main (574ad3fb9) vs test/pipeline-scan-nnx (574ad3fb9)| Metric | main 574ad3fb9 | test/pipeline-scan-nnx 574ad3fb9 | Diff (test/pipeline-scan-nnx − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 5.3910 | 5.3910 | 0 |
| TFLOP/s | 92.310 | 91.990 | -0.32 |
| Tok/s | 13913.9 | 13865.7 | -48.231 |
| Avg s/step | 8.677 | 8.159 | -0.518 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.8.1 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Sat Apr 18 19:28:06 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-18 19:28:31.072764: 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) I0418 19:28:31.253209 132355044058944 max_utils.py:273] Attempting to initialize the jax distributed system... I0418 19:28:40.294209 132355044058944 distributed.py:149] Starting JAX distributed service on [::]:8482 I0418 19:28:40.296422 132355044058944 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-true-rs-fhed3-slice-job-0-0.mt-15-ocdbt-true-rs-fhed3:8482 I0418 19:28:41.751239 132355044058944 max_utils.py:284] Jax distributed system initialized! I0418 19:28:46.987035 132355044058944 max_utils.py:800] System Information: Jax Version: 0.9.2 I0418 19:28:46.987138 132355044058944 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0418 19:28:46.987176 132355044058944 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0418 19:28:46.987209 132355044058944 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0418 19:28:47.684924 132355044058944 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0418 19:28:47.685195 132355044058944 checkpointing.py:677] Setting up checkpoint logger... I0418 19:28:47.685248 132355044058944 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0418 19:28:47.685291 132355044058944 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0418 19:28:47.685629 132355044058944 base_pytree_checkpoint_handler.py:441] 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 0x785fbb715040>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0418 19:28:49.811168 132355044058944 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0418 19:28:49.811394 132355044058944 checkpoint_manager.py:708] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x784c0052e930>}, handler_registry=None I0418 19:28:49.811643 132355044058944 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x784c0052e930>` 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`. I0418 19:28:49.811691 132355044058944 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x784c00326f90>` 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`. I0418 19:28:49.811726 132355044058944 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 0x784c0052e930>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x784c0052e930>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x784c00326f90>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x784c00326f90>}). I0418 19:28:49.812072 132355044058944 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0418 19:28:49.812146 132355044058944 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x784c0028fd80> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0418 19:28:52.153467 132234367452928 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776540361293379292, 'commit_timestamp_nsecs': 1776540408852754834, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints/9/_CHECKPOINT_METADATA I0418 19:28:52.408132 132234367452928 checkpoint_manager.py:1871] Missing metrics for step 9 I0418 19:28:52.818521 132355044058944 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints I0418 19:28:54.533291 132355044058944 checkpoint_manager.py:929] [process=5][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_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x784c00487b30> I0418 19:28:54.533489 132355044058944 checkpointing.py:301] Checkpoint manager created! I0418 19:28:55.453088 132355044058944 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0418 19:28:55.453202 132355044058944 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:28:55.829589 132355044058944 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 19:28:55.829678 132355044058944 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:28:55.845787 132355044058944 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 19:28:55.845841 132355044058944 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:28:55.869264 132355044058944 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 19:28:55.869331 132355044058944 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:28:55.885691 132355044058944 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 19:28:55.885752 132355044058944 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:28:55.902081 132355044058944 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 19:28:55.902141 132355044058944 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:28:55.927446 132355044058944 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0418 19:28:55.927527 132355044058944 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:28:55.948697 132355044058944 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0418 19:28:55.948765 132355044058944 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0418 19:28:56.174516 132355044058944 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0418 19:28:56.174620 132355044058944 checkpointing.py:581] restoring from this run's directory step 9 I0418 19:28:56.425709 132355044058944 event_tracking.py:70] [process=5] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints/9. I0418 19:28:57.232165 132355044058944 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints/9. I0418 19:28:57.232611 132355044058944 event_tracking.py:125] [process=5] [sync] Finished blocking load in 0.81 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints/9. I0418 19:28:59.291837 1445 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0418 19:29:08.376130 132355044058944 jax_array_handlers.py:843] [process=5] /jax/orbax/read/worker/io/requested throughput: 173.840 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 9.090365171432495 s) (per-host) I0418 19:29:08.381626 132355044058944 base_pytree_checkpoint_handler.py:130] [process=5] /jax/checkpoint/read/gbytes_per_sec: 166.705 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 9.493396997451782 s) (per-host) I0418 19:29:08.383747 132355044058944 event_tracking.py:138] [process=5] [sync] Finished load in 11.96 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints/9 I0418 19:29:08.383842 132355044058944 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints', 'checkpointer_start_time': 1776540536.4256046, 'checkpointer_duration_secs': 11.95820951461792, 'checkpoint_manager_start_time': 1776540536.17487, 'checkpoint_manager_duration_secs': 12.208945989608765} fsdp: 32 I0418 19:29:08.806889 132355044058944 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0418 19:29:08.807013 132355044058944 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0418 19:29:08.807071 132355044058944 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0418 19:29:08.807125 132355044058944 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0418 19:29:08.807175 132355044058944 maxtext_utils.py:1654] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0418 19:29:08.807212 132355044058944 maxtext_utils.py:1654] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0418 19:29:08.807280 132355044058944 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0418 19:29:08.807373 132355044058944 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0418 19:29:08.807450 132355044058944 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0418 19:29:08.807528 132355044058944 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0418 19:29:08.807608 132355044058944 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0418 19:29:08.807697 132355044058944 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0418 19:29:09.291992 132355044058944 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 19:29:09.292089 132355044058944 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0418 19:29:09.307698 132355044058944 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 19:29:09.307765 132355044058944 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0418 19:29:20.210910 132355044058944 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0418 19:29:20.211751 132355044058944 metric_logger.py:301] number parameters: 1.104 billion I0418 19:29:31.659871 132355044058944 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0418 19:29:31.814372 132355044058944 checkpointing.py:776] Waited 0.1544795036315918 seconds for step 10 to finish before starting checkpointing. I0418 19:29:31.816780 132355044058944 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 19:29:31.818386 132355044058944 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 10 I0418 19:29:31.819846 132355044058944 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints/10. I0418 19:29:33.010025 132355044058944 signaling_client.py:364] Using JaxDistributedSignalingClient I0418 19:29:33.010978 132355044058944 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 19:29:33.011033 132355044058944 replica_slices.py:424] 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 I0418 19:29:33.288638 132355044058944 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278691s I0418 19:29:33.288813 132355044058944 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.456 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2827270030975342 s) (per-host) I0418 19:29:33.288866 132355044058944 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.282791s (batch_requests_ready=0.002329s, total_serialization_initiated=0.280390s, others=0.000073s) I0418 19:29:33.288939 132355044058944 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.286809s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.286793) I0418 19:29:33.289632 132355044058944 event_tracking.py:125] [process=5] [async] Finished blocking save in 1.47 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints/10. I0418 19:29:33.290011 132232228361984 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 19:49:33.289969 I0418 19:29:33.294322 132355044058944 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0418 19:29:33.294680 132233855760128 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0418 19:29:33.294839 132355044058944 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776540571.8167553, 'wait_for_prev_duration_secs': 6.842613220214844e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776540571.818424, 'checkpointer_blocking_duration_secs': 1.471742868423462, 'get_old_steps_start_time': 1776540573.29019, 'get_old_steps_duration_secs': 3.361701965332031e-05, 'checkpoint_manager_blocking_start_time': 1776540571.814906, 'checkpoint_manager_blocking_duration_secs': 1.4798967838287354} I0418 19:29:33.294945 132355044058944 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0418 19:29:33.294995 132355044058944 max_utils.py:750] Memstats: After params initialized: I0418 19:29:33.295044 132355044058944 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0)) I0418 19:29:33.295076 132355044058944 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0)) I0418 19:29:33.295102 132355044058944 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0)) I0418 19:29:33.295126 132355044058944 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0)) I0418 19:29:33.613593 132355044058944 metric_logger.py:196] completed step: 10, seconds: 11.448, TFLOP/s/device: 1.187, Tokens/s/device: 178.896, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0418 19:29:33.793893 132355044058944 metric_logger.py:196] completed step: 11, seconds: 1.952, TFLOP/s/device: 6.960, Tokens/s/device: 1049.058, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0418 19:29:33.941802 132355044058944 metric_logger.py:196] completed step: 12, seconds: 0.033, TFLOP/s/device: 410.947, Tokens/s/device: 61942.352, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0418 19:29:34.089167 132355044058944 metric_logger.py:196] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.638, Tokens/s/device: 13360.559, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0418 19:29:34.093660 132355044058944 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0418 19:29:34.383551 132355044058944 checkpointing.py:776] Waited 0.2898554801940918 seconds for step 15 to finish before starting checkpointing. I0418 19:29:34.386543 132355044058944 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 19:29:39.474374 132233285318400 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints/10/items/array_metadatas/process_5 I0418 19:30:14.898998 132232228361984 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.706 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.8928759098053 s) (per-host) I0418 19:30:14.899122 132232228361984 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.608992s. I0418 19:30:23.475764 132232228361984 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 50.185618s. I0418 19:30:23.476056 132233855760128 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 19:30:23.476172 132233855760128 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0418 19:30:23.476221 132233855760128 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0418 19:30:23.477830 132233855760128 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0418 19:30:23.477956 132355044058944 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0418 19:30:23.478022 132355044058944 checkpoint_manager.py:1452] Waiting for previous save to complete took 49.091486 seconds. If this number is high, consider checkpointing less frequently. I0418 19:30:23.480166 132355044058944 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 15 I0418 19:30:23.482683 132355044058944 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints/15. I0418 19:30:24.192102 132355044058944 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 19:30:24.192192 132355044058944 replica_slices.py:424] 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 I0418 19:30:24.217782 132355044058944 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026762s I0418 19:30:24.217926 132355044058944 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 50.766 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03038620948791504 s) (per-host) I0418 19:30:24.217982 132355044058944 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030451s (batch_requests_ready=0.001930s, total_serialization_initiated=0.028449s, others=0.000072s) I0418 19:30:24.218058 132355044058944 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.035557s (all_items=0.000013s, per_item={'items': '0.00001311'}, temp_paths=0.035544) I0418 19:30:24.218775 132355044058944 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.74 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints/15. I0418 19:30:24.219119 132233855760128 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 19:50:24.219082 I0418 19:30:24.231419 132355044058944 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0418 19:30:24.231682 132227374360320 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0418 19:30:24.231816 132355044058944 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776540574.3865123, 'wait_for_prev_duration_secs': 49.09148597717285, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776540623.480205, 'checkpointer_blocking_duration_secs': 0.7390553951263428, 'get_old_steps_start_time': 1776540624.2192838, 'get_old_steps_duration_secs': 3.123283386230469e-05, 'checkpoint_manager_blocking_start_time': 1776540574.3841238, 'checkpoint_manager_blocking_duration_secs': 49.847657203674316} I0418 19:30:24.231933 132355044058944 checkpointing.py:408] Started an asynchronous checkpoint save for step 15 I0418 19:30:24.232731 132355044058944 metric_logger.py:196] completed step: 14, seconds: 0.148, TFLOP/s/device: 91.978, Tokens/s/device: 13863.973, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0418 19:30:24.238857 132355044058944 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.187, Tokens/s/device: 13895.391, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0418 19:30:32.385009 132233285318400 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints/15/items/array_metadatas/process_5 I0418 19:30:46.828865 132355044058944 metric_logger.py:196] completed step: 16, seconds: 50.144, TFLOP/s/device: 0.271, Tokens/s/device: 40.843, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0418 19:30:46.976232 132355044058944 metric_logger.py:196] completed step: 17, seconds: 22.444, TFLOP/s/device: 0.605, Tokens/s/device: 91.250, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0418 19:30:47.123577 132355044058944 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.190, Tokens/s/device: 13443.701, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0418 19:30:47.270324 132355044058944 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0418 19:30:47.271039 132355044058944 checkpointing.py:776] Waited 0.0007314682006835938 seconds for step 19 to finish before starting checkpointing. I0418 19:30:47.273076 132355044058944 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 19:31:07.799162 132233855760128 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 36.220 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.61159372329712 s) (per-host) I0418 19:31:07.799282 132233855760128 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.580052s. I0418 19:31:14.769452 132233855760128 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 50.550207s. I0418 19:31:14.769710 132227374360320 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 19:31:14.769768 132227374360320 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0418 19:31:14.769810 132227374360320 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0418 19:31:14.775332 132227374360320 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0418 19:31:14.775601 132355044058944 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0418 19:31:14.775743 132355044058944 checkpoint_manager.py:1452] Waiting for previous save to complete took 27.502671 seconds. If this number is high, consider checkpointing less frequently. I0418 19:31:14.777301 132355044058944 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 19 I0418 19:31:14.779512 132355044058944 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints/19. I0418 19:31:15.491492 132355044058944 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 19:31:15.491583 132355044058944 replica_slices.py:424] 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 I0418 19:31:15.517027 132355044058944 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026484s I0418 19:31:15.517159 132355044058944 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 51.337 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030048131942749023 s) (per-host) I0418 19:31:15.517207 132355044058944 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030104s (batch_requests_ready=0.001777s, total_serialization_initiated=0.028267s, others=0.000060s) I0418 19:31:15.517280 132355044058944 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034261s (all_items=0.000012s, per_item={'items': '0.00001168'}, temp_paths=0.034250) I0418 19:31:15.517949 132355044058944 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.74 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints/19. I0418 19:31:15.518269 132233285318400 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 19:51:15.518231 I0418 19:31:15.937484 132355044058944 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0418 19:31:15.937866 132227374360320 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0418 19:31:15.938045 132355044058944 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776540647.273046, 'wait_for_prev_duration_secs': 27.502670526504517, 'time_between_consecutive_saves_sec': 23.795177221298218, 'checkpointer_blocking_start_time': 1776540674.777338, 'checkpointer_blocking_duration_secs': 0.7410800457000732, 'get_old_steps_start_time': 1776540675.518437, 'get_old_steps_duration_secs': 3.552436828613281e-05, 'checkpoint_manager_blocking_start_time': 1776540647.2712762, 'checkpoint_manager_blocking_duration_secs': 28.666731119155884} I0418 19:31:15.938175 132355044058944 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0418 19:31:15.938223 132355044058944 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 19:31:21.817786 132227413309184 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_true/checkpoints/19/items/array_metadatas/process_5 I0418 19:31:56.855417 132233285318400 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 38.184 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.36826753616333 s) (per-host) I0418 19:31:56.855550 132233285318400 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.337165s. I0418 19:32:05.395843 132233285318400 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 49.877443s. I0418 19:32:05.396080 132227374360320 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 19:32:05.396138 132227374360320 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0418 19:32:05.396177 132227374360320 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0418 19:32:05.397910 132227374360320 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0418 19:32:05.400079 132355044058944 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0418 19:32:05.400227 132355044058944 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 19:32:05.401252 132355044058944 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.310, Tokens/s/device: 13913.894, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 18 19:32:17 UTC 2026 EXIT_CODE=0
XPK Start: Sun Apr 19 10:02:03 UTC 2026 2026-04-19 10:02:07.149009: 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 00:00:1776592927.161684 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776592927.165897 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776592927.177383 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776592927.177403 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776592927.177405 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776592927.177407 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-19 10:02:26.440003: 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) I0419 10:02:26.964879 136196921698112 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-19 10:02:36,003:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0419 10:02:36.003751 136196921698112 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-19 10:02:36,006:jax._src.distributed:157: Connecting to JAX distributed service on mt-15-ocdbt-true-rs-lfekf-slice-job-0-0.mt-15-ocdbt-true-rs-lfekf:8482 I0419 10:02:36.006201 136196921698112 distributed.py:157] Connecting to JAX distributed service on mt-15-ocdbt-true-rs-lfekf-slice-job-0-0.mt-15-ocdbt-true-rs-lfekf:8482 I0419 10:02:37.324380 136196921698112 max_utils.py:284] Jax distributed system initialized! I0419 10:02:43.538295 136196921698112 max_utils.py:800] System Information: Jax Version: 0.8.1 I0419 10:02:43.538400 136196921698112 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0419 10:02:43.538441 136196921698112 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 I0419 10:02:43.538476 136196921698112 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0419 10:02:44.157418 136196921698112 maxtext_utils.py:1398] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 10:02:44.157709 136196921698112 checkpointing.py:677] Setting up checkpoint logger... I0419 10:02:44.157765 136196921698112 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0419 10:02:44.157810 136196921698112 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0419 10:02:44.158203 136196921698112 base_pytree_checkpoint_handler.py:415] 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 0x7bde61550a70>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0419 10:02:45.415607 136196921698112 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0419 10:02:45.415840 136196921698112 checkpoint_manager.py:709] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bdd0c364740>}, handler_registry=None I0419 10:02:45.416095 136196921698112 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bdd0c364740>` 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`. I0419 10:02:45.416145 136196921698112 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bc9847e3a10>` 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`. I0419 10:02:45.416183 136196921698112 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 0x7bdd0c364740>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bdd0c364740>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bc9847e3a10>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bc9847e3a10>}). I0419 10:02:45.418849 136196921698112 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0419 10:02:45.418933 136196921698112 async_checkpointer.py:177] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7bde4f1f72e0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0419 10:02:47.728831 136077784381184 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776592687004325929, 'commit_timestamp_nsecs': 1776592729342035392, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints/0/_CHECKPOINT_METADATA I0419 10:02:47.982086 136077784381184 checkpoint_manager.py:1877] Missing metrics for step 0 I0419 10:02:49.276733 136196921698112 checkpoint_manager.py:1818] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints I0419 10:02:49.279030 136196921698112 checkpoint_manager.py:929] [process=4][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_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7bc9847dfd70> I0419 10:02:49.279145 136196921698112 checkpointing.py:301] Checkpoint manager created! I0419 10:02:50.168672 136196921698112 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 10:02:50.168781 136196921698112 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 10:02:50.541900 136196921698112 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 10:02:50.542004 136196921698112 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 10:02:50.558002 136196921698112 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 10:02:50.558063 136196921698112 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 10:02:50.581703 136196921698112 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 10:02:50.581774 136196921698112 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 10:02:50.597817 136196921698112 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 10:02:50.597877 136196921698112 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 10:02:50.613804 136196921698112 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 10:02:50.613867 136196921698112 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 10:02:50.638139 136196921698112 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0419 10:02:50.638212 136196921698112 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 10:02:50.657887 136196921698112 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0419 10:02:50.657962 136196921698112 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0419 10:02:50.850778 136196921698112 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0419 10:02:50.850885 136196921698112 checkpointing.py:581] restoring from this run's directory step 9 I0419 10:02:51.776334 136196921698112 checkpointer.py:304] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints/9. I0419 10:02:53.178696 1408 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0419 10:02:59.501052 136196921698112 jax_array_handlers.py:837] [process=4] /jax/orbax/read/worker/io/requested throughput: 249.746 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.327518939971924 s) (per-host) I0419 10:02:59.515219 136196921698112 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/read/gbytes_per_sec: 234.540 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.747647285461426 s) (per-host) I0419 10:03:00.996001 136196921698112 checkpointer.py:318] Finished restoring checkpoint in 9.90 seconds from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints/9. I0419 10:03:00.996158 136196921698112 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints', 'checkpointer_start_time': 1776592971.095323, 'checkpointer_duration_secs': 9.900803804397583, 'checkpoint_manager_start_time': 1776592970.851139, 'checkpoint_manager_duration_secs': 10.144989967346191} fsdp: 32 I0419 10:03:01.401054 136196921698112 maxtext_utils.py:1501] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0419 10:03:01.401188 136196921698112 maxtext_utils.py:1501] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0419 10:03:01.401242 136196921698112 maxtext_utils.py:1501] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0419 10:03:01.401317 136196921698112 maxtext_utils.py:1501] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0419 10:03:01.401389 136196921698112 maxtext_utils.py:1501] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0419 10:03:01.401432 136196921698112 maxtext_utils.py:1501] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0419 10:03:01.401488 136196921698112 maxtext_utils.py:1501] 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) I0419 10:03:01.401542 136196921698112 maxtext_utils.py:1501] 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') I0419 10:03:01.401583 136196921698112 maxtext_utils.py:1501] 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) I0419 10:03:01.401620 136196921698112 maxtext_utils.py:1501] 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) I0419 10:03:01.401669 136196921698112 maxtext_utils.py:1501] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0419 10:03:01.401715 136196921698112 maxtext_utils.py:1501] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0419 10:03:01.915139 136196921698112 train.py:156] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 10:03:01.915232 136196921698112 train.py:156] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0419 10:03:01.930026 136196921698112 train.py:163] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 10:03:01.930087 136196921698112 train.py:163] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0419 10:03:12.728718 136196921698112 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0419 10:03:12.729605 136196921698112 metric_logger.py:289] number parameters: 1.104 billion I0419 10:03:23.920701 136196921698112 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0419 10:03:24.074792 136196921698112 checkpointing.py:776] Waited 0.15407896041870117 seconds for step 10 to finish before starting checkpointing. I0419 10:03:24.077215 136196921698112 checkpoint_manager.py:2013] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 10:03:24.079185 136196921698112 checkpoint_manager.py:1518] [process=4] Saving checkpoint at step 10 I0419 10:03:24.082224 136196921698112 async_checkpointer.py:452] [process=4] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints/10. I0419 10:03:24.681837 136196921698112 signaling_client.py:364] Using JaxDistributedSignalingClient I0419 10:03:24.682839 136196921698112 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0419 10:03:24.682894 136196921698112 replica_slices.py:424] 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 I0419 10:03:24.929643 136196921698112 base_pytree_checkpoint_handler.py:153] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.247901s I0419 10:03:24.929813 136196921698112 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.945 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.523867130279541 s) (per-host) I0419 10:03:24.929866 136196921698112 base_pytree_checkpoint_handler.py:737] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.523930s (batch_requests_ready=0.273956s, total_serialization_initiated=0.249903s, others=0.000071s) I0419 10:03:24.929983 136196921698112 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.527881s (all_items=0.000019s, per_item={'items': '0.00001860'}, temp_paths=0.527863) I0419 10:03:24.930960 136077180401408 async_checkpointer.py:79] [process=4][thread=async_save] Background save thread started. I0419 10:03:24.931130 136196921698112 async_checkpointer.py:561] Finished blocking save. Time taken: 0.851825s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints/10. I0419 10:03:25.059142 136196921698112 checkpoint_manager.py:1566] [process=4][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0419 10:03:25.059562 136072835077888 async_checkpointer.py:265] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0419 10:03:25.059743 136196921698112 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776593004.0771954, 'wait_for_prev_duration_secs': 6.628036499023438e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776593004.0792692, 'checkpointer_blocking_duration_secs': 0.8519721031188965, 'get_old_steps_start_time': 1776593004.93126, 'get_old_steps_duration_secs': 3.0040740966796875e-05, 'checkpoint_manager_blocking_start_time': 1776593004.0752528, 'checkpoint_manager_blocking_duration_secs': 0.9844508171081543} I0419 10:03:25.059860 136196921698112 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0419 10:03:25.059914 136196921698112 max_utils.py:750] Memstats: After params initialized: I0419 10:03:25.059982 136196921698112 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0)) I0419 10:03:25.060018 136196921698112 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0)) I0419 10:03:25.060045 136196921698112 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0)) I0419 10:03:25.060071 136196921698112 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0)) I0419 10:03:25.322026 136196921698112 metric_logger.py:185] completed step: 10, seconds: 11.191, TFLOP/s/device: 1.214, Tokens/s/device: 183.004, total_weights: 65536, loss: 6.833 I0419 10:03:25.494648 136196921698112 metric_logger.py:185] completed step: 11, seconds: 1.400, TFLOP/s/device: 9.705, Tokens/s/device: 1462.840, total_weights: 65536, loss: 6.528 I0419 10:03:25.642332 136196921698112 metric_logger.py:185] completed step: 12, seconds: 0.025, TFLOP/s/device: 534.569, Tokens/s/device: 80575.992, total_weights: 65536, loss: 6.265 I0419 10:03:25.789609 136196921698112 metric_logger.py:185] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.696, Tokens/s/device: 13369.281, total_weights: 65536, loss: 6.045 I0419 10:03:25.794439 136196921698112 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0419 10:03:26.084324 136196921698112 checkpointing.py:776] Waited 0.28984904289245605 seconds for step 15 to finish before starting checkpointing. I0419 10:03:26.087069 136196921698112 checkpoint_manager.py:2024] [process=4][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 10:03:31.222517 136072851863296 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints/10/items/array_metadatas/process_4 I0419 10:04:07.067969 136077180401408 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/gbytes_per_sec: 37.026 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.661962032318115 s) (per-host) I0419 10:04:07.068094 136077180401408 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.137023s. I0419 10:04:15.494166 136077180401408 async_checkpointer.py:144] [process=4][thread=async_save] Background save thread done. Time taken: 50.563078s. I0419 10:04:15.494427 136072835077888 async_checkpointer.py:273] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 10:04:15.494555 136072835077888 async_checkpointer.py:283] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0419 10:04:15.494618 136072835077888 checkpoint_manager.py:2133] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0419 10:04:15.496217 136072835077888 checkpoint_manager.py:2142] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0419 10:04:15.496410 136196921698112 checkpoint_manager.py:2036] [process=4][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0419 10:04:15.496544 136196921698112 checkpoint_manager.py:1458] Waiting for previous save to complete took 49.409483 seconds. If this number is high, consider checkpointing less frequently. I0419 10:04:15.498684 136196921698112 checkpoint_manager.py:1518] [process=4] Saving checkpoint at step 15 I0419 10:04:15.502562 136196921698112 async_checkpointer.py:452] [process=4] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints/15. I0419 10:04:16.046500 136196921698112 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0419 10:04:16.046597 136196921698112 replica_slices.py:424] 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 I0419 10:04:16.089475 136196921698112 base_pytree_checkpoint_handler.py:153] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044224s I0419 10:04:16.089640 136196921698112 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.177 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2979464530944824 s) (per-host) I0419 10:04:16.089690 136196921698112 base_pytree_checkpoint_handler.py:737] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.298007s (batch_requests_ready=0.251800s, total_serialization_initiated=0.046139s, others=0.000068s) I0419 10:04:16.089779 136196921698112 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.302793s (all_items=0.000013s, per_item={'items': '0.00001335'}, temp_paths=0.302780) I0419 10:04:16.090728 136072851863296 async_checkpointer.py:79] [process=4][thread=async_save] Background save thread started. I0419 10:04:16.090888 136196921698112 async_checkpointer.py:561] Finished blocking save. Time taken: 0.592128s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints/15. I0419 10:04:16.524874 136196921698112 checkpoint_manager.py:1566] [process=4][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0419 10:04:16.525259 136072835077888 async_checkpointer.py:265] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0419 10:04:16.525429 136196921698112 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776593006.087038, 'wait_for_prev_duration_secs': 49.40948271751404, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776593055.498725, 'checkpointer_blocking_duration_secs': 0.5922820568084717, 'get_old_steps_start_time': 1776593056.091025, 'get_old_steps_duration_secs': 2.6226043701171875e-05, 'checkpoint_manager_blocking_start_time': 1776593006.0849702, 'checkpoint_manager_blocking_duration_secs': 50.44042205810547} I0419 10:04:16.525542 136196921698112 checkpointing.py:408] Started an asynchronous checkpoint save for step 15 I0419 10:04:16.526144 136196921698112 metric_logger.py:185] completed step: 14, seconds: 0.148, TFLOP/s/device: 92.034, Tokens/s/device: 13872.331, total_weights: 65536, loss: 5.864 I0419 10:04:16.532120 136196921698112 metric_logger.py:185] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.190, Tokens/s/device: 13895.862, total_weights: 65536, loss: 5.719 I0419 10:04:34.170993 136196921698112 metric_logger.py:185] completed step: 16, seconds: 50.737, TFLOP/s/device: 0.268, Tokens/s/device: 40.365, total_weights: 65536, loss: 5.605 I0419 10:04:34.318692 136196921698112 metric_logger.py:185] completed step: 17, seconds: 17.493, TFLOP/s/device: 0.777, Tokens/s/device: 117.077, total_weights: 65536, loss: 5.517 I0419 10:04:34.466192 136196921698112 metric_logger.py:185] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.347, Tokens/s/device: 13467.394, total_weights: 65536, loss: 5.448 I0419 10:04:34.613313 136196921698112 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0419 10:04:34.614005 136196921698112 checkpointing.py:776] Waited 0.0007123947143554688 seconds for step 19 to finish before starting checkpointing. I0419 10:04:34.615965 136196921698112 checkpoint_manager.py:2024] [process=4][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 10:04:38.430310 136072868648704 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints/15/items/array_metadatas/process_4 I0419 10:05:14.800790 136072851863296 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/gbytes_per_sec: 26.769 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 59.009044885635376 s) (per-host) I0419 10:05:14.800909 136072851863296 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 58.710069s. I0419 10:05:21.567890 136072851863296 async_checkpointer.py:144] [process=4][thread=async_save] Background save thread done. Time taken: 65.477036s. I0419 10:05:21.568142 136072835077888 async_checkpointer.py:273] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 10:05:21.568210 136072835077888 async_checkpointer.py:283] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0419 10:05:21.568264 136072835077888 checkpoint_manager.py:2133] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0419 10:05:21.570430 136072835077888 checkpoint_manager.py:2142] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0419 10:05:21.570610 136196921698112 checkpoint_manager.py:2036] [process=4][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0419 10:05:21.570734 136196921698112 checkpoint_manager.py:1458] Waiting for previous save to complete took 46.954796 seconds. If this number is high, consider checkpointing less frequently. I0419 10:05:21.572264 136196921698112 checkpoint_manager.py:1518] [process=4] Saving checkpoint at step 19 I0419 10:05:21.575500 136196921698112 async_checkpointer.py:452] [process=4] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints/19. I0419 10:05:22.538526 136196921698112 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0419 10:05:22.538624 136196921698112 replica_slices.py:424] 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 I0419 10:05:22.585374 136196921698112 base_pytree_checkpoint_handler.py:153] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.048120s I0419 10:05:22.585537 136196921698112 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.102 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.7338712215423584 s) (per-host) I0419 10:05:22.585588 136196921698112 base_pytree_checkpoint_handler.py:737] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.733932s (batch_requests_ready=0.683665s, total_serialization_initiated=0.050201s, others=0.000066s) I0419 10:05:22.585675 136196921698112 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.737957s (all_items=0.000012s, per_item={'items': '0.00001192'}, temp_paths=0.737946) I0419 10:05:22.586527 136077172008704 async_checkpointer.py:79] [process=4][thread=async_save] Background save thread started. I0419 10:05:22.586683 136196921698112 async_checkpointer.py:561] Finished blocking save. Time taken: 1.014348s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints/19. I0419 10:05:22.591870 136196921698112 checkpoint_manager.py:1566] [process=4][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0419 10:05:22.592118 136072835077888 async_checkpointer.py:265] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0419 10:05:22.592247 136196921698112 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776593074.6159217, 'wait_for_prev_duration_secs': 46.954795598983765, 'time_between_consecutive_saves_sec': 19.11966633796692, 'checkpointer_blocking_start_time': 1776593121.5723038, 'checkpointer_blocking_duration_secs': 1.0144894123077393, 'get_old_steps_start_time': 1776593122.5868056, 'get_old_steps_duration_secs': 2.288818359375e-05, 'checkpoint_manager_blocking_start_time': 1776593074.6142352, 'checkpoint_manager_blocking_duration_secs': 47.97797775268555} I0419 10:05:22.592351 136196921698112 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0419 10:05:22.592392 136196921698112 checkpoint_manager.py:2024] [process=4][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 10:05:27.726779 136077180401408 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260419_085454/linen_xpk_test_pipeline_scan_nnx_20260419_085454_15_ocdbt_true/checkpoints/19/items/array_metadatas/process_4 I0419 10:06:04.006919 136077172008704 base_pytree_checkpoint_handler.py:129] [process=4] /jax/checkpoint/write/gbytes_per_sec: 37.471 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.15522623062134 s) (per-host) I0419 10:06:04.007027 136077172008704 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.420390s. I0419 10:06:10.609816 136077172008704 async_checkpointer.py:144] [process=4][thread=async_save] Background save thread done. Time taken: 48.023163s. I0419 10:06:10.610064 136072835077888 async_checkpointer.py:273] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 10:06:10.610137 136072835077888 async_checkpointer.py:283] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0419 10:06:10.610187 136072835077888 checkpoint_manager.py:2133] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0419 10:06:10.611982 136072835077888 checkpoint_manager.py:2142] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0419 10:06:10.612175 136196921698112 checkpoint_manager.py:2036] [process=4][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0419 10:06:10.612329 136196921698112 checkpoint_manager.py:2013] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 10:06:10.613037 136196921698112 metric_logger.py:185] completed step: 19, seconds: 0.148, TFLOP/s/device: 91.990, Tokens/s/device: 13865.663, total_weights: 65536, loss: 5.391 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sun Apr 19 10:06:23 UTC 2026 EXIT_CODE=0