2026-04-16 20:53:07.428873: 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 20:53:07.548074 128756056390784 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0416 20:54:07.170606 128756056390784 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 20:54:07.170717 128756056390784 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 20:54:07.170754 128756056390784 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 20:54:07.170779 128756056390784 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system I0416 20:54:07.170801 128756056390784 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0416 20:54:07.170890 128756056390784 train.py:811] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 20:54:07.269301 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:54:07.674630 128756056390784 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 20:54:08.116912 128756056390784 checkpointing.py:688] Setting up checkpoint logger... I0416 20:54:08.117009 128756056390784 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0416 20:54:08.117059 128756056390784 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 20:54:08.117233 128756056390784 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 0x7519c41f8e90>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 20:54:09.772746 128756056390784 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 20:54:09.772928 128756056390784 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 0x7514c5eba6f0>}, handler_registry=None I0416 20:54:09.773155 128756056390784 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7514c5eba6f0>` 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 20:54:09.773195 128756056390784 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7514bf409b50>` 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 20:54:09.773224 128756056390784 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 0x7514c5eba6f0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7514c5eba6f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7514bf409b50>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7514bf409b50>}). I0416 20:54:09.773510 128756056390784 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 20:54:09.773559 128756056390784 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7513487a3240> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 20:54:10.196849 128649950922304 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776372617938124812, 'commit_timestamp_nsecs': 1776372652380878975, '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_true/checkpoints/0/_CHECKPOINT_METADATA I0416 20:54:10.276027 128649950922304 checkpoint_manager.py:1847] Missing metrics for step 0 I0416 20:54:10.317956 128756056390784 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_true/checkpoints I0416 20:54:10.318426 128756056390784 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_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7514bf5519d0> I0416 20:54:10.318509 128756056390784 checkpointing.py:302] Checkpoint manager created! W0416 20:54:10.339069 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:54:10.551649 128756056390784 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 20:54:10.551757 128756056390784 checkpointing.py:582] restoring from this run's directory step 9 I0416 20:54:10.716462 128756056390784 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_true/checkpoints/9. I0416 20:54:11.259658 4136723 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 20:54:11.259720 4136723 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 20:54:17.331882 128756056390784 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/read/gbytes_per_sec: 1.970 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host) I0416 20:54:17.332350 128756056390784 checkpointer.py:318] Finished restoring checkpoint in 6.70 seconds from gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_true/checkpoints/9. I0416 20:54:17.332486 128756056390784 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_true/checkpoints', 'checkpointer_start_time': 1776372850.6332881, 'checkpointer_duration_secs': 6.699134349822998, 'checkpoint_manager_start_time': 1776372850.5522847, 'checkpoint_manager_duration_secs': 6.780142307281494} [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 20:54:17.403004 128756056390784 maxtext_utils.py:1805] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0416 20:54:17.403121 128756056390784 maxtext_utils.py:1805] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0416 20:54:17.403162 128756056390784 maxtext_utils.py:1805] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0416 20:54:17.403194 128756056390784 maxtext_utils.py:1805] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0416 20:54:17.403223 128756056390784 maxtext_utils.py:1805] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0416 20:54:17.403249 128756056390784 maxtext_utils.py:1805] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0416 20:54:17.403277 128756056390784 maxtext_utils.py:1805] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 20:54:17.403303 128756056390784 maxtext_utils.py:1805] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0416 20:54:17.403331 128756056390784 maxtext_utils.py:1805] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 20:54:17.403355 128756056390784 maxtext_utils.py:1805] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 20:54:17.403379 128756056390784 maxtext_utils.py:1805] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0416 20:54:17.403402 128756056390784 maxtext_utils.py:1805] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0416 20:54:17.513430 128756056390784 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 20:54:17.513510 128756056390784 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 20:54:17.525689 128756056390784 nnx_decoders.py:465] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 20:54:17.525741 128756056390784 nnx_decoders.py:465] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 20:54:17.540606 128756056390784 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 20:54:17.540654 128756056390784 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 20:54:17.554472 128756056390784 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 20:54:17.554518 128756056390784 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 20:54:17.582650 128756056390784 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 20:54:17.582705 128756056390784 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 20:54:17.596695 128756056390784 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 20:54:17.596746 128756056390784 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 20:54:17.610815 128756056390784 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 20:54:17.610865 128756056390784 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 20:54:17.672313 128756056390784 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 20:54:17.672384 128756056390784 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 20:54:17.691679 128756056390784 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 20:54:17.691732 128756056390784 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). W0416 20:54:18.173404 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:54:18.239936 128756056390784 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 20:54:19.376472 128756056390784 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 20:54:19.378309 128756056390784 metric_logger.py:289] number parameters: 1.104 billion W0416 20:54:20.330779 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:54:20.404896 128756056390784 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0416 20:54:20.521001 128756056390784 checkpointing.py:798] Waited 0.11609864234924316 seconds for step 10 to finish before starting checkpointing. I0416 20:54:20.521498 128756056390784 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 20:54:20.521684 128756056390784 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 10 I0416 20:54:20.522055 128756056390784 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_true/checkpoints/10. I0416 20:54:20.608810 128756056390784 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 20:54:20.682128 128648411612736 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_true/checkpoints/10 I0416 20:54:20.694367 128756056390784 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 20:54:20.694458 128756056390784 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 20:54:21.360738 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:54:21.378658 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:54:21.391085 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:54:21.396395 128648401126976 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_true/checkpoints/10/items W0416 20:54:21.396709 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:54:21.401006 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:54:21.405057 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:54:21.409031 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:54:21.412967 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:54:21.531646 128648357086784 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776372861295237744, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776372861295237744, "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_true/checkpoints/10/_CHECKPOINT_METADATA I0416 20:54:27.554180 128756056390784 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 6.860835s I0416 20:54:27.560386 128756056390784 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.775 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host) I0416 20:54:27.560453 128756056390784 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 6.950694s (batch_requests_ready=0.076856s, total_serialization_initiated=6.867734s, others=0.006104s) I0416 20:54:27.560535 128756056390784 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 6.951248s (all_items=0.000032s, per_item={'items': '0.00003171'}, temp_paths=6.951216) I0416 20:54:27.561819 128648390641216 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 20:54:27.561932 128756056390784 async_checkpointer.py:561] Finished blocking save. Time taken: 7.040205s. 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_true/checkpoints/10. I0416 20:54:27.562119 128756056390784 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0416 20:54:27.562294 128648422098496 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 20:54:27.562390 128756056390784 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_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776372860.5214825, 'wait_for_prev_duration_secs': 4.220008850097656e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776372860.5217054, 'checkpointer_blocking_duration_secs': 7.040290117263794, 'get_old_steps_start_time': 1776372867.5620189, 'get_old_steps_duration_secs': 5.125999450683594e-05, 'checkpoint_manager_blocking_start_time': 1776372860.5213466, 'checkpoint_manager_blocking_duration_secs': 7.041015386581421} I0416 20:54:27.562512 128756056390784 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0416 20:54:27.562572 128756056390784 max_utils.py:750] Memstats: After params initialized: I0416 20:54:27.562623 128756056390784 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0416 20:54:27.562648 128756056390784 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0416 20:54:27.562668 128756056390784 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0416 20:54:27.562693 128756056390784 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0416 20:54:27.562711 128756056390784 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0416 20:54:27.562726 128756056390784 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0416 20:54:27.562743 128756056390784 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0416 20:54:27.562759 128756056390784 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0416 20:54:27.568616 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:54:27.575161 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:54:27.579931 4134752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:54:27.776203 128756056390784 metric_logger.py:185] completed step: 10, seconds: 1.025, TFLOP/s/device: 13.253, Tokens/s/device: 1997.564, total_weights: 16384, loss: 3.587 I0416 20:54:27.889759 128756056390784 metric_logger.py:185] completed step: 11, seconds: 7.369, TFLOP/s/device: 1.844, Tokens/s/device: 277.916, total_weights: 16384, loss: 3.086 I0416 20:54:28.008188 128756056390784 metric_logger.py:185] completed step: 12, seconds: 0.007, TFLOP/s/device: 2005.776, Tokens/s/device: 302332.448, total_weights: 16384, loss: 2.671 I0416 20:54:28.126667 128756056390784 metric_logger.py:185] completed step: 13, seconds: 0.115, TFLOP/s/device: 118.338, Tokens/s/device: 17837.235, total_weights: 16384, loss: 2.335 I0416 20:54:28.133093 128756056390784 checkpointing.py:794] Waiting for step 15 to finish before checkpoint... I0416 20:54:28.287016 128648380155456 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_true/checkpoints/10/items/array_metadatas/process_0 