2026-04-16 20:44:07.657649: 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 20:44:07.773061 138485273525376 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0416 20:45:03.780971 138485273525376 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 20:45:03.781102 138485273525376 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 20:45:03.781137 138485273525376 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 20:45:03.781165 138485273525376 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system I0416 20:45:03.781188 138485273525376 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0416 20:45:03.781263 138485273525376 train.py:811] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 20:45:03.873369 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:45:04.269746 138485273525376 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 20:45:04.368597 138485273525376 checkpointing.py:688] Setting up checkpoint logger... I0416 20:45:04.368693 138485273525376 checkpointing.py:234] Creating checkpoint manager with ocdbt=False and zarr3=True I0416 20:45:04.368733 138485273525376 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 20:45:04.368909 138485273525376 base_pytree_checkpoint_handler.py:411] Created BasePyTreeCheckpointHandler: use_ocdbt=False, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7df3087e82f0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 20:45:06.774927 138485273525376 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 20:45:06.775131 138485273525376 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 0x7dec8b923b30>}, handler_registry=None I0416 20:45:06.775380 138485273525376 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dec8b923b30>` 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 20:45:06.775419 138485273525376 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dec89b9ef30>` 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 20:45:06.775448 138485273525376 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 0x7dec8b923b30>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dec8b923b30>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dec89b9ef30>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dec89b9ef30>}). I0416 20:45:06.775837 138485273525376 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 20:45:06.775897 138485273525376 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7dec8ab79760> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 20:45:06.908684 138485273525376 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints I0416 20:45:06.908918 138485273525376 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_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7dec89b9f3e0> I0416 20:45:06.908997 138485273525376 checkpointing.py:302] Checkpoint manager created! W0416 20:45:06.924996 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:45:07.134086 138485273525376 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 20:45:07.134186 138485273525376 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. W0416 20:45:07.244338 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. [DECOUPLED NO-OP] gcs_storage: using stubs. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] workload_monitor: using stub. [DECOUPLED NO-OP] vertex_tensorboard: using stub. fsdp: 8 I0416 20:45:07.336679 138485273525376 maxtext_utils.py:1805] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0416 20:45:07.336867 138485273525376 maxtext_utils.py:1805] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0416 20:45:07.336913 138485273525376 maxtext_utils.py:1805] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0416 20:45:07.336946 138485273525376 maxtext_utils.py:1805] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0416 20:45:07.336977 138485273525376 maxtext_utils.py:1805] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0416 20:45:07.337002 138485273525376 maxtext_utils.py:1805] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0416 20:45:07.337029 138485273525376 maxtext_utils.py:1805] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 20:45:07.337068 138485273525376 maxtext_utils.py:1805] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0416 20:45:07.337095 138485273525376 maxtext_utils.py:1805] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 20:45:07.337119 138485273525376 maxtext_utils.py:1805] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 20:45:07.337141 138485273525376 maxtext_utils.py:1805] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0416 20:45:07.337164 138485273525376 maxtext_utils.py:1805] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0416 20:45:07.446063 138485273525376 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 20:45:07.446141 138485273525376 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 20:45:07.451762 138485273525376 nnx_decoders.py:465] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 20:45:07.451808 138485273525376 nnx_decoders.py:465] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 20:45:07.466113 138485273525376 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 20:45:07.466160 138485273525376 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 20:45:07.479791 138485273525376 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 20:45:07.479839 138485273525376 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 20:45:07.501010 138485273525376 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 20:45:07.501076 138485273525376 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 20:45:07.514806 138485273525376 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 20:45:07.514855 138485273525376 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 20:45:07.528567 138485273525376 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 20:45:07.528614 138485273525376 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 20:45:07.555693 138485273525376 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 20:45:07.555750 138485273525376 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 20:45:07.574122 138485273525376 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 20:45:07.574179 138485273525376 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). W0416 20:45:08.300097 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:45:08.419816 138485273525376 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 20:45:08.420324 138485273525376 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 20:45:08.421660 138485273525376 metric_logger.py:289] number parameters: 1.104 billion W0416 20:45:09.038116 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:45:09.134028 138485273525376 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0416 20:45:09.250693 138485273525376 checkpointing.py:798] Waited 0.11665534973144531 seconds for step 0 to finish before starting checkpointing. I0416 20:45:09.251135 138485273525376 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 20:45:09.251286 138485273525376 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0416 20:45:09.251662 138485273525376 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_15_ocdbt_false/checkpoints/0. I0416 20:45:09.357973 138485273525376 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 20:45:09.452522 138377795995200 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_15_ocdbt_false/checkpoints/0 I0416 20:45:09.456685 138485273525376 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 20:45:09.456774 138485273525376 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 20:45:10.101105 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:45:10.112831 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:45:10.120980 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:45:10.136948 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:45:10.137343 138377785509440 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_15_ocdbt_false/checkpoints/0/items W0416 20:45:10.146217 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:45:10.152282 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:45:10.157297 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:45:10.161300 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:45:10.280033 138377745663552 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776372310022678643, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776372310022678643, "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_15_ocdbt_false/checkpoints/0/_CHECKPOINT_METADATA I0416 20:45:10.588611 138485273525376 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.133013s I0416 20:45:10.589031 138485273525376 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 10.037 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0416 20:45:10.589096 138485273525376 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.229655s (batch_requests_ready=0.087960s, total_serialization_initiated=1.141356s, others=0.000338s) I0416 20:45:10.589158 138485273525376 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.230201s (all_items=0.000019s, per_item={'items': '0.00001884'}, temp_paths=1.230182) I0416 20:45:10.589920 138377724692032 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 20:45:10.590022 138485273525376 async_checkpointer.py:561] Finished blocking save. Time taken: 1.338696s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/0. I0416 20:45:10.590212 138485273525376 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0416 20:45:10.590395 138377806480960 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 20:45:10.590495 138485273525376 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776372309.25112, 'wait_for_prev_duration_secs': 4.458427429199219e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776372309.2513072, 'checkpointer_blocking_duration_secs': 1.3388359546661377, 'get_old_steps_start_time': 1776372310.5901606, 'get_old_steps_duration_secs': 1.9073486328125e-05, 'checkpoint_manager_blocking_start_time': 1776372309.2510233, 'checkpoint_manager_blocking_duration_secs': 1.3394453525543213} I0416 20:45:10.590610 138485273525376 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0416 20:45:10.590661 138485273525376 max_utils.py:750] Memstats: After params initialized: I0416 20:45:10.590759 138485273525376 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0416 20:45:10.590785 138485273525376 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0416 20:45:10.590805 138485273525376 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0416 20:45:10.590823 138485273525376 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0416 20:45:10.590840 138485273525376 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0416 20:45:10.590857 138485273525376 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0416 20:45:10.590875 138485273525376 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0416 20:45:10.590892 138485273525376 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0416 20:45:10.596523 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:45:10.602160 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 20:45:10.606908 4122908 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 20:45:10.668172 4125532 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 20:45:10.668225 4125532 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 20:45:10.788786 138485273525376 metric_logger.py:185] completed step: 0, seconds: 0.711, TFLOP/s/device: 19.107, Tokens/s/device: 2880.041, total_weights: 16384, loss: 10.881 I0416 20:45:10.789710 138485273525376 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/tensorboard/' I0416 20:45:10.902521 138485273525376 metric_logger.py:185] completed step: 1, seconds: 1.652, TFLOP/s/device: 8.223, Tokens/s/device: 1239.535, total_weights: 16384, loss: 10.881 I0416 20:45:11.021332 138485273525376 metric_logger.py:185] completed step: 2, seconds: 0.010, TFLOP/s/device: 1298.464, Tokens/s/device: 195718.654, total_weights: 16384, loss: 10.282 I0416 20:45:11.140270 138485273525376 metric_logger.py:185] completed step: 3, seconds: 0.112, TFLOP/s/device: 121.565, Tokens/s/device: 18323.670, total_weights: 16384, loss: 9.282 I0416 20:45:11.146387 138485273525376 checkpointing.py:794] Waiting for step 5 to finish before checkpoint... I0416 20:45:11.378179 138485273525376 checkpointing.py:798] Waited 0.23176145553588867 seconds for step 5 to finish before starting checkpointing. I0416 20:45:11.378614 138485273525376 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 20:45:11.403196 138377775023680 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_15_ocdbt_false/checkpoints/0/items/array_metadatas/process_0 I0416 20:45:18.053871 138377735177792 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 7.383084s (commit=6.991250s, array_metadata_write=0.391834s) I0416 20:45:18.054911 138377724692032 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 1.419 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 8 seconds) (per-host) I0416 20:45:18.055031 138377724692032 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 7.464947s. I0416 20:45:18.263920 138377724692032 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776372310022678643, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/0/_CHECKPOINT_METADATA W0416 20:45:18.367144 138377724692032 ocdbt_utils.py:175] [process=0][thread=async_save] Skipping merge of OCDBT checkpoints: No per-process OCDBT checkpoint subdirs found in gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/0/items, I0416 20:45:18.459309 138377724692032 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 20:45:18.459417 138377724692032 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.149621s. 