2026-04-16 12:55:56.179475: 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 12:55:56.299409 124679516093568 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0416 12:56:50.237361 124679516093568 max_utils.py:800] System Information: Jax Version: 0.9.2 I0416 12:56:50.237465 124679516093568 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0416 12:56:50.237499 124679516093568 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 12:56:50.237524 124679516093568 train_utils.py:334] WARNING: 'dataset_path' might be pointing your local file system I0416 12:56:50.237545 124679516093568 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0416 12:56:50.237640 124679516093568 train.py:683] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0416 12:56:50.328370 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 12:56:50.733000 124679516093568 maxtext_utils.py:1517] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0416 12:56:50.733273 124679516093568 checkpointing.py:677] Setting up checkpoint logger... I0416 12:56:50.733319 124679516093568 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0416 12:56:50.733362 124679516093568 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0416 12:56:50.733878 124679516093568 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 0x71649eb04740>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0416 12:56:53.072395 124679516093568 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0416 12:56:53.072568 124679516093568 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 0x715e2333a060>}, handler_registry=None I0416 12:56:53.072822 124679516093568 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x715e2333a060>` 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 12:56:53.072864 124679516093568 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x715e23338c50>` 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 12:56:53.072893 124679516093568 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 0x715e2333a060>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x715e2333a060>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x715e23338c50>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x715e23338c50>}). I0416 12:56:53.073201 124679516093568 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0416 12:56:53.073253 124679516093568 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x715e23331260> timeout: 600 secs and primary_host=0 for async checkpoint writes I0416 12:56:53.200447 124679516093568 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints I0416 12:56:53.200667 124679516093568 checkpoint_manager.py:921] [process=0][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_hns=False, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x715e23b205c0> I0416 12:56:53.200749 124679516093568 checkpointing.py:301] Checkpoint manager created! W0416 12:56:53.216757 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 12:56:53.395946 124679516093568 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0416 12:56:53.396060 124679516093568 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 12:56:53.485845 124679516093568 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 12:56:53.485978 124679516093568 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 12:56:53.500068 124679516093568 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0416 12:56:53.500117 124679516093568 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0416 12:56:53.520258 124679516093568 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 12:56:53.520312 124679516093568 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 12:56:53.534345 124679516093568 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 12:56:53.534392 124679516093568 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 12:56:53.548306 124679516093568 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 12:56:53.548352 124679516093568 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 12:56:53.569711 124679516093568 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0416 12:56:53.569767 124679516093568 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0416 12:56:53.587841 124679516093568 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0416 12:56:53.587893 124679516093568 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0416 12:56:53.769802 124679516093568 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0416 12:56:53.769892 124679516093568 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. W0416 12:56:53.820258 3841531 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 12:56:54.197758 124679516093568 maxtext_utils.py:1620] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0416 12:56:54.197868 124679516093568 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0416 12:56:54.197910 124679516093568 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0416 12:56:54.197953 124679516093568 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0416 12:56:54.197992 124679516093568 maxtext_utils.py:1620] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0416 12:56:54.198026 124679516093568 maxtext_utils.py:1620] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0416 12:56:54.198083 124679516093568 maxtext_utils.py:1620] 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, None, None) I0416 12:56:54.198126 124679516093568 maxtext_utils.py:1620] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0416 12:56:54.198156 124679516093568 maxtext_utils.py:1620] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0416 12:56:54.198182 124679516093568 maxtext_utils.py:1620] 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, None, None) I0416 12:56:54.198216 124679516093568 maxtext_utils.py:1620] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0416 12:56:54.198250 124679516093568 maxtext_utils.py:1620] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0416 12:56:54.616008 124679516093568 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 12:56:54.616097 124679516093568 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0416 12:56:54.629337 124679516093568 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0416 12:56:54.629384 124679516093568 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). W0416 12:56:55.277653 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 12:56:55.394830 124679516093568 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 12:56:55.395362 124679516093568 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0416 12:56:55.395510 124679516093568 metric_logger.py:289] number parameters: 1.104 billion W0416 12:56:56.220409 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 12:56:56.316165 124679516093568 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0416 12:56:56.433160 124679516093568 checkpointing.py:776] Waited 0.11698651313781738 seconds for step 0 to finish before starting checkpointing. I0416 12:56:56.433597 124679516093568 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 12:56:56.433754 124679516093568 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0416 12:56:56.434171 124679516093568 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/0. I0416 12:56:56.531678 124679516093568 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0416 12:56:56.613182 124679516093568 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 12:56:56.613266 124679516093568 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 12:56:56.618891 124572047246912 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/0 W0416 12:56:56.826815 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 12:56:56.841241 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 12:56:56.861372 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 12:56:56.871749 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 12:56:56.881914 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 12:56:56.887995 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 12:56:56.892986 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 12:56:56.897686 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 12:56:57.231408 124679516093568 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.618678s I0416 12:56:57.231773 124679516093568 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 17.659 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 698 milliseconds) (per-host) I0416 12:56:57.231821 124679516093568 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.698937s (batch_requests_ready=0.076196s, total_serialization_initiated=0.622464s, others=0.000277s) I0416 12:56:57.231877 124679516093568 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.699736s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.699718) I0416 12:56:57.232577 124571984332352 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 12:56:57.232672 124679516093568 async_checkpointer.py:561] Finished blocking save. Time taken: 0.798880s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/0. I0416 12:56:57.232847 124679516093568 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0416 12:56:57.233070 124679516093568 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776344216.433583, 'wait_for_prev_duration_secs': 4.291534423828125e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776344216.4337742, 'checkpointer_blocking_duration_secs': 0.7990031242370605, 'get_old_steps_start_time': 1776344217.232794, 'get_old_steps_duration_secs': 2.0503997802734375e-05, 'checkpoint_manager_blocking_start_time': 1776344216.4334993, 'checkpoint_manager_blocking_duration_secs': 0.799534797668457} I0416 12:56:57.233144 124679516093568 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0416 12:56:57.233182 124679516093568 max_utils.py:750] Memstats: After params initialized: I0416 12:56:57.233252 124679516093568 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0416 12:56:57.233278 124679516093568 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0416 12:56:57.233297 124679516093568 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0416 12:56:57.233315 124679516093568 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0416 12:56:57.233333 124679516093568 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0416 12:56:57.233350 124679516093568 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0416 12:56:57.233367 124679516093568 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0416 12:56:57.233383 124679516093568 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0416 12:56:57.238834 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 12:56:57.243816 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0416 12:56:57.248357 3841531 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0416 12:56:57.264202 124572036761152 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/0/items I0416 12:56:57.420810 124679516093568 metric_logger.py:185] completed step: 0, seconds: 0.920, TFLOP/s/device: 14.767, Tokens/s/device: 2225.840, total_weights: 16384, loss: 10.874 I0416 12:56:57.421687 124679516093568 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/tensorboard/' I0416 12:56:57.431288 124572005303872 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776344217178442596, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776344217178442596, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/0/_CHECKPOINT_METADATA I0416 12:56:57.431464 124572057732672 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 12:56:57.537283 124679516093568 metric_logger.py:185] completed step: 1, seconds: 1.103, TFLOP/s/device: 12.318, Tokens/s/device: 1856.650, total_weights: 16384, loss: 10.874 I0416 12:56:57.655368 124679516093568 metric_logger.py:185] completed step: 2, seconds: 0.005, TFLOP/s/device: 2585.070, Tokens/s/device: 389649.924, total_weights: 16384, loss: 10.276 I0416 12:56:57.746074 3843745 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0416 12:56:57.746140 3843745 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0416 12:56:57.773370 124679516093568 metric_logger.py:185] completed step: 3, seconds: 0.117, TFLOP/s/device: 115.914, Tokens/s/device: 17471.868, total_weights: 16384, loss: 9.282 I0416 12:56:57.777743 124679516093568 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0416 12:56:58.009532 124679516093568 checkpointing.py:776] Waited 0.23175621032714844 seconds for step 5 to finish before starting checkpointing. I0416 12:56:58.009920 124679516093568 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 12:56:58.478188 124572026275392 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/0/items/array_metadatas/process_0 I0416 12:57:20.714078 124572047246912 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 22.964780s (commit=22.518839s, array_metadata_write=0.445941s) I0416 12:57:20.715116 124571984332352 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 522.566 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 24 seconds) (per-host) I0416 12:57:20.715219 124571984332352 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 23.482503s. I0416 12:57:20.956788 124571984332352 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776344217178442596, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/0/_CHECKPOINT_METADATA I0416 12:57:21.138892 124571984332352 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 12:57:21.337544 124572005303872 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776344217178442596, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/0/_CHECKPOINT_METADATA I0416 12:57:21.567165 124571984332352 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 12:57:21.567346 124571984332352 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.572069s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/0/items I0416 12:57:21.567969 124571984332352 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/0/items I0416 12:57:21.811321 124571984332352 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/0 I0416 12:57:22.452928 124571984332352 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/0`. I0416 12:57:22.453571 124571984332352 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 26.019785s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/0 I0416 12:57:22.453646 124571984332352 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 25.220932s. I0416 12:57:22.453787 124572057732672 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 12:57:22.453897 124572057732672 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 12:57:22.453962 124572057732672 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0416 12:57:22.454007 124572057732672 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0416 12:57:22.454231 124679516093568 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 12:57:22.454351 124679516093568 checkpoint_manager.py:1441] Waiting for previous save to complete took 24.444430 seconds. If this number is high, consider checkpointing less frequently. I0416 12:57:22.455099 124679516093568 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 5 I0416 12:57:22.455371 124679516093568 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/5. I0416 12:57:22.639696 124572057732672 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/5 I0416 12:57:22.646178 124679516093568 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 12:57:22.646268 124679516093568 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 12:57:23.319512 124571913029184 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/5/items I0416 12:57:35.096109 124679516093568 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 12.450335s I0416 12:57:35.100302 124679516093568 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1006.946 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 12 seconds) (per-host) I0416 12:57:35.100370 124679516093568 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 12.549714s (batch_requests_ready=0.092120s, total_serialization_initiated=12.453504s, others=0.004090s) I0416 12:57:35.100457 124679516093568 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 12.550502s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=12.550485) I0416 12:57:35.101856 124572057732672 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 12:57:35.101941 124679516093568 async_checkpointer.py:561] Finished blocking save. Time taken: 12.646794s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/5. I0416 12:57:35.102156 124679516093568 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0416 12:57:35.102320 124571984332352 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 12:57:35.102422 124679516093568 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776344218.009898, 'wait_for_prev_duration_secs': 24.444430351257324, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776344242.455125, 'checkpointer_blocking_duration_secs': 12.64693570137024, 'get_old_steps_start_time': 1776344255.102083, 'get_old_steps_duration_secs': 2.9325485229492188e-05, 'checkpoint_manager_blocking_start_time': 1776344218.009827, 'checkpoint_manager_blocking_duration_secs': 37.09256863594055} I0416 12:57:35.102558 124679516093568 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0416 12:57:35.103007 124679516093568 metric_logger.py:185] completed step: 4, seconds: 0.118, TFLOP/s/device: 115.041, Tokens/s/device: 17340.208, total_weights: 16384, loss: 8.327 I0416 12:57:35.109788 124679516093568 metric_logger.py:185] completed step: 5, seconds: 0.118, TFLOP/s/device: 115.345, Tokens/s/device: 17385.989, total_weights: 16384, loss: 7.367 I0416 12:57:35.225844 124679516093568 metric_logger.py:185] completed step: 6, seconds: 37.330, TFLOP/s/device: 0.364, Tokens/s/device: 54.861, total_weights: 16384, loss: 6.459 I0416 12:57:35.344326 124679516093568 metric_logger.py:185] completed step: 7, seconds: 0.005, TFLOP/s/device: 2617.440, Tokens/s/device: 394528.992, total_weights: 16384, loss: 5.622 I0416 12:57:35.462697 124679516093568 metric_logger.py:185] completed step: 8, seconds: 0.117, TFLOP/s/device: 116.068, Tokens/s/device: 17495.003, total_weights: 16384, loss: 4.864 I0416 12:57:35.580683 124679516093568 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0416 12:57:35.581434 124679516093568 checkpointing.py:776] Waited 0.0007710456848144531 seconds for step 9 to finish before starting checkpointing. I0416 12:57:35.581699 124679516093568 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 12:57:35.829377 124572026275392 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/5/items/array_metadatas/process_0 I0416 12:58:10.734577 124571913029184 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 35.633560s (commit=35.163168s, array_metadata_write=0.470392s) I0416 12:58:10.735653 124572057732672 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 262.256 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 48 seconds) (per-host) I0416 12:58:10.735720 124572057732672 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 35.633736s. I0416 12:58:11.163125 124572057732672 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 12:58:11.599347 124572057732672 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 12:58:11.599506 124572057732672 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.578924s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/5/items I0416 12:58:11.600131 124572057732672 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/5/items I0416 12:58:11.834872 124572057732672 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/5 I0416 12:58:12.512002 124572057732672 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/5`. I0416 12:58:12.512656 124572057732672 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 50.057517s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/5 I0416 12:58:12.512737 124572057732672 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 37.410753s. I0416 12:58:12.512921 124571984332352 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 12:58:12.513035 124571984332352 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 12:58:12.513115 124571984332352 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0416 12:58:12.513157 124571984332352 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0416 12:58:12.513887 124679516093568 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0416 12:58:12.514007 124679516093568 checkpoint_manager.py:1441] Waiting for previous save to complete took 36.932316 seconds. If this number is high, consider checkpointing less frequently. I0416 12:58:12.514235 124679516093568 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0416 12:58:12.514520 124679516093568 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/9. I0416 12:58:12.690128 124679516093568 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0416 12:58:12.690212 124679516093568 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 12:58:12.806214 124571984332352 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/9 I0416 12:58:13.518022 124571913029184 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/9/items I0416 12:58:18.646297 124679516093568 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 5.956554s I0416 12:58:18.650454 124679516093568 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.041 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host) I0416 12:58:18.650510 124679516093568 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 6.046096s (batch_requests_ready=0.081993s, total_serialization_initiated=5.960064s, others=0.004039s) I0416 12:58:18.650588 124679516093568 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 6.046829s (all_items=0.000012s, per_item={'items': '0.00001240'}, temp_paths=6.046816) I0416 12:58:18.652274 124572057732672 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0416 12:58:18.652357 124679516093568 async_checkpointer.py:561] Finished blocking save. Time taken: 6.138077s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/9. I0416 12:58:18.652523 124679516093568 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0416 12:58:18.652670 124571984332352 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0416 12:58:18.652772 124679516093568 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776344255.5816746, 'wait_for_prev_duration_secs': 36.932315826416016, 'time_between_consecutive_saves_sec': 13.127651929855347, 'checkpointer_blocking_start_time': 1776344292.5142589, 'checkpointer_blocking_duration_secs': 6.138195753097534, 'get_old_steps_start_time': 1776344298.6524677, 'get_old_steps_duration_secs': 2.1457672119140625e-05, 'checkpoint_manager_blocking_start_time': 1776344255.581642, 'checkpoint_manager_blocking_duration_secs': 43.071104764938354} I0416 12:58:18.652906 124679516093568 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0416 12:58:18.652937 124679516093568 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0416 12:58:19.416603 124572047246912 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/9/items/array_metadatas/process_0 I0416 12:58:48.882146 124571913029184 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 30.230733s (commit=29.760419s, array_metadata_write=0.470314s) I0416 12:58:48.883188 124572057732672 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 348.325 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 36 seconds) (per-host) I0416 12:58:48.883241 124572057732672 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 30.230844s. I0416 12:58:49.267569 124572057732672 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 12:58:49.670392 124572057732672 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0416 12:58:49.670560 124572057732672 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.545572s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/9/items I0416 12:58:49.671267 124572057732672 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/9/items I0416 12:58:49.908305 124572057732672 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/9 I0416 12:58:50.579279 124572057732672 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/9`. I0416 12:58:50.579916 124572057732672 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 38.065645s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260416_123638/linen_main_20260416_123638_08_checkpoint_async_true/checkpoints/9 I0416 12:58:50.579982 124572057732672 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 31.927587s. I0416 12:58:50.580142 124571984332352 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0416 12:58:50.580242 124571984332352 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0416 12:58:50.580293 124571984332352 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0416 12:58:50.580332 124571984332352 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0416 12:58:50.580474 124679516093568 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 12:58:50.580613 124679516093568 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0416 12:58:50.581679 124679516093568 metric_logger.py:185] completed step: 9, seconds: 0.118, TFLOP/s/device: 114.684, Tokens/s/device: 17286.493, total_weights: 16384, loss: 4.189 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops