main (b341212ce) vs main (172d0f10a)| Metric | main b341212ce | main 172d0f10a | Diff (main − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 5.9150 | 5.9150 | 0 |
| TFLOP/s | 113.087 | 138.592 | +25.505 |
| Tok/s | 17045.6 | 20890.1 | +3844.428 |
| Avg s/step | 1.869 | 2.162 | +0.293 |
| Memory % | 5.09 | 5.09 | 0 |
| JAX | 0.8.1 | 0.8.1 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
2026-04-21 15:34:17.842277: 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 00:00:1776785657.852425 1525000 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776785657.855515 1525000 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776785657.865507 1525000 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776785657.865520 1525000 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776785657.865521 1525000 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776785657.865523 1525000 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-21 15:34:21.280269: 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 15:34:21.395263 134565803121792 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0421 15:34:49.072427 134565803121792 max_utils.py:800] System Information: Jax Version: 0.8.1 I0421 15:34:49.072709 134565803121792 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0421 15:34:49.072765 134565803121792 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 15:34:49.072799 134565803121792 train_utils.py:335] WARNING: 'dataset_path' might be pointing your local file system I0421 15:34:49.072827 134565803121792 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0421 15:34:49.072983 134565803121792 train.py:703] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0421 15:34:49.487452 134565803121792 maxtext_utils.py:1551] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 15:34:49.487821 134565803121792 checkpointing.py:677] Setting up checkpoint logger... I0421 15:34:49.487868 134565803121792 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0421 15:34:49.487962 134565803121792 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0421 15:34:49.488808 134565803121792 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 0x7a62be9a5df0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 15:34:51.875835 134565803121792 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0421 15:34:51.876134 134565803121792 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 0x7a6117d42840>}, handler_registry=None I0421 15:34:51.876512 134565803121792 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a6117d42840>` 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 15:34:51.876556 134565803121792 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a6116d82c00>` 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 15:34:51.876586 134565803121792 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 0x7a6117d42840>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a6117d42840>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a6116d82c00>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a6116d82c00>}). I0421 15:34:51.876874 134565803121792 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0421 15:34:51.876937 134565803121792 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7a6116da51c0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0421 15:34:51.997241 134565803121792 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints I0421 15:34:51.997511 134565803121792 checkpoint_manager.py:921] [process=0][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_hns=False, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7a6116d81910> I0421 15:34:51.997615 134565803121792 checkpointing.py:301] Checkpoint manager created! I0421 15:34:52.645408 134565803121792 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 15:34:52.645501 134565803121792 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 15:34:52.925575 134565803121792 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 15:34:52.925661 134565803121792 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 15:34:52.937516 134565803121792 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 15:34:52.937573 134565803121792 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 15:34:52.956252 134565803121792 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 15:34:52.956309 134565803121792 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 15:34:52.968248 134565803121792 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 15:34:52.968299 134565803121792 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 15:34:52.980201 134565803121792 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 15:34:52.980251 134565803121792 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 15:34:52.999208 134565803121792 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 15:34:52.999267 134565803121792 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 15:34:53.014330 134565803121792 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 15:34:53.014383 134565803121792 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0421 15:34:53.161458 134565803121792 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0421 15:34:53.161562 134565803121792 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 15:34:53.749204 134565803121792 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0421 15:34:53.749315 134565803121792 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 15:34:53.749357 134565803121792 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 15:34:53.749402 134565803121792 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 15:34:53.749442 134565803121792 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 15:34:53.749469 134565803121792 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 15:34:53.749509 134565803121792 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 15:34:53.749550 134565803121792 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 15:34:53.749578 134565803121792 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 15:34:53.749604 134565803121792 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 15:34:53.749638 134565803121792 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0421 15:34:53.749672 134565803121792 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0421 15:34:54.154502 134565803121792 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 15:34:54.154593 134565803121792 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0421 15:34:54.165787 134565803121792 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 15:34:54.165837 134565803121792 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0421 15:34:54.899113 134565803121792 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 15:34:54.899637 134565803121792 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0421 15:34:54.899790 134565803121792 metric_logger.py:301] number parameters: 1.104 billion I0421 15:34:57.579752 134565803121792 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0421 15:34:57.696699 134565803121792 checkpointing.py:776] Waited 0.1169283390045166 seconds for step 0 to finish before starting checkpointing. I0421 15:34:57.697190 134565803121792 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 15:34:57.697341 134565803121792 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0421 15:34:57.697762 134565803121792 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/0. I0421 15:34:57.788701 134565803121792 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0421 15:34:57.875600 134490236126784 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/0 I0421 15:34:57.885178 134565803121792 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 15:34:57.885277 134565803121792 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 15:34:58.638796 134490225641024 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/0/items I0421 15:34:58.700387 134565803121792 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.815658s I0421 15:34:58.700736 134565803121792 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 13.548 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 910 milliseconds) (per-host) I0421 15:34:58.700787 134565803121792 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.911001s (batch_requests_ready=0.091516s, total_serialization_initiated=0.819244s, others=0.000241s) I0421 15:34:58.700851 134565803121792 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.911580s (all_items=0.000021s, per_item={'items': '0.00002098'}, temp_paths=0.911559) I0421 15:34:58.701657 134490173212224 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 15:34:58.701750 134565803121792 async_checkpointer.py:561] Finished blocking save. Time taken: 1.004371s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/0. I0421 15:34:58.701938 134565803121792 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0421 15:34:58.702209 134565803121792 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776785697.6971776, 'wait_for_prev_duration_secs': 4.124641418457031e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776785697.6973617, 'checkpointer_blocking_duration_secs': 1.004504680633545, 'get_old_steps_start_time': 1776785698.7018807, 'get_old_steps_duration_secs': 2.1696090698242188e-05, 'checkpoint_manager_blocking_start_time': 1776785697.6970928, 'checkpoint_manager_blocking_duration_secs': 1.0050945281982422} I0421 15:34:58.702280 134565803121792 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0421 15:34:58.702316 134565803121792 max_utils.py:750] Memstats: After params initialized: I0421 15:34:58.702377 134565803121792 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0421 15:34:58.702403 134565803121792 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0421 15:34:58.702424 134565803121792 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0421 15:34:58.702442 134565803121792 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0421 15:34:58.702458 134565803121792 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0421 15:34:58.702475 134565803121792 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0421 15:34:58.702491 134565803121792 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0421 15:34:58.702508 134565803121792 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0421 15:34:58.798754 134490194183744 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776785698552749892, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776785698552749892, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/0/_CHECKPOINT_METADATA I0421 15:34:58.799024 134490246612544 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 15:34:59.095152 134565803121792 metric_logger.py:196] completed step: 0, seconds: 2.679, TFLOP/s/device: 5.071, Tokens/s/device: 764.344, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0421 15:34:59.096147 134565803121792 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/tensorboard/' I0421 15:34:59.130640 1526994 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0421 15:34:59.130711 1526994 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0421 15:34:59.420022 134565803121792 metric_logger.py:196] completed step: 1, seconds: 1.513, TFLOP/s/device: 8.979, Tokens/s/device: 1353.441, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0421 15:34:59.439491 134565803121792 metric_logger.py:196] completed step: 2, seconds: 0.007, TFLOP/s/device: 1873.828, Tokens/s/device: 282443.801, total_weights: 16384, loss: 9.775, lm_loss: 9.775, perplexity: 17589.287 I0421 15:34:59.557372 134565803121792 metric_logger.py:196] completed step: 3, seconds: 0.338, TFLOP/s/device: 40.152, Tokens/s/device: 6052.081, total_weights: 16384, loss: 8.838, lm_loss: 8.838, perplexity: 6892.933 I0421 15:35:14.137184 134565803121792 metric_logger.py:196] completed step: 4, seconds: 0.006, TFLOP/s/device: 2354.788, Tokens/s/device: 354939.341, total_weights: 16384, loss: 7.965, lm_loss: 7.965, perplexity: 2878.644 I0421 15:35:14.137874 134490236126784 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_13_scan_layers_true/checkpoints/0/items/array_metadatas/process_0 I0421 15:35:14.160544 134565803121792 metric_logger.py:196] completed step: 5, seconds: 0.119, TFLOP/s/device: 114.334, Tokens/s/device: 17233.690, total_weights: 16384, loss: 7.233, lm_loss: 7.233, perplexity: 1383.750 I0421 15:35:14.277992 134565803121792 metric_logger.py:196] completed step: 6, seconds: 14.596, TFLOP/s/device: 0.931, Tokens/s/device: 140.315, total_weights: 16384, loss: 6.676, lm_loss: 6.676, perplexity: 793.531 I0421 15:35:14.396824 134565803121792 metric_logger.py:196] completed step: 7, seconds: 0.006, TFLOP/s/device: 2270.197, Tokens/s/device: 342188.805, total_weights: 16384, loss: 6.293, lm_loss: 6.293, perplexity: 540.860 I0421 15:35:14.515821 134565803121792 metric_logger.py:196] completed step: 8, seconds: 0.119, TFLOP/s/device: 113.972, Tokens/s/device: 17179.046, total_weights: 16384, loss: 6.055, lm_loss: 6.055, perplexity: 426.371 I0421 15:35:14.634190 134565803121792 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0421 15:35:14.635152 134565803121792 checkpointing.py:776] Waited 0.0009887218475341797 seconds for step 9 to finish before starting checkpointing. I0421 15:35:14.635529 134565803121792 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 15:35:43.189436 134490183697984 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 44.064826s (commit=43.595703s, array_metadata_write=0.469122s) I0421 15:35:43.190707 134490173212224 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 278.338 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 45 seconds) (per-host) I0421 15:35:43.190778 134490173212224 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 44.488977s. I0421 15:35:43.441694 134490173212224 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776785698552749892, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/0/_CHECKPOINT_METADATA I0421 15:35:43.634561 134490173212224 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0421 15:35:43.820635 134490194183744 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776785698552749892, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/0/_CHECKPOINT_METADATA I0421 15:35:44.080978 134490173212224 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 15:35:44.081614 134490173212224 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.600482s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/0/items I0421 15:35:44.082388 134490173212224 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/0/items I0421 15:35:44.338294 134490173212224 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/0 I0421 15:35:45.045135 134490173212224 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/0`. I0421 15:35:45.045862 134490173212224 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 47.348484s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/0 I0421 15:35:45.045947 134490173212224 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 46.344151s. I0421 15:35:45.046157 134490246612544 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 15:35:45.046282 134490246612544 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 15:35:45.046354 134490246612544 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0421 15:35:45.046408 134490246612544 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0421 15:35:45.046504 134565803121792 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0421 15:35:45.046581 134565803121792 checkpoint_manager.py:1441] Waiting for previous save to complete took 30.411112 seconds. If this number is high, consider checkpointing less frequently. I0421 15:35:45.047805 134565803121792 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0421 15:35:45.048145 134565803121792 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/9. I0421 15:35:45.223637 134490246612544 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/9 I0421 15:35:45.233519 134565803121792 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 15:35:45.233627 134565803121792 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 15:35:45.956093 134490104006208 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/9/items I0421 15:35:48.043730 134565803121792 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 2.810709s I0421 15:35:48.047980 134565803121792 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.249 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 2 seconds) (per-host) I0421 15:35:48.048063 134565803121792 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 2.904406s (batch_requests_ready=0.086145s, total_serialization_initiated=2.814133s, others=0.004128s) I0421 15:35:48.048142 134565803121792 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.905033s (all_items=0.000027s, per_item={'items': '0.00002718'}, temp_paths=2.905006) I0421 15:35:48.049859 134490173212224 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 15:35:48.049948 134565803121792 async_checkpointer.py:561] Finished blocking save. Time taken: 3.002096s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/9. I0421 15:35:48.050194 134565803121792 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0421 15:35:48.050368 134490183697984 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 15:35:48.050473 134565803121792 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776785714.635442, 'wait_for_prev_duration_secs': 30.41111159324646, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776785745.04783, 'checkpointer_blocking_duration_secs': 3.002258539199829, 'get_old_steps_start_time': 1776785748.0501015, 'get_old_steps_duration_secs': 4.7206878662109375e-05, 'checkpoint_manager_blocking_start_time': 1776785714.6353848, 'checkpoint_manager_blocking_duration_secs': 33.41506099700928} I0421 15:35:48.050593 134565803121792 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0421 15:35:48.050627 134565803121792 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 15:35:48.819517 134490236126784 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_13_scan_layers_true/checkpoints/9/items/array_metadatas/process_0 I0421 15:36:22.586090 134490104006208 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 34.537146s (commit=34.049749s, array_metadata_write=0.487397s) I0421 15:36:22.587270 134490173212224 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 337.489 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 37 seconds) (per-host) I0421 15:36:22.587329 134490173212224 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 34.537334s. I0421 15:36:23.048931 134490173212224 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0421 15:36:23.489158 134490173212224 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 15:36:23.489806 134490173212224 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.579014s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/9/items I0421 15:36:23.490526 134490173212224 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/9/items I0421 15:36:23.741516 134490173212224 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/9 I0421 15:36:24.446659 134490173212224 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/9`. I0421 15:36:24.447404 134490173212224 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 39.399559s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_13_scan_layers_true/checkpoints/9 I0421 15:36:24.447479 134490173212224 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 36.397486s. I0421 15:36:24.447691 134490183697984 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 15:36:24.447820 134490183697984 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 15:36:24.447893 134490183697984 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0421 15:36:24.447935 134490183697984 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0421 15:36:24.448347 134565803121792 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0421 15:36:24.449081 134565803121792 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 15:36:24.450506 134565803121792 metric_logger.py:196] completed step: 9, seconds: 0.120, TFLOP/s/device: 113.087, Tokens/s/device: 17045.644, total_weights: 16384, loss: 5.915, lm_loss: 5.915, perplexity: 370.653 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops
2026-04-23 02:05:05.947458: 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 00:00:1776909905.960822 2303225 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776909905.964663 2303225 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776909905.976133 2303225 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776909905.976147 2303225 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776909905.976149 2303225 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776909905.976151 2303225 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'} 2026-04-23 02:05:09.628665: 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) I0423 02:05:09.763131 128802303108928 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0423 02:06:03.046648 128802303108928 max_utils.py:800] System Information: Jax Version: 0.8.1 I0423 02:06:03.046921 128802303108928 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0423 02:06:03.046973 128802303108928 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 I0423 02:06:03.047005 128802303108928 train_utils.py:348] WARNING: 'dataset_path' might be pointing your local file system I0423 02:06:03.047027 128802303108928 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0423 02:06:03.047206 128802303108928 train.py:703] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0423 02:06:03.703437 128802303108928 maxtext_utils.py:1604] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 02:06:03.703701 128802303108928 checkpointing.py:677] Setting up checkpoint logger... I0423 02:06:03.703744 128802303108928 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0423 02:06:03.703787 128802303108928 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0423 02:06:03.704239 128802303108928 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 0x7524d5fab110>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0423 02:06:06.065889 128802303108928 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0423 02:06:06.066286 128802303108928 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 0x75237cf36a80>}, handler_registry=None I0423 02:06:06.066524 128802303108928 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x75237cf36a80>` 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`. I0423 02:06:06.066565 128802303108928 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x752306fd8f50>` 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`. I0423 02:06:06.066594 128802303108928 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 0x75237cf36a80>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x75237cf36a80>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x752306fd8f50>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x752306fd8f50>}). I0423 