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_15_ocdbt_false/checkpoints/0/items I0416 20:45:18.460087 138377724692032 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/0/items I0416 20:45:18.644624 138377745663552 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776372310022678643, '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_15_ocdbt_false/checkpoints/0/_CHECKPOINT_METADATA I0416 20:45:18.709192 138377724692032 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/0 I0416 20:45:19.363100 138377724692032 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_15_ocdbt_false/checkpoints/0`. I0416 20:45:19.363728 138377724692032 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 10.112406s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/0 I0416 20:45:19.363805 138377724692032 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 8.773727s. I0416 20:45:19.363996 138377806480960 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 20:45:19.364118 138377806480960 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 20:45:19.364215 138377806480960 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0416 20:45:19.364266 138377806480960 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0416 20:45:19.364401 138485273525376 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0416 20:45:19.364512 138485273525376 checkpoint_manager.py:1441] Waiting for previous save to complete took 7.985928 seconds. If this number is high, consider checkpointing less frequently. I0416 20:45:19.365745 138485273525376 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 5 I0416 20:45:19.366035 138485273525376 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_15_ocdbt_false/checkpoints/5. I0416 20:45:19.531686 138485273525376 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 20:45:19.532121 138485273525376 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 20:45:19.535861 138377806480960 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_15_ocdbt_false/checkpoints/5 I0416 20:45:20.231696 138377653388864 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_15_ocdbt_false/checkpoints/5/items I0416 20:45:30.618423 138485273525376 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 11.087728s I0416 20:45:30.623951 138485273525376 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.104 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 11 seconds) (per-host) I0416 20:45:30.624229 138485273525376 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 11.179081s (batch_requests_ready=0.078300s, total_serialization_initiated=11.095137s, others=0.005644s) I0416 20:45:30.624319 138485273525376 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 11.179858s (all_items=0.000037s, per_item={'items': '0.00003743'}, temp_paths=11.179821) I0416 20:45:30.625514 138377735177792 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 20:45:30.625600 138485273525376 async_checkpointer.py:561] Finished blocking save. Time taken: 11.259806s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/5. I0416 20:45:30.625809 138485273525376 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0416 20:45:30.626024 138376164410944 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 20:45:30.626144 138485273525376 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776372311.378565, 'wait_for_prev_duration_secs': 7.985927581787109, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776372319.365769, 'checkpointer_blocking_duration_secs': 11.2599356174469, 'get_old_steps_start_time': 1776372330.6257288, 'get_old_steps_duration_secs': 3.647804260253906e-05, 'checkpoint_manager_blocking_start_time': 1776372311.3785, 'checkpoint_manager_blocking_duration_secs': 19.247618913650513} I0416 20:45:30.626296 138485273525376 checkpointing.py:409] Started an asynchronous checkpoint save for step 5 I0416 20:45:30.626743 138485273525376 metric_logger.py:185] completed step: 4, seconds: 0.119, TFLOP/s/device: 113.847, Tokens/s/device: 17160.190, total_weights: 16384, loss: 8.323 I0416 20:45:30.635206 138485273525376 metric_logger.py:185] completed step: 5, seconds: 0.119, TFLOP/s/device: 114.652, Tokens/s/device: 17281.534, total_weights: 16384, loss: 7.360 I0416 20:45:30.749485 138485273525376 metric_logger.py:185] completed step: 6, seconds: 19.486, TFLOP/s/device: 0.697, Tokens/s/device: 105.100, total_weights: 16384, loss: 6.449 I0416 20:45:30.868015 138485273525376 metric_logger.py:185] completed step: 7, seconds: 0.007, TFLOP/s/device: 1927.799, Tokens/s/device: 290578.888, total_weights: 16384, loss: 5.610 I0416 20:45:30.986229 138485273525376 metric_logger.py:185] completed step: 8, seconds: 0.116, TFLOP/s/device: 117.210, Tokens/s/device: 17667.204, total_weights: 16384, loss: 4.852 I0416 20:45:31.105874 138485273525376 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0416 20:45:31.107628 138485273525376 checkpointing.py:798] Waited 0.0018193721771240234 seconds for step 9 to finish before starting checkpointing. I0416 20:45:31.108156 138485273525376 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 20:45:31.373775 138377775023680 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_15_ocdbt_false/checkpoints/5/items/array_metadatas/process_0 I0416 20:45:38.096225 138377653388864 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 7.471582s (commit=7.082234s, array_metadata_write=0.389348s) I0416 20:45:38.097222 138377735177792 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 677.503 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 18 seconds) (per-host) I0416 20:45:38.097332 138377735177792 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 7.471682s. W0416 20:45:38.421358 138377735177792 ocdbt_utils.py:175] [process=0][thread=async_save] Skipping merge of OCDBT checkpoints: No per-process OCDBT checkpoint subdirs found in gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/5/items, I0416 20:45:38.502586 138377735177792 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 20:45:38.502689 138377735177792 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.155604s. 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_15_ocdbt_false/checkpoints/5/items I0416 20:45:38.503343 138377735177792 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/5/items I0416 20:45:38.739340 138377735177792 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/5 I0416 20:45:39.425817 138377735177792 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_15_ocdbt_false/checkpoints/5`. I0416 20:45:39.426382 138377735177792 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 20.060595s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/5 I0416 20:45:39.426462 138377735177792 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 8.800817s. I0416 20:45:39.426635 138376164410944 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 20:45:39.426739 138376164410944 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 20:45:39.426816 138376164410944 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0416 20:45:39.426862 138376164410944 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0416 20:45:39.427177 138485273525376 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0416 20:45:39.427300 138485273525376 checkpoint_manager.py:1441] Waiting for previous save to complete took 8.319171 seconds. If this number is high, consider checkpointing less frequently. I0416 20:45:39.427542 138485273525376 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0416 20:45:39.427833 138485273525376 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_15_ocdbt_false/checkpoints/9. I0416 20:45:39.592485 138377735177792 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_15_ocdbt_false/checkpoints/9 I0416 20:45:39.611520 138485273525376 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 20:45:39.611616 138485273525376 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 20:45:40.267917 138377653388864 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_15_ocdbt_false/checkpoints/9/items I0416 20:45:49.843232 138485273525376 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 10.232658s I0416 20:45:49.849153 138485273525376 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.193 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 10 seconds) (per-host) I0416 20:45:49.849217 138485273525376 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 10.342047s (batch_requests_ready=0.096428s, total_serialization_initiated=10.239797s, others=0.005821s) I0416 20:45:49.849292 138485273525376 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 10.342802s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=10.342784) I0416 20:45:49.850552 138377806480960 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 20:45:49.850636 138485273525376 async_checkpointer.py:561] Finished blocking save. Time taken: 10.423051s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/9. I0416 20:45:49.850845 138485273525376 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0416 20:45:49.851061 138376080524864 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 20:45:49.851168 138485273525376 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776372331.108111, 'wait_for_prev_duration_secs': 8.319170951843262, 'time_between_consecutive_saves_sec': 11.743828535079956, 'checkpointer_blocking_start_time': 1776372339.4275658, 'checkpointer_blocking_duration_secs': 10.423187971115112, 'get_old_steps_start_time': 1776372349.8507733, 'get_old_steps_duration_secs': 2.7894973754882812e-05, 'checkpoint_manager_blocking_start_time': 1776372331.1080618, 'checkpoint_manager_blocking_duration_secs': 18.743075132369995} I0416 20:45:49.851280 138485273525376 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0416 20:45:49.851322 138485273525376 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 20:45:50.593765 138377775023680 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_15_ocdbt_false/checkpoints/9/items/array_metadatas/process_0 I0416 20:45:57.424165 138377653388864 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 7.574266s (commit=7.137819s, array_metadata_write=0.436447s) I0416 20:45:57.425073 138377806480960 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 705.262 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 17 seconds) (per-host) I0416 20:45:57.425122 138377806480960 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 7.574434s. W0416 20:45:57.756345 138377806480960 ocdbt_utils.py:175] [process=0][thread=async_save] Skipping merge of OCDBT checkpoints: No per-process OCDBT checkpoint subdirs found in gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/9/items, I0416 20:45:57.849394 138377806480960 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 20:45:57.849510 138377806480960 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.153812s. 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_15_ocdbt_false/checkpoints/9/items I0416 20:45:57.850166 138377806480960 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/9/items I0416 20:45:58.885860 138377806480960 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/9 I0416 20:45:59.730144 138377806480960 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_15_ocdbt_false/checkpoints/9`. I0416 20:45:59.730760 138377806480960 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 20.303182s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_15_ocdbt_false/checkpoints/9 I0416 20:45:59.730826 138377806480960 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 9.880139s. I0416 20:45:59.730989 138376080524864 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 20:45:59.731106 138376080524864 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 20:45:59.731172 138376080524864 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0416 20:45:59.731214 138376080524864 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0416 20:45:59.731350 138485273525376 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0416 20:45:59.731492 138485273525376 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 20:45:59.732005 138485273525376 metric_logger.py:185] completed step: 9, seconds: 0.129, TFLOP/s/device: 105.131, Tokens/s/device: 15846.487, total_weights: 16384, loss: 4.176 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops