XPK Start: Tue Apr 21 13:03:02 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 13:03:26.401795: 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 13:03:26.937578 140330045351744 max_utils.py:273] Attempting to initialize the jax distributed system... I0421 13:03:35.977944 140330045351744 distributed.py:149] Starting JAX distributed service on [::]:8482 I0421 13:03:35.980306 140330045351744 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-true-rs-27xzy-slice-job-0-0.mt-16-zarr3-true-rs-27xzy:8482 I0421 13:03:36.898154 140330045351744 max_utils.py:284] Jax distributed system initialized! I0421 13:03:42.931503 140330045351744 max_utils.py:800] System Information: Jax Version: 0.9.2 I0421 13:03:42.931604 140330045351744 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0421 13:03:42.931643 140330045351744 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 13:03:42.931678 140330045351744 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0421 13:03:43.630390 140330045351744 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 13:03:43.630995 140330045351744 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 13:03:43.631174 140330045351744 checkpointing.py:688] Setting up checkpoint logger... I0421 13:03:43.631224 140330045351744 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0421 13:03:43.631266 140330045351744 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0421 13:03:43.631597 140330045351744 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7fa098913980>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 13:03:44.982082 140330045351744 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0421 13:03:44.982318 140330045351744 checkpoint_manager.py:708] [process=7][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f9f5ed6e540>}, handler_registry=None I0421 13:03:44.982550 140330045351744 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f9f5ed6e540>` 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 13:03:44.982597 140330045351744 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f8c70187ad0>` 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 13:03:44.982629 140330045351744 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 0x7f9f5ed6e540>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f9f5ed6e540>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f8c70187ad0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f8c70187ad0>}). I0421 13:03:44.982956 140330045351744 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0421 13:03:44.983032 140330045351744 async_checkpointer.py:192] [process=7][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f8c2016c180> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0421 13:03:46.844453 140208762758912 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776776363854911832, 'commit_timestamp_nsecs': 1776776406610370124, '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_true/checkpoints/0/_CHECKPOINT_METADATA I0421 13:03:47.101309 140208762758912 checkpoint_manager.py:1871] Missing metrics for step 0 I0421 13:03:47.277208 140330045351744 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_true/checkpoints I0421 13:03:47.711530 140330045351744 checkpoint_manager.py:929] [process=7][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_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f8c701819a0> I0421 13:03:47.711697 140330045351744 checkpointing.py:302] Checkpoint manager created! I0421 13:03:48.641284 140330045351744 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 13:03:48.641385 140330045351744 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 13:03:49.029948 140330045351744 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 13:03:49.030032 140330045351744 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 13:03:49.046743 140330045351744 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 13:03:49.046799 140330045351744 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 13:03:49.071360 140330045351744 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 13:03:49.071433 140330045351744 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 13:03:49.088072 140330045351744 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 13:03:49.088140 140330045351744 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 13:03:49.104608 140330045351744 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 13:03:49.104665 140330045351744 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 13:03:49.129690 140330045351744 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 13:03:49.129768 140330045351744 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 13:03:49.150946 140330045351744 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 13:03:49.151014 140330045351744 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0421 13:03:49.367078 140330045351744 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0421 13:03:49.367182 140330045351744 checkpointing.py:582] restoring from this run's directory step 9 I0421 13:03:49.617540 140330045351744 event_tracking.py:70] [process=7] [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_true/checkpoints/9. I0421 13:03:49.868822 140330045351744 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_true/checkpoints/9. I0421 13:03:49.869259 140330045351744 event_tracking.py:125] [process=7] [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_true/checkpoints/9. I0421 13:03:51.277978 1451 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0421 13:03:57.643600 140330045351744 jax_array_handlers.py:843] [process=7] /jax/orbax/read/worker/io/requested throughput: 248.015 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.371695518493652 s) (per-host) I0421 13:03:57.648765 140330045351744 base_pytree_checkpoint_handler.py:130] [process=7] /jax/checkpoint/read/gbytes_per_sec: 233.194 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 6.786590576171875 s) (per-host) I0421 13:04:00.009630 140330045351744 event_tracking.py:138] [process=7] [sync] Finished load in 10.39 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_true/checkpoints/9 I0421 13:04:00.009818 140330045351744 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_true/checkpoints', 'checkpointer_start_time': 1776776629.6174433, 'checkpointer_duration_secs': 10.39234209060669, 'checkpoint_manager_start_time': 1776776629.3674796, 'checkpoint_manager_duration_secs': 10.642308235168457} fsdp: 32 I0421 13:04:00.427360 140330045351744 maxtext_utils.py:1827] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0421 13:04:00.427484 140330045351744 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 13:04:00.427538 140330045351744 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 13:04:00.427595 140330045351744 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 13:04:00.427645 140330045351744 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 13:04:00.427693 140330045351744 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 13:04:00.427760 140330045351744 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 13:04:00.427817 140330045351744 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 13:04:00.427858 140330045351744 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 13:04:00.427896 140330045351744 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 13:04:00.427944 140330045351744 maxtext_utils.py:1827] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0421 13:04:00.427992 140330045351744 maxtext_utils.py:1827] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0421 13:04:00.917177 140330045351744 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0421 13:04:00.917274 140330045351744 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0421 13:04:00.932928 140330045351744 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0421 13:04:00.932986 140330045351744 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0421 13:04:11.915920 140330045351744 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 13:04:11.916772 140330045351744 metric_logger.py:301] number parameters: 1.104 billion I0421 13:04:23.127513 140330045351744 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0421 13:04:23.285013 140330045351744 checkpointing.py:798] Waited 0.15747618675231934 seconds for step 10 to finish before starting checkpointing. I0421 13:04:23.287970 140330045351744 checkpoint_manager.py:2009] [process=7][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 13:04:23.289835 140330045351744 checkpoint_manager.py:1512] [process=7] Saving checkpoint at step 10 I0421 13:04:23.291481 140330045351744 event_tracking.py:70] [process=7] [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_true/checkpoints/10. I0421 13:04:23.928671 140330045351744 signaling_client.py:364] Using JaxDistributedSignalingClient I0421 13:04:23.929624 140330045351744 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0421 13:04:23.929681 140330045351744 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 13:04:24.210729 140330045351744 base_pytree_checkpoint_handler.py:154] [process=7][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.282131s I0421 13:04:24.210930 140330045351744 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/blocking_gbytes_per_sec: 5.389 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.286271333694458 s) (per-host) I0421 13:04:24.210982 140330045351744 base_pytree_checkpoint_handler.py:768] [process=7][thread=MainThread] Initiated Pytree async_save. Time taken: 0.286335s (batch_requests_ready=0.002275s, total_serialization_initiated=0.283988s, others=0.000072s) I0421 13:04:24.211072 140330045351744 composite_checkpoint_handler.py:715] [process=7][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.290562s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=0.290545) I0421 13:04:24.211829 140330045351744 event_tracking.py:125] [process=7] [async] Finished blocking save in 0.92 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_true/checkpoints/10. I0421 13:04:24.212123 140208167171840 async_checkpointer.py:76] [process=7][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-21 13:24:24.212091 I0421 13:04:24.214279 140330045351744 checkpoint_manager.py:1560] [process=7][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0421 13:04:24.214570 140204422031104 async_checkpointer.py:280] [process=7][thread=save_finalize] Waiting for background save thread=async_save. I0421 13:04:24.214721 140330045351744 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_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776776663.2879505, 'wait_for_prev_duration_secs': 6.127357482910156e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776776663.289876, 'checkpointer_blocking_duration_secs': 0.9223968982696533, 'get_old_steps_start_time': 1776776664.2122965, 'get_old_steps_duration_secs': 3.266334533691406e-05, 'checkpoint_manager_blocking_start_time': 1776776663.2855413, 'checkpoint_manager_blocking_duration_secs': 0.9291448593139648} I0421 13:04:24.214845 140330045351744 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0421 13:04:24.214897 140330045351744 max_utils.py:750] Memstats: After params initialized: I0421 13:04:24.214944 140330045351744 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_26(process=7,(2,6,0,0)) I0421 13:04:24.214977 140330045351744 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_27(process=7,(3,6,0,0)) I0421 13:04:24.215003 140330045351744 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_30(process=7,(2,7,0,0)) I0421 13:04:24.215032 140330045351744 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_31(process=7,(3,7,0,0)) I0421 13:04:24.528994 140330045351744 metric_logger.py:196] completed step: 10, seconds: 11.211, TFLOP/s/device: 1.212, Tokens/s/device: 182.684, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0421 13:04:24.707540 140330045351744 metric_logger.py:196] completed step: 11, seconds: 1.400, TFLOP/s/device: 9.705, Tokens/s/device: 1462.915, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0421 13:04:24.855823 140330045351744 metric_logger.py:196] completed step: 12, seconds: 0.031, TFLOP/s/device: 432.946, Tokens/s/device: 65258.261, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0421 13:04:25.002851 140330045351744 metric_logger.py:196] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.610, Tokens/s/device: 13356.202, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0421 13:04:25.007502 140330045351744 checkpointing.py:794] Waiting for step 15 to finish before checkpoint... I0421 13:04:25.296685 140330045351744 checkpointing.py:798] Waited 0.2891542911529541 seconds for step 15 to finish before starting checkpointing. I0421 13:04:25.299900 140330045351744 checkpoint_manager.py:2020] [process=7][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 13:04:29.562546 140206564951808 array_metadata_store.py:203] [process=7][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_true/checkpoints/10/items/array_metadatas/process_7 I0421 13:05:04.662416 140208167171840 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/gbytes_per_sec: 38.775 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 40.737728118896484 s) (per-host) I0421 13:05:04.662526 140208167171840 async_checkpointer.py:90] [process=7][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.450294s. I0421 13:05:13.868016 140208167171840 async_checkpointer.py:160] [process=7][thread=async_save] Background save thread done. Time taken: 49.655769s. I0421 13:05:13.868282 140204422031104 async_checkpointer.py:288] [process=7][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 13:05:13.868399 140204422031104 async_checkpointer.py:298] [process=7][thread=save_finalize] No errors found in background save thread=async_save. I0421 13:05:13.868447 140204422031104 checkpoint_manager.py:2137] [process=7][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0421 13:05:13.870059 140204422031104 checkpoint_manager.py:2146] [process=7][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0421 13:05:13.870318 140330045351744 checkpoint_manager.py:2032] [process=7][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0421 13:05:13.870449 140330045351744 checkpoint_manager.py:1452] Waiting for previous save to complete took 48.570550 seconds. If this number is high, consider checkpointing less frequently. I0421 13:05:13.872672 140330045351744 checkpoint_manager.py:1512] [process=7] Saving checkpoint at step 15 I0421 13:05:13.875218 140330045351744 event_tracking.py:70] [process=7] [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_true/checkpoints/15. I0421 13:05:14.210692 140330045351744 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0421 13:05:14.210806 140330045351744 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 13:05:14.237150 140330045351744 base_pytree_checkpoint_handler.py:154] [process=7][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027627s I0421 13:05:14.237284 140330045351744 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/blocking_gbytes_per_sec: 49.594 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.031103849411010742 s) (per-host) I0421 13:05:14.237331 140330045351744 base_pytree_checkpoint_handler.py:768] [process=7][thread=MainThread] Initiated Pytree async_save. Time taken: 0.031159s (batch_requests_ready=0.001884s, total_serialization_initiated=0.029214s, others=0.000062s) I0421 13:05:14.237409 140330045351744 composite_checkpoint_handler.py:715] [process=7][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.036011s (all_items=0.000013s, per_item={'items': '0.00001287'}, temp_paths=0.035998) I0421 13:05:14.238097 140330045351744 event_tracking.py:125] [process=7] [async] Finished blocking save in 0.37 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_true/checkpoints/15. I0421 13:05:14.238353 140204422031104 async_checkpointer.py:76] [process=7][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-21 13:25:14.238329 I0421 13:05:14.250412 140330045351744 checkpoint_manager.py:1560] [process=7][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0421 13:05:14.250660 140206564951808 async_checkpointer.py:280] [process=7][thread=save_finalize] Waiting for background save thread=async_save. I0421 13:05:14.250795 140330045351744 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_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776776665.29987, 'wait_for_prev_duration_secs': 48.570549726486206, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776776713.8727105, 'checkpointer_blocking_duration_secs': 0.3657419681549072, 'get_old_steps_start_time': 1776776714.2384756, 'get_old_steps_duration_secs': 3.337860107421875e-05, 'checkpoint_manager_blocking_start_time': 1776776665.2973416, 'checkpoint_manager_blocking_duration_secs': 48.95341944694519} I0421 13:05:14.250910 140330045351744 checkpointing.py:409] Started an asynchronous checkpoint save for step 15 I0421 13:05:14.251695 140330045351744 metric_logger.py:196] completed step: 14, seconds: 0.148, TFLOP/s/device: 91.750, Tokens/s/device: 13829.615, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0421 13:05:14.257795 140330045351744 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.350, Tokens/s/device: 13919.947, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0421 13:05:19.359883 140202286769920 array_metadata_store.py:203] [process=7][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_true/checkpoints/15/items/array_metadatas/process_7 I0421 13:05:35.685761 140330045351744 metric_logger.py:196] completed step: 16, seconds: 49.249, TFLOP/s/device: 0.276, Tokens/s/device: 41.585, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0421 13:05:35.832994 140330045351744 metric_logger.py:196] completed step: 17, seconds: 21.282, TFLOP/s/device: 0.638, Tokens/s/device: 96.232, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0421 13:05:35.980274 140330045351744 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.188, Tokens/s/device: 13443.436, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0421 13:05:36.127015 140330045351744 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0421 13:05:36.127693 140330045351744 checkpointing.py:798] Waited 0.0006997585296630859 seconds for step 19 to finish before starting checkpointing. I0421 13:05:36.129752 140330045351744 checkpoint_manager.py:2020] [process=7][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 13:05:55.480597 140204422031104 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/gbytes_per_sec: 38.271 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.274381160736084 s) (per-host) I0421 13:05:55.480706 140204422031104 async_checkpointer.py:90] [process=7][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.242276s. I0421 13:06:04.571887 140204422031104 async_checkpointer.py:160] [process=7][thread=async_save] Background save thread done. Time taken: 50.333441s. I0421 13:06:04.572160 140206564951808 async_checkpointer.py:288] [process=7][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 13:06:04.572227 140206564951808 async_checkpointer.py:298] [process=7][thread=save_finalize] No errors found in background save thread=async_save. I0421 13:06:04.572282 140206564951808 checkpoint_manager.py:2137] [process=7][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0421 13:06:04.573851 140206564951808 checkpoint_manager.py:2146] [process=7][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0421 13:06:04.574002 140330045351744 checkpoint_manager.py:2032] [process=7][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0421 13:06:04.574130 140330045351744 checkpoint_manager.py:1452] Waiting for previous save to complete took 28.444400 seconds. If this number is high, consider checkpointing less frequently. I0421 13:06:04.575617 140330045351744 checkpoint_manager.py:1512] [process=7] Saving checkpoint at step 19 I0421 13:06:04.577648 140330045351744 event_tracking.py:70] [process=7] [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_true/checkpoints/19. I0421 13:06:05.301966 140330045351744 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0421 13:06:05.302065 140330045351744 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 13:06:05.328207 140330045351744 base_pytree_checkpoint_handler.py:154] [process=7][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027158s I0421 13:06:05.328372 140330045351744 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/blocking_gbytes_per_sec: 50.059 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03081536293029785 s) (per-host) I0421 13:06:05.328423 140330045351744 base_pytree_checkpoint_handler.py:768] [process=7][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030877s (batch_requests_ready=0.001786s, total_serialization_initiated=0.029024s, others=0.000067s) I0421 13:06:05.328510 140330045351744 composite_checkpoint_handler.py:715] [process=7][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034959s (all_items=0.000011s, per_item={'items': '0.00001121'}, temp_paths=0.034948) I0421 13:06:05.329285 140330045351744 event_tracking.py:125] [process=7] [async] Finished blocking save in 0.75 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_true/checkpoints/19. I0421 13:06:05.329569 140206564951808 async_checkpointer.py:76] [process=7][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-21 13:26:05.329542 I0421 13:06:05.332813 140330045351744 checkpoint_manager.py:1560] [process=7][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0421 13:06:05.333069 140208158779136 async_checkpointer.py:280] [process=7][thread=save_finalize] Waiting for background save thread=async_save. I0421 13:06:05.333225 140330045351744 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_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776776736.129706, 'wait_for_prev_duration_secs': 28.44440007209778, 'time_between_consecutive_saves_sec': 22.259597301483154, 'checkpointer_blocking_start_time': 1776776764.5756562, 'checkpointer_blocking_duration_secs': 0.7540380954742432, 'get_old_steps_start_time': 1776776765.3297124, 'get_old_steps_duration_secs': 4.076957702636719e-05, 'checkpoint_manager_blocking_start_time': 1776776736.1279483, 'checkpoint_manager_blocking_duration_secs': 29.205244064331055} I0421 13:06:05.333336 140330045351744 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0421 13:06:05.333378 140330045351744 checkpoint_manager.py:2020] [process=7][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 13:06:10.500540 140203332237056 array_metadata_store.py:203] [process=7][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_true/checkpoints/19/items/array_metadatas/process_7 I0421 13:06:46.344329 140206564951808 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/gbytes_per_sec: 38.483 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.046743392944336 s) (per-host) I0421 13:06:46.344437 140206564951808 async_checkpointer.py:90] [process=7][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.014775s. I0421 13:06:54.891143 140206564951808 async_checkpointer.py:160] [process=7][thread=async_save] Background save thread done. Time taken: 49.561465s. I0421 13:06:54.891456 140208158779136 async_checkpointer.py:288] [process=7][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 13:06:54.891577 140208158779136 async_checkpointer.py:298] [process=7][thread=save_finalize] No errors found in background save thread=async_save. I0421 13:06:54.891618 140208158779136 checkpoint_manager.py:2137] [process=7][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0421 13:06:54.893020 140208158779136 checkpoint_manager.py:2146] [process=7][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0421 13:06:54.893203 140330045351744 checkpoint_manager.py:2032] [process=7][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0421 13:06:54.893353 140330045351744 checkpoint_manager.py:2009] [process=7][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 13:06:54.894281 140330045351744 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.380, Tokens/s/device: 13924.490, 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 13:07:07 UTC 2026 EXIT_CODE=0