21:03:58.726532 125984858703424 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 21:03:58.726609 125984858703424 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0416 21:03:58.726679 125984858703424 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0416 21:03:58.727026 126092468206720 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 21:03:58.727147 126092468206720 checkpoint_manager.py:1441] Waiting for previous save to complete took 38.283956 seconds. If this number is high, consider checkpointing less frequently. I0416 21:03:58.727381 126092468206720 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 19 I0416 21:03:58.727714 126092468206720 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19. I0416 21:03:58.888705 125984858703424 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19 I0416 21:03:58.902227 126092468206720 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 21:03:58.902322 126092468206720 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 21:03:59.602681 125984837731904 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19/items I0416 21:04:01.980584 126092468206720 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 3.079509s I0416 21:04:01.981122 126092468206720 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.898 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 3 seconds) (per-host) I0416 21:04:01.981173 126092468206720 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 3.165758s (batch_requests_ready=0.077567s, total_serialization_initiated=3.087763s, others=0.000428s) I0416 21:04:01.981247 126092468206720 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 3.166286s (all_items=0.000015s, per_item={'items': '0.00001454'}, temp_paths=3.166271) I0416 21:04:01.987938 125986274281024 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 21:04:01.988054 126092468206720 async_checkpointer.py:561] Finished blocking save. Time taken: 3.260613s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19. I0416 21:04:01.988294 126092468206720 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0416 21:04:01.988440 125986316224064 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 21:04:01.988522 126092468206720 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_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': 1776373400.4431744, 'wait_for_prev_duration_secs': 38.28395628929138, 'time_between_consecutive_saves_sec': 10.042203187942505, 'checkpointer_blocking_start_time': 1776373438.7274055, 'checkpointer_blocking_duration_secs': 3.260789632797241, 'get_old_steps_start_time': 1776373441.98823, 'get_old_steps_duration_secs': 2.5510787963867188e-05, 'checkpoint_manager_blocking_start_time': 1776373400.4431434, 'checkpoint_manager_blocking_duration_secs': 41.54535269737244} I0416 21:04:01.988626 126092468206720 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0416 21:04:01.988658 126092468206720 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 21:04:02.784454 125986326709824 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19/items/array_metadatas/process_0 I0416 21:04:43.424289 125984837731904 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.437389s (commit=40.984199s, array_metadata_write=0.453190s) I0416 21:04:43.425606 125986274281024 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 283.272 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 44 seconds) (per-host) I0416 21:04:43.425724 125986274281024 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.437610s. I0416 21:04:43.870877 125986274281024 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 21:04:44.374072 125986274281024 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.647506s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19/items I0416 21:04:44.374799 125986274281024 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19/items I0416 21:04:44.627913 125986274281024 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19 I0416 21:04:45.323506 125986274281024 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19`. I0416 21:04:45.324177 125986274281024 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 46.596760s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_false/checkpoints/19 I0416 21:04:45.324247 125986274281024 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 43.336136s. I0416 21:04:45.325070 125986316224064 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 21:04:45.325120 125986316224064 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 21:04:45.325172 125986316224064 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0416 21:04:45.325218 125986316224064 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0416 21:04:45.325361 126092468206720 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 21:04:45.325587 126092468206720 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 21:04:45.326703 126092468206720 metric_logger.py:185] completed step: 19, seconds: 0.118, TFLOP/s/device: 115.529, Tokens/s/device: 17413.781, total_weights: 16384, loss: 1.441 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops