XPK Start: Tue Apr 21 12:46:10 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-21 12:46:34.614792: 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) I0421 12:46:35.152488 137812296066880 max_utils.py:273] Attempting to initialize the jax distributed system... I0421 12:46:44.195884 137812296066880 distributed.py:149] Starting JAX distributed service on [::]:8482 I0421 12:46:44.198264 137812296066880 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-true-rs-7uo5q-slice-job-0-0.mt-15-ocdbt-true-rs-7uo5q:8482 I0421 12:46:45.365009 137812296066880 max_utils.py:284] Jax distributed system initialized! I0421 12:46:51.574627 137812296066880 max_utils.py:800] System Information: Jax Version: 0.9.2 I0421 12:46:51.574744 137812296066880 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0421 12:46:51.574786 137812296066880 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 I0421 12:46:51.574821 137812296066880 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0421 12:46:52.281629 137812296066880 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 12:46:52.282239 137812296066880 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 12:46:52.282416 137812296066880 checkpointing.py:688] Setting up checkpoint logger... I0421 12:46:52.282463 137812296066880 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0421 12:46:52.282505 137812296066880 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0421 12:46:52.282851 137812296066880 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 0x7d564bf13710>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 12:46:53.567190 137812296066880 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0421 12:46:53.567526 137812296066880 checkpoint_manager.py:708] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d4138170b00>}, handler_registry=None I0421 12:46:53.567817 137812296066880 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d4138170b00>` 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`. I0421 12:46:53.567887 137812296066880 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d411872f9b0>` 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`. I0421 12:46:53.567934 137812296066880 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 0x7d4138170b00>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d4138170b00>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d411872f9b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d411872f9b0>}). I0421 12:46:53.568284 137812296066880 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0421 12:46:53.568357 137812296066880 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7d41186a4180> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0421 12:46:55.921580 137691148580608 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776775402816001121, 'commit_timestamp_nsecs': 1776775448635831878, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints/5/_CHECKPOINT_METADATA I0421 12:46:56.183984 137691148580608 checkpoint_manager.py:1871] Missing metrics for step 5 I0421 12:46:56.603928 137812296066880 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints I0421 12:46:57.023239 137812296066880 checkpoint_manager.py:929] [process=4][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/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7d41187325a0> I0421 12:46:57.023407 137812296066880 checkpointing.py:302] Checkpoint manager created! I0421 12:46:57.951992 137812296066880 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 12:46:57.952097 137812296066880 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 12:46:58.331685 137812296066880 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 12:46:58.331786 137812296066880 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 12:46:58.348061 137812296066880 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 12:46:58.348119 137812296066880 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 12:46:58.372517 137812296066880 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 12:46:58.372582 137812296066880 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 12:46:58.388883 137812296066880 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 12:46:58.388943 137812296066880 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 12:46:58.405130 137812296066880 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 12:46:58.405190 137812296066880 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 12:46:58.429822 137812296066880 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 12:46:58.429894 137812296066880 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 12:46:58.450794 137812296066880 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 12:46:58.450855 137812296066880 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0421 12:46:58.667143 137812296066880 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0421 12:46:58.667247 137812296066880 checkpointing.py:582] restoring from this run's directory step 9 I0421 12:46:58.921163 137812296066880 event_tracking.py:70] [process=4] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints/9. I0421 12:46:59.171089 137812296066880 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints/9. I0421 12:46:59.171530 137812296066880 event_tracking.py:125] [process=4] [sync] Finished blocking load in 0.25 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints/9. I0421 12:47:00.578369 1448 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0421 12:47:06.766261 137812296066880 jax_array_handlers.py:843] [process=4] /jax/orbax/read/worker/io/requested throughput: 255.130 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.193985223770142 s) (per-host) I0421 12:47:06.772358 137812296066880 base_pytree_checkpoint_handler.py:130] [process=4] /jax/checkpoint/read/gbytes_per_sec: 239.406 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.610501289367676 s) (per-host) I0421 12:47:08.338788 137812296066880 event_tracking.py:138] [process=4] [sync] Finished load in 9.42 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints/9 I0421 12:47:08.338939 137812296066880 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints', 'checkpointer_start_time': 1776775618.921064, 'checkpointer_duration_secs': 9.417842864990234, 'checkpoint_manager_start_time': 1776775618.667496, 'checkpoint_manager_duration_secs': 9.671413660049438} fsdp: 32 I0421 12:47:08.767283 137812296066880 maxtext_utils.py:1827] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0421 12:47:08.767408 137812296066880 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0421 12:47:08.767467 137812296066880 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0421 12:47:08.767524 137812296066880 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0421 12:47:08.767576 137812296066880 maxtext_utils.py:1827] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0421 12:47:08.767621 137812296066880 maxtext_utils.py:1827] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0421 12:47:08.767675 137812296066880 maxtext_utils.py:1827] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0421 12:47:08.767742 137812296066880 maxtext_utils.py:1827] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0421 12:47:08.767786 137812296066880 maxtext_utils.py:1827] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0421 12:47:08.767823 137812296066880 maxtext_utils.py:1827] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0421 12:47:08.767871 137812296066880 maxtext_utils.py:1827] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0421 12:47:08.767918 137812296066880 maxtext_utils.py:1827] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0421 12:47:09.262622 137812296066880 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0421 12:47:09.262728 137812296066880 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0421 12:47:09.278346 137812296066880 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0421 12:47:09.278409 137812296066880 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0421 12:47:20.190443 137812296066880 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. I0421 12:47:20.191250 137812296066880 metric_logger.py:301] number parameters: 1.104 billion I0421 12:47:31.366074 137812296066880 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0421 12:47:31.519760 137812296066880 checkpointing.py:798] Waited 0.15364480018615723 seconds for step 10 to finish before starting checkpointing. I0421 12:47:31.522140 137812296066880 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 12:47:31.524055 137812296066880 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 10 I0421 12:47:31.525632 137812296066880 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints/10. I0421 12:47:31.885905 137812296066880 signaling_client.py:364] Using JaxDistributedSignalingClient I0421 12:47:31.886779 137812296066880 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0421 12:47:31.886844 137812296066880 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 I0421 12:47:32.163172 137812296066880 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.277343s I0421 12:47:32.163346 137812296066880 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 5.480 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28149890899658203 s) (per-host) I0421 12:47:32.163397 137812296066880 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.281561s (batch_requests_ready=0.002241s, total_serialization_initiated=0.279249s, others=0.000070s) I0421 12:47:32.163494 137812296066880 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.285858s (all_items=0.000017s, per_item={'items': '0.00001740'}, temp_paths=0.285841) I0421 12:47:32.164250 137812296066880 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.64 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints/10. I0421 12:47:32.164534 137691106617088 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-21 13:07:32.164503 I0421 12:47:32.173285 137812296066880 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0421 12:47:32.173585 137688640374528 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0421 12:47:32.173747 137812296066880 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776775651.5221188, 'wait_for_prev_duration_secs': 7.963180541992188e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776775651.5241022, 'checkpointer_blocking_duration_secs': 0.6405758857727051, 'get_old_steps_start_time': 1776775652.1647172, 'get_old_steps_duration_secs': 3.24249267578125e-05, 'checkpoint_manager_blocking_start_time': 1776775651.5202727, 'checkpoint_manager_blocking_duration_secs': 0.6534392833709717} I0421 12:47:32.173846 137812296066880 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0421 12:47:32.173895 137812296066880 max_utils.py:750] Memstats: After params initialized: I0421 12:47:32.173941 137812296066880 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0)) I0421 12:47:32.173975 137812296066880 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0)) I0421 12:47:32.174004 137812296066880 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0)) I0421 12:47:32.174029 137812296066880 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0)) I0421 12:47:32.481342 137812296066880 metric_logger.py:196] completed step: 10, seconds: 11.175, TFLOP/s/device: 1.216, Tokens/s/device: 183.271, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0421 12:47:32.666740 137812296066880 metric_logger.py:196] completed step: 11, seconds: 1.114, TFLOP/s/device: 12.199, Tokens/s/device: 1838.694, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0421 12:47:32.813805 137812296066880 metric_logger.py:196] completed step: 12, seconds: 0.038, TFLOP/s/device: 357.500, Tokens/s/device: 53886.228, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0421 12:47:32.961361 137812296066880 metric_logger.py:196] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.833, Tokens/s/device: 13389.822, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0421 12:47:32.965574 137812296066880 checkpointing.py:794] Waiting for step 15 to finish before checkpoint... I0421 12:47:33.255472 137812296066880 checkpointing.py:798] Waited 0.28986454010009766 seconds for step 15 to finish before starting checkpointing. I0421 12:47:33.258334 137812296066880 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 12:47:37.139682 137689177245440 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints/10/items/array_metadatas/process_4 I0421 12:48:14.716996 137691106617088 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 36.876 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.83512091636658 s) (per-host) I0421 12:48:14.717104 137691106617088 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.552465s. I0421 12:48:21.719846 137691106617088 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 49.555189s. I0421 12:48:21.720123 137688640374528 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 12:48:21.720237 137688640374528 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0421 12:48:21.720287 137688640374528 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0421 12:48:21.721749 137688640374528 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0421 12:48:21.721919 137812296066880 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0421 12:48:21.722045 137812296066880 checkpoint_manager.py:1452] Waiting for previous save to complete took 48.463712 seconds. If this number is high, consider checkpointing less frequently. I0421 12:48:21.723982 137812296066880 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 15 I0421 12:48:21.726540 137812296066880 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints/15. I0421 12:48:22.442618 137812296066880 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0421 12:48:22.442725 137812296066880 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 I0421 12:48:22.467512 137812296066880 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026025s I0421 12:48:22.467653 137812296066880 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 52.244 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.029526472091674805 s) (per-host) I0421 12:48:22.467712 137812296066880 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.029584s (batch_requests_ready=0.001900s, total_serialization_initiated=0.027620s, others=0.000064s) I0421 12:48:22.467791 137812296066880 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034542s (all_items=0.000013s, per_item={'items': '0.00001287'}, temp_paths=0.034529) I0421 12:48:22.468516 137812296066880 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.74 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints/15. I0421 12:48:22.468849 137689684752128 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-21 13:08:22.468809 I0421 12:48:22.482745 137812296066880 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0421 12:48:22.482996 137688640374528 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0421 12:48:22.483131 137812296066880 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776775653.2583046, 'wait_for_prev_duration_secs': 48.46371245384216, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776775701.7240229, 'checkpointer_blocking_duration_secs': 0.7449686527252197, 'get_old_steps_start_time': 1776775702.4690166, 'get_old_steps_duration_secs': 2.8848648071289062e-05, 'checkpoint_manager_blocking_start_time': 1776775653.2560647, 'checkpoint_manager_blocking_duration_secs': 49.22703218460083} I0421 12:48:22.483240 137812296066880 checkpointing.py:409] Started an asynchronous checkpoint save for step 15 I0421 12:48:22.484090 137812296066880 metric_logger.py:196] completed step: 14, seconds: 0.147, TFLOP/s/device: 92.336, Tokens/s/device: 13917.960, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0421 12:48:22.490612 137812296066880 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.135, Tokens/s/device: 13887.665, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0421 12:48:28.102974 137683816929024 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints/15/items/array_metadatas/process_4 I0421 12:48:43.305827 137812296066880 metric_logger.py:196] completed step: 16, seconds: 49.524, TFLOP/s/device: 0.274, Tokens/s/device: 41.354, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0421 12:48:43.453136 137812296066880 metric_logger.py:196] completed step: 17, seconds: 20.669, TFLOP/s/device: 0.657, Tokens/s/device: 99.085, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0421 12:48:43.600323 137812296066880 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.270, Tokens/s/device: 13455.802, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0421 12:48:43.746960 137812296066880 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0421 12:48:43.747648 137812296066880 checkpointing.py:798] Waited 0.0007081031799316406 seconds for step 19 to finish before starting checkpointing. I0421 12:48:43.749621 137812296066880 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 12:49:04.873808 137689684752128 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 37.223 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.43565011024475 s) (per-host) I0421 12:49:04.873914 137689684752128 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.404953s. I0421 12:49:12.658536 137689684752128 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.189557s. I0421 12:49:12.658827 137688640374528 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 12:49:12.658949 137688640374528 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0421 12:49:12.659000 137688640374528 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0421 12:49:12.660644 137688640374528 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0421 12:49:12.660822 137812296066880 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0421 12:49:12.660948 137812296066880 checkpoint_manager.py:1452] Waiting for previous save to complete took 28.911334 seconds. If this number is high, consider checkpointing less frequently. I0421 12:49:12.662385 137812296066880 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 19 I0421 12:49:12.664510 137812296066880 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints/19. I0421 12:49:13.243267 137812296066880 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0421 12:49:13.243357 137812296066880 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 I0421 12:49:13.269651 137812296066880 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027355s I0421 12:49:13.269823 137812296066880 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 49.872 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030930519104003906 s) (per-host) I0421 12:49:13.269876 137812296066880 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030993s (batch_requests_ready=0.001792s, total_serialization_initiated=0.029133s, others=0.000067s) I0421 12:49:13.269961 137812296066880 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.035084s (all_items=0.000010s, per_item={'items': '0.00000978'}, temp_paths=0.035074) I0421 12:49:13.270710 137812296066880 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints/19. I0421 12:49:13.271047 137688640374528 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-21 13:09:13.271011 I0421 12:49:13.273632 137812296066880 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0421 12:49:13.273920 137683816929024 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0421 12:49:13.274080 137812296066880 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776775723.749592, 'wait_for_prev_duration_secs': 28.911333560943604, 'time_between_consecutive_saves_sec': 22.02779459953308, 'checkpointer_blocking_start_time': 1776775752.662424, 'checkpointer_blocking_duration_secs': 0.6087696552276611, 'get_old_steps_start_time': 1776775753.2712133, 'get_old_steps_duration_secs': 2.7894973754882812e-05, 'checkpoint_manager_blocking_start_time': 1776775723.7478964, 'checkpoint_manager_blocking_duration_secs': 29.526150465011597} I0421 12:49:13.274189 137812296066880 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0421 12:49:13.274235 137812296066880 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 12:49:18.397070 137683808536320 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_15_ocdbt_true/checkpoints/19/items/array_metadatas/process_4 I0421 12:49:54.364487 137688640374528 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 38.409 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.125553607940674 s) (per-host) I0421 12:49:54.364616 137688640374528 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.093457s. I0421 12:50:03.162590 137688640374528 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 49.891414s. I0421 12:50:03.162889 137683816929024 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 12:50:03.163005 137683816929024 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0421 12:50:03.163048 137683816929024 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0421 12:50:03.164393 137683816929024 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0421 12:50:03.164567 137812296066880 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0421 12:50:03.164733 137812296066880 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 12:50:03.165708 137812296066880 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.301, Tokens/s/device: 13912.665, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Tue Apr 21 12:50:11 UTC 2026 EXIT_CODE=0