XPK Start: Wed Apr 22 14:12:22 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 14:12:46.342020: 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 14:12:46.521249 139585633957696 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 14:12:55.564342 139585633957696 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 14:12:55.568801 139585633957696 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-rs-alh10-slice-job-0-0.mt-16-zarr3-false-rs-alh10:8482 I0422 14:12:57.108508 139585633957696 max_utils.py:284] Jax distributed system initialized! I0422 14:13:02.168604 139585633957696 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 14:13:02.168708 139585633957696 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 14:13:02.168749 139585633957696 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 14:13:02.168783 139585633957696 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 14:13:02.868474 139585633957696 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 14:13:02.869055 139585633957696 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 14:13:02.869257 139585633957696 checkpointing.py:688] Setting up checkpoint logger... I0422 14:13:02.869310 139585633957696 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=False I0422 14:13:02.869352 139585633957696 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 14:13:02.869685 139585633957696 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 0x7ef32f5f3800>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 14:13:04.200593 139585633957696 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0422 14:13:04.200822 139585633957696 checkpoint_manager.py:708] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7edfa45e4cb0>}, handler_registry=None I0422 14:13:04.201058 139585633957696 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7edfa45e4cb0>` 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 14:13:04.201118 139585633957696 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7edfa43dbb00>` 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 14:13:04.201154 139585633957696 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 0x7edfa45e4cb0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7edfa45e4cb0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7edfa43dbb00>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7edfa43dbb00>}). I0422 14:13:04.201469 139585633957696 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 14:13:04.201539 139585633957696 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7edfa4210540> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 14:13:05.908831 139466958161664 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776866920433518276, 'commit_timestamp_nsecs': 1776866967424170153, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0/_CHECKPOINT_METADATA I0422 14:13:06.160267 139466958161664 checkpoint_manager.py:1871] Missing metrics for step 0 I0422 14:13:06.192222 139585633957696 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints I0422 14:13:07.438899 139585633957696 checkpoint_manager.py:929] [process=6][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_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7edfa43db110> I0422 14:13:07.439069 139585633957696 checkpointing.py:302] Checkpoint manager created! I0422 14:13:08.410808 139585633957696 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 14:13:08.410912 139585633957696 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 14:13:08.790033 139585633957696 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 14:13:08.790162 139585633957696 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 14:13:08.806229 139585633957696 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 14:13:08.806285 139585633957696 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 14:13:08.829556 139585633957696 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 14:13:08.829621 139585633957696 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 14:13:08.845776 139585633957696 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 14:13:08.845844 139585633957696 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 14:13:08.862136 139585633957696 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 14:13:08.862196 139585633957696 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 14:13:08.886410 139585633957696 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 14:13:08.886479 139585633957696 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 14:13:08.907999 139585633957696 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 14:13:08.908062 139585633957696 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 14:13:09.122182 139585633957696 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0422 14:13:09.122292 139585633957696 checkpointing.py:582] restoring from this run's directory step 9 I0422 14:13:09.372195 139585633957696 event_tracking.py:70] [process=6] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9. I0422 14:13:09.623385 139585633957696 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9. I0422 14:13:09.623802 139585633957696 event_tracking.py:125] [process=6] [sync] Finished blocking load in 0.25 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9. I0422 14:13:11.464207 1412 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 14:13:18.208270 139585633957696 jax_array_handlers.py:843] [process=6] /jax/orbax/read/worker/io/requested throughput: 234.114 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.750018119812012 s) (per-host) I0422 14:13:18.214219 139585633957696 base_pytree_checkpoint_handler.py:130] [process=6] /jax/checkpoint/read/gbytes_per_sec: 221.090 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 7.158139228820801 s) (per-host) I0422 14:13:19.533025 139585633957696 event_tracking.py:138] [process=6] [sync] Finished load in 10.16 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9 I0422 14:13:19.533199 139585633957696 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints', 'checkpointer_start_time': 1776867189.3720767, 'checkpointer_duration_secs': 10.161089658737183, 'checkpoint_manager_start_time': 1776867189.1225455, 'checkpoint_manager_duration_secs': 10.410622596740723} fsdp: 32 I0422 14:13:19.948800 139585633957696 maxtext_utils.py:1827] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0422 14:13:19.948953 139585633957696 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) I0422 14:13:19.949025 139585633957696 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) I0422 14:13:19.949127 139585633957696 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') I0422 14:13:19.949208 139585633957696 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) I0422 14:13:19.949266 139585633957696 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) I0422 14:13:19.949349 139585633957696 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) I0422 14:13:19.949429 139585633957696 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') I0422 14:13:19.949486 139585633957696 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) I0422 14:13:19.949546 139585633957696 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) I0422 14:13:19.949613 139585633957696 maxtext_utils.py:1827] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 14:13:19.949693 139585633957696 maxtext_utils.py:1827] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 14:13:20.440219 139585633957696 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 14:13:20.440311 139585633957696 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 14:13:20.455705 139585633957696 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 14:13:20.455766 139585633957696 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 14:13:31.578887 139585633957696 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 14:13:31.579562 139585633957696 metric_logger.py:301] number parameters: 1.104 billion I0422 14:13:43.024846 139585633957696 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0422 14:13:43.178025 139585633957696 checkpointing.py:798] Waited 0.15316152572631836 seconds for step 10 to finish before starting checkpointing. I0422 14:13:43.180743 139585633957696 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 14:13:43.182642 139585633957696 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 10 I0422 14:13:43.184082 139585633957696 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/10. I0422 14:13:43.515665 139585633957696 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 14:13:43.516744 139585633957696 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 14:13:43.516802 139585633957696 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 14:13:43.800257 139585633957696 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.284667s I0422 14:13:43.800434 139585633957696 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.347 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28849053382873535 s) (per-host) I0422 14:13:43.800487 139585633957696 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.288555s (batch_requests_ready=0.002119s, total_serialization_initiated=0.286362s, others=0.000073s) I0422 14:13:43.800576 139585633957696 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.292813s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.292796) I0422 14:13:43.801331 139585633957696 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.62 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/10. I0422 14:13:43.801687 139466345789184 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 14:33:43.801648 I0422 14:13:43.803552 139585633957696 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0422 14:13:43.803802 139464215090944 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 14:13:43.803926 139585633957696 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_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': 1776867223.1807237, 'wait_for_prev_duration_secs': 5.9604644775390625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776867223.1826804, 'checkpointer_blocking_duration_secs': 0.6191625595092773, 'get_old_steps_start_time': 1776867223.8018656, 'get_old_steps_duration_secs': 3.24249267578125e-05, 'checkpoint_manager_blocking_start_time': 1776867223.1785357, 'checkpoint_manager_blocking_duration_secs': 0.6253571510314941} I0422 14:13:43.804081 139585633957696 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0422 14:13:43.804144 139585633957696 max_utils.py:750] Memstats: After params initialized: I0422 14:13:43.804190 139585633957696 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0422 14:13:43.804221 139585633957696 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0422 14:13:43.804252 139585633957696 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0422 14:13:43.804276 139585633957696 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0422 14:13:44.116034 139585633957696 metric_logger.py:196] completed step: 10, seconds: 11.445, TFLOP/s/device: 1.187, Tokens/s/device: 178.940, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0422 14:13:44.698614 139585633957696 metric_logger.py:196] completed step: 11, seconds: 1.090, TFLOP/s/device: 12.469, Tokens/s/device: 1879.435, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0422 14:13:44.846170 139585633957696 metric_logger.py:196] completed step: 12, seconds: 0.436, TFLOP/s/device: 31.151, Tokens/s/device: 4695.471, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0422 14:13:44.993505 139585633957696 metric_logger.py:196] completed step: 13, seconds: 0.152, TFLOP/s/device: 89.099, Tokens/s/device: 13429.949, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0422 14:13:44.998230 139585633957696 checkpointing.py:794] Waiting for step 15 to finish before checkpoint... I0422 14:13:45.287620 139585633957696 checkpointing.py:798] Waited 0.2893550395965576 seconds for step 15 to finish before starting checkpointing. I0422 14:13:45.290897 139585633957696 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 14:13:49.164558 139465259468544 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/10/items/array_metadatas/process_6 I0422 14:14:26.141746 139466345789184 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.054 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.6297709941864 s) (per-host) I0422 14:14:26.141852 139466345789184 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.340054s. I0422 14:14:33.689957 139466345789184 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 49.888140s. I0422 14:14:33.690241 139464215090944 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 14:14:33.690364 139464215090944 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 14:14:33.690417 139464215090944 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0422 14:14:33.692066 139464215090944 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0422 14:14:33.692278 139585633957696 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0422 14:14:33.692411 139585633957696 checkpoint_manager.py:1452] Waiting for previous save to complete took 48.401515 seconds. If this number is high, consider checkpointing less frequently. I0422 14:14:33.694288 139585633957696 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 15 I0422 14:14:33.696764 139585633957696 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/15. I0422 14:14:33.989324 139585633957696 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 14:14:33.989418 139585633957696 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 14:14:34.036879 139585633957696 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.048703s I0422 14:14:34.037044 139585633957696 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 29.602 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.052111148834228516 s) (per-host) I0422 14:14:34.037105 139585633957696 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.052170s (batch_requests_ready=0.001771s, total_serialization_initiated=0.050334s, others=0.000065s) I0422 14:14:34.037195 139585633957696 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.056988s (all_items=0.000012s, per_item={'items': '0.00001240'}, temp_paths=0.056975) I0422 14:14:34.037831 139585633957696 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.34 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/15. I0422 14:14:34.038224 139465259468544 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 14:34:34.038187 I0422 14:14:34.040204 139585633957696 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0422 14:14:34.040488 139464215090944 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 14:14:34.040619 139585633957696 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_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': 1776867225.2908664, 'wait_for_prev_duration_secs': 48.401514768600464, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776867273.694327, 'checkpointer_blocking_duration_secs': 0.3440437316894531, 'get_old_steps_start_time': 1776867274.0383942, 'get_old_steps_duration_secs': 3.337860107421875e-05, 'checkpoint_manager_blocking_start_time': 1776867225.2882266, 'checkpoint_manager_blocking_duration_secs': 48.7523627281189} I0422 14:14:34.040772 139585633957696 checkpointing.py:409] Started an asynchronous checkpoint save for step 15 I0422 14:14:34.041541 139585633957696 metric_logger.py:196] completed step: 14, seconds: 0.147, TFLOP/s/device: 92.117, Tokens/s/device: 13884.840, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0422 14:14:34.047515 139585633957696 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.169, Tokens/s/device: 13892.658, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0422 14:14:40.164765 139460967872256 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/15/items/array_metadatas/process_6 I0422 14:14:55.206538 139585633957696 metric_logger.py:196] completed step: 16, seconds: 49.048, TFLOP/s/device: 0.277, Tokens/s/device: 41.755, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0422 14:14:55.353757 139585633957696 metric_logger.py:196] completed step: 17, seconds: 21.013, TFLOP/s/device: 0.647, Tokens/s/device: 97.464, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0422 14:14:55.501337 139585633957696 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.146, Tokens/s/device: 13436.998, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0422 14:14:55.647957 139585633957696 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0422 14:14:55.648702 139585633957696 checkpointing.py:798] Waited 0.0007631778717041016 seconds for step 19 to finish before starting checkpointing. I0422 14:14:55.650745 139585633957696 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 14:15:16.058542 139465259468544 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.544 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.073570728302 s) (per-host) I0422 14:15:16.058666 139465259468544 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.020323s. I0422 14:15:25.006237 139465259468544 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.967876s. I0422 14:15:25.006565 139464215090944 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 14:15:25.006693 139464215090944 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 14:15:25.006743 139464215090944 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0422 14:15:25.008138 139464215090944 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0422 14:15:25.008332 139585633957696 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0422 14:15:25.008459 139585633957696 checkpoint_manager.py:1452] Waiting for previous save to complete took 29.357721 seconds. If this number is high, consider checkpointing less frequently. I0422 14:15:25.010009 139585633957696 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 19 I0422 14:15:25.011984 139585633957696 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/19. I0422 14:15:25.341754 139585633957696 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 14:15:25.341849 139585633957696 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 14:15:25.390433 139585633957696 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.049587s I0422 14:15:25.390618 139585633957696 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 29.075 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.0530545711517334 s) (per-host) I0422 14:15:25.390676 139585633957696 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.053123s (batch_requests_ready=0.001650s, total_serialization_initiated=0.051399s, others=0.000074s) I0422 14:15:25.390791 139585633957696 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.057345s (all_items=0.000011s, per_item={'items': '0.00001073'}, temp_paths=0.057335) I0422 14:15:25.391572 139585633957696 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.38 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/19. I0422 14:15:25.391867 139464215090944 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 14:35:25.391835 I0422 14:15:25.393735 139585633957696 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0422 14:15:25.394015 139460967872256 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 14:15:25.394159 139585633957696 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_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': 1776867295.6507154, 'wait_for_prev_duration_secs': 29.357721090316772, 'time_between_consecutive_saves_sec': 21.9585862159729, 'checkpointer_blocking_start_time': 1776867325.010048, 'checkpointer_blocking_duration_secs': 0.3819575309753418, 'get_old_steps_start_time': 1776867325.3920228, 'get_old_steps_duration_secs': 2.7894973754882812e-05, 'checkpoint_manager_blocking_start_time': 1776867295.6489353, 'checkpoint_manager_blocking_duration_secs': 29.74519181251526} I0422 14:15:25.394332 139585633957696 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0422 14:15:25.394376 139585633957696 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 14:15:31.122307 139460431816448 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/19/items/array_metadatas/process_6 I0422 14:16:08.756797 139464215090944 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 36.380 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.41919946670532 s) (per-host) I0422 14:16:08.756906 139464215090944 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.364928s. I0422 14:16:15.539936 139464215090944 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.147942s. I0422 14:16:15.540243 139460967872256 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 14:16:15.540380 139460967872256 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 14:16:15.540422 139460967872256 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0422 14:16:15.541882 139460967872256 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0422 14:16:15.542039 139585633957696 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0422 14:16:15.542205 139585633957696 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 14:16:15.543020 139585633957696 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.342, Tokens/s/device: 13918.811, 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: Wed Apr 22 14:16:27 UTC 2026 EXIT_CODE=0