XPK Start: Mon Apr 20 18:53:59 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 18:54:23.357908: 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 18:54:23.535125 134313187862336 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 18:54:32.576968 134313187862336 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 18:54:32.579315 134313187862336 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-sv-v7lvn-slice-job-0-0.mt-16-zarr3-false-sv-v7lvn:8482 I0420 18:54:33.741909 134313187862336 max_utils.py:284] Jax distributed system initialized! I0420 18:54:38.962374 134313187862336 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 18:54:38.962476 134313187862336 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 18:54:38.962516 134313187862336 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 18:54:38.962550 134313187862336 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 18:54:39.655492 134313187862336 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 18:54:39.767966 134313187862336 checkpointing.py:688] Setting up checkpoint logger... I0420 18:54:39.768095 134313187862336 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=False I0420 18:54:39.768141 134313187862336 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 18:54:39.768335 134313187862336 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 0x7a27a09766c0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 18:54:42.820461 134313187862336 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0420 18:54:42.820697 134313187862336 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 0x7a13045c5b80>}, handler_registry=None I0420 18:54:42.820940 134313187862336 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a13045c5b80>` 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 18:54:42.820988 134313187862336 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a130457bc50>` 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 18:54:42.821021 134313187862336 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 0x7a13045c5b80>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a13045c5b80>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a130457bc50>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a130457bc50>}). I0420 18:54:42.821347 134313187862336 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 18:54:42.821418 134313187862336 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7a13ec1f4fe0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 18:54:43.926095 134313187862336 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints I0420 18:54:44.366986 134313187862336 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_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7a13ec43b8f0> I0420 18:54:44.367191 134313187862336 checkpointing.py:302] Checkpoint manager created! I0420 18:54:45.313662 134313187862336 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0420 18:54:45.313776 134313187862336 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0420 18:54:46.673130 134313187862336 maxtext_utils.py:1836] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0420 18:54:46.673249 134313187862336 maxtext_utils.py:1836] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0420 18:54:46.673302 134313187862336 maxtext_utils.py:1836] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0420 18:54:46.673345 134313187862336 maxtext_utils.py:1836] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0420 18:54:46.673384 134313187862336 maxtext_utils.py:1836] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0420 18:54:46.673419 134313187862336 maxtext_utils.py:1836] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0420 18:54:46.673455 134313187862336 maxtext_utils.py:1836] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0420 18:54:46.673490 134313187862336 maxtext_utils.py:1836] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0420 18:54:46.673525 134313187862336 maxtext_utils.py:1836] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0420 18:54:46.673557 134313187862336 maxtext_utils.py:1836] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0420 18:54:46.673589 134313187862336 maxtext_utils.py:1836] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0420 18:54:46.673626 134313187862336 maxtext_utils.py:1836] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0420 18:54:47.073888 134313187862336 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 18:54:47.073989 134313187862336 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 18:54:47.079286 134313187862336 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 18:54:47.079341 134313187862336 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 18:54:47.095633 134313187862336 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 18:54:47.095692 134313187862336 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 18:54:47.111253 134313187862336 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 18:54:47.111310 134313187862336 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 18:54:47.136472 134313187862336 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 18:54:47.136546 134313187862336 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 18:54:47.152287 134313187862336 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 18:54:47.152349 134313187862336 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 18:54:47.168017 134313187862336 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 18:54:47.168089 134313187862336 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 18:54:47.198227 134313187862336 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 18:54:47.198299 134313187862336 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 18:54:47.217963 134313187862336 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 18:54:47.218027 134313187862336 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 18:54:58.578099 134313187862336 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 18:54:58.587483 134313187862336 metric_logger.py:301] number parameters: 1.104 billion I0420 18:55:10.067850 134313187862336 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0420 18:55:10.219833 134313187862336 checkpointing.py:798] Waited 0.15196800231933594 seconds for step 0 to finish before starting checkpointing. I0420 18:55:10.222518 134313187862336 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 18:55:10.224300 134313187862336 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0420 18:55:10.225787 134313187862336 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/0. I0420 18:55:10.571537 134313187862336 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 18:55:10.590627 134313187862336 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0420 18:55:10.590694 134313187862336 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 18:55:10.848101 134187365783296 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/0 I0420 18:55:10.858026 134313187862336 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.268818s I0420 18:55:10.858855 134313187862336 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.519 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.27950501441955566 s) (per-host) I0420 18:55:10.858920 134313187862336 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.279584s (batch_requests_ready=0.003004s, total_serialization_initiated=0.275851s, others=0.000730s) I0420 18:55:10.859017 134313187862336 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.283629s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.283611) I0420 18:55:10.859755 134313187862336 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.64 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/0. I0420 18:55:10.860114 134187374176000 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 19:15:10.860063 I0420 18:55:10.877954 134313187862336 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0420 18:55:10.878230 134185784178432 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 18:55:10.878371 134313187862336 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776711310.2224984, 'wait_for_prev_duration_secs': 6.222724914550781e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776711310.224339, 'checkpointer_blocking_duration_secs': 0.6359262466430664, 'get_old_steps_start_time': 1776711310.8602896, 'get_old_steps_duration_secs': 2.7418136596679688e-05, 'checkpoint_manager_blocking_start_time': 1776711310.220248, 'checkpoint_manager_blocking_duration_secs': 0.6580839157104492} I0420 18:55:10.878477 134313187862336 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0420 18:55:10.878546 134313187862336 max_utils.py:750] Memstats: After params initialized: I0420 18:55:10.878596 134313187862336 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0420 18:55:10.878630 134313187862336 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0420 18:55:10.878661 134313187862336 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0420 18:55:10.878686 134313187862336 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0420 18:55:11.227753 134313187862336 metric_logger.py:196] completed step: 0, seconds: 11.141, TFLOP/s/device: 1.220, Tokens/s/device: 183.822, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52786.469 I0420 18:55:11.232332 134313187862336 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/tensorboard/' I0420 18:55:11.653725 134313187862336 metric_logger.py:196] completed step: 1, seconds: 1.156, TFLOP/s/device: 11.751, Tokens/s/device: 1771.262, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52786.469 I0420 18:55:11.795429 134313187862336 metric_logger.py:196] completed step: 2, seconds: 0.427, TFLOP/s/device: 31.831, Tokens/s/device: 4797.916, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38565.637 I0420 18:55:11.942632 134313187862336 metric_logger.py:196] completed step: 3, seconds: 0.013, TFLOP/s/device: 1038.454, Tokens/s/device: 156527.056, total_weights: 65536, loss: 9.979, lm_loss: 9.979, perplexity: 21564.932 I0420 18:55:11.954929 134313187862336 checkpointing.py:794] Waiting for step 5 to finish before checkpoint... I0420 18:55:12.237431 134313187862336 checkpointing.py:798] Waited 0.2824726104736328 seconds for step 5 to finish before starting checkpointing. I0420 18:55:12.240335 134313187862336 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 18:55:13.272843 134187348997888 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/0/items I0420 18:55:13.462630 134184670713600 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776711313023996821, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776711313023996821, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/0/_CHECKPOINT_METADATA I0420 18:55:14.458737 2754 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 18:55:16.263197 134186272159488 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_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/0/items/array_metadatas/process_0 I0420 18:55:55.693614 134186263766784 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.249821s (commit=39.981348s, array_metadata_write=1.268474s) I0420 18:55:55.695196 134187374176000 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 35.012 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.11581206321716 s) (per-host) I0420 18:55:55.695361 134187374176000 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 44.835126s. I0420 18:55:56.390992 134187374176000 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776711313023996821, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/0/_CHECKPOINT_METADATA I0420 18:55:57.142669 134184670713600 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776711313023996821, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/0/_CHECKPOINT_METADATA I0420 18:55:57.203186 134187374176000 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000845s. I0420 18:55:58.753031 134187374176000 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.243322s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/0/items I0420 18:55:58.755400 134187374176000 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/0/items I0420 18:55:59.343528 134187374176000 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/0 I0420 18:56:00.781240 134187374176000 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_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/0`. I0420 18:56:00.782396 134187374176000 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 50.56 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/0 I0420 18:56:00.784161 134187374176000 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 49.923922s. I0420 18:56:00.784431 134185784178432 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 18:56:00.784629 134185784178432 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 18:56:00.784784 134185784178432 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0420 18:56:00.787375 134185784178432 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0420 18:56:00.787687 134313187862336 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. W0420 18:56:00.787866 134313187862336 checkpoint_manager.py:1452] Waiting for previous save to complete took 48.547532 seconds. If this number is high, consider checkpointing less frequently. I0420 18:56:00.789689 134313187862336 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 5 I0420 18:56:00.792103 134313187862336 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/5. I0420 18:56:01.141972 134313187862336 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0420 18:56:01.142138 134313187862336 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 18:56:01.183573 134313187862336 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044586s I0420 18:56:01.184552 134313187862336 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 24.079 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.06406235694885254 s) (per-host) I0420 18:56:01.184654 134313187862336 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.064187s (batch_requests_ready=0.006358s, total_serialization_initiated=0.056962s, others=0.000866s) I0420 18:56:01.184778 134313187862336 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.069049s (all_items=0.000024s, per_item={'items': '0.00002432'}, temp_paths=0.069025) I0420 18:56:01.185825 134313187862336 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.40 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/5. I0420 18:56:01.186204 134185784178432 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 19:16:01.186167 I0420 18:56:01.188675 134313187862336 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0420 18:56:01.188982 134186263766784 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 18:56:01.189194 134313187862336 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776711312.2402918, 'wait_for_prev_duration_secs': 48.547531604766846, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776711360.7897558, 'checkpointer_blocking_duration_secs': 0.3966236114501953, 'get_old_steps_start_time': 1776711361.1864083, 'get_old_steps_duration_secs': 4.220008850097656e-05, 'checkpoint_manager_blocking_start_time': 1776711312.2379005, 'checkpoint_manager_blocking_duration_secs': 48.95124077796936} I0420 18:56:01.189410 134313187862336 checkpointing.py:409] Started an asynchronous checkpoint save for step 5 I0420 18:56:01.370853 134187374176000 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/5 I0420 18:56:03.883388 134159477458688 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/5/items I0420 18:56:07.544231 134186810488576 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_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/5/items/array_metadatas/process_0 I0420 18:56:24.838136 134313187862336 metric_logger.py:196] completed step: 4, seconds: 0.143, TFLOP/s/device: 95.089, Tokens/s/device: 14332.804, total_weights: 65536, loss: 9.457, lm_loss: 9.457, perplexity: 12802.436 I0420 18:56:24.849167 134313187862336 metric_logger.py:196] completed step: 5, seconds: 0.146, TFLOP/s/device: 92.787, Tokens/s/device: 13985.918, total_weights: 65536, loss: 8.955, lm_loss: 8.955, perplexity: 7748.778 I0420 18:56:24.993509 134313187862336 metric_logger.py:196] completed step: 6, seconds: 72.894, TFLOP/s/device: 0.186, Tokens/s/device: 28.096, total_weights: 65536, loss: 8.465, lm_loss: 8.465, perplexity: 4743.984 I0420 18:56:25.140887 134313187862336 metric_logger.py:196] completed step: 7, seconds: 0.008, TFLOP/s/device: 1612.524, Tokens/s/device: 243057.204, total_weights: 65536, loss: 7.998, lm_loss: 7.998, perplexity: 2976.266 I0420 18:56:25.288507 134313187862336 metric_logger.py:196] completed step: 8, seconds: 0.146, TFLOP/s/device: 93.107, Tokens/s/device: 14034.030, total_weights: 65536, loss: 7.567, lm_loss: 7.567, perplexity: 1933.479 I0420 18:56:25.436436 134313187862336 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0420 18:56:25.437189 134313187862336 checkpointing.py:798] Waited 0.0007672309875488281 seconds for step 9 to finish before starting checkpointing. I0420 18:56:25.439131 134313187862336 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 18:56:48.288616 134186272159488 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.838480s (commit=41.590240s, array_metadata_write=1.248240s) I0420 18:56:48.289719 134185784178432 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 33.488 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 47.169211626052856 s) (per-host) I0420 18:56:48.289848 134185784178432 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 47.103501s. I0420 18:56:49.828639 134185784178432 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000417s. I0420 18:56:51.137235 134185784178432 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.907656s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/5/items I0420 18:56:51.139108 134185784178432 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/5/items I0420 18:56:51.726593 134185784178432 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/5 I0420 18:56:53.166707 134185784178432 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_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/5`. I0420 18:56:53.167630 134185784178432 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 52.38 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/5 I0420 18:56:53.169168 134185784178432 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 51.982820s. I0420 18:56:53.169308 134186263766784 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 18:56:53.169389 134186263766784 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 18:56:53.169456 134186263766784 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0420 18:56:53.171359 134186263766784 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0420 18:56:53.171506 134313187862336 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. W0420 18:56:53.171585 134313187862336 checkpoint_manager.py:1452] Waiting for previous save to complete took 27.732465 seconds. If this number is high, consider checkpointing less frequently. I0420 18:56:53.173120 134313187862336 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0420 18:56:53.175183 134313187862336 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9. I0420 18:56:53.487465 134313187862336 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0420 18:56:53.487563 134313187862336 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 18:56:53.515166 134313187862336 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.029137s I0420 18:56:53.515997 134313187862336 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 40.922 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03769516944885254 s) (per-host) I0420 18:56:53.516059 134313187862336 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.037774s (batch_requests_ready=0.002537s, total_serialization_initiated=0.034494s, others=0.000743s) I0420 18:56:53.516156 134313187862336 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.041787s (all_items=0.000015s, per_item={'items': '0.00001526'}, temp_paths=0.041772) I0420 18:56:53.516892 134313187862336 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.34 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9. I0420 18:56:53.517191 134185762408192 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 19:16:53.517154 I0420 18:56:53.533539 134313187862336 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0420 18:56:53.533790 134186263766784 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 18:56:53.533956 134313187862336 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776711385.439097, 'wait_for_prev_duration_secs': 27.732465028762817, 'time_between_consecutive_saves_sec': 24.651649951934814, 'checkpointer_blocking_start_time': 1776711413.1731591, 'checkpointer_blocking_duration_secs': 0.34417057037353516, 'get_old_steps_start_time': 1776711413.517349, 'get_old_steps_duration_secs': 2.8848648071289062e-05, 'checkpoint_manager_blocking_start_time': 1776711385.4374185, 'checkpoint_manager_blocking_duration_secs': 28.09650468826294} I0420 18:56:53.534065 134313187862336 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0420 18:56:53.534117 134313187862336 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 18:56:53.744222 134185784178432 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9 I0420 18:56:55.944100 134187374176000 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9/items I0420 18:56:58.888046 134186810488576 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_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9/items/array_metadatas/process_0 I0420 18:57:40.122763 134186272159488 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 43.032938s (commit=41.802137s, array_metadata_write=1.230801s) I0420 18:57:40.123899 134185762408192 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 33.864 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 46.64556860923767 s) (per-host) I0420 18:57:40.124022 134185762408192 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 46.606717s. I0420 18:57:41.911258 134185762408192 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000457s. I0420 18:57:43.259315 134185762408192 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.230206s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9/items I0420 18:57:43.260995 134185762408192 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9/items I0420 18:57:43.934307 134185762408192 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9 I0420 18:57:45.367128 134185762408192 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_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9`. I0420 18:57:45.367832 134185762408192 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 52.19 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_16_zarr3_false/checkpoints/9 I0420 18:57:45.368967 134185762408192 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 51.851659s. I0420 18:57:45.369161 134186263766784 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 18:57:45.369280 134186263766784 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 18:57:45.369331 134186263766784 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0420 18:57:45.371146 134186263766784 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0420 18:57:45.371263 134313187862336 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. I0420 18:57:45.371371 134313187862336 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 18:57:45.372395 134313187862336 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.402, Tokens/s/device: 13927.899, 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: Mon Apr 20 18:57:54 UTC 2026 EXIT_CODE=0