2026-04-16 04:55:43.227860: 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 04:55:43.343902 126593953442944 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0416 04:56:15.542333 126593953442944 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 04:56:15.542456 126593953442944 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 04:56:15.542492 126593953442944 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 04:56:15.542518 126593953442944 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system I0416 04:56:15.542616 126593953442944 train.py:818] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 04:56:15.637505 3707468 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 04:56:16.054530 126593953442944 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 04:56:16.154268 126593953442944 checkpointing.py:688] Setting up checkpoint logger... I0416 04:56:16.154378 126593953442944 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0416 04:56:16.154422 126593953442944 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 04:56:16.154628 126593953442944 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 0x73225d5f8c50>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 04:56:18.828626 126593953442944 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0416 04:56:18.828869 126593953442944 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 0x731bdfd1f0e0>}, handler_registry=None I0416 04:56:18.829168 126593953442944 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x731bdfd1f0e0>` 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 04:56:18.829212 126593953442944 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x731bdfdcd7f0>` 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 04:56:18.829241 126593953442944 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 0x731bdfd1f0e0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x731bdfd1f0e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x731bdfdcd7f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x731bdfdcd7f0>}). I0416 04:56:18.829569 126593953442944 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 04:56:18.829624 126593953442944 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x731be0d89940> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 04:56:18.966695 126593953442944 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints I0416 04:56:18.966961 126593953442944 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_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x731bdfdcdc40> I0416 04:56:18.967061 126593953442944 checkpointing.py:302] Checkpoint manager created! I0416 04:56:19.047508 126593953442944 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0416 04:56:19.050676 126593953442944 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 04:56:19.112146 126593953442944 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 04:56:19.143620 126593953442944 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2 I0416 04:56:19.143689 126593953442944 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2 I0416 04:56:20.169569 126593953442944 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0416 04:56:20.169688 126593953442944 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. W0416 04:56:20.282489 3707468 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 04:56:20.490331 126593953442944 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 04:56:20.490435 126593953442944 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 04:56:20.495555 126593953442944 nnx_decoders.py:465] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 04:56:20.495603 126593953442944 nnx_decoders.py:465] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 04:56:20.511251 126593953442944 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 04:56:20.511305 126593953442944 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 04:56:20.526349 126593953442944 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 04:56:20.526401 126593953442944 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 04:56:20.548876 126593953442944 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 04:56:20.548950 126593953442944 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 04:56:20.564030 126593953442944 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 04:56:20.564099 126593953442944 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 04:56:20.579120 126593953442944 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 04:56:20.579172 126593953442944 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 04:56:20.607540 126593953442944 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 04:56:20.607606 126593953442944 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 04:56:20.626309 126593953442944 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 04:56:20.626364 126593953442944 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0416 04:56:28.396030 126593953442944 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 04:56:28.396892 126593953442944 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 04:56:28.398828 126593953442944 metric_logger.py:289] number parameters: 1.104 billion I0416 04:56:39.029376 126593953442944 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0416 04:56:39.158928 126593953442944 checkpointing.py:798] Waited 0.12952947616577148 seconds for step 0 to finish before starting checkpointing. I0416 04:56:39.159552 126593953442944 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 04:56:39.159734 126593953442944 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0416 04:56:39.160229 126593953442944 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/0. I0416 04:56:39.243339 126593953442944 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 04:56:39.349202 126479887238720 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/0 I0416 04:56:39.349621 126593953442944 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 04:56:39.349735 126593953442944 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 04:56:40.052072 126479876752960 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/0/items W0416 04:56:40.159682 3707468 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 04:56:40.178693 3707468 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 04:56:40.190091 