XPK Start: Fri Apr 17 05:17:57 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-17 05:18:21.517386: 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) I0417 05:18:22.049295 139703165269824 max_utils.py:273] Attempting to initialize the jax distributed system... I0417 05:18:31.090751 139703165269824 distributed.py:149] Starting JAX distributed service on [::]:8482 I0417 05:18:31.093132 139703165269824 distributed.py:172] Connecting to JAX distributed service on mt-09-pdb-lt-1-4kugj-slice-job-0-0.mt-09-pdb-lt-1-4kugj:8482 I0417 05:18:32.166475 139703165269824 max_utils.py:284] Jax distributed system initialized! I0417 05:18:38.336328 139703165269824 max_utils.py:800] System Information: Jax Version: 0.9.2 I0417 05:18:38.336429 139703165269824 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0417 05:18:38.336468 139703165269824 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 I0417 05:18:38.336502 139703165269824 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0417 05:18:38.999139 139703165269824 maxtext_utils.py:1687] Num_devices: 32, shape (1, 1, 1, 8, 1, 1, 1, 1, 4, 1, 1, 1, 1) I0417 05:18:39.112027 139703165269824 checkpointing.py:688] Setting up checkpoint logger... I0417 05:18:39.112141 139703165269824 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0417 05:18:39.112187 139703165269824 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0417 05:18:39.112378 139703165269824 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 0x7f0e921180b0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0417 05:18:42.432976 139703165269824 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0417 05:18:42.433210 139703165269824 checkpoint_manager.py:708] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7efa645e4da0>}, handler_registry=None I0417 05:18:42.433447 139703165269824 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7efa645e4da0>` 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`. I0417 05:18:42.433495 139703165269824 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7efa645920c0>` 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`. I0417 05:18:42.433530 139703165269824 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 0x7efa645e4da0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7efa645e4da0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7efa645920c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7efa645920c0>}). I0417 05:18:42.433860 139703165269824 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0417 05:18:42.433933 139703165269824 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7efa64521ee0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0417 05:18:43.540656 139703165269824 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints I0417 05:18:44.422399 139703165269824 checkpoint_manager.py:929] [process=0][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7efa645bd130> I0417 05:18:44.422567 139703165269824 checkpointing.py:302] Checkpoint manager created! I0417 05:18:45.374580 139703165269824 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0417 05:18:45.374698 139703165269824 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 8 tensor: 4 I0417 05:18:46.775247 139703165269824 maxtext_utils.py:1805] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: ('tensor',) I0417 05:18:46.775359 139703165269824 maxtext_utils.py:1805] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, 'tensor') I0417 05:18:46.775412 139703165269824 maxtext_utils.py:1805] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, 'tensor') I0417 05:18:46.775455 139703165269824 maxtext_utils.py:1805] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: ('tensor', None, 'fsdp') I0417 05:18:46.775492 139703165269824 maxtext_utils.py:1805] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: ('tensor', None) I0417 05:18:46.775527 139703165269824 maxtext_utils.py:1805] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: ('tensor', None) I0417 05:18:46.775563 139703165269824 maxtext_utils.py:1805] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, 'tensor', None) I0417 05:18:46.775599 139703165269824 maxtext_utils.py:1805] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: ('tensor', None, None, 'fsdp') I0417 05:18:46.775635 139703165269824 maxtext_utils.py:1805] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, 'tensor', None) I0417 05:18:46.775669 139703165269824 maxtext_utils.py:1805] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, 'tensor', None) I0417 05:18:46.775702 139703165269824 maxtext_utils.py:1805] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', 'tensor') I0417 05:18:46.775733 139703165269824 maxtext_utils.py:1805] token_embedder/embedding/value Shape: float32[32000,2048] Physical: ('tensor', 'fsdp') I0417 05:18:47.176245 139703165269824 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 05:18:47.176336 139703165269824 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor'). I0417 05:18:47.181624 139703165269824 nnx_decoders.py:465] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 05:18:47.181679 139703165269824 nnx_decoders.py:465] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor'). I0417 05:18:47.197983 139703165269824 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 05:18:47.198043 139703165269824 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor'). I0417 05:18:47.213701 139703165269824 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 05:18:47.213759 139703165269824 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, 'tensor'). I0417 05:18:47.237506 139703165269824 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 05:18:47.237575 139703165269824 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0417 05:18:47.253277 139703165269824 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 05:18:47.253340 139703165269824 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0417 05:18:47.268820 139703165269824 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 05:18:47.268889 139703165269824 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0417 05:18:47.298210 139703165269824 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0417 05:18:47.298284 139703165269824 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, 'tensor', None). I0417 05:18:47.317981 139703165269824 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0417 05:18:47.318044 139703165269824 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, 'tensor'). I0417 05:19:02.504500 139703165269824 max_utils.py:791] Total memory size: 0.9 GB, Output size: 0.4 GB, Temp size: 0.5 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0417 05:19:02.513460 139703165269824 metric_logger.py:289] number parameters: 1.104 billion I0417 05:19:03.780448 139703165269824 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0417 05:19:17.794085 139703165269824 checkpointing.py:798] Waited 14.013618469238281 seconds for step 0 to finish before starting checkpointing. I0417 05:19:17.796447 139703165269824 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 05:19:17.798172 139703165269824 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0417 05:19:17.799707 139703165269824 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/0. I0417 05:19:18.546342 139703165269824 signaling_client.py:364] Using JaxDistributedSignalingClient I0417 05:19:18.565865 139703165269824 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0417 05:19:18.565937 139703165269824 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 I0417 05:19:18.824137 139703165269824 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.259763s I0417 05:19:18.825030 139703165269824 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.700 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2706174850463867 s) (per-host) I0417 05:19:18.825114 139703165269824 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.270722s (batch_requests_ready=0.003014s, total_serialization_initiated=0.266881s, others=0.000827s) I0417 05:19:18.825205 139703165269824 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.274817s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=0.274798) I0417 05:19:18.826009 139703165269824 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.03 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/0. I0417 05:19:18.826274 139575557089024 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 05:39:18.826250 I0417 05:19:18.829237 139575548696320 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/0 I0417 05:19:18.836743 139703165269824 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0417 05:19:18.837041 139575028610816 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 05:19:18.837199 139703165269824 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776403157.7964296, 'wait_for_prev_duration_secs': 5.984306335449219e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776403157.798211, 'checkpointer_blocking_duration_secs': 1.0281658172607422, 'get_old_steps_start_time': 1776403158.8264039, 'get_old_steps_duration_secs': 2.8848648071289062e-05, 'checkpoint_manager_blocking_start_time': 1776403157.7945006, 'checkpoint_manager_blocking_duration_secs': 1.0426561832427979} I0417 05:19:18.837307 139703165269824 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0417 05:19:18.837375 139703165269824 max_utils.py:750] Memstats: After params initialized: I0417 05:19:18.837427 139703165269824 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0417 05:19:18.837460 139703165269824 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0417 05:19:18.837487 139703165269824 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0417 05:19:18.837512 139703165269824 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0417 05:19:19.187070 139703165269824 metric_logger.py:185] completed step: 0, seconds: 0.909, TFLOP/s/device: 3.737, Tokens/s/device: 563.332, total_weights: 16384, loss: 10.864 I0417 05:19:19.190315 139703165269824 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/tensorboard/' I0417 05:19:19.612495 139703165269824 metric_logger.py:185] completed step: 1, seconds: 15.403, TFLOP/s/device: 0.221, Tokens/s/device: 33.240, total_weights: 16384, loss: 10.864 I0417 05:19:19.665166 139703165269824 metric_logger.py:185] completed step: 2, seconds: 0.426, TFLOP/s/device: 7.967, Tokens/s/device: 1200.798, total_weights: 16384, loss: 9.760 I0417 05:19:19.723772 139703165269824 metric_logger.py:185] completed step: 3, seconds: 0.012, TFLOP/s/device: 280.147, Tokens/s/device: 42226.804, total_weights: 16384, loss: 8.820 I0417 05:19:20.608518 139575531910912 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/0/items I0417 05:19:20.755268 139575011051264 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776403160358306146, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776403160358306146, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/0/_CHECKPOINT_METADATA I0417 05:19:22.172324 2582 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0417 05:19:24.370251 139578222335744 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/0/items/array_metadatas/process_0 I0417 05:19:33.878317 139703165269824 metric_logger.py:185] completed step: 4, seconds: 0.054, TFLOP/s/device: 62.612, Tokens/s/device: 9437.614, total_weights: 16384, loss: 7.947 I0417 05:19:33.888674 139703165269824 metric_logger.py:185] completed step: 5, seconds: 0.058, TFLOP/s/device: 58.747, Tokens/s/device: 8854.914, total_weights: 16384, loss: 7.209 I0417 05:19:33.944078 139703165269824 metric_logger.py:185] completed step: 6, seconds: 14.153, TFLOP/s/device: 0.240, Tokens/s/device: 36.175, total_weights: 16384, loss: 6.651 I0417 05:19:34.002735 139703165269824 metric_logger.py:185] completed step: 7, seconds: 0.008, TFLOP/s/device: 430.517, Tokens/s/device: 64892.269, total_weights: 16384, loss: 6.267 I0417 05:19:34.061443 139703165269824 metric_logger.py:185] completed step: 8, seconds: 0.057, TFLOP/s/device: 60.050, Tokens/s/device: 9051.374, total_weights: 16384, loss: 6.030 I0417 05:19:34.120959 139703165269824 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0417 05:19:34.121629 139703165269824 checkpointing.py:798] Waited 0.0006783008575439453 seconds for step 9 to finish before starting checkpointing. I0417 05:19:34.123644 139703165269824 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 05:19:57.901782 139575515125504 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 35.735093s (commit=34.538898s, array_metadata_write=1.196195s) I0417 05:19:57.902991 139575557089024 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 40.144 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 39.348551511764526 s) (per-host) I0417 05:19:57.903127 139575557089024 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 39.076776s. I0417 05:19:58.637953 139575557089024 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776403160358306146, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/0/_CHECKPOINT_METADATA I0417 05:19:59.793063 139575557089024 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000423s. I0417 05:19:59.816983 139575011051264 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776403160358306146, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/0/_CHECKPOINT_METADATA I0417 05:20:00.013176 139575557089024 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 05:20:00.929709 139575557089024 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.147846s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/0/items I0417 05:20:00.931340 139575557089024 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/0/items I0417 05:20:01.476865 139575557089024 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/0 I0417 05:20:03.282862 139575557089024 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/0`. I0417 05:20:03.283721 139575557089024 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 45.49 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/0 I0417 05:20:03.284931 139575557089024 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 44.458578s. I0417 05:20:03.285051 139575028610816 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 05:20:03.285126 139575028610816 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 05:20:03.285204 139575028610816 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0417 05:20:03.288505 139575028610816 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0417 05:20:03.288700 139703165269824 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0417 05:20:03.288858 139703165269824 checkpoint_manager.py:1452] Waiting for previous save to complete took 29.165197 seconds. If this number is high, consider checkpointing less frequently. I0417 05:20:03.290713 139703165269824 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0417 05:20:03.292669 139703165269824 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/9. I0417 05:20:03.595798 139703165269824 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0417 05:20:03.595902 139703165269824 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 I0417 05:20:03.623133 139703165269824 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.028684s I0417 05:20:03.623902 139703165269824 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 41.373 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03728485107421875 s) (per-host) I0417 05:20:03.623971 139703165269824 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.037369s (batch_requests_ready=0.002606s, total_serialization_initiated=0.034044s, others=0.000718s) I0417 05:20:03.624045 139703165269824 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.041349s (all_items=0.000016s, per_item={'items': '0.00001574'}, temp_paths=0.041333) I0417 05:20:03.624739 139703165269824 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/9. I0417 05:20:03.625082 139574483347200 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 05:40:03.625042 I0417 05:20:03.627080 139703165269824 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0417 05:20:03.627315 139575028610816 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 05:20:03.627418 139703165269824 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776403174.1236107, 'wait_for_prev_duration_secs': 29.165197372436523, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776403203.29075, 'checkpointer_blocking_duration_secs': 0.33447742462158203, 'get_old_steps_start_time': 1776403203.6252549, 'get_old_steps_duration_secs': 3.266334533691406e-05, 'checkpoint_manager_blocking_start_time': 1776403174.1218653, 'checkpoint_manager_blocking_duration_secs': 29.505520820617676} I0417 05:20:03.627587 139703165269824 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0417 05:20:03.627632 139703165269824 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 05:20:03.824485 139575557089024 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/9 I0417 05:20:05.987712 139574491739904 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/9/items I0417 05:20:09.983926 139575531910912 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/9/items/array_metadatas/process_0 I0417 05:20:47.374976 139575515125504 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 40.084048s (commit=38.623690s, array_metadata_write=1.460358s) I0417 05:20:47.376104 139574483347200 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 36.073 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.789456605911255 s) (per-host) I0417 05:20:47.376246 139574483347200 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.751046s. I0417 05:20:49.208945 139574483347200 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000416s. I0417 05:20:49.440427 139574483347200 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 05:20:50.344908 139574483347200 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.138089s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/9/items I0417 05:20:50.346548 139574483347200 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/9/items I0417 05:20:50.872693 139574483347200 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/9 I0417 05:20:52.727250 139574483347200 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/9`. I0417 05:20:52.728065 139574483347200 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 49.44 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_023710/nnx_xpk_feat_nnx_post_train_fixes_20260417_023710_09_pdb_lt_1/checkpoints/9 I0417 05:20:52.729511 139574483347200 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 49.104309s. I0417 05:20:52.729721 139575028610816 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 05:20:52.729848 139575028610816 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 05:20:52.729922 139575028610816 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0417 05:20:52.732664 139575028610816 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0417 05:20:52.732855 139703165269824 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0417 05:20:52.733025 139703165269824 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 05:20:52.733754 139703165269824 metric_logger.py:185] completed step: 9, seconds: 0.058, TFLOP/s/device: 58.175, Tokens/s/device: 8768.775, total_weights: 16384, loss: 5.890 Per train step: Total TFLOPs: 3.40 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Fri Apr 17 05:21:03 UTC 2026 EXIT_CODE=0