I0416 20:54:28.363959 128756056390784 checkpointing.py:798] Waited 0.23084068298339844 seconds for step 15 to finish before starting checkpointing. I0416 20:54:28.364355 128756056390784 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 20:55:09.840370 128648401126976 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.279366s (commit=41.809224s, array_metadata_write=0.470142s) I0416 20:55:09.841622 128648390641216 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 256.679 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 49 seconds) (per-host) I0416 20:55:09.841732 128648390641216 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.279835s. I0416 20:55:10.289791 128648390641216 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 20:55:10.475615 128648357086784 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776372861295237744, '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_true/checkpoints/10/_CHECKPOINT_METADATA I0416 20:55:10.740136 128648390641216 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 20:55:10.740761 128648390641216 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.602937s. use_zarr3=True, 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_true/checkpoints/10/items I0416 20:55:10.741424 128648390641216 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_true/checkpoints/10/items I0416 20:55:10.981913 128648390641216 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_true/checkpoints/10 I0416 20:55:11.659091 128648390641216 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_true/checkpoints/10`. I0416 20:55:11.659724 128648390641216 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 51.138003s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_true/checkpoints/10 I0416 20:55:11.659805 128648390641216 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 44.097909s. I0416 20:55:11.659986 128648422098496 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 20:55:11.660105 128648422098496 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 20:55:11.660174 128648422098496 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0416 20:55:11.660226 128648422098496 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0416 20:55:11.660746 128756056390784 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 20:55:11.661560 128756056390784 checkpoint_manager.py:1441] Waiting for previous save to complete took 43.297206 seconds. If this number is high, consider checkpointing less frequently. I0416 20:55:11.662365 128756056390784 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 15 I0416 20:55:11.662647 128756056390784 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_true/checkpoints/15. I0416 20:55:11.847776 128648422098496 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_true/checkpoints/15 I0416 20:55:11.847947 128756056390784 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 20:55:11.848382 128756056390784 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 20:55:12.546830 128648346601024 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_true/checkpoints/15/items I0416 20:55:19.221673 128756056390784 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 7.374977s I0416 20:55:19.227954 128756056390784 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.651 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 7 seconds) (per-host) I0416 20:55:19.228015 128756056390784 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 7.476894s (batch_requests_ready=0.087702s, total_serialization_initiated=7.383020s, others=0.006172s) I0416 20:55:19.228107 128756056390784 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 7.477651s (all_items=0.000013s, per_item={'items': '0.00001311'}, temp_paths=7.477638) I0416 20:55:19.229511 128649898493504 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 20:55:19.229609 128756056390784 async_checkpointer.py:561] Finished blocking save. Time taken: 7.567201s. 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_true/checkpoints/15. I0416 20:55:19.229871 128756056390784 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0416 20:55:19.230112 128649856550464 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 20:55:19.230242 128756056390784 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_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776372868.3643239, 'wait_for_prev_duration_secs': 43.29720616340637, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776372911.6623888, 'checkpointer_blocking_duration_secs': 7.567392349243164, 'get_old_steps_start_time': 1776372919.229801, 'get_old_steps_duration_secs': 2.7894973754882812e-05, 'checkpoint_manager_blocking_start_time': 1776372868.3642614, 'checkpoint_manager_blocking_duration_secs': 50.865952014923096} I0416 20:55:19.230350 128756056390784 checkpointing.py:409] Started an asynchronous checkpoint save for step 15 I0416 20:55:19.230822 128756056390784 metric_logger.py:185] completed step: 14, seconds: 0.118, TFLOP/s/device: 114.989, Tokens/s/device: 17332.431, total_weights: 16384, loss: 2.072 I0416 20:55:19.240994 128756056390784 metric_logger.py:185] completed step: 15, seconds: 0.119, TFLOP/s/device: 114.130, Tokens/s/device: 17202.856, total_weights: 16384, loss: 1.869 I0416 20:55:19.354582 128756056390784 metric_logger.py:185] completed step: 16, seconds: 51.105, TFLOP/s/device: 0.266, Tokens/s/device: 40.074, total_weights: 16384, loss: 1.716 I0416 20:55:19.472763 128756056390784 metric_logger.py:185] completed step: 17, seconds: 0.008, TFLOP/s/device: 1795.339, Tokens/s/device: 270613.108, total_weights: 16384, loss: 1.601 I0416 20:55:19.591246 128756056390784 metric_logger.py:185] completed step: 18, seconds: 0.116, TFLOP/s/device: 117.526, Tokens/s/device: 17714.731, total_weights: 16384, loss: 1.513 I0416 20:55:19.710231 128756056390784 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0416 20:55:19.711500 128756056390784 checkpointing.py:798] Waited 0.001283884048461914 seconds for step 19 to finish before starting checkpointing. I0416 20:55:19.712418 128756056390784 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 20:55:19.983013 128649908979264 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_true/checkpoints/15/items/array_metadatas/process_0 I0416 20:55:56.673227 128648401126976 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 37.444489s (commit=36.998938s, array_metadata_write=0.445550s) I0416 20:55:56.674413 128649898493504 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 281.297 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 44 seconds) (per-host) I0416 20:55:56.674516 128649898493504 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.444860s. I0416 20:55:57.132143 128649898493504 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 20:55:57.523021 128649898493504 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 20:55:57.523694 128649898493504 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.529998s. use_zarr3=True, 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_true/checkpoints/15/items I0416 20:55:57.524332 128649898493504 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_true/checkpoints/15/items I0416 20:55:57.773253 128649898493504 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_true/checkpoints/15 I0416 20:55:58.453000 128649898493504 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_true/checkpoints/15`. I0416 20:55:58.453615 128649898493504 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 46.791214s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_true/checkpoints/15 I0416 20:55:58.453689 128649898493504 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 39.224037s. I0416 20:55:58.453861 128649856550464 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 20:55:58.453960 128649856550464 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 20:55:58.454024 128649856550464 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0416 20:55:58.454095 128649856550464 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0416 20:55:58.454247 128756056390784 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 20:55:58.454365 128756056390784 checkpoint_manager.py:1441] Waiting for previous save to complete took 38.741950 seconds. If this number is high, consider checkpointing less frequently. I0416 20:55:58.454581 128756056390784 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 19 I0416 20:55:58.454873 128756056390784 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_true/checkpoints/19. I0416 20:55:58.629861 128756056390784 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 20:55:58.629950 128756056390784 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 20:55:58.635924 128649856550464 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_true/checkpoints/19 I0416 20:55:59.361648 128649867036224 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_true/checkpoints/19/items I0416 20:56:10.857915 128756056390784 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 12.229045s I0416 20:56:10.864376 128756056390784 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.001 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 12 seconds) (per-host) I0416 20:56:10.864434 128756056390784 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 12.327851s (batch_requests_ready=0.084551s, total_serialization_initiated=12.236966s, others=0.006334s) I0416 20:56:10.864523 128756056390784 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 12.328604s (all_items=0.000014s, per_item={'items': '0.00001383'}, temp_paths=12.328591) I0416 20:56:10.865916 128649846064704 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 20:56:10.866017 128756056390784 async_checkpointer.py:561] Finished blocking save. Time taken: 12.411396s. 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_true/checkpoints/19. I0416 20:56:10.866208 128756056390784 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0416 20:56:10.866425 128649825093184 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 20:56:10.866672 128756056390784 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_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776372919.7123947, 'wait_for_prev_duration_secs': 38.741950035095215, 'time_between_consecutive_saves_sec': 8.05215334892273, 'checkpointer_blocking_start_time': 1776372958.4546034, 'checkpointer_blocking_duration_secs': 12.411533117294312, 'get_old_steps_start_time': 1776372970.8661513, 'get_old_steps_duration_secs': 2.2172927856445312e-05, 'checkpoint_manager_blocking_start_time': 1776372919.7123654, 'checkpoint_manager_blocking_duration_secs': 51.154282331466675} I0416 20:56:10.866790 128756056390784 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0416 20:56:10.866821 128756056390784 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 20:56:11.614290 128649908979264 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_true/checkpoints/19/items/array_metadatas/process_0 I0416 20:56:52.994450 128649856550464 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.129331s (commit=41.655680s, array_metadata_write=0.473651s) I0416 20:56:52.995644 128649846064704 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 232.042 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 54 seconds) (per-host) I0416 20:56:52.995743 128649846064704 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.129672s. I0416 20:56:53.426025 128649846064704 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 20:56:53.874556 128649846064704 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 20:56:53.875231 128649846064704 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.588584s. use_zarr3=True, 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_true/checkpoints/19/items I0416 20:56:53.875857 128649846064704 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_true/checkpoints/19/items I0416 20:56:54.125930 128649846064704 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_true/checkpoints/19 I0416 20:56:54.796548 128649846064704 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_true/checkpoints/19`. I0416 20:56:54.797176 128649846064704 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 56.342562s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_16_zarr3_true/checkpoints/19 I0416 20:56:54.797243 128649846064704 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 43.931172s. I0416 20:56:54.797410 128649825093184 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 20:56:54.797509 128649825093184 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 20:56:54.797563 128649825093184 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0416 20:56:54.797617 128649825093184 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0416 20:56:54.798086 128756056390784 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 20:56:54.798259 128756056390784 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 20:56:54.799389 128756056390784 metric_logger.py:185] completed step: 19, seconds: 0.117, TFLOP/s/device: 115.804, Tokens/s/device: 17455.190, 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