main (574ad3fb9) vs test/pipeline-scan-nnx (e27fc1e97)| Metric | main 574ad3fb9 | test/pipeline-scan-nnx e27fc1e97 | Diff (test/pipeline-scan-nnx − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 5.3910 | 5.3910 | 0 |
| TFLOP/s | 92.127 | 92.132 | +0.005 |
| Tok/s | 13886.3 | 13887.2 | +0.848 |
| Avg s/step | 4.989 | 4.881 | -0.108 |
| 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: Mon Apr 20 07:12:41 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-20 07:13:05.425208: 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) I0420 07:13:05.960439 137912096147264 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 07:13:15.002393 137912096147264 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 07:13:15.004789 137912096147264 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-false-rs-o5i2c-slice-job-0-0.mt-15-ocdbt-false-rs-o5i2c:8482 I0420 07:13:16.368739 137912096147264 max_utils.py:284] Jax distributed system initialized! I0420 07:13:22.655984 137912096147264 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 07:13:22.656089 137912096147264 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 07:13:22.656128 137912096147264 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 I0420 07:13:22.656162 137912096147264 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 07:13:23.350818 137912096147264 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 07:13:23.351091 137912096147264 checkpointing.py:677] Setting up checkpoint logger... I0420 07:13:23.351145 137912096147264 checkpointing.py:233] Creating checkpoint manager with ocdbt=False and zarr3=True I0420 07:13:23.351188 137912096147264 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 07:13:23.351554 137912096147264 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=False, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7d6d9d3141a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 07:13:24.596007 137912096147264 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0420 07:13:24.596238 137912096147264 checkpoint_manager.py:708] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d58402f3770>}, handler_registry=None I0420 07:13:24.596504 137912096147264 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d58402f3770>` 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`. I0420 07:13:24.596557 137912096147264 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d5840452120>` 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`. I0420 07:13:24.596596 137912096147264 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 0x7d58402f3770>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d58402f3770>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d5840452120>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d5840452120>}). I0420 07:13:24.596916 137912096147264 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 07:13:24.596988 137912096147264 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7d57fc617880> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 07:13:26.510511 137789773432576 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776669032372857120, 'commit_timestamp_nsecs': 1776669046354114586, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints/0/_CHECKPOINT_METADATA I0420 07:13:26.758313 137789781825280 checkpoint_manager.py:1871] Missing metrics for step 5 I0420 07:13:27.122997 137912096147264 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints I0420 07:13:27.494912 137912096147264 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_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7d58402f2f00> I0420 07:13:27.495085 137912096147264 checkpointing.py:301] Checkpoint manager created! I0420 07:13:28.412562 137912096147264 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 07:13:28.412679 137912096147264 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 07:13:28.792001 137912096147264 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 07:13:28.792090 137912096147264 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 07:13:28.808526 137912096147264 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 07:13:28.808589 137912096147264 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 07:13:28.833073 137912096147264 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 07:13:28.833139 137912096147264 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:13:28.849595 137912096147264 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 07:13:28.849655 137912096147264 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:13:28.865938 137912096147264 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 07:13:28.865996 137912096147264 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:13:28.896257 137912096147264 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 07:13:28.896366 137912096147264 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 07:13:28.917832 137912096147264 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 07:13:28.917894 137912096147264 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 07:13:29.133817 137912096147264 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0420 07:13:29.133920 137912096147264 checkpointing.py:581] restoring from this run's directory step 9 I0420 07:13:29.385881 137912096147264 event_tracking.py:70] [process=4] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints/9. I0420 07:13:29.636987 137912096147264 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints/9. I0420 07:13:29.637416 137912096147264 event_tracking.py:125] [process=4] [sync] Finished blocking load in 0.25 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints/9. I0420 07:13:32.366500 1495 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 07:13:41.594278 137912096147264 jax_array_handlers.py:843] [process=4] /jax/orbax/read/worker/io/requested throughput: 152.202 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.382753133773804 s) (per-host) I0420 07:13:41.611308 137912096147264 base_pytree_checkpoint_handler.py:130] [process=4] /jax/checkpoint/read/gbytes_per_sec: 146.064 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.834927558898926 s) (per-host) I0420 07:13:42.786175 137912096147264 event_tracking.py:138] [process=4] [sync] Finished load in 13.40 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints/9 I0420 07:13:42.786326 137912096147264 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints', 'checkpointer_start_time': 1776669209.3857725, 'checkpointer_duration_secs': 13.400522470474243, 'checkpoint_manager_start_time': 1776669209.1341562, 'checkpoint_manager_duration_secs': 13.652140855789185} fsdp: 32 I0420 07:13:43.204629 137912096147264 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 07:13:43.204761 137912096147264 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) I0420 07:13:43.204825 137912096147264 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) I0420 07:13:43.204953 137912096147264 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') I0420 07:13:43.205019 137912096147264 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) I0420 07:13:43.205066 137912096147264 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) I0420 07:13:43.205140 137912096147264 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) I0420 07:13:43.205198 137912096147264 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') I0420 07:13:43.205238 137912096147264 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) I0420 07:13:43.205276 137912096147264 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) I0420 07:13:43.205322 137912096147264 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 07:13:43.205384 137912096147264 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 07:13:43.688929 137912096147264 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 07:13:43.689021 137912096147264 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0420 07:13:43.704538 137912096147264 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 07:13:43.704594 137912096147264 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0420 07:13:54.935581 137912096147264 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. I0420 07:13:54.936494 137912096147264 metric_logger.py:301] number parameters: 1.104 billion I0420 07:14:06.362269 137912096147264 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0420 07:14:06.515065 137912096147264 checkpointing.py:776] Waited 0.1527724266052246 seconds for step 10 to finish before starting checkpointing. I0420 07:14:06.517972 137912096147264 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 07:14:06.519549 137912096147264 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 10 I0420 07:14:06.520965 137912096147264 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints/10. I0420 07:14:06.857804 137912096147264 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 07:14:06.858740 137912096147264 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 07:14:06.858806 137912096147264 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 I0420 07:14:07.133141 137912096147264 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.275414s I0420 07:14:07.133314 137912096147264 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 5.519 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2795276641845703 s) (per-host) I0420 07:14:07.133379 137912096147264 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.279604s (batch_requests_ready=0.002355s, total_serialization_initiated=0.277165s, others=0.000084s) I0420 07:14:07.133474 137912096147264 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.283838s (all_items=0.000022s, per_item={'items': '0.00002193'}, temp_paths=0.283816) I0420 07:14:07.134257 137912096147264 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints/10. I0420 07:14:07.134571 137786929682176 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:34:07.134525 I0420 07:14:07.145684 137912096147264 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0420 07:14:07.146007 137786308949760 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 07:14:07.146159 137912096147264 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776669246.5179543, 'wait_for_prev_duration_secs': 6.103515625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776669246.5195916, 'checkpointer_blocking_duration_secs': 0.6151297092437744, 'get_old_steps_start_time': 1776669247.1347458, 'get_old_steps_duration_secs': 3.1948089599609375e-05, 'checkpoint_manager_blocking_start_time': 1776669246.5155675, 'checkpoint_manager_blocking_duration_secs': 0.6305558681488037} I0420 07:14:07.146268 137912096147264 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0420 07:14:07.146319 137912096147264 max_utils.py:750] Memstats: After params initialized: I0420 07:14:07.146385 137912096147264 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0)) I0420 07:14:07.146420 137912096147264 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0)) I0420 07:14:07.146447 137912096147264 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0)) I0420 07:14:07.146473 137912096147264 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0)) I0420 07:14:07.456778 137912096147264 metric_logger.py:196] completed step: 10, seconds: 11.426, TFLOP/s/device: 1.189, Tokens/s/device: 179.246, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0420 07:14:07.627836 137912096147264 metric_logger.py:196] completed step: 11, seconds: 1.093, TFLOP/s/device: 12.431, Tokens/s/device: 1873.811, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0420 07:14:07.775385 137912096147264 metric_logger.py:196] completed step: 12, seconds: 0.024, TFLOP/s/device: 577.439, Tokens/s/device: 87037.824, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0420 07:14:07.922723 137912096147264 metric_logger.py:196] completed step: 13, seconds: 0.154, TFLOP/s/device: 88.412, Tokens/s/device: 13326.393, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0420 07:14:07.927505 137912096147264 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0420 07:14:08.217010 137912096147264 checkpointing.py:776] Waited 0.28946828842163086 seconds for step 15 to finish before starting checkpointing. I0420 07:14:08.220222 137912096147264 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 07:14:12.281544 137786392811264 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints/10/items/array_metadatas/process_4 I0420 07:14:17.610084 137786929682176 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 146.854 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.756263017654419 s) (per-host) I0420 07:14:17.610194 137786929682176 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.475514s. I0420 07:14:22.994287 137786929682176 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 15.859588s. I0420 07:14:22.994661 137786308949760 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 07:14:22.994775 137786308949760 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 07:14:22.994824 137786308949760 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0420 07:14:22.997181 137786308949760 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0420 07:14:22.997308 137912096147264 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0420 07:14:22.997386 137912096147264 checkpoint_manager.py:1452] Waiting for previous save to complete took 14.777172 seconds. If this number is high, consider checkpointing less frequently. I0420 07:14:22.999552 137912096147264 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 15 I0420 07:14:23.002081 137912096147264 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints/15. I0420 07:14:23.295837 137912096147264 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 07:14:23.295936 137912096147264 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 I0420 07:14:23.336930 137912096147264 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042139s I0420 07:14:23.337095 137912096147264 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 33.793 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04564714431762695 s) (per-host) I0420 07:14:23.337143 137912096147264 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.045705s (batch_requests_ready=0.001849s, total_serialization_initiated=0.043791s, others=0.000065s) I0420 07:14:23.337229 137912096147264 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.050584s (all_items=0.000029s, per_item={'items': '0.00002933'}, temp_paths=0.050555) I0420 07:14:23.337964 137912096147264 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.34 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints/15. I0420 07:14:23.338260 137786308949760 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:34:23.338226 I0420 07:14:23.752180 137912096147264 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0420 07:14:23.752562 137782131414784 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 07:14:23.752743 137912096147264 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776669248.2201936, 'wait_for_prev_duration_secs': 14.777171611785889, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776669262.9995904, 'checkpointer_blocking_duration_secs': 0.3387877941131592, 'get_old_steps_start_time': 1776669263.338403, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776669248.2176168, 'checkpoint_manager_blocking_duration_secs': 15.535088777542114} I0420 07:14:23.752872 137912096147264 checkpointing.py:408] Started an asynchronous checkpoint save for step 15 I0420 07:14:23.754656 137912096147264 metric_logger.py:196] completed step: 14, seconds: 0.147, TFLOP/s/device: 92.241, Tokens/s/device: 13903.504, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0420 07:14:23.761005 137912096147264 metric_logger.py:196] completed step: 15, seconds: 0.148, TFLOP/s/device: 92.020, Tokens/s/device: 13870.264, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0420 07:14:28.486447 137786392811264 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints/15/items/array_metadatas/process_4 I0420 07:14:33.502850 137786308949760 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 154.690 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.211363554000854 s) (per-host) I0420 07:14:33.502969 137786308949760 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.164055s. I0420 07:14:39.946513 137786308949760 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 16.607583s. I0420 07:14:39.946799 137782131414784 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 07:14:39.946934 137782131414784 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 07:14:39.946999 137782131414784 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0420 07:14:39.948459 137782131414784 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0420 07:14:44.675718 137912096147264 metric_logger.py:196] completed step: 16, seconds: 15.832, TFLOP/s/device: 0.858, Tokens/s/device: 129.358, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0420 07:14:44.823311 137912096147264 metric_logger.py:196] completed step: 17, seconds: 20.768, TFLOP/s/device: 0.654, Tokens/s/device: 98.611, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0420 07:14:44.970898 137912096147264 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.110, Tokens/s/device: 13431.622, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0420 07:14:45.117680 137912096147264 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0420 07:14:45.118372 137912096147264 checkpointing.py:776] Waited 0.0007126331329345703 seconds for step 19 to finish before starting checkpointing. I0420 07:14:45.120511 137912096147264 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 07:14:45.122262 137912096147264 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 19 I0420 07:14:45.123707 137912096147264 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints/19. I0420 07:14:45.435712 137912096147264 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 07:14:45.435806 137912096147264 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 I0420 07:14:45.477113 137912096147264 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042369s I0420 07:14:45.477279 137912096147264 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 33.632 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04586672782897949 s) (per-host) I0420 07:14:45.477329 137912096147264 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.045927s (batch_requests_ready=0.001710s, total_serialization_initiated=0.044152s, others=0.000065s) I0420 07:14:45.477440 137912096147264 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.050033s (all_items=0.000011s, per_item={'items': '0.00001121'}, temp_paths=0.050022) I0420 07:14:45.478147 137912096147264 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.36 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints/19. I0420 07:14:45.478854 137782131414784 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:34:45.478830 I0420 07:14:45.480868 137912096147264 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0420 07:14:45.481113 137786300557056 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 07:14:45.481264 137912096147264 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776669285.1204789, 'wait_for_prev_duration_secs': 7.414817810058594e-05, 'time_between_consecutive_saves_sec': 5.171980381011963, 'checkpointer_blocking_start_time': 1776669285.1223, 'checkpointer_blocking_duration_secs': 0.3566701412200928, 'get_old_steps_start_time': 1776669285.4789896, 'get_old_steps_duration_secs': 2.8848648071289062e-05, 'checkpoint_manager_blocking_start_time': 1776669285.1185963, 'checkpoint_manager_blocking_duration_secs': 0.3626370429992676} I0420 07:14:45.481393 137912096147264 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0420 07:14:45.481441 137912096147264 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 07:14:50.574261 137786308949760 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_15_ocdbt_false/checkpoints/19/items/array_metadatas/process_4 I0420 07:14:55.660848 137782131414784 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 154.417 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.229405641555786 s) (per-host) I0420 07:14:55.660965 137782131414784 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.182022s. I0420 07:15:00.534375 137782131414784 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 15.055393s. I0420 07:15:00.534669 137786300557056 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 07:15:00.534780 137786300557056 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 07:15:00.534820 137786300557056 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0420 07:15:00.536514 137786300557056 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0420 07:15:00.536681 137912096147264 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0420 07:15:00.536824 137912096147264 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 07:15:00.537755 137912096147264 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.127, Tokens/s/device: 13886.346, 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: Mon Apr 20 07:15:12 UTC 2026 EXIT_CODE=0
XPK Start: Tue Apr 21 02:08:12 UTC 2026 2026-04-21 02:08:16.379586: 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 00:00:1776737296.392361 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776737296.396039 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776737296.407340 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776737296.407358 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776737296.407361 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776737296.407363 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-21 02:08:35.752191: 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 02:08:36.263068 139103314573120 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-21 02:08:45,304:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0421 02:08:45.304523 139103314573120 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-21 02:08:45,306:jax._src.distributed:157: Connecting to JAX distributed service on mt-15-ocdbt-false-rs-nn8ey-slice-job-0-0.mt-15-ocdbt-false-rs-nn8ey:8482 I0421 02:08:45.306916 139103314573120 distributed.py:157] Connecting to JAX distributed service on mt-15-ocdbt-false-rs-nn8ey-slice-job-0-0.mt-15-ocdbt-false-rs-nn8ey:8482 I0421 02:08:46.162934 139103314573120 max_utils.py:284] Jax distributed system initialized! I0421 02:08:53.206022 139103314573120 max_utils.py:800] System Information: Jax Version: 0.8.1 I0421 02:08:53.206125 139103314573120 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0421 02:08:53.206166 139103314573120 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 02:08:53.206201 139103314573120 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0421 02:08:53.830069 139103314573120 maxtext_utils.py:1398] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 02:08:53.830355 139103314573120 checkpointing.py:677] Setting up checkpoint logger... I0421 02:08:53.830426 139103314573120 checkpointing.py:233] Creating checkpoint manager with ocdbt=False and zarr3=True I0421 02:08:53.830473 139103314573120 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0421 02:08:53.830885 139103314573120 base_pytree_checkpoint_handler.py:415] Created BasePyTreeCheckpointHandler: use_ocdbt=False, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7e82f8579340>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 02:08:55.110300 139103314573120 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0421 02:08:55.110551 139103314573120 checkpoint_manager.py:709] [process=1][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e709a1ddbe0>}, handler_registry=None I0421 02:08:55.110780 139103314573120 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e709a1ddbe0>` 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 02:08:55.110828 139103314573120 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e70996bb020>` 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 02:08:55.110865 139103314573120 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 0x7e709a1ddbe0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e709a1ddbe0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e70996bb020>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e70996bb020>}). I0421 02:08:55.111414 139103314573120 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0421 02:08:55.111497 139103314573120 async_checkpointer.py:177] [process=1][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7e6fa322c040> timeout: 600 secs and primary_host=0 for async checkpoint writes I0421 02:08:57.841493 138982918711040 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776737183401969848, 'commit_timestamp_nsecs': 1776737198365249227, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints/5/_CHECKPOINT_METADATA I0421 02:08:58.088435 138982918711040 checkpoint_manager.py:1877] Missing metrics for step 5 I0421 02:08:58.655040 139103314573120 checkpoint_manager.py:1818] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints I0421 02:08:58.936076 139103314573120 checkpoint_manager.py:929] [process=1][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_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e70996b6a20> I0421 02:08:58.936250 139103314573120 checkpointing.py:301] Checkpoint manager created! I0421 02:08:59.823739 139103314573120 nnx_wrappers.py:453] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 02:08:59.823842 139103314573120 nnx_wrappers.py:453] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 02:09:00.197824 139103314573120 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 02:09:00.197925 139103314573120 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 02:09:00.213841 139103314573120 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 02:09:00.213906 139103314573120 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 02:09:00.237010 139103314573120 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 02:09:00.237085 139103314573120 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 02:09:00.253120 139103314573120 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 02:09:00.253186 139103314573120 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 02:09:00.269131 139103314573120 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 02:09:00.269198 139103314573120 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 02:09:00.294373 139103314573120 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 02:09:00.294456 139103314573120 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 02:09:00.314038 139103314573120 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 02:09:00.314104 139103314573120 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0421 02:09:00.506298 139103314573120 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0421 02:09:00.506426 139103314573120 checkpointing.py:581] restoring from this run's directory step 9 I0421 02:09:01.442171 139103314573120 checkpointer.py:304] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints/9. I0421 02:09:04.444691 1445 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0421 02:09:14.381864 139103314573120 jax_array_handlers.py:837] [process=1] /jax/orbax/read/worker/io/requested throughput: 138.415 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 11.41690969467163 s) (per-host) I0421 02:09:14.398856 139103314573120 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/read/gbytes_per_sec: 133.893 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 11.819809436798096 s) (per-host) I0421 02:09:14.934773 139103314573120 checkpointer.py:318] Finished restoring checkpoint in 14.17 seconds from gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints/9. I0421 02:09:14.934946 139103314573120 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints', 'checkpointer_start_time': 1776737340.7638013, 'checkpointer_duration_secs': 14.171112298965454, 'checkpoint_manager_start_time': 1776737340.5066702, 'checkpoint_manager_duration_secs': 14.428245067596436} fsdp: 32 I0421 02:09:15.337688 139103314573120 maxtext_utils.py:1501] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0421 02:09:15.337816 139103314573120 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) I0421 02:09:15.337874 139103314573120 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) I0421 02:09:15.337930 139103314573120 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') I0421 02:09:15.337984 139103314573120 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) I0421 02:09:15.338023 139103314573120 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) I0421 02:09:15.338078 139103314573120 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) I0421 02:09:15.338131 139103314573120 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') I0421 02:09:15.338171 139103314573120 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) I0421 02:09:15.338206 139103314573120 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) I0421 02:09:15.338252 139103314573120 maxtext_utils.py:1501] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0421 02:09:15.338296 139103314573120 maxtext_utils.py:1501] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0421 02:09:15.844451 139103314573120 train.py:156] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0421 02:09:15.844547 139103314573120 train.py:156] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0421 02:09:15.859428 139103314573120 train.py:163] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0421 02:09:15.859486 139103314573120 train.py:163] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0421 02:09:26.653321 139103314573120 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. I0421 02:09:26.654219 139103314573120 metric_logger.py:289] number parameters: 1.104 billion I0421 02:09:37.811668 139103314573120 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0421 02:09:37.973112 139103314573120 checkpointing.py:776] Waited 0.16142868995666504 seconds for step 10 to finish before starting checkpointing. I0421 02:09:37.975603 139103314573120 checkpoint_manager.py:2013] [process=1][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 02:09:37.977336 139103314573120 checkpoint_manager.py:1518] [process=1] Saving checkpoint at step 10 I0421 02:09:37.980659 139103314573120 async_checkpointer.py:452] [process=1] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints/10. I0421 02:09:38.998803 139103314573120 signaling_client.py:364] Using JaxDistributedSignalingClient I0421 02:09:38.999775 139103314573120 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0421 02:09:38.999834 139103314573120 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 I0421 02:09:39.256636 139103314573120 base_pytree_checkpoint_handler.py:153] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.257926s I0421 02:09:39.256798 139103314573120 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.904 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5312676429748535 s) (per-host) I0421 02:09:39.256850 139103314573120 base_pytree_checkpoint_handler.py:737] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.531330s (batch_requests_ready=0.271317s, total_serialization_initiated=0.259942s, others=0.000071s) I0421 02:09:39.256920 139103314573120 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.535278s (all_items=0.000020s, per_item={'items': '0.00001979'}, temp_paths=0.535258) I0421 02:09:39.257851 138981148702464 async_checkpointer.py:79] [process=1][thread=async_save] Background save thread started. I0421 02:09:39.258021 139103314573120 async_checkpointer.py:561] Finished blocking save. Time taken: 1.280615s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints/10. I0421 02:09:39.259860 139103314573120 checkpoint_manager.py:1566] [process=1][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0421 02:09:39.260084 138978037208832 async_checkpointer.py:265] [process=1][thread=save_finalize] Waiting for background save thread=async_save. I0421 02:09:39.260178 139103314573120 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776737377.9755857, 'wait_for_prev_duration_secs': 6.198883056640625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776737377.9773731, 'checkpointer_blocking_duration_secs': 1.2807562351226807, 'get_old_steps_start_time': 1776737379.2581484, 'get_old_steps_duration_secs': 2.8133392333984375e-05, 'checkpoint_manager_blocking_start_time': 1776737377.9735687, 'checkpoint_manager_blocking_duration_secs': 1.286576271057129} I0421 02:09:39.260325 139103314573120 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0421 02:09:39.260375 139103314573120 max_utils.py:750] Memstats: After params initialized: I0421 02:09:39.260434 139103314573120 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_2(process=1,(2,0,0,0)) I0421 02:09:39.260469 139103314573120 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_3(process=1,(3,0,0,0)) I0421 02:09:39.260495 139103314573120 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_6(process=1,(2,1,0,0)) I0421 02:09:39.260519 139103314573120 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_7(process=1,(3,1,0,0)) I0421 02:09:39.517140 139103314573120 metric_logger.py:185] completed step: 10, seconds: 11.157, TFLOP/s/device: 1.218, Tokens/s/device: 183.556, total_weights: 65536, loss: 6.833 I0421 02:09:39.706095 139103314573120 metric_logger.py:185] completed step: 11, seconds: 1.704, TFLOP/s/device: 7.973, Tokens/s/device: 1201.767, total_weights: 65536, loss: 6.528 I0421 02:09:39.853716 139103314573120 metric_logger.py:185] completed step: 12, seconds: 0.042, TFLOP/s/device: 324.267, Tokens/s/device: 48877.115, total_weights: 65536, loss: 6.265 I0421 02:09:40.001300 139103314573120 metric_logger.py:185] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.847, Tokens/s/device: 13391.923, total_weights: 65536, loss: 6.045 I0421 02:09:40.006148 139103314573120 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0421 02:09:40.296008 139103314573120 checkpointing.py:776] Waited 0.2898240089416504 seconds for step 15 to finish before starting checkpointing. I0421 02:09:40.298895 139103314573120 checkpoint_manager.py:2024] [process=1][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 02:09:44.976019 138978606958336 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints/10/items/array_metadatas/process_1 I0421 02:09:50.863846 138981148702464 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/gbytes_per_sec: 130.134 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 12.138274192810059 s) (per-host) I0421 02:09:50.863963 138981148702464 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 11.606001s. I0421 02:09:56.082474 138981148702464 async_checkpointer.py:144] [process=1][thread=async_save] Background save thread done. Time taken: 16.824495s. I0421 02:09:56.082744 138978037208832 async_checkpointer.py:273] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 02:09:56.082860 138978037208832 async_checkpointer.py:283] [process=1][thread=save_finalize] No errors found in background save thread=async_save. I0421 02:09:56.082909 138978037208832 checkpoint_manager.py:2133] [process=1][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0421 02:09:56.084698 138978037208832 checkpoint_manager.py:2142] [process=1][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0421 02:09:56.084869 139103314573120 checkpoint_manager.py:2036] [process=1][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0421 02:09:56.085009 139103314573120 checkpoint_manager.py:1458] Waiting for previous save to complete took 15.786118 seconds. If this number is high, consider checkpointing less frequently. I0421 02:09:56.087084 139103314573120 checkpoint_manager.py:1518] [process=1] Saving checkpoint at step 15 I0421 02:09:56.091300 139103314573120 async_checkpointer.py:452] [process=1] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints/15. I0421 02:09:57.468970 139103314573120 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0421 02:09:57.469069 139103314573120 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 I0421 02:09:57.514996 139103314573120 base_pytree_checkpoint_handler.py:153] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.047420s I0421 02:09:57.515177 139103314573120 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.119 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.7280905246734619 s) (per-host) I0421 02:09:57.515233 139103314573120 base_pytree_checkpoint_handler.py:737] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.728158s (batch_requests_ready=0.678752s, total_serialization_initiated=0.049329s, others=0.000076s) I0421 02:09:57.515324 139103314573120 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.732999s (all_items=0.000014s, per_item={'items': '0.00001383'}, temp_paths=0.732985) I0421 02:09:57.516247 138978606958336 async_checkpointer.py:79] [process=1][thread=async_save] Background save thread started. I0421 02:09:57.516420 139103314573120 async_checkpointer.py:561] Finished blocking save. Time taken: 1.429250s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints/15. I0421 02:09:57.518236 139103314573120 checkpoint_manager.py:1566] [process=1][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0421 02:09:57.518495 138978037208832 async_checkpointer.py:265] [process=1][thread=save_finalize] Waiting for background save thread=async_save. I0421 02:09:57.518622 139103314573120 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776737380.298867, 'wait_for_prev_duration_secs': 15.786118268966675, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776737396.0871239, 'checkpointer_blocking_duration_secs': 1.4294054508209229, 'get_old_steps_start_time': 1776737397.5165486, 'get_old_steps_duration_secs': 2.6226043701171875e-05, 'checkpoint_manager_blocking_start_time': 1776737380.2967203, 'checkpoint_manager_blocking_duration_secs': 17.221871376037598} I0421 02:09:57.518781 139103314573120 checkpointing.py:408] Started an asynchronous checkpoint save for step 15 I0421 02:09:57.519263 139103314573120 metric_logger.py:185] completed step: 14, seconds: 0.148, TFLOP/s/device: 91.866, Tokens/s/device: 13847.007, total_weights: 65536, loss: 5.864 I0421 02:09:57.524852 139103314573120 metric_logger.py:185] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.128, Tokens/s/device: 13886.535, total_weights: 65536, loss: 5.719 I0421 02:10:15.318651 139103314573120 metric_logger.py:185] completed step: 16, seconds: 17.517, TFLOP/s/device: 0.776, Tokens/s/device: 116.912, total_weights: 65536, loss: 5.605 I0421 02:10:15.466226 139103314573120 metric_logger.py:185] completed step: 17, seconds: 17.648, TFLOP/s/device: 0.770, Tokens/s/device: 116.049, total_weights: 65536, loss: 5.517 I0421 02:10:15.613791 139103314573120 metric_logger.py:185] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.267, Tokens/s/device: 13455.272, total_weights: 65536, loss: 5.448 I0421 02:10:15.760856 139103314573120 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0421 02:10:15.761547 139103314573120 checkpointing.py:776] Waited 0.0007104873657226562 seconds for step 19 to finish before starting checkpointing. I0421 02:10:15.763494 139103314573120 checkpoint_manager.py:2024] [process=1][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 02:10:20.141654 138980217571072 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints/15/items/array_metadatas/process_1 I0421 02:10:25.498853 138978606958336 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/gbytes_per_sec: 55.016 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 28.711719751358032 s) (per-host) I0421 02:10:25.498973 138978606958336 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 27.982616s. I0421 02:10:32.956008 138978606958336 async_checkpointer.py:144] [process=1][thread=async_save] Background save thread done. Time taken: 35.439637s. I0421 02:10:32.956259 138978037208832 async_checkpointer.py:273] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 02:10:32.956373 138978037208832 async_checkpointer.py:283] [process=1][thread=save_finalize] No errors found in background save thread=async_save. I0421 02:10:32.956446 138978037208832 checkpoint_manager.py:2133] [process=1][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0421 02:10:32.958197 138978037208832 checkpoint_manager.py:2142] [process=1][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0421 02:10:32.958368 139103314573120 checkpoint_manager.py:2036] [process=1][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0421 02:10:32.958501 139103314573120 checkpoint_manager.py:1458] Waiting for previous save to complete took 17.195018 seconds. If this number is high, consider checkpointing less frequently. I0421 02:10:32.960363 139103314573120 checkpoint_manager.py:1518] [process=1] Saving checkpoint at step 19 I0421 02:10:32.963909 139103314573120 async_checkpointer.py:452] [process=1] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints/19. I0421 02:10:33.917510 139103314573120 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0421 02:10:33.917605 139103314573120 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 I0421 02:10:33.964962 139103314573120 base_pytree_checkpoint_handler.py:153] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.048423s I0421 02:10:33.965132 139103314573120 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.113 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.30167651176452637 s) (per-host) I0421 02:10:33.965184 139103314573120 base_pytree_checkpoint_handler.py:737] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.301739s (batch_requests_ready=0.251149s, total_serialization_initiated=0.050522s, others=0.000067s) I0421 02:10:33.965319 139103314573120 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.305790s (all_items=0.000012s, per_item={'items': '0.00001240'}, temp_paths=0.305777) I0421 02:10:33.966250 138976434296576 async_checkpointer.py:79] [process=1][thread=async_save] Background save thread started. I0421 02:10:33.966429 139103314573120 async_checkpointer.py:561] Finished blocking save. Time taken: 1.005971s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints/19. I0421 02:10:34.371382 139103314573120 checkpoint_manager.py:1566] [process=1][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0421 02:10:34.371761 138978037208832 async_checkpointer.py:265] [process=1][thread=save_finalize] Waiting for background save thread=async_save. I0421 02:10:34.371934 139103314573120 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776737415.763467, 'wait_for_prev_duration_secs': 17.19501781463623, 'time_between_consecutive_saves_sec': 19.678728818893433, 'checkpointer_blocking_start_time': 1776737432.960416, 'checkpointer_blocking_duration_secs': 1.0061240196228027, 'get_old_steps_start_time': 1776737433.9665525, 'get_old_steps_duration_secs': 2.1696090698242188e-05, 'checkpoint_manager_blocking_start_time': 1776737415.7617762, 'checkpoint_manager_blocking_duration_secs': 18.610122680664062} I0421 02:10:34.372040 139103314573120 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0421 02:10:34.372082 139103314573120 checkpoint_manager.py:2024] [process=1][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 02:10:39.550290 138980217571072 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_test_pipeline_scan_nnx_20260421_005327/linen_xpk_test_pipeline_scan_nnx_20260421_005327_15_ocdbt_false/checkpoints/19/items/array_metadatas/process_1 I0421 02:10:44.675843 138976434296576 base_pytree_checkpoint_handler.py:129] [process=1] /jax/checkpoint/write/gbytes_per_sec: 143.439 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 11.012347221374512 s) (per-host) I0421 02:10:44.675957 138976434296576 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.709595s. I0421 02:10:50.187849 138976434296576 async_checkpointer.py:144] [process=1][thread=async_save] Background save thread done. Time taken: 16.221472s. I0421 02:10:50.188075 138978037208832 async_checkpointer.py:273] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 02:10:50.188133 138978037208832 async_checkpointer.py:283] [process=1][thread=save_finalize] No errors found in background save thread=async_save. I0421 02:10:50.188198 138978037208832 checkpoint_manager.py:2133] [process=1][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0421 02:10:50.190311 138978037208832 checkpoint_manager.py:2142] [process=1][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0421 02:10:50.190476 139103314573120 checkpoint_manager.py:2036] [process=1][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0421 02:10:50.190617 139103314573120 checkpoint_manager.py:2013] [process=1][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 02:10:50.191281 139103314573120 metric_logger.py:185] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.132, Tokens/s/device: 13887.194, 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: Tue Apr 21 02:11:02 UTC 2026 EXIT_CODE=0