3707468 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 04:56:40.198842 3707468 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 04:56:40.205401 126479836907072 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776315399961068216, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776315399961068216, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/0/_CHECKPOINT_METADATA W0416 04:56:40.205740 3707468 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 04:56:40.211279 3707468 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 04:56:40.216405 3707468 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 04:56:40.221720 3707468 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 04:56:40.551333 126593953442944 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.203229s I0416 04:56:40.551839 126593953442944 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 9.446 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0416 04:56:40.551893 126593953442944 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.306514s (batch_requests_ready=0.093986s, total_serialization_initiated=1.212150s, others=0.000378s) I0416 04:56:40.551956 126593953442944 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.308007s (all_items=0.000027s, per_item={'items': '0.00002718'}, temp_paths=1.307980) I0416 04:56:40.552956 126479815935552 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 04:56:40.553089 126593953442944 async_checkpointer.py:561] Finished blocking save. Time taken: 1.393312s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/0. I0416 04:56:40.553294 126593953442944 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0416 04:56:40.553613 126479805449792 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 04:56:40.553740 126593953442944 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776315399.1595376, 'wait_for_prev_duration_secs': 4.4345855712890625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776315399.159755, 'checkpointer_blocking_duration_secs': 1.3934566974639893, 'get_old_steps_start_time': 1776315400.5532308, 'get_old_steps_duration_secs': 2.8133392333984375e-05, 'checkpoint_manager_blocking_start_time': 1776315399.1594315, 'checkpoint_manager_blocking_duration_secs': 1.3942797183990479} I0416 04:56:40.553878 126593953442944 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0416 04:56:40.553932 126593953442944 max_utils.py:750] Memstats: After params initialized: I0416 04:56:40.554087 126593953442944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0416 04:56:40.554118 126593953442944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0416 04:56:40.554139 126593953442944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0416 04:56:40.554160 126593953442944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0416 04:56:40.554178 126593953442944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0416 04:56:40.554198 126593953442944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0416 04:56:40.554215 126593953442944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0416 04:56:40.554234 126593953442944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0416 04:56:40.561120 3707468 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 04:56:40.567521 3707468 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 04:56:40.572915 3707468 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 04:56:40.606583 3710524 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 04:56:40.606674 3710524 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 04:56:40.942271 126593953442944 metric_logger.py:185] completed step: 0, seconds: 10.629, TFLOP/s/device: 1.278, Tokens/s/device: 192.681, total_weights: 13328, loss: 10.880 I0416 04:56:40.943268 126593953442944 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/tensorboard/' I0416 04:56:41.054278 126593953442944 metric_logger.py:185] completed step: 1, seconds: 1.908, TFLOP/s/device: 7.121, Tokens/s/device: 1073.412, total_weights: 12332, loss: 10.862 I0416 04:56:41.171937 126593953442944 metric_logger.py:185] completed step: 2, seconds: 0.023, TFLOP/s/device: 591.336, Tokens/s/device: 89132.611, total_weights: 15161, loss: 9.926 I0416 04:56:41.289644 126593953442944 metric_logger.py:185] completed step: 3, seconds: 0.105, TFLOP/s/device: 129.432, Tokens/s/device: 19509.407, total_weights: 13327, loss: 9.396 I0416 04:56:41.600501 126479866267200 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/0/items/array_metadatas/process_0 I0416 04:57:01.601918 126593953442944 metric_logger.py:185] completed step: 4, seconds: 0.118, TFLOP/s/device: 115.256, Tokens/s/device: 17372.569, total_weights: 11939, loss: 9.005 I0416 04:57:01.616021 126593953442944 metric_logger.py:185] completed step: 5, seconds: 0.118, TFLOP/s/device: 115.404, Tokens/s/device: 17394.997, total_weights: 15502, loss: 8.861 I0416 04:57:01.730457 126593953442944 metric_logger.py:185] completed step: 6, seconds: 20.313, TFLOP/s/device: 0.669, Tokens/s/device: 100.821, total_weights: 13864, loss: 8.738 I0416 04:57:01.847896 126593953442944 metric_logger.py:185] completed step: 7, seconds: 0.009, TFLOP/s/device: 1547.333, Tokens/s/device: 233230.839, total_weights: 12988, loss: 8.651 I0416 04:57:01.965618 126593953442944 metric_logger.py:185] completed step: 8, seconds: 0.116, TFLOP/s/device: 117.081, Tokens/s/device: 17647.718, total_weights: 13820, loss: 8.671 I0416 04:57:02.083621 126593953442944 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0416 04:57:02.084784 126593953442944 checkpointing.py:798] Waited 0.0011796951293945312 seconds for step 9 to finish before starting checkpointing. I0416 04:57:02.085779 126593953442944 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 04:57:07.145154 126479887238720 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 26.542466s (commit=26.082773s, array_metadata_write=0.459692s) I0416 04:57:07.146743 126479815935552 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 452.911 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 27 seconds) (per-host) I0416 04:57:07.146904 126479815935552 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 26.593766s. I0416 04:57:07.357596 126479815935552 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776315399961068216, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/0/_CHECKPOINT_METADATA I0416 04:57:07.538908 126479815935552 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 04:57:07.729574 126479836907072 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776315399961068216, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/0/_CHECKPOINT_METADATA I0416 04:57:07.982755 126479815935552 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 04:57:07.983468 126479815935552 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.585485s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/0/items I0416 04:57:07.984246 126479815935552 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/0/items I0416 04:57:08.217881 126479815935552 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/0 I0416 04:57:08.885067 126479815935552 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/0`. I0416 04:57:08.885862 126479815935552 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 29.726093s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/0 I0416 04:57:08.885945 126479815935552 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 28.332810s. I0416 04:57:08.886142 126479805449792 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 04:57:08.886267 126479805449792 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 04:57:08.886330 126479805449792 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0416 04:57:08.886375 126479805449792 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0416 04:57:08.886468 126593953442944 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 04:57:08.886537 126593953442944 checkpoint_manager.py:1441] Waiting for previous save to complete took 6.800791 seconds. If this number is high, consider checkpointing less frequently. I0416 04:57:08.887336 126593953442944 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0416 04:57:08.887634 126593953442944 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/9. I0416 04:57:09.066328 126479805449792 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/9 I0416 04:57:09.076963 126593953442944 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0416 04:57:09.077090 126593953442944 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 04:57:09.736999 126479897724480 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/9/items I0416 04:57:15.916799 126593953442944 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 6.840929s I0416 04:57:15.923278 126593953442944 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.777 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host) I0416 04:57:15.923345 126593953442944 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 6.945726s (batch_requests_ready=0.090503s, total_serialization_initiated=6.848884s, others=0.006339s) I0416 04:57:15.923423 126593953442944 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 6.946360s (all_items=0.000021s, per_item={'items': '0.00002098'}, temp_paths=6.946339) I0416 04:57:15.924791 126479887238720 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 04:57:15.924867 126593953442944 async_checkpointer.py:561] Finished blocking save. Time taken: 7.037480s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/9. I0416 04:57:15.925129 126593953442944 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0416 04:57:15.925308 126479815935552 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 04:57:15.925410 126593953442944 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776315422.0857203, 'wait_for_prev_duration_secs': 6.800791263580322, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776315428.88736, 'checkpointer_blocking_duration_secs': 7.037649631500244, 'get_old_steps_start_time': 1776315435.9250395, 'get_old_steps_duration_secs': 4.7206878662109375e-05, 'checkpoint_manager_blocking_start_time': 1776315422.0856771, 'checkpoint_manager_blocking_duration_secs': 13.8397057056427} I0416 04:57:15.925499 126593953442944 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0416 04:57:15.925531 126593953442944 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 04:57:16.664685 126479866267200 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/9/items/array_metadatas/process_0 I0416 04:57:44.804703 126479897724480 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 28.880592s (commit=28.465147s, array_metadata_write=0.415445s) I0416 04:57:44.805998 126479887238720 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 352.704 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 35 seconds) (per-host) I0416 04:57:44.806153 126479887238720 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 28.881228s. I0416 04:57:45.224563 126479887238720 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 04:57:45.652949 126479887238720 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 04:57:45.653675 126479887238720 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.573001s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/9/items I0416 04:57:45.654465 126479887238720 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/9/items I0416 04:57:45.888153 126479887238720 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/9 I0416 04:57:46.557346 126479887238720 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/9`. I0416 04:57:46.558169 126479887238720 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 37.670797s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260416_033532/nnx_feat_nnx_trainstate_and_training_loop_20260416_033532_01_base/checkpoints/9 I0416 04:57:46.558254 126479887238720 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 30.633331s. I0416 04:57:46.558401 126479815935552 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 04:57:46.558450 126479815935552 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 04:57:46.558500 126479815935552 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0416 04:57:46.558544 126479815935552 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0416 04:57:46.558681 126593953442944 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 04:57:46.558869 126593953442944 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 04:57:46.559644 126593953442944 metric_logger.py:185] completed step: 9, seconds: 0.117, TFLOP/s/device: 116.050, Tokens/s/device: 17492.313, total_weights: 12300, loss: 8.668 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops