XPK Start: Wed Apr 22 15:36:59 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-22 15:37:23.390661: 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) I0422 15:37:23.925364 132315162629952 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 15:37:32.965552 132315162629952 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 15:37:32.967800 132315162629952 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-sv-b2zdt-slice-job-0-0.mt-16-zarr3-false-sv-b2zdt:8482 I0422 15:37:34.512470 132315162629952 max_utils.py:284] Jax distributed system initialized! I0422 15:37:39.490809 132315162629952 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 15:37:39.490911 132315162629952 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 15:37:39.491074 132315162629952 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 I0422 15:37:39.491112 132315162629952 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 15:37:40.192766 132315162629952 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 15:37:40.305625 132315162629952 checkpointing.py:688] Setting up checkpoint logger... I0422 15:37:40.305738 132315162629952 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=False I0422 15:37:40.305784 132315162629952 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 15:37:40.305987 132315162629952 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 0x7856657f2480>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 15:37:43.308194 132315162629952 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0422 15:37:43.308417 132315162629952 checkpoint_manager.py:708] [process=3][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x785542e10e30>}, handler_registry=None I0422 15:37:43.308659 132315162629952 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x785542e10e30>` 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`. I0422 15:37:43.308712 132315162629952 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7841c029c6b0>` 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`. I0422 15:37:43.308748 132315162629952 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 0x785542e10e30>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x785542e10e30>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7841c029c6b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7841c029c6b0>}). I0422 15:37:43.309084 132315162629952 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 15:37:43.309154 132315162629952 async_checkpointer.py:192] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7841c02e19e0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 15:37:44.431608 132315162629952 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints I0422 15:37:44.864089 132315162629952 checkpoint_manager.py:929] [process=3][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_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7841c029c740> I0422 15:37:44.864261 132315162629952 checkpointing.py:302] Checkpoint manager created! I0422 15:37:45.843618 132315162629952 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0422 15:37:45.843739 132315162629952 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 15:37:47.217545 132315162629952 maxtext_utils.py:1836] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0422 15:37:47.217660 132315162629952 maxtext_utils.py:1836] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0422 15:37:47.217712 132315162629952 maxtext_utils.py:1836] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0422 15:37:47.217755 132315162629952 maxtext_utils.py:1836] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0422 15:37:47.217793 132315162629952 maxtext_utils.py:1836] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0422 15:37:47.217828 132315162629952 maxtext_utils.py:1836] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0422 15:37:47.217863 132315162629952 maxtext_utils.py:1836] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0422 15:37:47.217897 132315162629952 maxtext_utils.py:1836] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0422 15:37:47.217941 132315162629952 maxtext_utils.py:1836] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0422 15:37:47.217977 132315162629952 maxtext_utils.py:1836] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0422 15:37:47.218010 132315162629952 maxtext_utils.py:1836] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0422 15:37:47.218042 132315162629952 maxtext_utils.py:1836] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0422 15:37:47.614632 132315162629952 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 15:37:47.614726 132315162629952 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 15:37:47.619981 132315162629952 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 15:37:47.620034 132315162629952 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 15:37:47.636616 132315162629952 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 15:37:47.636678 132315162629952 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 15:37:47.652809 132315162629952 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 15:37:47.652869 132315162629952 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 15:37:47.677109 132315162629952 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 15:37:47.677184 132315162629952 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 15:37:47.693644 132315162629952 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 15:37:47.693705 132315162629952 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 15:37:47.710136 132315162629952 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 15:37:47.710201 132315162629952 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 15:37:47.740642 132315162629952 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 15:37:47.740717 132315162629952 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 15:37:47.760910 132315162629952 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 15:37:47.760989 132315162629952 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 15:37:59.197719 132315162629952 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. I0422 15:37:59.200168 132315162629952 metric_logger.py:301] number parameters: 1.104 billion I0422 15:38:10.340429 132315162629952 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0422 15:38:10.493540 132315162629952 checkpointing.py:798] Waited 0.15308356285095215 seconds for step 0 to finish before starting checkpointing. I0422 15:38:10.496112 132315162629952 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 15:38:10.497642 132315162629952 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 0 I0422 15:38:10.499131 132315162629952 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0. I0422 15:38:10.835906 132315162629952 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 15:38:10.836909 132315162629952 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0422 15:38:10.836983 132315162629952 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 I0422 15:38:11.109674 132315162629952 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.274530s I0422 15:38:11.109844 132315162629952 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 5.437 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28369951248168945 s) (per-host) I0422 15:38:11.109893 132315162629952 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.283759s (batch_requests_ready=0.003162s, total_serialization_initiated=0.280530s, others=0.000067s) I0422 15:38:11.110006 132315162629952 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.287895s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.287878) I0422 15:38:11.110826 132315162629952 event_tracking.py:125] [process=3] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0. I0422 15:38:11.111162 132187988342528 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 15:58:11.111122 I0422 15:38:11.127336 132315162629952 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 15:38:11.127618 132186355250944 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 15:38:11.127779 132315162629952 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776872290.4960945, 'wait_for_prev_duration_secs': 5.984306335449219e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776872290.4976838, 'checkpointer_blocking_duration_secs': 0.613635778427124, 'get_old_steps_start_time': 1776872291.1113431, 'get_old_steps_duration_secs': 3.0994415283203125e-05, 'checkpoint_manager_blocking_start_time': 1776872290.493976, 'checkpoint_manager_blocking_duration_secs': 0.6337616443634033} I0422 15:38:11.127888 132315162629952 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0422 15:38:11.127968 132315162629952 max_utils.py:750] Memstats: After params initialized: I0422 15:38:11.128017 132315162629952 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_10(process=3,(2,2,0,0)) I0422 15:38:11.128050 132315162629952 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_11(process=3,(3,2,0,0)) I0422 15:38:11.128076 132315162629952 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_14(process=3,(2,3,0,0)) I0422 15:38:11.128101 132315162629952 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_15(process=3,(3,3,0,0)) I0422 15:38:11.447281 132315162629952 metric_logger.py:196] completed step: 0, seconds: 11.139, TFLOP/s/device: 1.220, Tokens/s/device: 183.857, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52786.469 I0422 15:38:11.632302 132315162629952 metric_logger.py:196] completed step: 1, seconds: 1.105, TFLOP/s/device: 12.295, Tokens/s/device: 1853.259, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52786.469 I0422 15:38:12.068708 132315162629952 metric_logger.py:196] completed step: 2, seconds: 0.039, TFLOP/s/device: 350.880, Tokens/s/device: 52888.464, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38565.637 I0422 15:38:12.216299 132315162629952 metric_logger.py:196] completed step: 3, seconds: 0.426, TFLOP/s/device: 31.911, Tokens/s/device: 4809.962, total_weights: 65536, loss: 9.979, lm_loss: 9.979, perplexity: 21564.932 I0422 15:38:12.221964 132315162629952 checkpointing.py:794] Waiting for step 5 to finish before checkpoint... I0422 15:38:12.511036 132315162629952 checkpointing.py:798] Waited 0.2890431880950928 seconds for step 5 to finish before starting checkpointing. I0422 15:38:12.513807 132315162629952 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 15:38:14.113405 2777 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 15:38:16.679093 132186363643648 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0/items/array_metadatas/process_3 I0422 15:38:52.354691 132187988342528 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 38.036 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.52850699424744 s) (per-host) I0422 15:38:52.354812 132187988342528 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.243532s. I0422 15:39:01.456612 132187988342528 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 50.345316s. I0422 15:39:01.456834 132186355250944 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 15:39:01.456898 132186355250944 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 15:39:01.456965 132186355250944 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 15:39:01.460289 132186355250944 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 15:39:01.460472 132315162629952 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0422 15:39:01.460605 132315162629952 checkpoint_manager.py:1452] Waiting for previous save to complete took 48.946796 seconds. If this number is high, consider checkpointing less frequently. I0422 15:39:01.462716 132315162629952 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 5 I0422 15:39:01.465317 132315162629952 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/5. I0422 15:39:01.765339 132315162629952 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0422 15:39:01.765434 132315162629952 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 I0422 15:39:01.800438 132315162629952 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036714s I0422 15:39:01.800612 132315162629952 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 35.220 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.043798208236694336 s) (per-host) I0422 15:39:01.800665 132315162629952 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.043861s (batch_requests_ready=0.002746s, total_serialization_initiated=0.041044s, others=0.000071s) I0422 15:39:01.800762 132315162629952 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.049159s (all_items=0.000015s, per_item={'items': '0.00001502'}, temp_paths=0.049144) I0422 15:39:01.801500 132315162629952 event_tracking.py:125] [process=3] [async] Finished blocking save in 0.34 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/5. I0422 15:39:01.801826 132186355250944 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 15:59:01.801790 I0422 15:39:01.820898 132315162629952 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0422 15:39:01.821223 132161688553216 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 15:39:01.821388 132315162629952 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776872292.5137784, 'wait_for_prev_duration_secs': 48.94679641723633, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776872341.4627554, 'checkpointer_blocking_duration_secs': 0.33922696113586426, 'get_old_steps_start_time': 1776872341.802005, 'get_old_steps_duration_secs': 3.147125244140625e-05, 'checkpoint_manager_blocking_start_time': 1776872292.5113869, 'checkpoint_manager_blocking_duration_secs': 49.30996584892273} I0422 15:39:01.821498 132315162629952 checkpointing.py:409] Started an asynchronous checkpoint save for step 5 I0422 15:39:01.822633 132315162629952 metric_logger.py:196] completed step: 4, seconds: 0.165, TFLOP/s/device: 82.404, Tokens/s/device: 12420.778, total_weights: 65536, loss: 9.457, lm_loss: 9.457, perplexity: 12802.436 I0422 15:39:01.830652 132315162629952 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.368, Tokens/s/device: 13922.786, total_weights: 65536, loss: 8.955, lm_loss: 8.955, perplexity: 7748.778 I0422 15:39:07.065217 132186363643648 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/5/items/array_metadatas/process_3 I0422 15:39:26.352961 132315162629952 metric_logger.py:196] completed step: 6, seconds: 49.607, TFLOP/s/device: 0.274, Tokens/s/device: 41.285, total_weights: 65536, loss: 8.465, lm_loss: 8.465, perplexity: 4743.984 I0422 15:39:26.500229 132315162629952 metric_logger.py:196] completed step: 7, seconds: 24.377, TFLOP/s/device: 0.557, Tokens/s/device: 84.014, total_weights: 65536, loss: 7.998, lm_loss: 7.998, perplexity: 2976.266 I0422 15:39:26.647698 132315162629952 metric_logger.py:196] completed step: 8, seconds: 0.154, TFLOP/s/device: 88.483, Tokens/s/device: 13337.067, total_weights: 65536, loss: 7.567, lm_loss: 7.567, perplexity: 1933.479 I0422 15:39:26.796113 132315162629952 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0422 15:39:26.796797 132315162629952 checkpointing.py:798] Waited 0.0006966590881347656 seconds for step 9 to finish before starting checkpointing. I0422 15:39:26.798758 132315162629952 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 15:39:44.502919 132186355250944 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 36.953 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.74606728553772 s) (per-host) I0422 15:39:44.503057 132186355250944 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.701110s. I0422 15:39:52.795494 132186355250944 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 50.993532s. I0422 15:39:52.795775 132161688553216 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 15:39:52.795899 132161688553216 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 15:39:52.795973 132161688553216 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0422 15:39:52.797479 132161688553216 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0422 15:39:52.797662 132315162629952 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0422 15:39:52.797784 132315162629952 checkpoint_manager.py:1452] Waiting for previous save to complete took 25.999034 seconds. If this number is high, consider checkpointing less frequently. I0422 15:39:52.799370 132315162629952 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 9 I0422 15:39:52.801448 132315162629952 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9. I0422 15:39:53.105899 132315162629952 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0422 15:39:53.106006 132315162629952 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 I0422 15:39:53.149091 132315162629952 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044737s I0422 15:39:53.149260 132315162629952 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 29.815 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.05173802375793457 s) (per-host) I0422 15:39:53.149309 132315162629952 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.051798s (batch_requests_ready=0.002829s, total_serialization_initiated=0.048904s, others=0.000065s) I0422 15:39:53.149404 132315162629952 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.055707s (all_items=0.000010s, per_item={'items': '0.00000978'}, temp_paths=0.055698) I0422 15:39:53.150142 132315162629952 event_tracking.py:125] [process=3] [async] Finished blocking save in 0.35 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9. I0422 15:39:53.150486 132161688553216 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 15:59:53.150445 I0422 15:39:53.152459 132315162629952 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 15:39:53.152702 132184751838976 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0422 15:39:53.152860 132315162629952 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776872366.798729, 'wait_for_prev_duration_secs': 25.9990336894989, 'time_between_consecutive_saves_sec': 25.338399648666382, 'checkpointer_blocking_start_time': 1776872392.7994087, 'checkpointer_blocking_duration_secs': 0.3512270450592041, 'get_old_steps_start_time': 1776872393.1506543, 'get_old_steps_duration_secs': 2.5272369384765625e-05, 'checkpoint_manager_blocking_start_time': 1776872366.7970486, 'checkpoint_manager_blocking_duration_secs': 26.355780363082886} I0422 15:39:53.152988 132315162629952 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0422 15:39:53.153030 132315162629952 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 15:39:58.535688 132186363643648 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9/items/array_metadatas/process_3 I0422 15:40:36.490546 132161688553216 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 36.402 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.39298176765442 s) (per-host) I0422 15:40:36.490662 132161688553216 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.340063s. I0422 15:40:42.501983 132161688553216 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 49.351367s. I0422 15:40:42.502246 132184751838976 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 15:40:42.502362 132184751838976 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0422 15:40:42.502405 132184751838976 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 15:40:42.503826 132184751838976 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 15:40:42.504003 132315162629952 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0422 15:40:42.504173 132315162629952 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 15:40:42.505326 132315162629952 metric_logger.py:196] completed step: 9, seconds: 0.148, TFLOP/s/device: 91.880, Tokens/s/device: 13849.160, 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: Wed Apr 22 15:40:54 UTC 2026 EXIT_CODE=0