2026-04-22 18:57:09.816613: 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 00:00:1776884229.829570 2090796 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776884229.833385 2090796 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776884229.844853 2090796 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776884229.844870 2090796 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776884229.844872 2090796 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776884229.844873 2090796 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'} 2026-04-22 18:57:13.757344: 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) I0422 18:57:13.893701 137400093873984 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0422 18:57:43.958764 137400093873984 max_utils.py:800] System Information: Jax Version: 0.8.1 I0422 18:57:43.958880 137400093873984 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0422 18:57:43.958915 137400093873984 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 I0422 18:57:43.958941 137400093873984 train_utils.py:348] WARNING: 'dataset_path' might be pointing your local file system I0422 18:57:43.958965 137400093873984 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 18:57:43.959081 137400093873984 train.py:705] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0422 18:57:44.638863 137400093873984 maxtext_utils.py:1604] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 18:57:44.639151 137400093873984 checkpointing.py:677] Setting up checkpoint logger... I0422 18:57:44.639200 137400093873984 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 18:57:44.639234 137400093873984 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0422 18:57:44.639669 137400093873984 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 0x7cf6af7e3cb0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 18:57:47.027481 137400093873984 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0422 18:57:47.028714 137400093873984 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 0x7cf4dfb19ac0>}, handler_registry=None I0422 18:57:47.029109 137400093873984 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cf4dfb19ac0>` 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`. I0422 18:57:47.029159 137400093873984 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cf4dfb19f10>` 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`. I0422 18:57:47.029192 137400093873984 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 0x7cf4dfb19ac0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7cf4dfb19ac0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cf4dfb19f10>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cf4dfb19f10>}). I0422 18:57:47.029492 137400093873984 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0422 18:57:47.029549 137400093873984 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7cf4dfb30900> timeout: 600 secs and primary_host=0 for async checkpoint writes I0422 18:57:47.145452 137400093873984 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints I0422 18:57:47.145686 137400093873984 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_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7cf4e40d1ac0> I0422 18:57:47.145961 137400093873984 checkpointing.py:301] Checkpoint manager created! I0422 18:57:47.540274 137400093873984 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 18:57:47.540375 137400093873984 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0422 18:57:47.822680 137400093873984 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 18:57:47.822772 137400093873984 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0422 18:57:47.834983 137400093873984 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 18:57:47.835034 137400093873984 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0422 18:57:47.854196 137400093873984 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 18:57:47.854261 137400093873984 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0422 18:57:47.866600 137400093873984 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 18:57:47.866654 137400093873984 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0422 18:57:47.879835 137400093873984 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 18:57:47.879889 137400093873984 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0422 18:57:47.899387 137400093873984 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 18:57:47.899450 137400093873984 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0422 18:57:47.914899 137400093873984 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 18:57:47.914956 137400093873984 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0422 18:57:48.066380 137400093873984 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0422 18:57:48.066481 137400093873984 checkpointing.py:665] 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 I0422 18:57:48.693320 137400093873984 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0422 18:57:48.693430 137400093873984 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 18:57:48.693473 137400093873984 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 18:57:48.693517 137400093873984 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 18:57:48.693556 137400093873984 maxtext_utils.py:1707] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0422 18:57:48.693588 137400093873984 maxtext_utils.py:1707] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0422 18:57:48.693629 137400093873984 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 18:57:48.693668 137400093873984 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: PartitionSpec('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0422 18:57:48.693696 137400093873984 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0422 18:57:48.693723 137400093873984 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 18:57:48.693759 137400093873984 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 18:57:48.693793 137400093873984 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 18:57:49.101890 137400093873984 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0422 18:57:49.101988 137400093873984 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0422 18:57:49.113744 137400093873984 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0422 18:57:49.113793 137400093873984 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0422 18:57:49.884918 137400093873984 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. I0422 18:57:49.885493 137400093873984 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0422 18:57:49.885649 137400093873984 metric_logger.py:301] number parameters: 1.104 billion I0422 18:57:52.366340 137400093873984 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0422 18:57:52.484273 137400093873984 checkpointing.py:776] Waited 0.11790800094604492 seconds for step 0 to finish before starting checkpointing. I0422 18:57:52.484918 137400093873984 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 18:57:52.485085 137400093873984 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0422 18:57:52.485486 137400093873984 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/0. I0422 18:57:52.572033 137400093873984 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0422 18:57:52.657634 137369080563264 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/0 I0422 18:57:52.663664 137400093873984 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0422 18:57:52.663769 137400093873984 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 I0422 18:57:53.396307 137369070077504 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/0/items I0422 18:57:53.583338 137369038620224 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776884273318634917, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776884273318634917, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0422 18:57:53.888175 2092885 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0422 18:57:53.888236 2092885 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0422 18:57:57.180532 137400093873984 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 4.517314s I0422 18:57:57.185343 137400093873984 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.676 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 4 seconds) (per-host) I0422 18:57:57.185412 137400093873984 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 4.612291s (batch_requests_ready=0.086200s, total_serialization_initiated=4.521402s, others=0.004689s) I0422 18:57:57.185493 137400093873984 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 4.612904s (all_items=0.000020s, per_item={'items': '0.00001979'}, temp_paths=4.612884) I0422 18:57:57.186996 137368250091072 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0422 18:57:57.187095 137400093873984 async_checkpointer.py:561] Finished blocking save. Time taken: 4.701969s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/0. I0422 18:57:57.187287 137400093873984 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 18:57:57.187493 137369091049024 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0422 18:57:57.187609 137400093873984 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776884272.484905, 'wait_for_prev_duration_secs': 4.2438507080078125e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776884272.4851072, 'checkpointer_blocking_duration_secs': 4.702098846435547, 'get_old_steps_start_time': 1776884277.187223, 'get_old_steps_duration_secs': 2.86102294921875e-05, 'checkpoint_manager_blocking_start_time': 1776884272.484823, 'checkpoint_manager_blocking_duration_secs': 4.702755928039551} I0422 18:57:57.187723 137400093873984 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0422 18:57:57.187765 137400093873984 max_utils.py:750] Memstats: After params initialized: I0422 18:57:57.187814 137400093873984 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0422 18:57:57.187839 137400093873984 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0422 18:57:57.187859 137400093873984 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0422 18:57:57.187876 137400093873984 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0422 18:57:57.187893 137400093873984 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0422 18:57:57.187909 137400093873984 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0422 18:57:57.187925 137400093873984 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0422 18:57:57.187941 137400093873984 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0422 18:57:57.443594 137400093873984 metric_logger.py:196] completed step: 0, seconds: 2.480, TFLOP/s/device: 5.478, Tokens/s/device: 825.757, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0422 18:57:57.444472 137400093873984 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/tensorboard/' I0422 18:57:57.560437 137400093873984 metric_logger.py:196] completed step: 1, seconds: 5.075, TFLOP/s/device: 2.677, Tokens/s/device: 403.528, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0422 18:57:57.679279 137400093873984 metric_logger.py:196] completed step: 2, seconds: 0.006, TFLOP/s/device: 2149.182, Tokens/s/device: 323948.118, total_weights: 16384, loss: 10.276, lm_loss: 10.276, perplexity: 29019.973 I0422 18:57:57.798186 137400093873984 metric_logger.py:196] completed step: 3, seconds: 0.118, TFLOP/s/device: 115.556, Tokens/s/device: 17417.928, total_weights: 16384, loss: 9.282, lm_loss: 9.282, perplexity: 10739.224 I0422 18:57:57.803438 137400093873984 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0422 18:57:57.938124 137368642258496 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/0/items/array_metadatas/process_0 I0422 18:57:58.035342 137400093873984 checkpointing.py:776] Waited 0.23186874389648438 seconds for step 5 to finish before starting checkpointing. I0422 18:57:58.036315 137400093873984 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 18:58:31.744459 137368451417664 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 34.558147s (commit=34.085743s, array_metadata_write=0.472404s) I0422 18:58:31.745655 137368250091072 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 322.593 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 39 seconds) (per-host) I0422 18:58:31.745709 137368250091072 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 34.558569s. I0422 18:58:32.026832 137368250091072 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776884273318634917, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0422 18:58:32.250654 137368250091072 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0422 18:58:32.455571 137369038620224 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776884273318634917, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0422 18:58:32.650259 137368250091072 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0422 18:58:32.650446 137368250091072 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.586297s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/0/items I0422 18:58:32.651272 137368250091072 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/0/items I0422 18:58:32.908816 137368250091072 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/0 I0422 18:58:33.628684 137368250091072 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/0`. I0422 18:58:33.629432 137368250091072 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 41.144311s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/0 I0422 18:58:33.629515 137368250091072 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 36.442377s. I0422 18:58:33.629780 137369091049024 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 18:58:33.629904 137369091049024 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0422 18:58:33.629976 137369091049024 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 18:58:33.630026 137369091049024 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 18:58:33.630198 137400093873984 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. W0422 18:58:33.630342 137400093873984 checkpoint_manager.py:1441] Waiting for previous save to complete took 35.594023 seconds. If this number is high, consider checkpointing less frequently. I0422 18:58:33.631514 137400093873984 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 5 I0422 18:58:33.631829 137400093873984 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/5. I0422 18:58:33.805166 137400093873984 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0422 18:58:33.805265 137400093873984 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 I0422 18:58:33.806981 137368250091072 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/5 I0422 18:58:34.497097 137368642258496 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/5/items I0422 18:58:36.972234 137400093873984 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 3.167576s I0422 18:58:36.976435 137400093873984 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.783 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 3 seconds) (per-host) I0422 18:58:36.976511 137400093873984 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 3.262209s (batch_requests_ready=0.086846s, total_serialization_initiated=3.171275s, others=0.004088s) I0422 18:58:36.976604 137400093873984 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 3.262844s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=3.262827) I0422 18:58:36.978456 137368250091072 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0422 18:58:36.978541 137400093873984 async_checkpointer.py:561] Finished blocking save. Time taken: 3.346985s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/5. I0422 18:58:36.978737 137400093873984 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0422 18:58:36.978934 137369091049024 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0422 18:58:36.979073 137400093873984 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776884278.0362916, 'wait_for_prev_duration_secs': 35.59402322769165, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776884313.6315382, 'checkpointer_blocking_duration_secs': 3.3471083641052246, 'get_old_steps_start_time': 1776884316.9786663, 'get_old_steps_duration_secs': 3.075599670410156e-05, 'checkpoint_manager_blocking_start_time': 1776884278.036224, 'checkpoint_manager_blocking_duration_secs': 38.94281053543091} I0422 18:58:36.979212 137400093873984 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0422 18:58:36.979888 137400093873984 metric_logger.py:196] completed step: 4, seconds: 0.119, TFLOP/s/device: 114.246, Tokens/s/device: 17220.358, total_weights: 16384, loss: 8.327, lm_loss: 8.327, perplexity: 4134.812 I0422 18:58:36.988822 137400093873984 metric_logger.py:196] completed step: 5, seconds: 0.119, TFLOP/s/device: 114.347, Tokens/s/device: 17235.575, total_weights: 16384, loss: 7.367, lm_loss: 7.367, perplexity: 1583.522 I0422 18:58:37.105394 137400093873984 metric_logger.py:196] completed step: 6, seconds: 39.183, TFLOP/s/device: 0.347, Tokens/s/device: 52.267, total_weights: 16384, loss: 6.459, lm_loss: 6.459, perplexity: 638.363 I0422 18:58:37.224336 137400093873984 metric_logger.py:196] completed step: 7, seconds: 0.006, TFLOP/s/device: 2193.241, Tokens/s/device: 330589.185, total_weights: 16384, loss: 5.622, lm_loss: 5.622, perplexity: 276.577 I0422 18:58:37.343431 137400093873984 metric_logger.py:196] completed step: 8, seconds: 0.118, TFLOP/s/device: 115.141, Tokens/s/device: 17355.344, total_weights: 16384, loss: 4.864, lm_loss: 4.864, perplexity: 129.523 I0422 18:58:37.461609 137400093873984 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0422 18:58:37.462725 137400093873984 checkpointing.py:776] Waited 0.0011446475982666016 seconds for step 9 to finish before starting checkpointing. I0422 18:58:37.463003 137400093873984 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 18:58:37.711487 137368889722432 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/5/items/array_metadatas/process_0 I0422 18:59:17.376584 137368642258496 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 40.399199s (commit=39.903176s, array_metadata_write=0.496023s) I0422 18:59:17.381273 137368250091072 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 289.390 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 43 seconds) (per-host) I0422 18:59:17.381332 137368250091072 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.402747s. I0422 18:59:17.897120 137368250091072 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0422 18:59:18.315005 137368250091072 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0422 18:59:18.315255 137368250091072 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.594330s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/5/items I0422 18:59:18.316080 137368250091072 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/5/items I0422 18:59:18.555677 137368250091072 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/5 I0422 18:59:19.272927 137368250091072 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/5`. I0422 18:59:19.273650 137368250091072 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 45.642100s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/5 I0422 18:59:19.273723 137368250091072 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 42.295138s. I0422 18:59:19.273959 137369091049024 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 18:59:19.274101 137369091049024 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0422 18:59:19.274170 137369091049024 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0422 18:59:19.274217 137369091049024 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0422 18:59:19.274329 137400093873984 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. W0422 18:59:19.274408 137400093873984 checkpoint_manager.py:1441] Waiting for previous save to complete took 41.811415 seconds. If this number is high, consider checkpointing less frequently. I0422 18:59:19.274833 137400093873984 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0422 18:59:19.275162 137400093873984 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/9. I0422 18:59:19.439707 137368250091072 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/9 I0422 18:59:19.449020 137400093873984 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0422 18:59:19.449136 137400093873984 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 I0422 18:59:20.175935 137368889722432 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/9/items I0422 18:59:25.675419 137400093873984 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 6.226763s I0422 18:59:25.679813 137400093873984 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.952 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host) I0422 18:59:25.679876 137400093873984 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 6.322969s (batch_requests_ready=0.088046s, total_serialization_initiated=6.230651s, others=0.004273s) I0422 18:59:25.679949 137400093873984 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 6.323546s (all_items=0.000014s, per_item={'items': '0.00001407'}, temp_paths=6.323532) I0422 18:59:25.681828 137369091049024 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0422 18:59:25.681914 137400093873984 async_checkpointer.py:561] Finished blocking save. Time taken: 6.407038s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/9. I0422 18:59:25.682110 137400093873984 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 18:59:25.682257 137368250091072 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0422 18:59:25.682354 137400093873984 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776884317.4629767, 'wait_for_prev_duration_secs': 41.81141495704651, 'time_between_consecutive_saves_sec': 3.832937002182007, 'checkpointer_blocking_start_time': 1776884359.2748587, 'checkpointer_blocking_duration_secs': 6.40716552734375, 'get_old_steps_start_time': 1776884365.6820376, 'get_old_steps_duration_secs': 3.838539123535156e-05, 'checkpoint_manager_blocking_start_time': 1776884317.462938, 'checkpoint_manager_blocking_duration_secs': 48.21938490867615} I0422 18:59:25.682461 137400093873984 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0422 18:59:25.682500 137400093873984 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 18:59:26.408112 137368879236672 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/9/items/array_metadatas/process_0 I0422 18:59:57.196778 137368889722432 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 31.516269s (commit=31.041089s, array_metadata_write=0.475180s) I0422 18:59:57.197724 137369091049024 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 333.946 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 37 seconds) (per-host) I0422 18:59:57.197777 137369091049024 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 31.515820s. I0422 18:59:57.708720 137369091049024 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0422 18:59:58.161700 137369091049024 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0422 18:59:58.161881 137369091049024 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.624590s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/9/items I0422 18:59:58.162619 137369091049024 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/9/items I0422 18:59:58.416893 137369091049024 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/9 I0422 18:59:59.117873 137369091049024 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/9`. I0422 18:59:59.118609 137369091049024 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 39.843739s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260422_174509/linen_main_20260422_174509_16_zarr3_true/checkpoints/9 I0422 18:59:59.118670 137369091049024 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 33.436714s. I0422 18:59:59.118783 137368250091072 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 18:59:59.118831 137368250091072 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0422 18:59:59.118893 137368250091072 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 18:59:59.118930 137368250091072 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 18:59:59.119084 137400093873984 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. I0422 18:59:59.119235 137400093873984 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 18:59:59.120182 137400093873984 metric_logger.py:196] completed step: 9, seconds: 0.119, TFLOP/s/device: 114.110, Tokens/s/device: 17199.966, total_weights: 16384, loss: 4.189, lm_loss: 4.189, perplexity: 65.979 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops