2026-04-17 13:18:27.397305: 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) I0417 13:18:27.512783 129407805738112 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0417 13:19:17.771620 129407805738112 max_utils.py:800] System Information: Jax Version: 0.9.2 I0417 13:19:17.771729 129407805738112 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0417 13:19:17.771767 129407805738112 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 I0417 13:19:17.771790 129407805738112 train_utils.py:335] WARNING: 'dataset_path' might be pointing your local file system I0417 13:19:17.771889 129407805738112 train.py:671] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0417 13:19:17.864825 449583 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:19:18.266916 129407805738112 maxtext_utils.py:1548] Num_devices: 8, shape (1, 1, 1, 2, 1, 1, 1, 1, 4, 1, 1, 1, 1) I0417 13:19:18.267271 129407805738112 checkpointing.py:677] Setting up checkpoint logger... I0417 13:19:18.267318 129407805738112 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0417 13:19:18.267393 129407805738112 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0417 13:19:18.267930 129407805738112 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 0x75b1839f8c80>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0417 13:19:20.640692 129407805738112 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0417 13:19:20.640898 129407805738112 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 0x75ab077326f0>}, handler_registry=None I0417 13:19:20.641242 129407805738112 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x75ab077326f0>` 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`. I0417 13:19:20.641286 129407805738112 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x75ab07731430>` 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`. I0417 13:19:20.641314 129407805738112 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 0x75ab077326f0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x75ab077326f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x75ab07731430>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x75ab07731430>}). I0417 13:19:20.641609 129407805738112 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0417 13:19:20.641665 129407805738112 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x75ab0772e0c0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0417 13:19:20.764643 129407805738112 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints I0417 13:19:20.764883 129407805738112 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/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x75abd7bbf0b0> I0417 13:19:20.764970 129407805738112 checkpointing.py:301] Checkpoint manager created! I0417 13:19:21.223458 129407805738112 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0417 13:19:21.226366 129407805738112 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. I0417 13:19:21.278614 129407805738112 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 I0417 13:19:21.301847 129407805738112 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2 I0417 13:19:21.301916 129407805738112 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2 I0417 13:19:21.910574 129407805738112 nnx_wrappers.py:437] Unknown Logical: bfloat16[2,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 13:19:21.910673 129407805738112 nnx_wrappers.py:437] Unknown Physical: bfloat16[2,2048,2048]....................................... ('fsdp', None, 'tensor'). I0417 13:19:22.014188 129407805738112 attentions.py:1088] attentions/inputs_q Logical: bfloat16[2,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 13:19:22.014261 129407805738112 attentions.py:1088] attentions/inputs_q Physical: bfloat16[2,2048,2048]....................................... ('fsdp', None, 'tensor'). I0417 13:19:22.029454 129407805738112 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[2,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 13:19:22.029501 129407805738112 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[2,2048,2048]....................................... ('fsdp', None, 'tensor'). I0417 13:19:22.051914 129407805738112 attentions.py:1154] attentions/query Logical: bfloat16[2,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 13:19:22.051970 129407805738112 attentions.py:1154] attentions/query Physical: bfloat16[2,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0417 13:19:22.067361 129407805738112 attentions.py:1155] attentions/key Logical: bfloat16[2,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 13:19:22.067411 129407805738112 attentions.py:1155] attentions/key Physical: bfloat16[2,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0417 13:19:22.082682 129407805738112 attentions.py:1156] attentions/value Logical: bfloat16[2,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 13:19:22.082731 129407805738112 attentions.py:1156] attentions/value Physical: bfloat16[2,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0417 13:19:22.105382 129407805738112 attentions.py:1197] attentions/out Logical: bfloat16[2,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0417 13:19:22.105441 129407805738112 attentions.py:1197] attentions/out Physical: bfloat16[2,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0417 13:19:22.124455 129407805738112 linears.py:525] linears/x Logical: bfloat16[2,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0417 13:19:22.124506 129407805738112 linears.py:525] linears/x Physical: bfloat16[2,2048,7168]....................................... ('fsdp', None, 'tensor'). I0417 13:19:22.319089 129407805738112 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0417 13:19:22.319182 129407805738112 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. W0417 13:19:22.369636 449583 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: 2 tensor: 4 I0417 13:19:22.767282 129407805738112 maxtext_utils.py:1651] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: ('tensor',) I0417 13:19:22.767389 129407805738112 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, 'tensor') I0417 13:19:22.767431 129407805738112 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, 'tensor') I0417 13:19:22.767474 129407805738112 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: ('tensor', None, 'fsdp') I0417 13:19:22.767513 129407805738112 maxtext_utils.py:1651] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: ('tensor', None) I0417 13:19:22.767540 129407805738112 maxtext_utils.py:1651] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: ('tensor', None) I0417 13:19:22.767579 129407805738112 maxtext_utils.py:1651] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, 'tensor', None) I0417 13:19:22.767616 129407805738112 maxtext_utils.py:1651] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: ('tensor', None, None, 'fsdp') I0417 13:19:22.767643 129407805738112 maxtext_utils.py:1651] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, 'tensor', None) I0417 13:19:22.767668 129407805738112 maxtext_utils.py:1651] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, 'tensor', None) I0417 13:19:22.767702 129407805738112 maxtext_utils.py:1651] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', 'tensor') I0417 13:19:22.767735 129407805738112 maxtext_utils.py:1651] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: ('tensor', 'fsdp') I0417 13:19:23.198657 129407805738112 train.py:155] train/xent Logical: float32[2,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0417 13:19:23.198737 129407805738112 train.py:155] train/xent Physical: float32[2,2048]............................................. ('fsdp', None). I0417 13:19:23.213016 129407805738112 train.py:162] train/z_loss Logical: float32[2,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0417 13:19:23.213078 129407805738112 train.py:162] train/z_loss Physical: float32[2,2048]............................................. ('fsdp', None). I0417 13:19:38.189071 129407805738112 max_utils.py:791] Total memory size: 2.5 GB, Output size: 1.5 GB, Temp size: 0.9 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0417 13:19:38.189860 129407805738112 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0417 13:19:38.190076 129407805738112 metric_logger.py:301] number parameters: 1.104 billion W0417 13:19:41.778845 449583 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:19:41.850080 129407805738112 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0417 13:19:41.904791 129407805738112 checkpointing.py:776] Waited 0.05471682548522949 seconds for step 0 to finish before starting checkpointing. I0417 13:19:41.905220 129407805738112 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 13:19:41.905393 129407805738112 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0417 13:19:41.905750 129407805738112 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/0. I0417 13:19:41.984027 129407805738112 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0417 13:19:42.078117 129407805738112 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0417 13:19:42.078254 129293887735360 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/0 I0417 13:19:42.078704 129407805738112 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 I0417 13:19:42.844216 129293877249600 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/0/items I0417 13:19:43.005582 129293829015104 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776431982758780597, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776431982758780597, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/0/_CHECKPOINT_METADATA I0417 13:19:43.403084 452523 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0417 13:19:43.403139 452523 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0417 13:19:43.485182 129407805738112 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.407832s I0417 13:19:43.485755 129407805738112 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 8.223 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0417 13:19:43.485808 129407805738112 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.500818s (batch_requests_ready=0.088900s, total_serialization_initiated=1.411437s, others=0.000481s) I0417 13:19:43.485870 129407805738112 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.501371s (all_items=0.000019s, per_item={'items': '0.00001907'}, temp_paths=1.501352) I0417 13:19:43.492181 129293755614784 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0417 13:19:43.492268 129407805738112 async_checkpointer.py:561] Finished blocking save. Time taken: 1.586833s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/0. I0417 13:19:43.492452 129407805738112 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0417 13:19:43.492691 129293898221120 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 13:19:43.492797 129407805738112 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776431981.9052072, 'wait_for_prev_duration_secs': 4.124641418457031e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776431981.9054148, 'checkpointer_blocking_duration_secs': 1.5869648456573486, 'get_old_steps_start_time': 1776431983.492398, 'get_old_steps_duration_secs': 1.9788742065429688e-05, 'checkpoint_manager_blocking_start_time': 1776431981.9051256, 'checkpoint_manager_blocking_duration_secs': 1.587646245956421} I0417 13:19:43.492914 129407805738112 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0417 13:19:43.492953 129407805738112 max_utils.py:750] Memstats: After params initialized: I0417 13:19:43.492996 129407805738112 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_0(process=0,(0,0,0,0)) I0417 13:19:43.493021 129407805738112 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_1(process=0,(1,0,0,0)) I0417 13:19:43.493040 129407805738112 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_2(process=0,(0,1,0,0)) I0417 13:19:43.493074 129407805738112 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_3(process=0,(1,1,0,0)) I0417 13:19:43.493091 129407805738112 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_4(process=0,(0,2,0,0)) I0417 13:19:43.493108 129407805738112 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_5(process=0,(1,2,0,0)) I0417 13:19:43.493124 129407805738112 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_6(process=0,(0,3,0,0)) I0417 13:19:43.493140 129407805738112 max_utils.py:756] Using (GB) 1.58 / 31.25 (5.056000%) on TPU_7(process=0,(1,3,0,0)) W0417 13:19:43.499302 449583 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:19:43.505575 449583 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:19:43.510186 449583 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:19:43.879394 129407805738112 metric_logger.py:196] completed step: 0, seconds: 3.659, TFLOP/s/device: 0.928, Tokens/s/device: 139.918, total_weights: 3650, loss: 10.896, lm_loss: 10.896, perplexity: 53934.855 I0417 13:19:43.880278 129407805738112 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/tensorboard/' I0417 13:19:43.929796 129407805738112 metric_logger.py:196] completed step: 1, seconds: 2.027, TFLOP/s/device: 1.676, Tokens/s/device: 252.584, total_weights: 2676, loss: 10.833, lm_loss: 10.833, perplexity: 50674.457 I0417 13:19:43.982093 129407805738112 metric_logger.py:196] completed step: 2, seconds: 0.013, TFLOP/s/device: 258.369, Tokens/s/device: 38944.246, total_weights: 3783, loss: 10.136, lm_loss: 10.136, perplexity: 25231.213 I0417 13:19:44.034417 129407805738112 metric_logger.py:196] completed step: 3, seconds: 0.051, TFLOP/s/device: 66.341, Tokens/s/device: 9999.609, total_weights: 3680, loss: 9.403, lm_loss: 9.403, perplexity: 12122.400 I0417 13:19:44.246693 129293776586304 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/0/items/array_metadatas/process_0 I0417 13:19:54.125198 129407805738112 metric_logger.py:196] completed step: 4, seconds: 0.053, TFLOP/s/device: 64.456, Tokens/s/device: 9715.554, total_weights: 3736, loss: 9.172, lm_loss: 9.172, perplexity: 9620.940 I0417 13:19:54.137083 129407805738112 metric_logger.py:196] completed step: 5, seconds: 0.052, TFLOP/s/device: 65.281, Tokens/s/device: 9839.909, total_weights: 4096, loss: 9.075, lm_loss: 9.075, perplexity: 8737.906 I0417 13:19:54.188858 129407805738112 metric_logger.py:196] completed step: 6, seconds: 10.090, TFLOP/s/device: 0.337, Tokens/s/device: 50.744, total_weights: 2931, loss: 8.899, lm_loss: 8.899, perplexity: 7322.388 I0417 13:19:54.241189 129407805738112 metric_logger.py:196] completed step: 7, seconds: 0.009, TFLOP/s/device: 378.851, Tokens/s/device: 57104.617, total_weights: 3526, loss: 8.818, lm_loss: 8.818, perplexity: 6752.802 I0417 13:19:54.293432 129407805738112 metric_logger.py:196] completed step: 8, seconds: 0.053, TFLOP/s/device: 64.662, Tokens/s/device: 9746.626, total_weights: 3012, loss: 8.750, lm_loss: 8.750, perplexity: 6310.135 I0417 13:19:54.345424 129407805738112 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0417 13:19:54.346212 129407805738112 checkpointing.py:776] Waited 0.0008018016815185547 seconds for step 9 to finish before starting checkpointing. I0417 13:19:54.346465 129407805738112 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 13:20:12.796166 129293766100544 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 29.304798s (commit=28.902297s, array_metadata_write=0.402501s) I0417 13:20:12.797476 129293755614784 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 410.120 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 30 seconds) (per-host) I0417 13:20:12.797588 129293755614784 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 29.305273s. I0417 13:20:13.014952 129293755614784 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776431982758780597, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/0/_CHECKPOINT_METADATA I0417 13:20:13.205742 129293755614784 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0417 13:20:13.400367 129293829015104 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776431982758780597, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/0/_CHECKPOINT_METADATA I0417 13:20:13.596569 129293755614784 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0417 13:20:13.596758 129293755614784 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.537738s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/0/items I0417 13:20:13.597402 129293755614784 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/0/items I0417 13:20:13.818708 129293755614784 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/0 I0417 13:20:14.527785 129293755614784 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/0`. I0417 13:20:14.528444 129293755614784 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 32.623017s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/0 I0417 13:20:14.528519 129293755614784 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 31.036207s. I0417 13:20:14.528691 129293898221120 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 13:20:14.528793 129293898221120 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 13:20:14.528851 129293898221120 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0417 13:20:14.528898 129293898221120 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0417 13:20:14.528979 129407805738112 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. W0417 13:20:14.529057 129407805738112 checkpoint_manager.py:1441] Waiting for previous save to complete took 20.182594 seconds. If this number is high, consider checkpointing less frequently. I0417 13:20:14.529823 129407805738112 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0417 13:20:14.530119 129407805738112 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/9. I0417 13:20:14.707181 129407805738112 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0417 13:20:14.707321 129293898221120 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/9 I0417 13:20:14.707404 129407805738112 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 I0417 13:20:15.376088 129293734643264 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/9/items I0417 13:20:19.495447 129407805738112 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 4.788988s I0417 13:20:19.500431 129407805738112 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.527 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 4 seconds) (per-host) I0417 13:20:19.500493 129407805738112 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 4.883645s (batch_requests_ready=0.085973s, total_serialization_initiated=4.792792s, others=0.004880s) I0417 13:20:19.500567 129407805738112 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 4.884159s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=4.884141) I0417 13:20:19.501992 129293766100544 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0417 13:20:19.502088 129407805738112 async_checkpointer.py:561] Finished blocking save. Time taken: 4.972218s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/9. I0417 13:20:19.502283 129407805738112 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0417 13:20:19.502478 129293787072064 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 13:20:19.502581 129407805738112 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776431994.34643, 'wait_for_prev_duration_secs': 20.18259358406067, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776432014.5298507, 'checkpointer_blocking_duration_secs': 4.972344398498535, 'get_old_steps_start_time': 1776432019.502211, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776431994.3463893, 'checkpoint_manager_blocking_duration_secs': 25.156163692474365} I0417 13:20:19.502710 129407805738112 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0417 13:20:19.502745 129407805738112 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 13:20:20.227172 129293808043584 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/9/items/array_metadatas/process_0 I0417 13:20:52.399420 129293734643264 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 32.898505s (commit=32.442121s, array_metadata_write=0.456384s) I0417 13:20:52.400369 129293766100544 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 334.452 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 37 seconds) (per-host) I0417 13:20:52.400433 129293766100544 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 32.898296s. I0417 13:20:52.825963 129293766100544 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0417 13:20:53.251430 129293766100544 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0417 13:20:53.251619 129293766100544 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.559212s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/9/items I0417 13:20:53.252278 129293766100544 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/9/items I0417 13:20:53.499067 129293766100544 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/9 I0417 13:20:54.200431 129293766100544 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/9`. I0417 13:20:54.201070 129293766100544 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 39.671206s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_09_pdb_lt_1/checkpoints/9 I0417 13:20:54.201147 129293766100544 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 34.699014s. I0417 13:20:54.201314 129293787072064 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 13:20:54.201421 129293787072064 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 13:20:54.201473 129293787072064 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0417 13:20:54.201513 129293787072064 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0417 13:20:54.201746 129407805738112 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. I0417 13:20:54.202221 129407805738112 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 13:20:54.202910 129407805738112 metric_logger.py:196] completed step: 9, seconds: 0.052, TFLOP/s/device: 65.096, Tokens/s/device: 9812.001, total_weights: 3004, loss: 8.888, lm_loss: 8.888, perplexity: 7244.713 Per train step: Total TFLOPs: 3.40 split as 93.93% learnable weight flops and 6.07% attention flops