XPK Start: Mon Apr 20 18:06:21 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 18:06:45.519701: 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 18:06:45.697219 138632011491136 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 18:06:54.739916 138632011491136 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 18:06:54.742254 138632011491136 distributed.py:172] Connecting to JAX distributed service on mt-11-optimizer-offload-3wne9-slice-job-0-0.mt-11-optimizer-offload-3wne9:8482 I0420 18:06:56.009775 138632011491136 max_utils.py:284] Jax distributed system initialized! I0420 18:07:01.038217 138632011491136 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 18:07:01.038318 138632011491136 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 18:07:01.038358 138632011491136 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 18:07:01.038395 138632011491136 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 18:07:02.079788 138632011491136 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 18:07:02.193839 138632011491136 checkpointing.py:688] Setting up checkpoint logger... I0420 18:07:02.193951 138632011491136 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0420 18:07:02.193996 138632011491136 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 18:07:02.194201 138632011491136 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 0x7e1526feb920>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 18:07:04.778890 138632011491136 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0420 18:07:04.779140 138632011491136 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 0x7e1403f32f90>}, handler_registry=None I0420 18:07:04.779376 138632011491136 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e1403f32f90>` 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 18:07:04.779423 138632011491136 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e1403ead940>` 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 18:07:04.779457 138632011491136 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 0x7e1403f32f90>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e1403f32f90>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e1403ead940>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e1403ead940>}). I0420 18:07:04.779799 138632011491136 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 18:07:04.779869 138632011491136 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7e01885ac4a0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 18:07:05.741021 138632011491136 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_11_optimizer_offload_false/checkpoints I0420 18:07:06.617436 138632011491136 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_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e1403f59cd0> I0420 18:07:06.617614 138632011491136 checkpointing.py:302] Checkpoint manager created! I0420 18:07:07.562914 138632011491136 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0420 18:07:07.563033 138632011491136 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0420 18:07:08.953346 138632011491136 maxtext_utils.py:1836] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0420 18:07:08.953463 138632011491136 maxtext_utils.py:1836] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0420 18:07:08.953515 138632011491136 maxtext_utils.py:1836] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0420 18:07:08.953558 138632011491136 maxtext_utils.py:1836] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0420 18:07:08.953596 138632011491136 maxtext_utils.py:1836] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0420 18:07:08.953629 138632011491136 maxtext_utils.py:1836] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0420 18:07:08.953664 138632011491136 maxtext_utils.py:1836] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0420 18:07:08.953699 138632011491136 maxtext_utils.py:1836] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0420 18:07:08.953730 138632011491136 maxtext_utils.py:1836] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0420 18:07:08.953760 138632011491136 maxtext_utils.py:1836] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0420 18:07:08.953792 138632011491136 maxtext_utils.py:1836] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0420 18:07:08.953822 138632011491136 maxtext_utils.py:1836] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0420 18:07:09.355069 138632011491136 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 18:07:09.355176 138632011491136 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 18:07:09.361335 138632011491136 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 18:07:09.361391 138632011491136 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 18:07:09.377596 138632011491136 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 18:07:09.377654 138632011491136 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 18:07:09.393173 138632011491136 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 18:07:09.393232 138632011491136 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 18:07:09.416913 138632011491136 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 18:07:09.416976 138632011491136 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 18:07:09.432588 138632011491136 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 18:07:09.432647 138632011491136 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 18:07:09.448211 138632011491136 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 18:07:09.448271 138632011491136 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 18:07:09.482333 138632011491136 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 18:07:09.482436 138632011491136 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 18:07:09.502472 138632011491136 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 18:07:09.502533 138632011491136 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 18:07:20.958153 138632011491136 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 18:07:20.968272 138632011491136 metric_logger.py:301] number parameters: 1.104 billion I0420 18:07:32.392173 138632011491136 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0420 18:07:32.543218 138632011491136 checkpointing.py:798] Waited 0.15103459358215332 seconds for step 0 to finish before starting checkpointing. I0420 18:07:32.545825 138632011491136 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 18:07:32.547771 138632011491136 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0420 18:07:32.549025 138632011491136 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_11_optimizer_offload_false/checkpoints/0. I0420 18:07:32.890257 138632011491136 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 18:07:32.909374 138632011491136 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0420 18:07:32.909443 138632011491136 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 18:07:33.176108 138632011491136 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.268157s I0420 18:07:33.176948 138632011491136 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.530 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2789306640625 s) (per-host) I0420 18:07:33.177010 138632011491136 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.279010s (batch_requests_ready=0.003042s, total_serialization_initiated=0.275212s, others=0.000757s) I0420 18:07:33.177094 138632011491136 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.282995s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.282977) I0420 18:07:33.177904 138632011491136 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.63 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_11_optimizer_offload_false/checkpoints/0. I0420 18:07:33.178203 138507072964352 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 18:27:33.178171 I0420 18:07:33.192655 138632011491136 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0420 18:07:33.192903 138502293419776 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 18:07:33.193024 138632011491136 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_11_optimizer_offload_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776708452.5458055, 'wait_for_prev_duration_secs': 6.222724914550781e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776708452.54781, 'checkpointer_blocking_duration_secs': 0.6305379867553711, 'get_old_steps_start_time': 1776708453.1783736, 'get_old_steps_duration_secs': 2.8133392333984375e-05, 'checkpoint_manager_blocking_start_time': 1776708452.543655, 'checkpoint_manager_blocking_duration_secs': 0.649329423904419} I0420 18:07:33.193152 138632011491136 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0420 18:07:33.193224 138632011491136 max_utils.py:750] Memstats: After params initialized: I0420 18:07:33.193277 138632011491136 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0420 18:07:33.193310 138632011491136 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0420 18:07:33.193336 138632011491136 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0420 18:07:33.193359 138632011491136 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0420 18:07:33.201939 138506991040256 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_11_optimizer_offload_false/checkpoints/0 I0420 18:07:33.540342 138632011491136 metric_logger.py:196] completed step: 0, seconds: 11.121, TFLOP/s/device: 1.222, Tokens/s/device: 184.163, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52786.469 I0420 18:07:33.543843 138632011491136 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_11_optimizer_offload_false/tensorboard/' I0420 18:07:33.987687 138632011491136 metric_logger.py:196] completed step: 1, seconds: 1.144, TFLOP/s/device: 11.873, Tokens/s/device: 1789.673, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52786.469 I0420 18:07:34.129442 138632011491136 metric_logger.py:196] completed step: 2, seconds: 0.448, TFLOP/s/device: 30.297, Tokens/s/device: 4566.689, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28632.133 I0420 18:07:34.276975 138632011491136 metric_logger.py:196] completed step: 3, seconds: 0.013, TFLOP/s/device: 1065.908, Tokens/s/device: 160665.255, total_weights: 65536, loss: 9.732, lm_loss: 9.732, perplexity: 16855.789 I0420 18:07:35.019681 138506974254848 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_11_optimizer_offload_false/checkpoints/0/items I0420 18:07:35.168435 138502244529920 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776708454755363236, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776708454755363236, "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_11_optimizer_offload_false/checkpoints/0/_CHECKPOINT_METADATA I0420 18:07:36.207161 2767 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 18:07:38.049961 138504922457856 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_11_optimizer_offload_false/checkpoints/0/items/array_metadatas/process_0 I0420 18:07:57.421433 138632011491136 metric_logger.py:196] completed step: 4, seconds: 0.143, TFLOP/s/device: 95.012, Tokens/s/device: 14321.178, total_weights: 65536, loss: 9.274, lm_loss: 9.274, perplexity: 10661.594 I0420 18:07:57.431926 138632011491136 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.385, Tokens/s/device: 13925.342, total_weights: 65536, loss: 8.888, lm_loss: 8.888, perplexity: 7245.978 I0420 18:07:57.576605 138632011491136 metric_logger.py:196] completed step: 6, seconds: 23.143, TFLOP/s/device: 0.587, Tokens/s/device: 88.494, total_weights: 65536, loss: 8.589, lm_loss: 8.589, perplexity: 5371.156 I0420 18:07:57.724313 138632011491136 metric_logger.py:196] completed step: 7, seconds: 0.008, TFLOP/s/device: 1677.838, Tokens/s/device: 252901.951, total_weights: 65536, loss: 8.380, lm_loss: 8.380, perplexity: 4360.465 I0420 18:07:57.871754 138632011491136 metric_logger.py:196] completed step: 8, seconds: 0.146, TFLOP/s/device: 92.920, Tokens/s/device: 14005.909, total_weights: 65536, loss: 8.251, lm_loss: 8.251, perplexity: 3831.256 I0420 18:07:58.020057 138632011491136 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0420 18:07:58.020732 138632011491136 checkpointing.py:798] Waited 0.0006897449493408203 seconds for step 9 to finish before starting checkpointing. I0420 18:07:58.022701 138632011491136 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 18:08:13.883558 138503845975808 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 37.673993s (commit=36.466148s, array_metadata_write=1.207845s) I0420 18:08:13.885795 138507072964352 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 38.538 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 40.98774766921997 s) (per-host) I0420 18:08:13.885926 138507072964352 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.707614s. I0420 18:08:14.621152 138507072964352 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776708454755363236, '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_11_optimizer_offload_false/checkpoints/0/_CHECKPOINT_METADATA I0420 18:08:15.376272 138502244529920 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776708454755363236, '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_11_optimizer_offload_false/checkpoints/0/_CHECKPOINT_METADATA I0420 18:08:15.781272 138507072964352 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000465s. I0420 18:08:16.131178 138507072964352 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0420 18:08:17.078246 138507072964352 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.338426s. use_zarr3=True, 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_11_optimizer_offload_false/checkpoints/0/items I0420 18:08:17.081619 138507072964352 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_11_optimizer_offload_false/checkpoints/0/items I0420 18:08:17.610478 138507072964352 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_11_optimizer_offload_false/checkpoints/0 I0420 18:08:19.050059 138507072964352 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_11_optimizer_offload_false/checkpoints/0`. I0420 18:08:19.050952 138507072964352 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 46.50 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_false/checkpoints/0 I0420 18:08:19.052588 138507072964352 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 45.874274s. I0420 18:08:19.052729 138502293419776 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 18:08:19.052820 138502293419776 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 18:08:19.052882 138502293419776 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0420 18:08:19.055827 138502293419776 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0420 18:08:19.055992 138632011491136 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 18:08:19.056140 138632011491136 checkpoint_manager.py:1452] Waiting for previous save to complete took 21.033439 seconds. If this number is high, consider checkpointing less frequently. I0420 18:08:19.058070 138632011491136 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0420 18:08:19.060333 138632011491136 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_11_optimizer_offload_false/checkpoints/9. I0420 18:08:19.775571 138632011491136 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0420 18:08:19.775662 138632011491136 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 18:08:19.803822 138632011491136 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.029737s I0420 18:08:19.804821 138632011491136 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 40.035 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.038530826568603516 s) (per-host) I0420 18:08:19.804901 138632011491136 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.038629s (batch_requests_ready=0.002633s, total_serialization_initiated=0.035071s, others=0.000925s) I0420 18:08:19.805007 138632011491136 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.042755s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.042738) I0420 18:08:19.806701 138632011491136 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_20260420_153543/nnx_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_false/checkpoints/9. I0420 18:08:19.807111 138502293419776 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 18:28:19.807016 I0420 18:08:19.817735 138632011491136 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0420 18:08:19.818001 138504943429376 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 18:08:19.818152 138632011491136 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_11_optimizer_offload_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776708478.0226688, 'wait_for_prev_duration_secs': 21.03343915939331, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776708499.058128, 'checkpointer_blocking_duration_secs': 0.7491393089294434, 'get_old_steps_start_time': 1776708499.807291, 'get_old_steps_duration_secs': 3.0994415283203125e-05, 'checkpoint_manager_blocking_start_time': 1776708478.020959, 'checkpoint_manager_blocking_duration_secs': 21.79716181755066} I0420 18:08:19.818311 138632011491136 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0420 18:08:19.818355 138632011491136 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 18:08:20.009369 138507072964352 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_11_optimizer_offload_false/checkpoints/9 I0420 18:08:21.857456 138504922457856 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_11_optimizer_offload_false/checkpoints/9/items I0420 18:08:25.265606 138506009552640 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_11_optimizer_offload_false/checkpoints/9/items/array_metadatas/process_0 I0420 18:09:06.162468 138504951822080 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 43.146284s (commit=41.875983s, array_metadata_write=1.270301s) I0420 18:09:06.163632 138502293419776 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.045 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 46.397313594818115 s) (per-host) I0420 18:09:06.163757 138502293419776 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 46.356534s. I0420 18:09:07.767466 138502293419776 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000425s. I0420 18:09:07.962357 138502293419776 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0420 18:09:09.037288 138502293419776 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.928015s. use_zarr3=True, 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_11_optimizer_offload_false/checkpoints/9/items I0420 18:09:09.039144 138502293419776 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_11_optimizer_offload_false/checkpoints/9/items I0420 18:09:09.562146 138502293419776 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_11_optimizer_offload_false/checkpoints/9 I0420 18:09:11.441341 138502293419776 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_11_optimizer_offload_false/checkpoints/9`. I0420 18:09:11.442200 138502293419776 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_11_optimizer_offload_false/checkpoints/9 I0420 18:09:11.443804 138502293419776 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 51.636580s. I0420 18:09:11.443959 138504943429376 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 18:09:11.444020 138504943429376 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 18:09:11.444111 138504943429376 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0420 18:09:11.446515 138504943429376 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0420 18:09:11.446707 138632011491136 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 18:09:11.446874 138632011491136 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 18:09:11.447960 138632011491136 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.202, Tokens/s/device: 13897.654, total_weights: 65536, loss: 8.175, lm_loss: 8.175, perplexity: 3551.438 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Mon Apr 20 18:09:22 UTC 2026 EXIT_CODE=0