2026-04-21 17:58:43.941463: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered WARNING: All log messages before absl::InitializeLog() is called are written to STDERR E0000 00:00:1776794323.951641 1684627 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776794323.954730 1684627 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776794323.964778 1684627 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776794323.964794 1684627 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776794323.964796 1684627 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776794323.964797 1684627 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} `rope_scaling`'s factor field must be a float >= 1, got 40 `rope_scaling`'s beta_fast field must be a float, got 32 `rope_scaling`'s beta_slow field must be a float, got 1 Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} 2026-04-21 17:58:47.645168: 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) I0421 17:58:47.782958 129142307708032 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0421 17:59:17.516319 129142307708032 max_utils.py:800] System Information: Jax Version: 0.8.1 I0421 17:59:17.516583 129142307708032 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0421 17:59:17.516629 129142307708032 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709 I0421 17:59:17.516659 129142307708032 train_utils.py:378] WARNING: 'dataset_path' might be pointing your local file system I0421 17:59:17.516842 129142307708032 train.py:806] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0421 17:59:18.147281 129142307708032 maxtext_utils.py:1732] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 17:59:18.247494 129142307708032 checkpointing.py:688] Setting up checkpoint logger... I0421 17:59:18.247608 129142307708032 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0421 17:59:18.247648 129142307708032 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0421 17:59:18.247821 129142307708032 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 0x7573ffbae270>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 17:59:20.687746 129142307708032 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0421 17:59:20.687943 129142307708032 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 0x7572539a9c10>}, handler_registry=None I0421 17:59:20.688183 129142307708032 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7572539a9c10>` 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`. I0421 17:59:20.688225 129142307708032 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x75725395d9a0>` 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`. I0421 17:59:20.688254 129142307708032 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 0x7572539a9c10>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7572539a9c10>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x75725395d9a0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x75725395d9a0>}). I0421 17:59:20.688534 129142307708032 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0421 17:59:20.688588 129142307708032 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x757254dacc20> timeout: 600 secs and primary_host=0 for async checkpoint writes I0421 17:59:20.811501 129142307708032 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints I0421 17:59:20.811740 129142307708032 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_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7572539abc20> I0421 17:59:20.811823 129142307708032 checkpointing.py:302] Checkpoint manager created! I0421 17:59:20.903606 129142307708032 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0421 17:59:20.906571 129142307708032 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. I0421 17:59:20.958677 129142307708032 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 I0421 17:59:20.981415 129142307708032 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2 I0421 17:59:20.981480 129142307708032 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2 I0421 17:59:21.537913 129142307708032 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0421 17:59:21.539590 129142307708032 reader.py:262] Creating a tf.data.Dataset reading 2 files located in folders: tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0. I0421 17:59:21.555801 129142307708032 logging_logger.py:49] Constructing tf.data.Dataset __local_c4_builder for split validation, from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0421 17:59:21.558436 129142307708032 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2 I0421 17:59:21.558486 129142307708032 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2 I0421 17:59:21.914984 129142307708032 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0421 17:59:21.915111 129142307708032 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. [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 I0421 17:59:22.796750 129142307708032 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 17:59:22.796988 129142307708032 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 17:59:22.802404 129142307708032 nnx_decoders.py:465] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 17:59:22.802558 129142307708032 nnx_decoders.py:465] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 17:59:22.815900 129142307708032 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 17:59:22.815982 129142307708032 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 17:59:22.828549 129142307708032 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 17:59:22.828630 129142307708032 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 17:59:22.849466 129142307708032 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 17:59:22.849558 129142307708032 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 17:59:22.862271 129142307708032 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 17:59:22.862354 129142307708032 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 17:59:22.875078 129142307708032 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 17:59:22.875161 129142307708032 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 17:59:22.897064 129142307708032 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 17:59:22.897161 129142307708032 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 17:59:22.915120 129142307708032 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 17:59:22.915209 129142307708032 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0421 17:59:24.017564 129142307708032 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. I0421 17:59:24.021185 129142307708032 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0421 17:59:24.022628 129142307708032 metric_logger.py:301] number parameters: 1.104 billion I0421 17:59:29.001352 129142307708032 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0421 17:59:29.121756 129142307708032 checkpointing.py:798] Waited 0.12037944793701172 seconds for step 0 to finish before starting checkpointing. I0421 17:59:29.122276 129142307708032 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 17:59:29.122425 129142307708032 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0421 17:59:29.122837 129142307708032 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/0. I0421 17:59:29.212773 129142307708032 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0421 17:59:29.301623 129142307708032 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0421 17:59:29.301777 129049758271040 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/0 I0421 17:59:29.302173 129142307708032 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 I0421 17:59:29.998943 129049747785280 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/0/items I0421 17:59:30.158835 129049716328000 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776794369915073599, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776794369915073599, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/0/_CHECKPOINT_METADATA I0421 17:59:30.477561 129142307708032 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.177590s I0421 17:59:30.478133 129142307708032 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 9.768 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0421 17:59:30.478192 129142307708032 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.263516s (batch_requests_ready=0.077805s, total_serialization_initiated=1.185293s, others=0.000418s) I0421 17:59:30.478268 129142307708032 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.264935s (all_items=0.000019s, per_item={'items': '0.00001931'}, temp_paths=1.264915) I0421 17:59:30.479486 129049695356480 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 17:59:30.479619 129142307708032 async_checkpointer.py:561] Finished blocking save. Time taken: 1.357148s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/0. I0421 17:59:30.479869 129142307708032 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0421 17:59:30.480082 129049768756800 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 17:59:30.480185 129142307708032 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776794369.1222456, 'wait_for_prev_duration_secs': 5.984306335449219e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776794369.1224465, 'checkpointer_blocking_duration_secs': 1.3573188781738281, 'get_old_steps_start_time': 1776794370.4797897, 'get_old_steps_duration_secs': 2.7179718017578125e-05, 'checkpoint_manager_blocking_start_time': 1776794369.1221611, 'checkpoint_manager_blocking_duration_secs': 1.3579862117767334} I0421 17:59:30.480340 129142307708032 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0421 17:59:30.480412 129142307708032 max_utils.py:750] Memstats: After params initialized: I0421 17:59:30.480532 129142307708032 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0421 17:59:30.480559 129142307708032 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0421 17:59:30.480580 129142307708032 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0421 17:59:30.480599 129142307708032 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0421 17:59:30.480618 129142307708032 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0421 17:59:30.480635 129142307708032 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0421 17:59:30.480652 129142307708032 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0421 17:59:30.480668 129142307708032 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0421 17:59:30.547647 1687274 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0421 17:59:30.547730 1687274 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0421 17:59:30.868298 129142307708032 metric_logger.py:196] completed step: 0, seconds: 4.977, TFLOP/s/device: 2.730, Tokens/s/device: 411.464, total_weights: 13328, loss: 10.880, lm_loss: 10.880, perplexity: 53108.402 I0421 17:59:30.869209 129142307708032 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/tensorboard/' I0421 17:59:30.981814 129142307708032 metric_logger.py:196] completed step: 1, seconds: 1.863, TFLOP/s/device: 7.293, Tokens/s/device: 1099.327, total_weights: 12332, loss: 10.862, lm_loss: 10.862, perplexity: 52152.828 I0421 17:59:31.099147 129142307708032 metric_logger.py:196] completed step: 2, seconds: 0.021, TFLOP/s/device: 648.829, Tokens/s/device: 97798.577, total_weights: 15161, loss: 9.926, lm_loss: 9.926, perplexity: 20464.070 I0421 17:59:31.749403 129049737299520 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_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/0/items/array_metadatas/process_0 I0421 17:59:37.297073 129142307708032 train.py:723] Completed eval step 0 I0421 17:59:37.340779 129142307708032 train.py:723] Completed eval step 1 I0421 17:59:37.346113 129142307708032 metric_logger.py:216] eval metrics after step: 4, loss=8.926, perplexity=7524.814, total_weights=26703.0 I0421 17:59:37.346647 129142307708032 metric_logger.py:196] completed step: 3, seconds: 0.107, TFLOP/s/device: 127.341, Tokens/s/device: 19194.182, total_weights: 13327, loss: 9.396, lm_loss: 9.396, perplexity: 12037.986 I0421 17:59:52.758758 129142307708032 metric_logger.py:196] completed step: 4, seconds: 0.117, TFLOP/s/device: 115.750, Tokens/s/device: 17447.160, total_weights: 11939, loss: 9.005, lm_loss: 9.005, perplexity: 8141.998 I0421 17:59:52.778392 129142307708032 metric_logger.py:196] completed step: 5, seconds: 6.247, TFLOP/s/device: 2.175, Tokens/s/device: 327.844, total_weights: 15502, loss: 8.861, lm_loss: 8.861, perplexity: 7054.059 I0421 17:59:52.894363 129142307708032 metric_logger.py:196] completed step: 6, seconds: 15.421, TFLOP/s/device: 0.881, Tokens/s/device: 132.808, total_weights: 13864, loss: 8.738, lm_loss: 8.738, perplexity: 6233.198 I0421 17:59:53.012491 129142307708032 metric_logger.py:196] completed step: 7, seconds: 0.013, TFLOP/s/device: 1029.328, Tokens/s/device: 155151.515, total_weights: 12988, loss: 8.651, lm_loss: 8.651, perplexity: 5718.487 I0421 17:59:53.707201 129142307708032 train.py:723] Completed eval step 0 I0421 17:59:53.747874 129142307708032 train.py:723] Completed eval step 1 I0421 17:59:53.751453 129142307708032 metric_logger.py:216] eval metrics after step: 9, loss=8.623, perplexity=5557.299, total_weights=26703.0 I0421 17:59:53.751935 129142307708032 metric_logger.py:196] completed step: 8, seconds: 0.112, TFLOP/s/device: 121.515, Tokens/s/device: 18316.132, total_weights: 13820, loss: 8.671, lm_loss: 8.671, perplexity: 5832.209 I0421 17:59:53.754936 129142307708032 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0421 17:59:53.755885 129142307708032 checkpointing.py:798] Waited 0.000949859619140625 seconds for step 9 to finish before starting checkpointing. I0421 17:59:53.756156 129142307708032 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 18:00:05.733314 129049705842240 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 35.182291s (commit=34.748939s, array_metadata_write=0.433352s) I0421 18:00:05.734553 129049695356480 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 346.025 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 36 seconds) (per-host) I0421 18:00:05.734611 129049695356480 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 35.254934s. I0421 18:00:05.978978 129049695356480 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776794369915073599, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/0/_CHECKPOINT_METADATA I0421 18:00:06.170627 129049695356480 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0421 18:00:06.384856 129049716328000 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776794369915073599, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/0/_CHECKPOINT_METADATA I0421 18:00:06.586705 129049695356480 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 18:00:06.586910 129049695356480 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.563758s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/0/items I0421 18:00:06.587675 129049695356480 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/0/items I0421 18:00:06.835011 129049695356480 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/0 I0421 18:00:07.518141 129049695356480 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_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/0`. I0421 18:00:07.518887 129049695356480 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 38.396429s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/0 I0421 18:00:07.518971 129049695356480 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 37.039294s. I0421 18:00:07.519121 129049768756800 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 18:00:07.519173 129049768756800 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 18:00:07.519225 129049768756800 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0421 18:00:07.519271 129049768756800 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0421 18:00:07.519415 129142307708032 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. W0421 18:00:07.519569 129142307708032 checkpoint_manager.py:1441] Waiting for previous save to complete took 13.763419 seconds. If this number is high, consider checkpointing less frequently. I0421 18:00:07.520586 129142307708032 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0421 18:00:07.520922 129142307708032 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/9. I0421 18:00:07.701500 129049768756800 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/9 I0421 18:00:07.721270 129142307708032 jax_array_handlers.py:347] Scheduling D2H of 69 prioritized jax.Array. I0421 18:00:07.721376 129142307708032 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 I0421 18:00:08.426614 129049632441920 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/9/items I0421 18:00:20.319363 129142307708032 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 12.599438s I0421 18:00:20.325938 129142307708032 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 993.917 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 12 seconds) (per-host) I0421 18:00:20.326000 129142307708032 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 12.714217s (batch_requests_ready=0.100974s, total_serialization_initiated=12.606791s, others=0.006451s) I0421 18:00:20.326106 129142307708032 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 12.714857s (all_items=0.000015s, per_item={'items': '0.00001502'}, temp_paths=12.714842) I0421 18:00:20.327528 129049768756800 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 18:00:20.327614 129142307708032 async_checkpointer.py:561] Finished blocking save. Time taken: 12.806981s. Continuing background save to gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/9. I0421 18:00:20.327810 129142307708032 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0421 18:00:20.327973 129049695356480 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 18:00:20.328092 129142307708032 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776794393.7561126, 'wait_for_prev_duration_secs': 13.763418912887573, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776794407.5206127, 'checkpointer_blocking_duration_secs': 12.807109117507935, 'get_old_steps_start_time': 1776794420.3277392, 'get_old_steps_duration_secs': 3.0279159545898438e-05, 'checkpoint_manager_blocking_start_time': 1776794393.7560732, 'checkpoint_manager_blocking_duration_secs': 26.571993350982666} I0421 18:00:20.328221 129142307708032 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0421 18:00:20.328254 129142307708032 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 18:00:21.118530 129047491249728 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_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/9/items/array_metadatas/process_0 I0421 18:01:14.071533 129049737299520 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 53.744748s (commit=53.220269s, array_metadata_write=0.524478s) I0421 18:01:14.072951 129049768756800 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 190.138 MiB/s (total gbytes: 12.3 GiB) (time elapsed: a minute) (per-host) I0421 18:01:14.073093 129049768756800 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 53.745433s. I0421 18:01:14.499278 129049768756800 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0421 18:01:14.970229 129049768756800 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 18:01:14.971068 129049768756800 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.618137s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/9/items I0421 18:01:14.971830 129049768756800 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/9/items I0421 18:01:15.220630 129049768756800 atomicity.py:608] Finalizing gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/9 I0421 18:01:15.963443 129049768756800 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_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/9`. I0421 18:01:15.964243 129049768756800 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 68.443616s. directory=gs://wanglance-maxtext/nnx_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/nnx_feat_nnx_trainstate_and_training_loop_20260421_161934_07_eval/checkpoints/9 I0421 18:01:15.964323 129049768756800 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 55.636666s. I0421 18:01:15.964585 129049695356480 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 18:01:15.964698 129049695356480 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 18:01:15.964757 129049695356480 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0421 18:01:15.964802 129049695356480 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0421 18:01:15.964996 129142307708032 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. I0421 18:01:15.965196 129142307708032 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 18:01:15.966007 129142307708032 metric_logger.py:196] completed step: 9, seconds: 0.120, TFLOP/s/device: 113.309, Tokens/s/device: 17079.191, total_weights: 12300, loss: 8.668, lm_loss: 8.668, perplexity: 5810.989 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops