XPK Start: Tue Apr 21 13:13:05 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 13:13:29.849841: 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 13:13:30.031383 139844782446400 max_utils.py:273] Attempting to initialize the jax distributed system... I0421 13:13:39.073195 139844782446400 distributed.py:149] Starting JAX distributed service on [::]:8482 I0421 13:13:39.075396 139844782446400 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-rs-ij7gw-slice-job-0-0.mt-16-zarr3-false-rs-ij7gw:8482 I0421 13:13:40.309631 139844782446400 max_utils.py:284] Jax distributed system initialized! I0421 13:13:45.506403 139844782446400 max_utils.py:800] System Information: Jax Version: 0.9.2 I0421 13:13:45.506505 139844782446400 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0421 13:13:45.506546 139844782446400 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 13:13:45.506580 139844782446400 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0421 13:13:46.210019 139844782446400 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 13:13:46.210592 139844782446400 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 13:13:46.210782 139844782446400 checkpointing.py:688] Setting up checkpoint logger... I0421 13:13:46.210832 139844782446400 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=False I0421 13:13:46.210874 139844782446400 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0421 13:13:46.211202 139844782446400 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 0x7f2f8611c110>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 13:13:47.559944 139844782446400 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0421 13:13:47.560176 139844782446400 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 0x7f2e62faca70>}, handler_registry=None I0421 13:13:47.560414 139844782446400 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f2e62faca70>` 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 13:13:47.560462 139844782446400 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f1bf80b6510>` 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 13:13:47.560497 139844782446400 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 0x7f2e62faca70>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f2e62faca70>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f1bf80b6510>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f1bf80b6510>}). I0421 13:13:47.560830 139844782446400 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0421 13:13:47.560900 139844782446400 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f1bd8588540> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0421 13:13:50.747765 139725872551680 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776777014059958170, 'commit_timestamp_nsecs': 1776777058784302973, '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_16_zarr3_false/checkpoints/5/_CHECKPOINT_METADATA I0421 13:13:51.000278 139725872551680 checkpoint_manager.py:1871] Missing metrics for step 5 I0421 13:13:51.013475 139844782446400 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_16_zarr3_false/checkpoints I0421 13:13:51.840641 139844782446400 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_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f1bf80b1400> I0421 13:13:51.840842 139844782446400 checkpointing.py:302] Checkpoint manager created! I0421 13:13:52.784333 139844782446400 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 13:13:52.784439 139844782446400 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 13:13:53.159173 139844782446400 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 13:13:53.159262 139844782446400 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 13:13:53.180784 139844782446400 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 13:13:53.180879 139844782446400 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 13:13:53.205062 139844782446400 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 13:13:53.205130 139844782446400 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 13:13:53.221605 139844782446400 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 13:13:53.221667 139844782446400 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 13:13:53.238026 139844782446400 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 13:13:53.238089 139844782446400 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 13:13:53.262788 139844782446400 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 13:13:53.262857 139844782446400 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 13:13:53.284614 139844782446400 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 13:13:53.284680 139844782446400 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0421 13:13:53.497831 139844782446400 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0421 13:13:53.497940 139844782446400 checkpointing.py:582] restoring from this run's directory step 9 I0421 13:13:53.741561 139844782446400 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_16_zarr3_false/checkpoints/9. I0421 13:13:53.991013 139844782446400 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_16_zarr3_false/checkpoints/9. I0421 13:13:53.991447 139844782446400 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_16_zarr3_false/checkpoints/9. I0421 13:13:55.405999 1411 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0421 13:14:01.802681 139844782446400 jax_array_handlers.py:843] [process=4] /jax/orbax/read/worker/io/requested throughput: 246.825 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.402402639389038 s) (per-host) I0421 13:14:01.808381 139844782446400 base_pytree_checkpoint_handler.py:130] [process=4] /jax/checkpoint/read/gbytes_per_sec: 232.202 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.815597772598267 s) (per-host) I0421 13:14:02.206251 139844782446400 event_tracking.py:138] [process=4] [sync] Finished load in 8.46 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260421_114057/linen_xpk_feat_nnx_post_train_fixes_20260421_114057_16_zarr3_false/checkpoints/9 I0421 13:14:02.206403 139844782446400 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_16_zarr3_false/checkpoints', 'checkpointer_start_time': 1776777233.7414656, 'checkpointer_duration_secs': 8.464905500411987, 'checkpoint_manager_start_time': 1776777233.4981892, 'checkpoint_manager_duration_secs': 8.70818305015564} fsdp: 32 I0421 13:14:02.623571 139844782446400 maxtext_utils.py:1827] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0421 13:14:02.623711 139844782446400 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 13:14:02.623766 139844782446400 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 13:14:02.623824 139844782446400 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 13:14:02.623876 139844782446400 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 13:14:02.623915 139844782446400 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 13:14:02.623968 139844782446400 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 13:14:02.624021 139844782446400 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 13:14:02.624060 139844782446400 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 13:14:02.624097 139844782446400 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 13:14:02.624144 139844782446400 maxtext_utils.py:1827] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0421 13:14:02.624190 139844782446400 maxtext_utils.py:1827] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0421 13:14:03.109654 139844782446400 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0421 13:14:03.109764 139844782446400 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0421 13:14:03.125260 139844782446400 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0421 13:14:03.125321 139844782446400 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0421 13:14:14.083427 139844782446400 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 13:14:14.084233 139844782446400 metric_logger.py:301] number parameters: 1.104 billion I0421 13:14:25.121783 139844782446400 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0421 13:14:25.647203 139844782446400 checkpointing.py:798] Waited 0.5253996849060059 seconds for step 10 to finish before starting checkpointing. I0421 13:14:25.649652 139844782446400 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 13:14:25.651616 139844782446400 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 10 I0421 13:14:25.653020 139844782446400 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_16_zarr3_false/checkpoints/10. I0421 13:14:26.392223 139844782446400 signaling_client.py:364] Using JaxDistributedSignalingClient I0421 13:14:26.393150 139844782446400 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0421 13:14:26.393209 139844782446400 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 13:14:26.674553 139844782446400 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.282405s I0421 13:14:26.674744 139844782446400 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 5.390 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2862098217010498 s) (per-host) I0421 13:14:26.674799 139844782446400 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.286276s (batch_requests_ready=0.002127s, total_serialization_initiated=0.284074s, others=0.000075s) I0421 13:14:26.674886 139844782446400 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.290539s (all_items=0.000017s, per_item={'items': '0.00001717'}, temp_paths=0.290522) I0421 13:14:26.675607 139844782446400 event_tracking.py:125] [process=4] [async] Finished blocking save in 1.02 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_16_zarr3_false/checkpoints/10. I0421 13:14:26.675986 139725847373568 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-21 13:34:26.675948 I0421 13:14:26.680396 139844782446400 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0421 13:14:26.680728 139722122848000 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0421 13:14:26.680885 139844782446400 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_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776777265.6496336, 'wait_for_prev_duration_secs': 7.510185241699219e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776777265.6516554, 'checkpointer_blocking_duration_secs': 1.0244836807250977, 'get_old_steps_start_time': 1776777266.6761625, 'get_old_steps_duration_secs': 3.314018249511719e-05, 'checkpoint_manager_blocking_start_time': 1776777265.6478596, 'checkpoint_manager_blocking_duration_secs': 1.032987356185913} I0421 13:14:26.680994 139844782446400 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0421 13:14:26.681044 139844782446400 max_utils.py:750] Memstats: After params initialized: I0421 13:14:26.681094 139844782446400 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_16(process=4,(0,4,0,0)) I0421 13:14:26.681126 139844782446400 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_17(process=4,(1,4,0,0)) I0421 13:14:26.681154 139844782446400 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_20(process=4,(0,5,0,0)) I0421 13:14:26.681178 139844782446400 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_21(process=4,(1,5,0,0)) I0421 13:14:26.990807 139844782446400 metric_logger.py:196] completed step: 10, seconds: 11.037, TFLOP/s/device: 1.231, Tokens/s/device: 185.550, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0421 13:14:27.170116 139844782446400 metric_logger.py:196] completed step: 11, seconds: 1.868, TFLOP/s/device: 7.275, Tokens/s/device: 1096.620, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0421 13:14:27.317564 139844782446400 metric_logger.py:196] completed step: 12, seconds: 0.032, TFLOP/s/device: 423.671, Tokens/s/device: 63860.306, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0421 13:14:27.464978 139844782446400 metric_logger.py:196] completed step: 13, seconds: 0.154, TFLOP/s/device: 88.480, Tokens/s/device: 13336.633, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0421 13:14:27.469529 139844782446400 checkpointing.py:794] Waiting for step 15 to finish before checkpoint... I0421 13:14:27.758848 139844782446400 checkpointing.py:798] Waited 0.28928542137145996 seconds for step 15 to finish before starting checkpointing. I0421 13:14:27.761949 139844782446400 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 13:14:31.622296 139722131240704 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_16_zarr3_false/checkpoints/10/items/array_metadatas/process_4 I0421 13:15:09.581148 139725847373568 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 36.571 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.19257926940918 s) (per-host) I0421 13:15:09.581275 139725847373568 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.905179s. I0421 13:15:17.117347 139725847373568 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.441236s. I0421 13:15:17.117605 139722122848000 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 13:15:17.117738 139722122848000 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0421 13:15:17.117788 139722122848000 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0421 13:15:17.120375 139722122848000 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0421 13:15:17.120553 139844782446400 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 13:15:17.120688 139844782446400 checkpoint_manager.py:1452] Waiting for previous save to complete took 49.358738 seconds. If this number is high, consider checkpointing less frequently. I0421 13:15:17.122810 139844782446400 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 15 I0421 13:15:17.125224 139844782446400 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_16_zarr3_false/checkpoints/15. I0421 13:15:17.439538 139844782446400 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0421 13:15:17.439628 139844782446400 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 13:15:17.482793 139844782446400 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044409s I0421 13:15:17.482960 139844782446400 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 32.285 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04778027534484863 s) (per-host) I0421 13:15:17.483012 139844782446400 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.047840s (batch_requests_ready=0.001743s, total_serialization_initiated=0.046030s, others=0.000067s) I0421 13:15:17.483099 139844782446400 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.052670s (all_items=0.000013s, per_item={'items': '0.00001311'}, temp_paths=0.052657) I0421 13:15:17.483845 139844782446400 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.36 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_16_zarr3_false/checkpoints/15. I0421 13:15:17.484238 139722122848000 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-21 13:35:17.484201 I0421 13:15:17.486198 139844782446400 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0421 13:15:17.486473 139722131240704 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0421 13:15:17.486628 139844782446400 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_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776777267.7619202, 'wait_for_prev_duration_secs': 49.35873770713806, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776777317.1228492, 'checkpointer_blocking_duration_secs': 0.36153221130371094, 'get_old_steps_start_time': 1776777317.4844055, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776777267.7594662, 'checkpoint_manager_blocking_duration_secs': 49.72713017463684} I0421 13:15:17.486747 139844782446400 checkpointing.py:409] Started an asynchronous checkpoint save for step 15 I0421 13:15:17.487529 139844782446400 metric_logger.py:196] completed step: 14, seconds: 0.147, TFLOP/s/device: 92.401, Tokens/s/device: 13927.709, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0421 13:15:17.493447 139844782446400 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.180, Tokens/s/device: 13894.354, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0421 13:15:23.158285 139721032320768 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_16_zarr3_false/checkpoints/15/items/array_metadatas/process_4 I0421 13:15:38.761690 139844782446400 metric_logger.py:196] completed step: 16, seconds: 50.023, TFLOP/s/device: 0.272, Tokens/s/device: 40.942, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0421 13:15:38.908967 139844782446400 metric_logger.py:196] completed step: 17, seconds: 21.122, TFLOP/s/device: 0.643, Tokens/s/device: 96.960, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0421 13:15:39.056139 139844782446400 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.387, Tokens/s/device: 13473.330, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0421 13:15:39.203116 139844782446400 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0421 13:15:39.203824 139844782446400 checkpointing.py:798] Waited 0.0007297992706298828 seconds for step 19 to finish before starting checkpointing. I0421 13:15:39.205863 139844782446400 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 13:15:58.958497 139722122848000 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 38.041 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.52327513694763 s) (per-host) I0421 13:15:58.958619 139722122848000 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.474271s. I0421 13:16:06.761617 139722122848000 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 49.277255s. I0421 13:16:06.761930 139722131240704 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 13:16:06.762050 139722131240704 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0421 13:16:06.762099 139722131240704 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0421 13:16:06.764509 139722131240704 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0421 13:16:06.764715 139844782446400 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 13:16:06.764844 139844782446400 checkpoint_manager.py:1452] Waiting for previous save to complete took 27.558985 seconds. If this number is high, consider checkpointing less frequently. I0421 13:16:06.766180 139844782446400 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 19 I0421 13:16:06.768169 139844782446400 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_16_zarr3_false/checkpoints/19. I0421 13:16:07.506259 139844782446400 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0421 13:16:07.506352 139844782446400 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 13:16:07.547074 139844782446400 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.041812s I0421 13:16:07.547238 139844782446400 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 34.093 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04524636268615723 s) (per-host) I0421 13:16:07.547288 139844782446400 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.045305s (batch_requests_ready=0.001628s, total_serialization_initiated=0.043614s, others=0.000063s) I0421 13:16:07.547378 139844782446400 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.049281s (all_items=0.000010s, per_item={'items': '0.00001049'}, temp_paths=0.049271) I0421 13:16:07.548015 139844782446400 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.78 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_16_zarr3_false/checkpoints/19. I0421 13:16:07.548335 139722131240704 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-21 13:36:07.548297 I0421 13:16:07.550276 139844782446400 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0421 13:16:07.550542 139721032320768 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0421 13:16:07.550705 139844782446400 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_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776777339.2058342, 'wait_for_prev_duration_secs': 27.558985471725464, 'time_between_consecutive_saves_sec': 22.085418701171875, 'checkpointer_blocking_start_time': 1776777366.7662196, 'checkpointer_blocking_duration_secs': 0.7822704315185547, 'get_old_steps_start_time': 1776777367.5485063, 'get_old_steps_duration_secs': 2.47955322265625e-05, 'checkpoint_manager_blocking_start_time': 1776777339.204047, 'checkpoint_manager_blocking_duration_secs': 28.346619367599487} I0421 13:16:07.550809 139844782446400 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0421 13:16:07.550851 139844782446400 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 13:16:13.786851 139719413274368 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_16_zarr3_false/checkpoints/19/items/array_metadatas/process_4 I0421 13:16:50.873270 139722131240704 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 36.420 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.371238231658936 s) (per-host) I0421 13:16:50.873392 139722131240704 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.324937s. I0421 13:16:58.257671 139722131240704 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 50.709202s. I0421 13:16:58.257965 139721032320768 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 13:16:58.258083 139721032320768 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0421 13:16:58.258125 139721032320768 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0421 13:16:58.259757 139721032320768 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0421 13:16:58.259926 139844782446400 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 13:16:58.260075 139844782446400 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 13:16:58.261080 139844782446400 metric_logger.py:196] completed step: 19, seconds: 0.148, TFLOP/s/device: 92.089, Tokens/s/device: 13880.699, 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 13:17:11 UTC 2026 EXIT_CODE=0