2026-04-16 19:45:11.314816: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0416 19:45:11.436641 124609502665856 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. ~/maxtext_venv/lib/python3.12/site-packages/jax/_src/xla_bridge.py:219: UserWarning: TPU backend initialization is taking more than 60.0 seconds. Did you run your code on all TPU hosts? See https://docs.jax.dev/en/latest/multi_process.html for more information. warnings.warn( I0416 19:46:13.675714 124609502665856 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 19:46:13.675826 124609502665856 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 19:46:13.675861 124609502665856 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Apr 6 2026 20:48:10 (1775533690) cl/895581894 I0416 19:46:13.675887 124609502665856 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system I0416 19:46:13.675963 124609502665856 train.py:811] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 19:46:13.768471 4056373 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:46:14.170143 124609502665856 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 19:46:14.267942 124609502665856 checkpointing.py:688] Setting up checkpoint logger... I0416 19:46:14.268039 124609502665856 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0416 19:46:14.268095 124609502665856 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 19:46:14.268274 124609502665856 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 0x715452dfd070>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 19:46:16.591111 124609502665856 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 19:46:16.591306 124609502665856 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 0x714dd3176cc0>}, handler_registry=None I0416 19:46:16.591584 124609502665856 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x714dd3176cc0>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0416 19:46:16.591624 124609502665856 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x714dd31beff0>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0416 19:46:16.591662 124609502665856 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 0x714dd3176cc0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x714dd3176cc0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x714dd31beff0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x714dd31beff0>}). I0416 19:46:16.592337 124609502665856 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 19:46:16.592391 124609502665856 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x714dd4195800> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 19:46:16.726495 124609502665856 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_03_dropout/checkpoints I0416 19:46:16.726727 124609502665856 checkpoint_manager.py:921] [process=0][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_hns=False, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_03_dropout/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x714dd31beb10> I0416 19:46:16.726813 124609502665856 checkpointing.py:302] Checkpoint manager created! I0416 19:46:16.809317 124609502665856 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0416 19:46:16.812369 124609502665856 reader.py:262] Creating a tf.data.Dataset reading 8 files located in folders: tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0. I0416 19:46:16.866809 124609502665856 logging_logger.py:49] Constructing tf.data.Dataset __local_c4_builder for split train, from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0416 19:46:16.896767 124609502665856 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2 I0416 19:46:16.896832 124609502665856 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2 I0416 19:46:17.914941 124609502665856 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 19:46:17.915075 124609502665856 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. W0416 19:46:18.025345 4056373 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 19:46:18.119446 124609502665856 maxtext_utils.py:1805] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0416 19:46:18.119549 124609502665856 maxtext_utils.py:1805] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0416 19:46:18.119589 124609502665856 maxtext_utils.py:1805] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0416 19:46:18.119621 124609502665856 maxtext_utils.py:1805] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0416 19:46:18.119655 124609502665856 maxtext_utils.py:1805] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0416 19:46:18.119681 124609502665856 maxtext_utils.py:1805] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0416 19:46:18.119707 124609502665856 maxtext_utils.py:1805] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 19:46:18.119733 124609502665856 maxtext_utils.py:1805] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0416 19:46:18.119757 124609502665856 maxtext_utils.py:1805] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 19:46:18.119781 124609502665856 maxtext_utils.py:1805] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0416 19:46:18.119805 124609502665856 maxtext_utils.py:1805] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0416 19:46:18.119828 124609502665856 maxtext_utils.py:1805] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0416 19:46:18.242266 124609502665856 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[8,128,2048]........................................ ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 19:46:18.242344 124609502665856 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[8,128,2048]........................................ ('fsdp', None, None). I0416 19:46:18.246796 124609502665856 nnx_decoders.py:465] Unknown Logical: bfloat16[8,128,2048]........................................ ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 19:46:18.246843 124609502665856 nnx_decoders.py:465] Unknown Physical: bfloat16[8,128,2048]........................................ ('fsdp', None, None). I0416 19:46:18.261785 124609502665856 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,128,2048]........................................ ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 19:46:18.261833 124609502665856 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,128,2048]........................................ ('fsdp', None, None). I0416 19:46:18.276230 124609502665856 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,128,2048]........................................ ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 19:46:18.276276 124609502665856 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,128,2048]........................................ ('fsdp', None, None). I0416 19:46:18.297890 124609502665856 attentions.py:1154] attentions/query Logical: bfloat16[8,128,16,128]...................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 19:46:18.297948 124609502665856 attentions.py:1154] attentions/query Physical: bfloat16[8,128,16,128]...................................... ('fsdp', None, None, None). I0416 19:46:18.312399 124609502665856 attentions.py:1155] attentions/key Logical: bfloat16[8,128,16,128]...................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 19:46:18.312447 124609502665856 attentions.py:1155] attentions/key Physical: bfloat16[8,128,16,128]...................................... ('fsdp', None, None, None). I0416 19:46:18.326862 124609502665856 attentions.py:1156] attentions/value Logical: bfloat16[8,128,16,128]...................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0416 19:46:18.326909 124609502665856 attentions.py:1156] attentions/value Physical: bfloat16[8,128,16,128]...................................... ('fsdp', None, None, None). I0416 19:46:18.354023 124609502665856 attentions.py:1197] attentions/out Logical: bfloat16[8,128,16,128]...................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 19:46:18.354093 124609502665856 attentions.py:1197] attentions/out Physical: bfloat16[8,128,16,128]...................................... ('fsdp', None, None, None). I0416 19:46:18.378591 124609502665856 linears.py:525] linears/x Logical: bfloat16[8,128,7168]........................................ ('activation_batch', 'activation_length', 'activation_mlp'). I0416 19:46:18.378653 124609502665856 linears.py:525] linears/x Physical: bfloat16[8,128,7168]........................................ ('fsdp', None, None). W0416 19:46:18.717732 4056373 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:46:18.799196 124609502665856 max_utils.py:791] Total memory size: 2.5 GB, Output size: 1.5 GB, Temp size: 1.0 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0416 19:46:18.799698 124609502665856 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 19:46:18.801190 124609502665856 metric_logger.py:289] number parameters: 1.104 billion W0416 19:46:21.595356 4056373 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:46:21.700197 124609502665856 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0416 19:46:21.765712 124609502665856 checkpointing.py:798] Waited 0.06550908088684082 seconds for step 0 to finish before starting checkpointing. I0416 19:46:21.766133 124609502665856 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 19:46:21.766283 124609502665856 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0416 19:46:21.766654 124609502665856 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_03_dropout/checkpoints/0. I0416 19:46:21.844578 124609502665856 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 19:46:21.945729 124609502665856 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 19:46:21.945870 124496285533760 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_03_dropout/checkpoints/0 I0416 19:46:21.946254 124609502665856 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False W0416 19:46:22.583389 4056373 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:46:22.601369 4056373 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:46:22.610139 4056373 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:46:22.617307 4056373 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:46:22.622714 4056373 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:46:22.627873 4056373 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:46:22.633713 4056373 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:46:22.641813 4056373 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:46:22.643158 124496275048000 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_03_dropout/checkpoints/0/items I0416 19:46:22.787882 124496226813504 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776368782556802396, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776368782556802396, "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_03_dropout/checkpoints/0/_CHECKPOINT_METADATA I0416 19:46:23.096481 124609502665856 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.152077s I0416 19:46:23.096941 124609502665856 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 9.864 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0416 19:46:23.097009 124609502665856 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.251245s (batch_requests_ready=0.090911s, total_serialization_initiated=1.159981s, others=0.000353s) I0416 19:46:23.097110 124609502665856 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.252091s (all_items=0.000019s, per_item={'items': '0.00001860'}, temp_paths=1.252072) I0416 19:46:23.098337 124496205841984 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 19:46:23.098438 124609502665856 async_checkpointer.py:561] Finished blocking save. Time taken: 1.332112s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_03_dropout/checkpoints/0. I0416 19:46:23.098643 124609502665856 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0416 19:46:23.098832 124496296019520 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 19:46:23.098950 124609502665856 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_03_dropout/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776368781.7661195, 'wait_for_prev_duration_secs': 4.315376281738281e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776368781.766304, 'checkpointer_blocking_duration_secs': 1.3322548866271973, 'get_old_steps_start_time': 1776368783.0985808, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1776368781.7660222, 'checkpoint_manager_blocking_duration_secs': 1.3328912258148193} I0416 19:46:23.099112 124609502665856 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0416 19:46:23.099191 124609502665856 max_utils.py:750] Memstats: After params initialized: I0416 19:46:23.099352 124609502665856 max_utils.py:756] Using (GB) 1.56 / 31.25 (4.992000%) on TPU_0(process=0,(0,0,0,0)) I0416 19:46:23.099383 124609502665856 max_utils.py:756] Using (GB) 1.56 / 31.25 (4.992000%) on TPU_1(process=0,(1,0,0,0)) I0416 19:46:23.099403 124609502665856 max_utils.py:756] Using (GB) 1.56 / 31.25 (4.992000%) on TPU_2(process=0,(0,1,0,0)) I0416 19:46:23.099421 124609502665856 max_utils.py:756] Using (GB) 1.56 / 31.25 (4.992000%) on TPU_3(process=0,(1,1,0,0)) I0416 19:46:23.099440 124609502665856 max_utils.py:756] Using (GB) 1.56 / 31.25 (4.992000%) on TPU_4(process=0,(0,2,0,0)) I0416 19:46:23.099457 124609502665856 max_utils.py:756] Using (GB) 1.56 / 31.25 (4.992000%) on TPU_5(process=0,(1,2,0,0)) I0416 19:46:23.099475 124609502665856 max_utils.py:756] Using (GB) 1.56 / 31.25 (4.992000%) on TPU_6(process=0,(0,3,0,0)) I0416 19:46:23.099493 124609502665856 max_utils.py:756] Using (GB) 1.56 / 31.25 (4.992000%) on TPU_7(process=0,(1,3,0,0)) W0416 19:46:23.106039 4056373 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:46:23.112509 4056373 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 19:46:23.118009 4056373 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 19:46:23.186842 4059167 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 19:46:23.186893 4059167 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 19:46:23.407420 124609502665856 metric_logger.py:185] completed step: 0, seconds: 2.898, TFLOP/s/device: 0.276, Tokens/s/device: 44.172, total_weights: 963, loss: 10.836 I0416 19:46:23.408433 124609502665856 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_03_dropout/tensorboard/' I0416 19:46:23.465869 124609502665856 metric_logger.py:185] completed step: 1, seconds: 1.703, TFLOP/s/device: 0.470, Tokens/s/device: 75.146, total_weights: 945, loss: 10.847 I0416 19:46:23.528943 124609502665856 metric_logger.py:185] completed step: 2, seconds: 0.021, TFLOP/s/device: 38.541, Tokens/s/device: 6159.769, total_weights: 962, loss: 10.296 I0416 19:46:23.592193 124609502665856 metric_logger.py:185] completed step: 3, seconds: 0.050, TFLOP/s/device: 15.861, Tokens/s/device: 2534.904, total_weights: 975, loss: 9.778 I0416 19:46:24.116647 124496285533760 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_03_dropout/checkpoints/0/items/array_metadatas/process_0 I0416 19:46:35.226812 124609502665856 metric_logger.py:185] completed step: 4, seconds: 0.063, TFLOP/s/device: 12.643, Tokens/s/device: 2020.744, total_weights: 962, loss: 9.422 I0416 19:46:35.236784 124609502665856 metric_logger.py:185] completed step: 5, seconds: 0.063, TFLOP/s/device: 12.635, Tokens/s/device: 2019.310, total_weights: 958, loss: 9.222 I0416 19:46:35.297499 124609502665856 metric_logger.py:185] completed step: 6, seconds: 11.634, TFLOP/s/device: 0.069, Tokens/s/device: 11.002, total_weights: 930, loss: 9.362 I0416 19:46:35.360654 124609502665856 metric_logger.py:185] completed step: 7, seconds: 0.008, TFLOP/s/device: 104.758, Tokens/s/device: 16742.969, total_weights: 935, loss: 9.333 I0416 19:46:35.423730 124609502665856 metric_logger.py:185] completed step: 8, seconds: 0.061, TFLOP/s/device: 13.119, Tokens/s/device: 2096.745, total_weights: 930, loss: 9.560 I0416 19:46:35.487622 124609502665856 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0416 19:46:35.488716 124609502665856 checkpointing.py:798] Waited 0.0011115074157714844 seconds for step 9 to finish before starting checkpointing. I0416 19:46:35.488978 124609502665856 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 19:46:48.847887 124496216327744 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 25.668299s (commit=25.260856s, array_metadata_write=0.407443s) I0416 19:46:48.848948 124496205841984 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 467.975 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 27 seconds) (per-host) I0416 19:46:48.849065 124496205841984 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 25.750580s. I0416 19:46:49.059621 124496205841984 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776368782556802396, '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_03_dropout/checkpoints/0/_CHECKPOINT_METADATA I0416 19:46:49.230644 124496205841984 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0416 19:46:49.427729 124496226813504 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776368782556802396, '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_03_dropout/checkpoints/0/_CHECKPOINT_METADATA I0416 19:46:49.650120 124496205841984 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 19:46:49.650283 124496205841984 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.553681s. 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_03_dropout/checkpoints/0/items I0416 19:46:49.650950 124496205841984 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_03_dropout/checkpoints/0/items I0416 19:46:49.887247 124496205841984 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_03_dropout/checkpoints/0 I0416 19:46:50.536053 124496205841984 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_03_dropout/checkpoints/0`. I0416 19:46:50.536666 124496205841984 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 28.770350s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_03_dropout/checkpoints/0 I0416 19:46:50.536741 124496205841984 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 27.438257s. I0416 19:46:50.536917 124496296019520 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 19:46:50.537027 124496296019520 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 19:46:50.537101 124496296019520 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0416 19:46:50.537153 124496296019520 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0416 19:46:50.537254 124609502665856 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 19:46:50.537336 124609502665856 checkpoint_manager.py:1441] Waiting for previous save to complete took 15.048365 seconds. If this number is high, consider checkpointing less frequently. I0416 19:46:50.538434 124609502665856 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0416 19:46:50.538669 124609502665856 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_03_dropout/checkpoints/9. I0416 19:46:50.707207 124609502665856 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 19:46:50.707298 124609502665856 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0416 19:46:50.719950 124496296019520 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_03_dropout/checkpoints/9 I0416 19:46:51.442982 124496165996096 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_03_dropout/checkpoints/9/items I0416 19:47:02.221911 124609502665856 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 11.515722s I0416 19:47:02.228953 124609502665856 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.063 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 11 seconds) (per-host) I0416 19:47:02.229020 124609502665856 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 11.611095s (batch_requests_ready=0.081184s, total_serialization_initiated=11.522990s, others=0.006922s) I0416 19:47:02.229114 124609502665856 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 11.611628s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=11.611610) I0416 19:47:02.230352 124496296019520 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 19:47:02.230431 124609502665856 async_checkpointer.py:561] Finished blocking save. Time taken: 11.691955s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_03_dropout/checkpoints/9. I0416 19:47:02.230638 124609502665856 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0416 19:47:02.230824 124496205841984 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 19:47:02.230941 124609502665856 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_03_dropout/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776368795.4889443, 'wait_for_prev_duration_secs': 15.048365116119385, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776368810.5384586, 'checkpointer_blocking_duration_secs': 11.692075967788696, 'get_old_steps_start_time': 1776368822.230553, 'get_old_steps_duration_secs': 3.2901763916015625e-05, 'checkpoint_manager_blocking_start_time': 1776368795.4889042, 'checkpoint_manager_blocking_duration_secs': 26.742007970809937} I0416 19:47:02.231084 124609502665856 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0416 19:47:02.231121 124609502665856 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 19:47:02.982058 124496285533760 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_03_dropout/checkpoints/9/items/array_metadatas/process_0 I0416 19:47:33.969710 124496165996096 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 31.740013s (commit=31.318510s, array_metadata_write=0.421503s) I0416 19:47:33.970717 124496296019520 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 291.487 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 43 seconds) (per-host) I0416 19:47:33.970767 124496296019520 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 31.740293s. I0416 19:47:34.421035 124496296019520 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0416 19:47:34.806841 124496296019520 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 19:47:34.807476 124496296019520 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.540315s. 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_03_dropout/checkpoints/9/items I0416 19:47:34.808106 124496296019520 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_03_dropout/checkpoints/9/items I0416 19:47:35.049397 124496296019520 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_03_dropout/checkpoints/9 I0416 19:47:35.725798 124496296019520 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_03_dropout/checkpoints/9`. I0416 19:47:35.726428 124496296019520 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 45.187959s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_post_train_fixes_20260416_181521/nnx_feat_nnx_post_train_fixes_20260416_181521_03_dropout/checkpoints/9 I0416 19:47:35.726503 124496296019520 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 33.496029s. I0416 19:47:35.726652 124496205841984 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 19:47:35.726757 124496205841984 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 19:47:35.726818 124496205841984 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0416 19:47:35.726866 124496205841984 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0416 19:47:35.727072 124609502665856 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 19:47:35.727802 124609502665856 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 19:47:35.728389 124609502665856 metric_logger.py:185] completed step: 9, seconds: 0.063, TFLOP/s/device: 12.642, Tokens/s/device: 2020.489, total_weights: 948, loss: 9.098 Per train step: Total TFLOPs: 0.80 split as 99.60% learnable weight flops and 0.40% attention flops