XPK Start: Sun Apr 19 02:48:04 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-19 02:48:28.385928: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0419 02:48:28.564122 138112449099584 max_utils.py:273] Attempting to initialize the jax distributed system... I0419 02:48:37.604912 138112449099584 distributed.py:149] Starting JAX distributed service on [::]:8482 I0419 02:48:37.607191 138112449099584 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-sv-9x21a-slice-job-0-0.mt-16-zarr3-false-sv-9x21a:8482 I0419 02:48:39.125155 138112449099584 max_utils.py:284] Jax distributed system initialized! I0419 02:48:45.429789 138112449099584 max_utils.py:800] System Information: Jax Version: 0.9.2 I0419 02:48:45.429906 138112449099584 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0419 02:48:45.429946 138112449099584 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 I0419 02:48:45.429982 138112449099584 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0419 02:48:46.125030 138112449099584 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 02:48:46.236570 138112449099584 checkpointing.py:688] Setting up checkpoint logger... I0419 02:48:46.236683 138112449099584 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=False I0419 02:48:46.236728 138112449099584 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0419 02:48:46.236933 138112449099584 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7d9c33f1b800>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0419 02:48:50.837779 138112449099584 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0419 02:48:50.838026 138112449099584 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 0x7d87d850e7b0>}, handler_registry=None I0419 02:48:50.838268 138112449099584 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d87d850e7b0>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0419 02:48:50.838316 138112449099584 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d87f0427f80>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0419 02:48:50.838351 138112449099584 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 0x7d87d850e7b0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d87d850e7b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d87f0427f80>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d87f0427f80>}). I0419 02:48:50.838689 138112449099584 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0419 02:48:50.838759 138112449099584 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7d87d8644fe0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0419 02:48:51.943088 138112449099584 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints I0419 02:48:52.389558 138112449099584 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=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/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7d87d86cb4a0> I0419 02:48:52.389728 138112449099584 checkpointing.py:302] Checkpoint manager created! I0419 02:48:53.338749 138112449099584 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0419 02:48:53.338876 138112449099584 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0419 02:48:54.708164 138112449099584 maxtext_utils.py:1836] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0419 02:48:54.708279 138112449099584 maxtext_utils.py:1836] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0419 02:48:54.708331 138112449099584 maxtext_utils.py:1836] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0419 02:48:54.708373 138112449099584 maxtext_utils.py:1836] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0419 02:48:54.708409 138112449099584 maxtext_utils.py:1836] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0419 02:48:54.708441 138112449099584 maxtext_utils.py:1836] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0419 02:48:54.708474 138112449099584 maxtext_utils.py:1836] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0419 02:48:54.708508 138112449099584 maxtext_utils.py:1836] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0419 02:48:54.708540 138112449099584 maxtext_utils.py:1836] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0419 02:48:54.708571 138112449099584 maxtext_utils.py:1836] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0419 02:48:54.708603 138112449099584 maxtext_utils.py:1836] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0419 02:48:54.708633 138112449099584 maxtext_utils.py:1836] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0419 02:48:55.120132 138112449099584 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 02:48:55.120224 138112449099584 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 02:48:55.125505 138112449099584 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 02:48:55.125564 138112449099584 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 02:48:55.141746 138112449099584 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 02:48:55.141829 138112449099584 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 02:48:55.157661 138112449099584 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 02:48:55.157726 138112449099584 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 02:48:55.182607 138112449099584 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 02:48:55.182680 138112449099584 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 02:48:55.198374 138112449099584 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 02:48:55.198446 138112449099584 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 02:48:55.214239 138112449099584 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 02:48:55.214305 138112449099584 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 02:48:55.243592 138112449099584 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0419 02:48:55.243669 138112449099584 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 02:48:55.263347 138112449099584 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0419 02:48:55.263416 138112449099584 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0419 02:49:06.385968 138112449099584 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0419 02:49:06.395116 138112449099584 metric_logger.py:301] number parameters: 1.104 billion I0419 02:49:18.070879 138112449099584 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0419 02:49:18.223493 138112449099584 checkpointing.py:798] Waited 0.15259957313537598 seconds for step 0 to finish before starting checkpointing. I0419 02:49:18.225960 138112449099584 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 02:49:18.227675 138112449099584 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0419 02:49:18.229079 138112449099584 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/0. I0419 02:49:18.951999 138112449099584 signaling_client.py:364] Using JaxDistributedSignalingClient I0419 02:49:18.971410 138112449099584 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0419 02:49:18.971482 138112449099584 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0419 02:49:19.244494 137983474792192 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/0 I0419 02:49:19.247319 138112449099584 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.277311s I0419 02:49:19.248235 138112449099584 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.348 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2884640693664551 s) (per-host) I0419 02:49:19.248311 138112449099584 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.288559s (batch_requests_ready=0.003243s, total_serialization_initiated=0.284486s, others=0.000830s) I0419 02:49:19.248405 138112449099584 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.292500s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=0.292484) I0419 02:49:19.249242 138112449099584 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.02 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/0. I0419 02:49:19.249522 137983961306880 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 03:09:19.249489 I0419 02:49:19.254147 138112449099584 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0419 02:49:19.254438 137979678938880 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 02:49:19.254593 138112449099584 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776566958.2259414, 'wait_for_prev_duration_secs': 5.984306335449219e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776566958.2277153, 'checkpointer_blocking_duration_secs': 1.0219488143920898, 'get_old_steps_start_time': 1776566959.249689, 'get_old_steps_duration_secs': 2.7894973754882812e-05, 'checkpoint_manager_blocking_start_time': 1776566958.2239475, 'checkpoint_manager_blocking_duration_secs': 1.0306057929992676} I0419 02:49:19.254696 138112449099584 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0419 02:49:19.254767 138112449099584 max_utils.py:750] Memstats: After params initialized: I0419 02:49:19.254829 138112449099584 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0419 02:49:19.254864 138112449099584 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0419 02:49:19.254892 138112449099584 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0419 02:49:19.254917 138112449099584 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0419 02:49:19.592961 138112449099584 metric_logger.py:196] completed step: 0, seconds: 11.375, TFLOP/s/device: 1.194, Tokens/s/device: 180.037, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52786.469 I0419 02:49:19.596451 138112449099584 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/tensorboard/' I0419 02:49:20.449919 138112449099584 metric_logger.py:196] completed step: 1, seconds: 1.518, TFLOP/s/device: 8.949, Tokens/s/device: 1348.836, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52786.469 I0419 02:49:20.591885 138112449099584 metric_logger.py:196] completed step: 2, seconds: 0.858, TFLOP/s/device: 15.837, Tokens/s/device: 2387.155, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38565.637 I0419 02:49:20.739487 138112449099584 metric_logger.py:196] completed step: 3, seconds: 0.013, TFLOP/s/device: 1041.479, Tokens/s/device: 156982.983, total_weights: 65536, loss: 9.979, lm_loss: 9.979, perplexity: 21564.932 I0419 02:49:20.752204 138112449099584 checkpointing.py:794] Waiting for step 5 to finish before checkpoint... I0419 02:49:21.034354 138112449099584 checkpointing.py:798] Waited 0.2821176052093506 seconds for step 5 to finish before starting checkpointing. I0419 02:49:21.037401 138112449099584 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 02:49:21.883719 137983458006784 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/0/items I0419 02:49:22.044750 137979129071360 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776566961626090500, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776566961626090500, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/0/_CHECKPOINT_METADATA I0419 02:49:23.015757 2902 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0419 02:49:25.166681 137980739872512 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_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/0/items/array_metadatas/process_0 I0419 02:50:04.805497 137979687331584 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.801755s (commit=40.153378s, array_metadata_write=1.648377s) I0419 02:50:04.806993 137983961306880 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.454 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.84717535972595 s) (per-host) I0419 02:50:04.807165 137983961306880 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.557529s. I0419 02:50:05.542371 137983961306880 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776566961626090500, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/0/_CHECKPOINT_METADATA I0419 02:50:06.287878 137979129071360 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776566961626090500, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/0/_CHECKPOINT_METADATA I0419 02:50:07.085091 137983961306880 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000776s. I0419 02:50:08.435165 137983961306880 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.767678s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/0/items I0419 02:50:08.437327 137983961306880 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/0/items I0419 02:50:09.419771 137983961306880 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/0 I0419 02:50:11.262035 137983961306880 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_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/0`. I0419 02:50:11.263183 137983961306880 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 53.04 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/0 I0419 02:50:11.264746 137983961306880 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 52.015107s. I0419 02:50:11.265020 137979678938880 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 02:50:11.265179 137979678938880 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 02:50:11.265289 137979678938880 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0419 02:50:11.268425 137979678938880 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0419 02:50:11.268764 138112449099584 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. W0419 02:50:11.268957 138112449099584 checkpoint_manager.py:1452] Waiting for previous save to complete took 50.231565 seconds. If this number is high, consider checkpointing less frequently. I0419 02:50:11.270852 138112449099584 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 5 I0419 02:50:11.273375 138112449099584 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/5. I0419 02:50:12.026695 138112449099584 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0419 02:50:12.026845 138112449099584 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0419 02:50:12.069595 138112449099584 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.045861s I0419 02:50:12.070788 138112449099584 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 23.410 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.06589317321777344 s) (per-host) I0419 02:50:12.070943 138112449099584 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.066075s (batch_requests_ready=0.006801s, total_serialization_initiated=0.058141s, others=0.001133s) I0419 02:50:12.071123 138112449099584 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.071127s (all_items=0.000026s, per_item={'items': '0.00002551'}, temp_paths=0.071102) I0419 02:50:12.072304 138112449099584 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.80 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/5. I0419 02:50:12.072739 137979678938880 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 03:10:12.072698 I0419 02:50:12.075248 138112449099584 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0419 02:50:12.075578 137983424435968 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 02:50:12.075803 138112449099584 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776566961.0373473, 'wait_for_prev_duration_secs': 50.23156476020813, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776567011.270922, 'checkpointer_blocking_duration_secs': 0.8020164966583252, 'get_old_steps_start_time': 1776567012.0729666, 'get_old_steps_duration_secs': 4.1484832763671875e-05, 'checkpoint_manager_blocking_start_time': 1776566961.0348778, 'checkpoint_manager_blocking_duration_secs': 51.04087162017822} I0419 02:50:12.075960 138112449099584 checkpointing.py:409] Started an asynchronous checkpoint save for step 5 I0419 02:50:12.247879 137983961306880 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/5 I0419 02:50:14.473202 137982929528576 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/5/items I0419 02:50:18.604228 137983441221376 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_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/5/items/array_metadatas/process_0 I0419 02:50:35.799700 138112449099584 metric_logger.py:196] completed step: 4, seconds: 0.143, TFLOP/s/device: 95.043, Tokens/s/device: 14325.886, total_weights: 65536, loss: 9.457, lm_loss: 9.457, perplexity: 12802.436 I0419 02:50:35.811134 138112449099584 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.145, Tokens/s/device: 13889.077, total_weights: 65536, loss: 8.955, lm_loss: 8.955, perplexity: 7748.778 I0419 02:50:35.955058 138112449099584 metric_logger.py:196] completed step: 6, seconds: 75.058, TFLOP/s/device: 0.181, Tokens/s/device: 27.285, total_weights: 65536, loss: 8.465, lm_loss: 8.465, perplexity: 4743.984 I0419 02:50:36.102677 138112449099584 metric_logger.py:196] completed step: 7, seconds: 0.010, TFLOP/s/device: 1317.859, Tokens/s/device: 198642.095, total_weights: 65536, loss: 7.998, lm_loss: 7.998, perplexity: 2976.266 I0419 02:50:36.250375 138112449099584 metric_logger.py:196] completed step: 8, seconds: 0.144, TFLOP/s/device: 94.277, Tokens/s/device: 14210.479, total_weights: 65536, loss: 7.567, lm_loss: 7.567, perplexity: 1933.479 I0419 02:50:36.398494 138112449099584 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0419 02:50:36.399311 138112449099584 checkpointing.py:798] Waited 0.0008285045623779297 seconds for step 9 to finish before starting checkpointing. I0419 02:50:36.401450 138112449099584 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 02:50:57.409682 137983432828672 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.409843s (commit=40.224915s, array_metadata_write=1.184928s) I0419 02:50:57.410828 137979678938880 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.788 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.40591311454773 s) (per-host) I0419 02:50:57.410955 137979678938880 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.338055s. I0419 02:50:59.687446 137979678938880 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000399s. I0419 02:51:01.165472 137979678938880 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.154544s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/5/items I0419 02:51:01.167045 137979678938880 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/5/items I0419 02:51:02.227260 137979678938880 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/5 I0419 02:51:04.049984 137979678938880 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_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/5`. I0419 02:51:04.050828 137979678938880 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 52.78 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/5 I0419 02:51:04.052208 137979678938880 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 51.979305s. I0419 02:51:04.052419 137983424435968 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 02:51:04.052535 137983424435968 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 02:51:04.052603 137983424435968 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0419 02:51:04.054499 137983424435968 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0419 02:51:04.054671 138112449099584 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0419 02:51:04.054818 138112449099584 checkpoint_manager.py:1452] Waiting for previous save to complete took 27.653359 seconds. If this number is high, consider checkpointing less frequently. I0419 02:51:04.056391 138112449099584 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0419 02:51:04.058567 138112449099584 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/9. I0419 02:51:04.778109 138112449099584 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0419 02:51:04.778204 138112449099584 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0419 02:51:04.804493 138112449099584 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027837s I0419 02:51:04.805337 138112449099584 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 42.290 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03647589683532715 s) (per-host) I0419 02:51:04.805414 138112449099584 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.036571s (batch_requests_ready=0.002778s, total_serialization_initiated=0.033032s, others=0.000761s) I0419 02:51:04.805500 138112449099584 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.040618s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=0.040601) I0419 02:51:04.806270 138112449099584 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/9. I0419 02:51:04.806537 137982921135872 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 03:11:04.806504 I0419 02:51:04.812898 138112449099584 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0419 02:51:04.813200 137983424435968 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 02:51:04.813360 138112449099584 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776567036.4014165, 'wait_for_prev_duration_secs': 27.653358697891235, 'time_between_consecutive_saves_sec': 25.132920742034912, 'checkpointer_blocking_start_time': 1776567064.0564296, 'checkpointer_blocking_duration_secs': 0.7502386569976807, 'get_old_steps_start_time': 1776567064.8066885, 'get_old_steps_duration_secs': 2.7179718017578125e-05, 'checkpoint_manager_blocking_start_time': 1776567036.3995686, 'checkpoint_manager_blocking_duration_secs': 28.413758993148804} I0419 02:51:04.813470 138112449099584 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0419 02:51:04.813513 138112449099584 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 02:51:05.005847 137979678938880 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/9 I0419 02:51:06.733892 137982937921280 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/9/items I0419 02:51:10.914009 137983432828672 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_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/9/items/array_metadatas/process_0 I0419 02:51:49.700797 137982929528576 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.834226s (commit=40.152076s, array_metadata_write=1.682150s) I0419 02:51:49.701903 137982921135872 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 35.155 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 44.933014154434204 s) (per-host) I0419 02:51:49.702030 137982921135872 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 44.895384s. I0419 02:51:51.430794 137982921135872 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000419s. I0419 02:51:52.459793 137982921135872 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.897251s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/9/items I0419 02:51:52.461521 137982921135872 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/9/items I0419 02:51:52.982901 137982921135872 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/9 I0419 02:51:54.400374 137982921135872 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_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/9`. I0419 02:51:54.401188 137982921135872 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 50.34 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/nnx_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_false/checkpoints/9 I0419 02:51:54.402625 137982921135872 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 49.595976s. I0419 02:51:54.402826 137983424435968 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 02:51:54.402939 137983424435968 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 02:51:54.402990 137983424435968 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0419 02:51:54.405833 137983424435968 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0419 02:51:54.405973 138112449099584 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. I0419 02:51:54.406096 138112449099584 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 02:51:54.407049 138112449099584 metric_logger.py:196] completed step: 9, seconds: 0.148, TFLOP/s/device: 91.643, Tokens/s/device: 13813.478, total_weights: 65536, loss: 7.176, lm_loss: 7.176, perplexity: 1307.366 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sun Apr 19 02:52:05 UTC 2026 EXIT_CODE=0