XPK Start: Wed Apr 22 13:53:54 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 13:54:18.201527: 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 13:54:18.741387 135866337167168 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 13:54:27.782782 135866337167168 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 13:54:27.785181 135866337167168 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-false-rs-khp96-slice-job-0-0.mt-15-ocdbt-false-rs-khp96:8482 I0422 13:54:29.416571 135866337167168 max_utils.py:284] Jax distributed system initialized! I0422 13:54:35.414807 135866337167168 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 13:54:35.414907 135866337167168 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 13:54:35.414948 135866337167168 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 13:54:35.414984 135866337167168 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 13:54:36.112704 135866337167168 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 13:54:36.113296 135866337167168 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 13:54:36.113474 135866337167168 checkpointing.py:688] Setting up checkpoint logger... I0422 13:54:36.113523 135866337167168 checkpointing.py:234] Creating checkpoint manager with ocdbt=False and zarr3=True I0422 13:54:36.113571 135866337167168 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 13:54:36.113909 135866337167168 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=False, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7b913fd14590>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 13:54:37.444987 135866337167168 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0422 13:54:37.445245 135866337167168 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 0x7b7d2863f230>}, handler_registry=None I0422 13:54:37.445482 135866337167168 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b7d2863f230>` 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 13:54:37.445530 135866337167168 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b7d2863e7b0>` 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 13:54:37.445571 135866337167168 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 0x7b7d2863f230>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b7d2863f230>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b7d2863e7b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b7d2863e7b0>}). I0422 13:54:37.445883 135866337167168 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 13:54:37.445953 135866337167168 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7b7d283dc180> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 13:54:39.177527 135745452893952 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776865950981550649, 'commit_timestamp_nsecs': 1776865964117560523, '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_15_ocdbt_false/checkpoints/9/_CHECKPOINT_METADATA I0422 13:54:39.432023 135745452893952 checkpoint_manager.py:1871] Missing metrics for step 9 I0422 13:54:39.724406 135866337167168 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_15_ocdbt_false/checkpoints I0422 13:54:39.867167 135866337167168 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_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7b7d2863e420> I0422 13:54:39.867338 135866337167168 checkpointing.py:302] Checkpoint manager created! I0422 13:54:40.803284 135866337167168 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 13:54:40.803391 135866337167168 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 13:54:41.183906 135866337167168 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 13:54:41.184000 135866337167168 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 13:54:41.200400 135866337167168 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 13:54:41.200458 135866337167168 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 13:54:41.225027 135866337167168 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 13:54:41.225101 135866337167168 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 13:54:41.241497 135866337167168 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 13:54:41.241558 135866337167168 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 13:54:41.257956 135866337167168 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 13:54:41.258016 135866337167168 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 13:54:41.282809 135866337167168 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 13:54:41.282877 135866337167168 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 13:54:41.303791 135866337167168 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 13:54:41.303855 135866337167168 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 13:54:41.519341 135866337167168 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0422 13:54:41.519450 135866337167168 checkpointing.py:582] restoring from this run's directory step 9 I0422 13:54:41.761349 135866337167168 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_15_ocdbt_false/checkpoints/9. I0422 13:54:42.013479 135866337167168 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_15_ocdbt_false/checkpoints/9. I0422 13:54:42.013916 135866337167168 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_15_ocdbt_false/checkpoints/9. I0422 13:54:44.591362 1476 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 13:54:53.639041 135866337167168 jax_array_handlers.py:843] [process=6] /jax/orbax/read/worker/io/requested throughput: 156.595 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.09144639968872 s) (per-host) I0422 13:54:53.646839 135866337167168 base_pytree_checkpoint_handler.py:130] [process=6] /jax/checkpoint/read/gbytes_per_sec: 150.664 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.50410771369934 s) (per-host) I0422 13:54:54.450146 135866337167168 event_tracking.py:138] [process=6] [sync] Finished load in 12.69 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_15_ocdbt_false/checkpoints/9 I0422 13:54:54.450315 135866337167168 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_15_ocdbt_false/checkpoints', 'checkpointer_start_time': 1776866081.7612474, 'checkpointer_duration_secs': 12.689035415649414, 'checkpoint_manager_start_time': 1776866081.5197132, 'checkpoint_manager_duration_secs': 12.930571556091309} fsdp: 32 I0422 13:54:54.866584 135866337167168 maxtext_utils.py:1827] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0422 13:54:54.866709 135866337167168 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 13:54:54.866761 135866337167168 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 13:54:54.866816 135866337167168 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 13:54:54.866867 135866337167168 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 13:54:54.866911 135866337167168 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 13:54:54.866964 135866337167168 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 13:54:54.867015 135866337167168 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 13:54:54.867055 135866337167168 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 13:54:54.867126 135866337167168 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 13:54:54.867186 135866337167168 maxtext_utils.py:1827] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 13:54:54.867236 135866337167168 maxtext_utils.py:1827] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 13:54:55.357442 135866337167168 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 13:54:55.357535 135866337167168 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 13:54:55.372984 135866337167168 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 13:54:55.373044 135866337167168 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 13:55:06.343988 135866337167168 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 13:55:06.344854 135866337167168 metric_logger.py:301] number parameters: 1.104 billion I0422 13:55:17.804648 135866337167168 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0422 13:55:17.959179 135866337167168 checkpointing.py:798] Waited 0.15451717376708984 seconds for step 10 to finish before starting checkpointing. I0422 13:55:17.961456 135866337167168 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 13:55:17.963274 135866337167168 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 10 I0422 13:55:17.964677 135866337167168 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_15_ocdbt_false/checkpoints/10. I0422 13:55:18.295830 135866337167168 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 13:55:18.296805 135866337167168 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 13:55:18.296861 135866337167168 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 13:55:18.585459 135866337167168 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.289704s I0422 13:55:18.585643 135866337167168 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.255 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2935354709625244 s) (per-host) I0422 13:55:18.585697 135866337167168 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.293600s (batch_requests_ready=0.002115s, total_serialization_initiated=0.291411s, others=0.000074s) I0422 13:55:18.585784 135866337167168 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.297920s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=0.297904) I0422 13:55:18.586558 135866337167168 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_15_ocdbt_false/checkpoints/10. I0422 13:55:18.586939 135742701430528 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 14:15:18.586895 I0422 13:55:18.588958 135866337167168 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0422 13:55:18.589276 135742659467008 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 13:55:18.589402 135866337167168 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_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776866117.9614365, 'wait_for_prev_duration_secs': 6.4849853515625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776866117.9633124, 'checkpointer_blocking_duration_secs': 0.6238288879394531, 'get_old_steps_start_time': 1776866118.587164, 'get_old_steps_duration_secs': 3.2901763916015625e-05, 'checkpoint_manager_blocking_start_time': 1776866117.9595504, 'checkpoint_manager_blocking_duration_secs': 0.6298160552978516} I0422 13:55:18.589563 135866337167168 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0422 13:55:18.589614 135866337167168 max_utils.py:750] Memstats: After params initialized: I0422 13:55:18.589661 135866337167168 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0422 13:55:18.589693 135866337167168 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0422 13:55:18.589720 135866337167168 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0422 13:55:18.589744 135866337167168 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0422 13:55:18.899263 135866337167168 metric_logger.py:196] completed step: 10, seconds: 11.460, TFLOP/s/device: 1.186, Tokens/s/device: 178.714, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0422 13:55:19.076797 135866337167168 metric_logger.py:196] completed step: 11, seconds: 1.093, TFLOP/s/device: 12.430, Tokens/s/device: 1873.598, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0422 13:55:19.224309 135866337167168 metric_logger.py:196] completed step: 12, seconds: 0.030, TFLOP/s/device: 451.505, Tokens/s/device: 68055.694, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0422 13:55:19.371829 135866337167168 metric_logger.py:196] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.578, Tokens/s/device: 13351.413, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0422 13:55:19.376318 135866337167168 checkpointing.py:794] Waiting for step 15 to finish before checkpoint... I0422 13:55:19.665852 135866337167168 checkpointing.py:798] Waited 0.28949785232543945 seconds for step 15 to finish before starting checkpointing. I0422 13:55:19.669179 135866337167168 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 13:55:23.773674 135742684645120 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_15_ocdbt_false/checkpoints/10/items/array_metadatas/process_6 I0422 13:55:29.280179 135742701430528 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 143.756 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.988041400909424 s) (per-host) I0422 13:55:29.280287 135742701430528 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.693208s. I0422 13:55:34.411146 135742701430528 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 15.824050s. I0422 13:55:34.411458 135742659467008 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 13:55:34.411579 135742659467008 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 13:55:34.411627 135742659467008 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0422 13:55:34.414078 135742659467008 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0422 13:55:34.414273 135866337167168 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 13:55:34.414412 135866337167168 checkpoint_manager.py:1452] Waiting for previous save to complete took 14.745232 seconds. If this number is high, consider checkpointing less frequently. I0422 13:55:34.416681 135866337167168 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 15 I0422 13:55:34.419200 135866337167168 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_15_ocdbt_false/checkpoints/15. I0422 13:55:34.705749 135866337167168 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 13:55:34.705839 135866337167168 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 13:55:34.752650 135866337167168 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.048024s I0422 13:55:34.752822 135866337167168 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 30.006 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.05140948295593262 s) (per-host) I0422 13:55:34.752873 135866337167168 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.051469s (batch_requests_ready=0.001740s, total_serialization_initiated=0.049661s, others=0.000068s) I0422 13:55:34.752963 135866337167168 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.056398s (all_items=0.000013s, per_item={'items': '0.00001264'}, temp_paths=0.056386) I0422 13:55:34.753689 135866337167168 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_15_ocdbt_false/checkpoints/15. I0422 13:55:34.754040 135742659467008 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 14:15:34.754003 I0422 13:55:34.756040 135866337167168 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0422 13:55:34.756282 135738398070528 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 13:55:34.756406 135866337167168 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_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776866119.6691484, 'wait_for_prev_duration_secs': 14.745231866836548, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776866134.4167216, 'checkpointer_blocking_duration_secs': 0.33748674392700195, 'get_old_steps_start_time': 1776866134.7542334, 'get_old_steps_duration_secs': 3.457069396972656e-05, 'checkpoint_manager_blocking_start_time': 1776866119.6664882, 'checkpoint_manager_blocking_duration_secs': 15.089884281158447} I0422 13:55:34.756515 135866337167168 checkpointing.py:409] Started an asynchronous checkpoint save for step 15 I0422 13:55:34.757472 135866337167168 metric_logger.py:196] completed step: 14, seconds: 0.148, TFLOP/s/device: 92.110, Tokens/s/device: 13883.804, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0422 13:55:34.763250 135866337167168 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.126, Tokens/s/device: 13886.252, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0422 13:55:39.854316 135742684645120 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_15_ocdbt_false/checkpoints/15/items/array_metadatas/process_6 I0422 13:55:44.915055 135742659467008 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 154.656 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.213612794876099 s) (per-host) I0422 13:55:44.915171 135742659467008 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.160993s. I0422 13:55:50.529408 135742659467008 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 15.775215s. I0422 13:55:50.529633 135738398070528 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 13:55:50.529688 135738398070528 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 13:55:50.529734 135738398070528 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0422 13:55:50.531291 135738398070528 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0422 13:55:55.683966 135866337167168 metric_logger.py:196] completed step: 16, seconds: 15.385, TFLOP/s/device: 0.883, Tokens/s/device: 133.112, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0422 13:55:55.831288 135866337167168 metric_logger.py:196] completed step: 17, seconds: 20.774, TFLOP/s/device: 0.654, Tokens/s/device: 98.583, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0422 13:55:55.978506 135866337167168 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.191, Tokens/s/device: 13443.789, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0422 13:55:56.125288 135866337167168 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0422 13:55:56.125951 135866337167168 checkpointing.py:798] Waited 0.0007045269012451172 seconds for step 19 to finish before starting checkpointing. I0422 13:55:56.128027 135866337167168 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 13:55:56.129644 135866337167168 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 19 I0422 13:55:56.131021 135866337167168 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_15_ocdbt_false/checkpoints/19. I0422 13:55:56.459511 135866337167168 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 13:55:56.459605 135866337167168 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 13:55:56.504513 135866337167168 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.046019s I0422 13:55:56.504659 135866337167168 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 31.220 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04940986633300781 s) (per-host) I0422 13:55:56.504707 135866337167168 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.049468s (batch_requests_ready=0.001616s, total_serialization_initiated=0.047789s, others=0.000062s) I0422 13:55:56.504791 135866337167168 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.053532s (all_items=0.000011s, per_item={'items': '0.00001121'}, temp_paths=0.053521) I0422 13:55:56.505507 135866337167168 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_15_ocdbt_false/checkpoints/19. I0422 13:55:56.505773 135738398070528 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 14:15:56.505750 I0422 13:55:56.507664 135866337167168 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0422 13:55:56.507963 135737844414208 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 13:55:56.508133 135866337167168 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_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776866156.1279979, 'wait_for_prev_duration_secs': 7.033348083496094e-05, 'time_between_consecutive_saves_sec': 5.596664905548096, 'checkpointer_blocking_start_time': 1776866156.1296828, 'checkpointer_blocking_duration_secs': 0.3761932849884033, 'get_old_steps_start_time': 1776866156.5058923, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1776866156.1261966, 'checkpoint_manager_blocking_duration_secs': 0.3818931579589844} I0422 13:55:56.508245 135866337167168 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0422 13:55:56.508291 135866337167168 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 13:56:01.601843 135742684645120 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_15_ocdbt_false/checkpoints/19/items/array_metadatas/process_6 I0422 13:56:06.523970 135738398070528 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 156.882 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.068679094314575 s) (per-host) I0422 13:56:06.524088 135738398070528 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.018234s. I0422 13:56:12.499679 135738398070528 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 15.993810s. I0422 13:56:12.499949 135737844414208 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 13:56:12.500070 135737844414208 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 13:56:12.500142 135737844414208 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0422 13:56:12.502448 135737844414208 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0422 13:56:12.502625 135866337167168 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 13:56:12.502770 135866337167168 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 13:56:12.503662 135866337167168 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.294, Tokens/s/device: 13911.531, 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 13:56:24 UTC 2026 EXIT_CODE=0