2026-04-21 15:36:45.217455: 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 00:00:1776785805.228628 1527506 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776785805.232127 1527506 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776785805.243291 1527506 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776785805.243303 1527506 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776785805.243305 1527506 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776785805.243306 1527506 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-21 15:36:48.884266: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0421 15:36:49.000272 127855866530944 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0421 15:37:46.248416 127855866530944 max_utils.py:800] System Information: Jax Version: 0.8.1 I0421 15:37:46.248710 127855866530944 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0421 15:37:46.248764 127855866530944 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709 I0421 15:37:46.248798 127855866530944 train_utils.py:335] WARNING: 'dataset_path' might be pointing your local file system I0421 15:37:46.248820 127855866530944 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0421 15:37:46.249001 127855866530944 train.py:703] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0421 15:37:46.613468 127855866530944 maxtext_utils.py:1551] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 15:37:46.613847 127855866530944 checkpointing.py:677] Setting up checkpoint logger... I0421 15:37:46.613895 127855866530944 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0421 15:37:46.613990 127855866530944 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0421 15:37:46.614823 127855866530944 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 0x7448773a33e0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 15:37:49.027144 127855866530944 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0421 15:37:49.027428 127855866530944 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 0x7446cff770e0>}, handler_registry=None I0421 15:37:49.027828 127855866530944 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7446cff770e0>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0421 15:37:49.027882 127855866530944 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7446cfbe3b90>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0421 15:37:49.027913 127855866530944 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 0x7446cff770e0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7446cff770e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7446cfbe3b90>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7446cfbe3b90>}). I0421 15:37:49.028235 127855866530944 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0421 15:37:49.151414 127855866530944 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints I0421 15:37:49.151684 127855866530944 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=False, 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_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7446cfbe3a40> I0421 15:37:49.151776 127855866530944 checkpointing.py:301] Checkpoint manager created! I0421 15:37:49.803997 127855866530944 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 15:37:49.804110 127855866530944 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 15:37:50.085216 127855866530944 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 15:37:50.085295 127855866530944 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 15:37:50.097719 127855866530944 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 15:37:50.097767 127855866530944 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 15:37:50.117065 127855866530944 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 15:37:50.117120 127855866530944 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 15:37:50.129525 127855866530944 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 15:37:50.129573 127855866530944 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 15:37:50.141878 127855866530944 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 15:37:50.141926 127855866530944 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 15:37:50.161451 127855866530944 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 15:37:50.161504 127855866530944 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 15:37:50.176977 127855866530944 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 15:37:50.177025 127855866530944 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0421 15:37:50.326969 127855866530944 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0421 15:37:50.327080 127855866530944 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 I0421 15:37:50.896449 127855866530944 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0421 15:37:50.896555 127855866530944 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0421 15:37:50.896605 127855866530944 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0421 15:37:50.896651 127855866530944 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0421 15:37:50.896690 127855866530944 maxtext_utils.py:1654] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0421 15:37:50.896718 127855866530944 maxtext_utils.py:1654] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0421 15:37:50.896757 127855866530944 maxtext_utils.py:1654] 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) I0421 15:37:50.896797 127855866530944 maxtext_utils.py:1654] 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') I0421 15:37:50.896825 127855866530944 maxtext_utils.py:1654] 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) I0421 15:37:50.896851 127855866530944 maxtext_utils.py:1654] 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) I0421 15:37:50.896885 127855866530944 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0421 15:37:50.896919 127855866530944 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0421 15:37:51.303819 127855866530944 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 15:37:51.303900 127855866530944 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0421 15:37:51.315625 127855866530944 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 15:37:51.315676 127855866530944 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0421 15:37:52.056086 127855866530944 max_utils.py:791] Total memory size: 3.6 GB, Output size: 1.5 GB, Temp size: 2.0 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0421 15:37:52.056632 127855866530944 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0421 15:37:52.056787 127855866530944 metric_logger.py:301] number parameters: 1.104 billion I0421 15:37:54.151057 127855866530944 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0421 15:37:54.268488 127855866530944 checkpointing.py:776] Waited 0.11743521690368652 seconds for step 0 to finish before starting checkpointing. I0421 15:37:54.268935 127855866530944 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 15:37:54.269114 127855866530944 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0421 15:37:54.269173 127855866530944 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/0. I0421 15:37:54.438445 127855866530944 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/0 I0421 15:37:55.267272 127855866530944 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776785875023187510, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776785875023187510, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA I0421 15:37:55.268365 127855866530944 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient W0421 15:37:55.268715 127855866530944 future.py:328] [process=0][thread=MainThread][operation_id=1] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0421 15:37:55.352931 127771711899200 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/0/items I0421 15:37:55.937271 127855866530944 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 15:37:55.937382 127855866530944 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0421 15:37:55.955884 1529308 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0421 15:37:55.955941 1529308 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0421 15:37:56.676824 127855866530944 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.739904s I0421 15:37:56.681341 127855866530944 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 14.430 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 855 milliseconds) (per-host) I0421 15:37:56.681402 127855866530944 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.855325s (batch_requests_ready=0.107482s, total_serialization_initiated=0.743441s, others=0.004402s) I0421 15:37:56.681483 127855866530944 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.413186s (all_items=0.000035s, per_item={'items': '0.00003481'}, temp_paths=1.413151) W0421 15:37:56.681522 127855866530944 future.py:328] [process=0][thread=MainThread][operation_id=1] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0421 15:37:57.455479 127771722384960 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/0/items/array_metadatas/process_0 I0421 15:38:34.553513 127771634304576 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 37.871766s (commit=37.393283s, array_metadata_write=0.478483s) I0421 15:38:34.554569 127855866530944 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 326.292 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 38 seconds) (per-host) I0421 15:38:34.806629 127855866530944 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776785875023187510, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA I0421 15:38:35.193246 127855866530944 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776785875023187510, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA I0421 15:38:35.332497 127855866530944 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0421 15:38:35.778204 127855866530944 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 15:38:35.778386 127855866530944 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.584774s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/0/items I0421 15:38:35.779098 127855866530944 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/0/items I0421 15:38:36.026362 127855866530944 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/0 I0421 15:38:36.731744 127855866530944 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/0`. I0421 15:38:36.732413 127855866530944 checkpointer.py:290] Finished synchronous save in 42.46 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/0 I0421 15:38:36.732557 127855866530944 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0421 15:38:36.732598 127855866530944 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=0] CheckpointManager Save Finalize is done on all hosts. I0421 15:38:36.732620 127855866530944 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=0] Finished synchronous save. I0421 15:38:36.732654 127855866530944 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776785874.2689226, 'wait_for_prev_duration_secs': 4.029273986816406e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776785874.269136, 'checkpointer_blocking_duration_secs': 42.463321685791016, 'get_old_steps_start_time': 1776785916.7324758, 'get_old_steps_duration_secs': 2.2172927856445312e-05, 'checkpoint_manager_blocking_start_time': 1776785874.268839, 'checkpoint_manager_blocking_duration_secs': 42.463794231414795} I0421 15:38:36.732726 127855866530944 checkpointing.py:410] Saved a checkpoint at step 0. I0421 15:38:36.732757 127855866530944 max_utils.py:750] Memstats: After params initialized: I0421 15:38:36.732806 127855866530944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0421 15:38:36.732830 127855866530944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0421 15:38:36.732851 127855866530944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0421 15:38:36.732877 127855866530944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0421 15:38:36.732897 127855866530944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0421 15:38:36.732914 127855866530944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0421 15:38:36.732932 127855866530944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0421 15:38:36.732949 127855866530944 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0421 15:38:36.999670 127855866530944 metric_logger.py:196] completed step: 0, seconds: 2.094, TFLOP/s/device: 6.490, Tokens/s/device: 978.171, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0421 15:38:37.000584 127855866530944 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/tensorboard/' I0421 15:38:37.115485 127855866530944 metric_logger.py:196] completed step: 1, seconds: 42.845, TFLOP/s/device: 0.317, Tokens/s/device: 47.800, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0421 15:38:37.234341 127855866530944 metric_logger.py:196] completed step: 2, seconds: 0.007, TFLOP/s/device: 1880.051, Tokens/s/device: 283381.763, total_weights: 16384, loss: 10.276, lm_loss: 10.276, perplexity: 29019.973 I0421 15:38:37.353788 127855866530944 metric_logger.py:196] completed step: 3, seconds: 0.116, TFLOP/s/device: 116.903, Tokens/s/device: 17620.842, total_weights: 16384, loss: 9.282, lm_loss: 9.282, perplexity: 10739.224 I0421 15:38:37.358793 127855866530944 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0421 15:38:37.590447 127855866530944 checkpointing.py:776] Waited 0.23162007331848145 seconds for step 5 to finish before starting checkpointing. I0421 15:38:37.590846 127855866530944 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 15:38:37.590991 127855866530944 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 5 I0421 15:38:37.591062 127855866530944 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/5. I0421 15:38:37.763756 127855866530944 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/5 W0421 15:38:38.594578 127855866530944 future.py:328] [process=0][thread=MainThread][operation_id=2] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0421 15:38:38.676606 127771711899200 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/5/items I0421 15:38:39.239777 127855866530944 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 15:38:39.239884 127855866530944 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0421 15:38:39.945991 127855866530944 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.706563s I0421 15:38:39.950137 127855866530944 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 15.463 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 798 milliseconds) (per-host) I0421 15:38:39.950206 127855866530944 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.798211s (batch_requests_ready=0.084216s, total_serialization_initiated=0.709965s, others=0.004030s) I0421 15:38:39.950301 127855866530944 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.356038s (all_items=0.000029s, per_item={'items': '0.00002933'}, temp_paths=1.356009) W0421 15:38:39.950338 127855866530944 future.py:328] [process=0][thread=MainThread][operation_id=2] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0421 15:38:40.716163 127771634304576 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/5/items/array_metadatas/process_0 I0421 15:39:14.206631 127771623818816 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 34.255911s (commit=33.771489s, array_metadata_write=0.484422s) I0421 15:39:14.207942 127855866530944 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 360.475 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 35 seconds) (per-host) I0421 15:39:14.944647 127855866530944 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0421 15:39:15.355169 127855866530944 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 15:39:15.355743 127855866530944 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.547873s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/5/items I0421 15:39:15.356407 127855866530944 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/5/items I0421 15:39:15.604903 127855866530944 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/5 I0421 15:39:16.302778 127855866530944 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/5`. I0421 15:39:16.303449 127855866530944 checkpointer.py:290] Finished synchronous save in 38.71 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/5 I0421 15:39:16.303622 127855866530944 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0421 15:39:16.303665 127855866530944 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=5] CheckpointManager Save Finalize is done on all hosts. I0421 15:39:16.303688 127855866530944 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=5] Finished synchronous save. I0421 15:39:16.303725 127855866530944 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776785917.5908294, 'wait_for_prev_duration_secs': 4.291534423828125e-05, 'time_between_consecutive_saves_sec': 0.8582167625427246, 'checkpointer_blocking_start_time': 1776785917.5910108, 'checkpointer_blocking_duration_secs': 38.71248722076416, 'get_old_steps_start_time': 1776785956.3035202, 'get_old_steps_duration_secs': 3.933906555175781e-05, 'checkpoint_manager_blocking_start_time': 1776785917.590765, 'checkpoint_manager_blocking_duration_secs': 38.71293807029724} I0421 15:39:16.303815 127855866530944 checkpointing.py:410] Saved a checkpoint at step 5. I0421 15:39:16.304798 127855866530944 metric_logger.py:196] completed step: 4, seconds: 0.119, TFLOP/s/device: 114.344, Tokens/s/device: 17235.140, total_weights: 16384, loss: 8.327, lm_loss: 8.327, perplexity: 4134.812 I0421 15:39:16.315318 127855866530944 metric_logger.py:196] completed step: 5, seconds: 0.120, TFLOP/s/device: 113.308, Tokens/s/device: 17079.049, total_weights: 16384, loss: 7.367, lm_loss: 7.367, perplexity: 1583.522 I0421 15:39:16.429930 127855866530944 metric_logger.py:196] completed step: 6, seconds: 38.953, TFLOP/s/device: 0.349, Tokens/s/device: 52.576, total_weights: 16384, loss: 6.459, lm_loss: 6.459, perplexity: 638.363 I0421 15:39:16.548462 127855866530944 metric_logger.py:196] completed step: 7, seconds: 0.007, TFLOP/s/device: 1858.450, Tokens/s/device: 280125.838, total_weights: 16384, loss: 5.622, lm_loss: 5.622, perplexity: 276.577 I0421 15:39:16.667217 127855866530944 metric_logger.py:196] completed step: 8, seconds: 0.116, TFLOP/s/device: 117.478, Tokens/s/device: 17707.532, total_weights: 16384, loss: 4.864, lm_loss: 4.864, perplexity: 129.523 I0421 15:39:16.785141 127855866530944 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0421 15:39:16.785943 127855866530944 checkpointing.py:776] Waited 0.0008237361907958984 seconds for step 9 to finish before starting checkpointing. I0421 15:39:16.786203 127855866530944 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 15:39:16.786347 127855866530944 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0421 15:39:16.786401 127855866530944 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/9. I0421 15:39:16.946606 127855866530944 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/9 W0421 15:39:17.799010 127855866530944 future.py:328] [process=0][thread=MainThread][operation_id=3] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0421 15:39:17.877913 127771711899200 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/9/items I0421 15:39:18.428062 127855866530944 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 15:39:18.428173 127855866530944 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0421 15:39:20.351020 127855866530944 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.923323s I0421 15:39:20.355542 127855866530944 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 6.137 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 2 seconds) (per-host) I0421 15:39:20.355602 127855866530944 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 2.010878s (batch_requests_ready=0.080073s, total_serialization_initiated=1.926414s, others=0.004391s) I0421 15:39:20.355674 127855866530944 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.556987s (all_items=0.000032s, per_item={'items': '0.00003242'}, temp_paths=2.556954) W0421 15:39:20.355707 127855866530944 future.py:328] [process=0][thread=MainThread][operation_id=3] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0421 15:39:21.123962 127771659470400 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/9/items/array_metadatas/process_0 I0421 15:39:57.300796 127771623818816 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 36.944855s (commit=36.440479s, array_metadata_write=0.504376s) I0421 15:39:57.301922 127855866530944 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 324.376 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 38 seconds) (per-host) I0421 15:39:58.030581 127855866530944 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0421 15:39:58.471614 127855866530944 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 15:39:58.471777 127855866530944 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.571059s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/9/items I0421 15:39:58.472488 127855866530944 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/9/items I0421 15:39:58.726723 127855866530944 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/9 I0421 15:39:59.422736 127855866530944 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/9`. I0421 15:39:59.423408 127855866530944 checkpointer.py:290] Finished synchronous save in 42.64 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints/9 I0421 15:39:59.423583 127855866530944 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0421 15:39:59.423625 127855866530944 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=9] CheckpointManager Save Finalize is done on all hosts. I0421 15:39:59.423650 127855866530944 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=9] Finished synchronous save. I0421 15:39:59.423688 127855866530944 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776785956.7861834, 'wait_for_prev_duration_secs': 4.673004150390625e-05, 'time_between_consecutive_saves_sec': 0.48250365257263184, 'checkpointer_blocking_start_time': 1776785956.7863677, 'checkpointer_blocking_duration_secs': 42.63709831237793, 'get_old_steps_start_time': 1776785999.4234886, 'get_old_steps_duration_secs': 3.24249267578125e-05, 'checkpoint_manager_blocking_start_time': 1776785956.7861497, 'checkpoint_manager_blocking_duration_secs': 42.637513875961304} I0421 15:39:59.423812 127855866530944 checkpointing.py:410] Saved a checkpoint at step 9. I0421 15:39:59.423840 127855866530944 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 15:39:59.423874 127855866530944 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 15:39:59.424973 127855866530944 metric_logger.py:196] completed step: 9, seconds: 0.119, TFLOP/s/device: 114.628, Tokens/s/device: 17278.035, 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