02:06:06.066886 128802303108928 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0423 02:06:06.066940 128802303108928 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x752305f00860> timeout: 600 secs and primary_host=0 for async checkpoint writes I0423 02:06:06.201435 128802303108928 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints I0423 02:06:06.201664 128802303108928 checkpoint_manager.py:921] [process=0][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_hns=False, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x752306fd8cb0> I0423 02:06:06.201944 128802303108928 checkpointing.py:301] Checkpoint manager created! I0423 02:06:06.585311 128802303108928 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0423 02:06:06.585413 128802303108928 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 02:06:06.865889 128802303108928 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 02:06:06.865996 128802303108928 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 02:06:06.878243 128802303108928 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 02:06:06.878293 128802303108928 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 02:06:06.897277 128802303108928 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 02:06:06.897334 128802303108928 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 02:06:06.909713 128802303108928 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 02:06:06.909763 128802303108928 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 02:06:06.922012 128802303108928 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 02:06:06.922080 128802303108928 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 02:06:06.941453 128802303108928 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0423 02:06:06.941516 128802303108928 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 02:06:06.956945 128802303108928 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0423 02:06:06.957003 128802303108928 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0423 02:06:07.107231 128802303108928 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0423 02:06:07.107336 128802303108928 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 I0423 02:06:07.737453 128802303108928 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0423 02:06:07.737566 128802303108928 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) I0423 02:06:07.737621 128802303108928 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) I0423 02:06:07.737667 128802303108928 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') I0423 02:06:07.737706 128802303108928 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) I0423 02:06:07.737739 128802303108928 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) I0423 02:06:07.737777 128802303108928 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) I0423 02:06:07.737817 128802303108928 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') I0423 02:06:07.737845 128802303108928 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) I0423 02:06:07.737872 128802303108928 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) I0423 02:06:07.737905 128802303108928 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0423 02:06:07.737939 128802303108928 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0423 02:06:08.144273 128802303108928 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 02:06:08.144363 128802303108928 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0423 02:06:08.155847 128802303108928 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 02:06:08.155895 128802303108928 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0423 02:06:08.952065 128802303108928 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. I0423 02:06:08.952586 128802303108928 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0423 02:06:08.952733 128802303108928 metric_logger.py:301] number parameters: 1.104 billion I0423 02:06:11.633074 128802303108928 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0423 02:06:11.751241 128802303108928 checkpointing.py:776] Waited 0.1181631088256836 seconds for step 0 to finish before starting checkpointing. I0423 02:06:11.751875 128802303108928 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 02:06:11.752034 128802303108928 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0423 02:06:11.752449 128802303108928 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/0. I0423 02:06:11.835991 128802303108928 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0423 02:06:11.920850 128802303108928 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 02:06:11.921008 128770969175616 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/0 I0423 02:06:11.921377 128802303108928 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 I0423 02:06:12.712412 128770958689856 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/0/items I0423 02:06:12.860789 128770927232576 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776909972624952526, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776909972624952526, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/0/_CHECKPOINT_METADATA I0423 02:06:13.149629 128802303108928 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.229599s I0423 02:06:13.150186 128802303108928 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 9.401 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0423 02:06:13.150247 128802303108928 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.312830s (batch_requests_ready=0.078893s, total_serialization_initiated=1.233513s, others=0.000424s) I0423 02:06:13.150318 128802303108928 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.313780s (all_items=0.000020s, per_item={'items': '0.00001979'}, temp_paths=1.313760) I0423 02:06:13.151258 128770732197440 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 02:06:13.151387 128802303108928 async_checkpointer.py:561] Finished blocking save. Time taken: 1.399287s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/0. I0423 02:06:13.151783 128802303108928 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0423 02:06:13.152201 128771118073408 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 02:06:13.152326 128802303108928 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776909971.7518623, 'wait_for_prev_duration_secs': 4.220008850097656e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776909971.7520764, 'checkpointer_blocking_duration_secs': 1.3994536399841309, 'get_old_steps_start_time': 1776909973.1515512, 'get_old_steps_duration_secs': 0.00014066696166992188, 'checkpoint_manager_blocking_start_time': 1776909971.7517786, 'checkpoint_manager_blocking_duration_secs': 1.4005117416381836} I0423 02:06:13.152471 128802303108928 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0423 02:06:13.152518 128802303108928 max_utils.py:750] Memstats: After params initialized: I0423 02:06:13.152626 128802303108928 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0423 02:06:13.152657 128802303108928 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0423 02:06:13.152682 128802303108928 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0423 02:06:13.152704 128802303108928 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0423 02:06:13.152725 128802303108928 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0423 02:06:13.152744 128802303108928 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0423 02:06:13.152763 128802303108928 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0423 02:06:13.152782 128802303108928 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0423 02:06:13.178257 2305299 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0423 02:06:13.178317 2305299 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0423 02:06:13.534509 128802303108928 metric_logger.py:196] completed step: 0, seconds: 2.680, TFLOP/s/device: 5.070, Tokens/s/device: 764.241, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0423 02:06:13.535583 128802303108928 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/tensorboard/' I0423 02:06:13.650946 128802303108928 metric_logger.py:196] completed step: 1, seconds: 1.899, TFLOP/s/device: 7.154, Tokens/s/device: 1078.364, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0423 02:06:13.770031 128802303108928 metric_logger.py:196] completed step: 2, seconds: 0.008, TFLOP/s/device: 1780.284, Tokens/s/device: 268343.816, total_weights: 16384, loss: 9.775, lm_loss: 9.775, perplexity: 17589.287 I0423 02:06:13.889463 128802303108928 metric_logger.py:196] completed step: 3, seconds: 0.118, TFLOP/s/device: 115.554, Tokens/s/device: 17417.484, total_weights: 16384, loss: 8.838, lm_loss: 8.838, perplexity: 6892.933 I0423 02:06:13.906393 128770916746816 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/0/items/array_metadatas/process_0 I0423 02:06:30.729987 128802303108928 metric_logger.py:196] completed step: 4, seconds: 0.119, TFLOP/s/device: 113.725, Tokens/s/device: 17141.805, total_weights: 16384, loss: 7.965, lm_loss: 7.965, perplexity: 2878.644 I0423 02:06:30.842912 128802303108928 metric_logger.py:196] completed step: 5, seconds: 0.119, TFLOP/s/device: 113.769, Tokens/s/device: 17148.551, total_weights: 16384, loss: 7.233, lm_loss: 7.233, perplexity: 1383.750 I0423 02:06:30.974920 128802303108928 metric_logger.py:196] completed step: 6, seconds: 16.945, TFLOP/s/device: 0.802, Tokens/s/device: 120.864, total_weights: 16384, loss: 6.676, lm_loss: 6.676, perplexity: 793.531 I0423 02:06:31.079681 128802303108928 metric_logger.py:196] completed step: 7, seconds: 0.109, TFLOP/s/device: 124.536, Tokens/s/device: 18771.425, total_weights: 16384, loss: 6.293, lm_loss: 6.293, perplexity: 540.860 I0423 02:06:31.200019 128802303108928 metric_logger.py:196] completed step: 8, seconds: 0.039, TFLOP/s/device: 352.226, Tokens/s/device: 53091.380, total_weights: 16384, loss: 6.055, lm_loss: 6.055, perplexity: 426.371 I0423 02:06:31.320532 128802303108928 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0423 02:06:31.323038 128802303108928 checkpointing.py:776] Waited 0.002913951873779297 seconds for step 9 to finish before starting checkpointing. I0423 02:06:31.329706 128802303108928 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 02:07:07.047163 128770948204096 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 53.867185s (commit=53.396641s, array_metadata_write=0.470544s) I0423 02:07:07.048334 128770732197440 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 228.882 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 55 seconds) (per-host) I0423 02:07:07.048403 128770732197440 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 53.896953s. I0423 02:07:07.321442 128770732197440 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776909972624952526, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/0/_CHECKPOINT_METADATA I0423 02:07:07.555279 128770732197440 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 02:07:07.751276 128770927232576 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776909972624952526, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/0/_CHECKPOINT_METADATA I0423 02:07:07.981494 128770732197440 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 02:07:07.981721 128770732197440 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.597638s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/0/items I0423 02:07:07.982488 128770732197440 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/0/items I0423 02:07:08.226520 128770732197440 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/0 I0423 02:07:08.897680 128770732197440 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/0`. I0423 02:07:08.898480 128770732197440 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 57.146387s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/0 I0423 02:07:08.898570 128770732197440 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 55.747124s. I0423 02:07:08.898814 128771118073408 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 02:07:08.898943 128771118073408 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 02:07:08.899008 128771118073408 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0423 02:07:08.899077 128771118073408 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0423 02:07:08.899259 128802303108928 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. W0423 02:07:08.899501 128802303108928 checkpoint_manager.py:1441] Waiting for previous save to complete took 37.574715 seconds. If this number is high, consider checkpointing less frequently. I0423 02:07:08.900952 128802303108928 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0423 02:07:08.901316 128802303108928 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/9. I0423 02:07:09.069487 128802303108928 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 02:07:09.069590 128802303108928 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 I0423 02:07:09.070639 128770732197440 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/9 I0423 02:07:09.730867 128770916746816 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/9/items I0423 02:07:13.594218 128802303108928 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 4.525387s I0423 02:07:13.598795 128802303108928 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.673 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 4 seconds) (per-host) I0423 02:07:13.598905 128802303108928 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 4.616731s (batch_requests_ready=0.083609s, total_serialization_initiated=4.528623s, others=0.004499s) I0423 02:07:13.598990 128802303108928 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 4.617309s (all_items=0.000024s, per_item={'items': '0.00002408'}, temp_paths=4.617285) I0423 02:07:13.600466 128771118073408 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 02:07:13.600550 128802303108928 async_checkpointer.py:561] Finished blocking save. Time taken: 4.699541s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/9. I0423 02:07:13.600774 128802303108928 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0423 02:07:13.600981 128770948204096 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 02:07:13.601115 128802303108928 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776909991.3246717, 'wait_for_prev_duration_secs': 37.57471489906311, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776910028.9009871, 'checkpointer_blocking_duration_secs': 4.6996684074401855, 'get_old_steps_start_time': 1776910033.6006775, 'get_old_steps_duration_secs': 5.3882598876953125e-05, 'checkpoint_manager_blocking_start_time': 1776909991.323349, 'checkpoint_manager_blocking_duration_secs': 42.27773857116699} I0423 02:07:13.601233 128802303108928 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0423 02:07:13.601269 128802303108928 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 02:07:14.314541 128770767849024 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/9/items/array_metadatas/process_0 I0423 02:07:50.811069 128770757363264 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 37.211571s (commit=36.769745s, array_metadata_write=0.441826s) I0423 02:07:50.812096 128771118073408 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 302.099 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 41 seconds) (per-host) I0423 02:07:50.812149 128771118073408 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.211555s. I0423 02:07:51.270830 128771118073408 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 02:07:51.652218 128771118073408 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 02:07:51.652395 128771118073408 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.549744s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/9/items I0423 02:07:51.653137 128771118073408 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/9/items I0423 02:07:51.864028 128771118073408 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/9 I0423 02:07:52.586236 128771118073408 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/9`. I0423 02:07:52.586973 128771118073408 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 43.685965s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_13_scan_layers_true/checkpoints/9 I0423 02:07:52.587054 128771118073408 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 38.986448s. I0423 02:07:52.587224 128770948204096 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 02:07:52.587338 128770948204096 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 02:07:52.587393 128770948204096 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0423 02:07:52.587437 128770948204096 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0423 02:07:52.587593 128802303108928 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. I0423 02:07:52.587744 128802303108928 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 02:07:52.588677 128802303108928 metric_logger.py:196] completed step: 9, seconds: 0.098, TFLOP/s/device: 138.592, Tokens/s/device: 20890.072, total_weights: 16384, loss: 5.915, lm_loss: 5.915, perplexity: 370.653 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops