XPK Start: Fri Apr 17 14:19:18 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-17 14:19:42.968398: 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) I0417 14:19:43.151908 136979417962304 max_utils.py:273] Attempting to initialize the jax distributed system... I0417 14:19:52.191416 136979417962304 distributed.py:149] Starting JAX distributed service on [::]:8482 I0417 14:19:52.193751 136979417962304 distributed.py:172] Connecting to JAX distributed service on mt-14-async-ckpt-false--zkmni-slice-job-0-0.mt-14-async-ckpt-false--zkmni:8482 I0417 14:19:53.599682 136979417962304 max_utils.py:284] Jax distributed system initialized! I0417 14:20:01.047735 136979417962304 max_utils.py:800] System Information: Jax Version: 0.9.2 I0417 14:20:01.047878 136979417962304 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0417 14:20:01.047927 136979417962304 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 I0417 14:20:01.047962 136979417962304 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0417 14:20:01.718691 136979417962304 maxtext_utils.py:1548] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0417 14:20:01.718967 136979417962304 checkpointing.py:677] Setting up checkpoint logger... I0417 14:20:01.719021 136979417962304 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0417 14:20:01.719068 136979417962304 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0417 14:20:01.719409 136979417962304 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 0x7c946d91b1d0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0417 14:20:04.775708 136979417962304 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0417 14:20:04.776084 136979417962304 checkpoint_manager.py:708] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c80e3e40890>}, handler_registry=None I0417 14:20:04.776343 136979417962304 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c80e3e40890>` 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`. I0417 14:20:04.776395 136979417962304 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c81b57c1bb0>` 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`. I0417 14:20:04.776433 136979417962304 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 0x7c80e3e40890>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c80e3e40890>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c81b57c1bb0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c81b57c1bb0>}). I0417 14:20:04.776776 136979417962304 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0417 14:20:06.041089 136979417962304 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints I0417 14:20:06.894392 136979417962304 checkpoint_manager.py:929] [process=0][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=False, 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_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7c80e3e43c20> I0417 14:20:06.894554 136979417962304 checkpointing.py:301] Checkpoint manager created! I0417 14:20:07.816074 136979417962304 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 14:20:07.816183 136979417962304 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 14:20:08.199443 136979417962304 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 14:20:08.199534 136979417962304 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 14:20:08.216035 136979417962304 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 14:20:08.216094 136979417962304 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 14:20:08.239944 136979417962304 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 14:20:08.240011 136979417962304 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 14:20:08.256495 136979417962304 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 14:20:08.256556 136979417962304 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 14:20:08.273083 136979417962304 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 14:20:08.273140 136979417962304 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 14:20:08.298006 136979417962304 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0417 14:20:08.298074 136979417962304 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 14:20:08.320107 136979417962304 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0417 14:20:08.320174 136979417962304 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0417 14:20:08.540191 136979417962304 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0417 14:20:08.540298 136979417962304 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0417 14:20:09.976377 136979417962304 maxtext_utils.py:1651] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0417 14:20:09.976513 136979417962304 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0417 14:20:09.976577 136979417962304 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0417 14:20:09.976656 136979417962304 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0417 14:20:09.976732 136979417962304 maxtext_utils.py:1651] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0417 14:20:09.976791 136979417962304 maxtext_utils.py:1651] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0417 14:20:09.976889 136979417962304 maxtext_utils.py:1651] 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) I0417 14:20:09.976973 136979417962304 maxtext_utils.py:1651] 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') I0417 14:20:09.977033 136979417962304 maxtext_utils.py:1651] 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) I0417 14:20:09.977092 136979417962304 maxtext_utils.py:1651] 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) I0417 14:20:09.977156 136979417962304 maxtext_utils.py:1651] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0417 14:20:09.977230 136979417962304 maxtext_utils.py:1651] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0417 14:20:10.473021 136979417962304 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0417 14:20:10.473118 136979417962304 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0417 14:20:10.488984 136979417962304 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0417 14:20:10.489050 136979417962304 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0417 14:20:21.209072 136979417962304 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. I0417 14:20:21.217718 136979417962304 metric_logger.py:301] number parameters: 1.104 billion I0417 14:20:32.539579 136979417962304 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0417 14:20:32.694833 136979417962304 checkpointing.py:776] Waited 0.15522217750549316 seconds for step 0 to finish before starting checkpointing. I0417 14:20:32.697116 136979417962304 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 14:20:32.698991 136979417962304 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0417 14:20:32.700390 136979417962304 event_tracking.py:70] [process=0] [sync] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/0. I0417 14:20:33.748735 136979417962304 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/0 I0417 14:20:36.622018 136979417962304 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776435635746690694, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776435635746690694, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA I0417 14:20:36.625342 136979417962304 signaling_client.py:364] Using JaxDistributedSignalingClient I0417 14:20:36.626363 136979417962304 future.py:372] [process=0][thread=MainThread][operation_id=1] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 14:20:36.896200 136851391694592 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/0/items I0417 14:20:38.075997 136979417962304 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0417 14:20:38.076093 136979417962304 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 I0417 14:20:38.107538 2663 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0417 14:20:38.349118 136979417962304 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.274395s I0417 14:20:38.354866 136979417962304 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.395 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.285947322845459 s) (per-host) I0417 14:20:38.354940 136979417962304 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.286052s (batch_requests_ready=0.002288s, total_serialization_initiated=0.278112s, others=0.005652s) I0417 14:20:38.355032 136979417962304 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.729787s (all_items=0.000033s, per_item={'items': '0.00003266'}, temp_paths=1.729754) I0417 14:20:38.355080 136979417962304 future.py:372] [process=0][thread=MainThread][operation_id=1] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 14:20:40.616471 136848677201664 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/0/items/array_metadatas/process_0 I0417 14:21:15.005420 136846038263552 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 36.650112s (commit=35.397786s, array_metadata_write=1.252326s) I0417 14:21:15.006547 136979417962304 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 42.764 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 36.93760108947754 s) (per-host) I0417 14:21:15.792171 136979417962304 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776435635746690694, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA I0417 14:21:17.013427 136979417962304 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776435635746690694, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA I0417 14:21:17.015478 136979417962304 event_tracking.py:125] [process=0] [sync] Finished blocking save in 44.32 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/0. I0417 14:21:17.631948 136979417962304 array_metadata_store.py:411] [process=0][thread=MainThread] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000281s. I0417 14:21:18.391649 136979417962304 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 14:21:19.330358 136979417962304 base_pytree_checkpoint_handler.py:1406] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.314602s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/0/items I0417 14:21:19.332073 136979417962304 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/0/items I0417 14:21:19.893988 136979417962304 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/0 I0417 14:21:21.808865 136979417962304 atomicity.py:847] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/0`. I0417 14:21:21.811596 136979417962304 event_tracking.py:138] [process=0] [sync] Finished save in 49.11 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/0 I0417 14:21:21.814610 136979417962304 checkpoint_manager.py:2137] [process=0][thread=MainThread][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0417 14:21:21.816402 136979417962304 checkpoint_manager.py:2146] [process=0][thread=MainThread][step=0] CheckpointManager Save Finalize is done on all hosts. I0417 14:21:21.816454 136979417962304 checkpoint_manager.py:1581] [process=0][thread=MainThread][step=0] Finished synchronous save. I0417 14:21:21.816509 136979417962304 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776435632.6970973, 'wait_for_prev_duration_secs': 5.888938903808594e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776435632.6990297, 'checkpointer_blocking_duration_secs': 49.112671852111816, 'get_old_steps_start_time': 1776435681.811722, 'get_old_steps_duration_secs': 2.4080276489257812e-05, 'checkpoint_manager_blocking_start_time': 1776435632.6953428, 'checkpoint_manager_blocking_duration_secs': 49.121140003204346} I0417 14:21:21.816594 136979417962304 checkpointing.py:410] Saved a checkpoint at step 0. I0417 14:21:21.816637 136979417962304 max_utils.py:750] Memstats: After params initialized: I0417 14:21:21.816684 136979417962304 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0417 14:21:21.816715 136979417962304 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0417 14:21:21.816742 136979417962304 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0417 14:21:21.816767 136979417962304 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0417 14:21:22.165360 136979417962304 metric_logger.py:196] completed step: 0, seconds: 11.017, TFLOP/s/device: 1.233, Tokens/s/device: 185.886, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0417 14:21:22.169037 136979417962304 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/tensorboard/' I0417 14:21:22.600298 136979417962304 metric_logger.py:196] completed step: 1, seconds: 49.609, TFLOP/s/device: 0.274, Tokens/s/device: 41.283, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0417 14:21:22.746254 136979417962304 metric_logger.py:196] completed step: 2, seconds: 0.451, TFLOP/s/device: 30.145, Tokens/s/device: 4543.760, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0417 14:21:22.893893 136979417962304 metric_logger.py:196] completed step: 3, seconds: 0.009, TFLOP/s/device: 1506.835, Tokens/s/device: 227126.539, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0417 14:21:22.903394 136979417962304 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0417 14:21:23.187926 136979417962304 checkpointing.py:776] Waited 0.2843747138977051 seconds for step 5 to finish before starting checkpointing. I0417 14:21:23.191044 136979417962304 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 14:21:23.192818 136979417962304 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 5 I0417 14:21:23.194569 136979417962304 event_tracking.py:70] [process=0] [sync] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/5. I0417 14:21:24.202062 136979417962304 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/5 I0417 14:21:27.019489 136979417962304 future.py:372] [process=0][thread=MainThread][operation_id=2] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 14:21:27.734208 136850844923648 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/5/items I0417 14:21:29.638769 136979417962304 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0417 14:21:29.638924 136979417962304 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 I0417 14:21:29.734946 136979417962304 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.097364s I0417 14:21:29.747519 136979417962304 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 13.073 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.11800074577331543 s) (per-host) I0417 14:21:29.747651 136979417962304 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.118167s (batch_requests_ready=0.004444s, total_serialization_initiated=0.101250s, others=0.012473s) I0417 14:21:29.747813 136979417962304 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.729456s (all_items=0.000042s, per_item={'items': '0.00004244'}, temp_paths=2.729413) I0417 14:21:29.747918 136979417962304 future.py:372] [process=0][thread=MainThread][operation_id=2] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 14:21:31.560338 136850836530944 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/5/items/array_metadatas/process_0 I0417 14:22:11.238445 136850853316352 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.486622s (commit=40.224978s, array_metadata_write=1.261643s) I0417 14:22:11.239822 136979417962304 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 37.962 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.61028742790222 s) (per-host) I0417 14:22:12.852511 136979417962304 event_tracking.py:125] [process=0] [sync] Finished blocking save in 49.66 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/5. I0417 14:22:13.499508 136979417962304 array_metadata_store.py:411] [process=0][thread=MainThread] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000382s. I0417 14:22:13.700062 136979417962304 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 14:22:14.777230 136979417962304 base_pytree_checkpoint_handler.py:1406] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.924257s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/5/items I0417 14:22:14.779507 136979417962304 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/5/items I0417 14:22:15.331670 136979417962304 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/5 I0417 14:22:17.209798 136979417962304 atomicity.py:847] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/5`. I0417 14:22:17.213329 136979417962304 event_tracking.py:138] [process=0] [sync] Finished save in 54.02 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/5 I0417 14:22:17.216291 136979417962304 checkpoint_manager.py:2137] [process=0][thread=MainThread][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0417 14:22:17.217999 136979417962304 checkpoint_manager.py:2146] [process=0][thread=MainThread][step=5] CheckpointManager Save Finalize is done on all hosts. I0417 14:22:17.218081 136979417962304 checkpoint_manager.py:1581] [process=0][thread=MainThread][step=5] Finished synchronous save. I0417 14:22:17.218178 136979417962304 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776435683.1910121, 'wait_for_prev_duration_secs': 0.00010251998901367188, 'time_between_consecutive_saves_sec': 1.3745698928833008, 'checkpointer_blocking_start_time': 1776435683.1928968, 'checkpointer_blocking_duration_secs': 54.02058792114258, 'get_old_steps_start_time': 1776435737.2135074, 'get_old_steps_duration_secs': 3.361701965332031e-05, 'checkpoint_manager_blocking_start_time': 1776435683.1885626, 'checkpoint_manager_blocking_duration_secs': 54.02957224845886} I0417 14:22:17.218295 136979417962304 checkpointing.py:410] Saved a checkpoint at step 5. I0417 14:22:39.853857 136979417962304 metric_logger.py:196] completed step: 4, seconds: 0.147, TFLOP/s/device: 92.389, Tokens/s/device: 13925.815, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0417 14:22:39.862887 136979417962304 metric_logger.py:196] completed step: 5, seconds: 0.148, TFLOP/s/device: 91.763, Tokens/s/device: 13831.483, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0417 14:22:40.008240 136979417962304 metric_logger.py:196] completed step: 6, seconds: 76.958, TFLOP/s/device: 0.177, Tokens/s/device: 26.612, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0417 14:22:40.155852 136979417962304 metric_logger.py:196] completed step: 7, seconds: 0.007, TFLOP/s/device: 1926.433, Tokens/s/device: 290372.891, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0417 14:22:40.303287 136979417962304 metric_logger.py:196] completed step: 8, seconds: 0.146, TFLOP/s/device: 92.811, Tokens/s/device: 13989.549, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0417 14:22:40.450054 136979417962304 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0417 14:22:40.450680 136979417962304 checkpointing.py:776] Waited 0.0006465911865234375 seconds for step 9 to finish before starting checkpointing. I0417 14:22:40.452991 136979417962304 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 14:22:40.454936 136979417962304 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0417 14:22:40.456400 136979417962304 event_tracking.py:70] [process=0] [sync] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/9. I0417 14:22:41.506449 136979417962304 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/9 I0417 14:22:43.598145 136979417962304 future.py:372] [process=0][thread=MainThread][operation_id=3] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 14:22:43.846079 136852531037952 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/9/items I0417 14:22:45.448065 136979417962304 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0417 14:22:45.448161 136979417962304 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 I0417 14:22:45.501773 136979417962304 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.054892s I0417 14:22:45.507448 136979417962304 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 23.899 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.064544677734375 s) (per-host) I0417 14:22:45.507522 136979417962304 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.064639s (batch_requests_ready=0.001902s, total_serialization_initiated=0.061871s, others=0.000865s) I0417 14:22:45.507608 136979417962304 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.910424s (all_items=0.000022s, per_item={'items': '0.00002241'}, temp_paths=1.910402) I0417 14:22:45.507651 136979417962304 future.py:372] [process=0][thread=MainThread][operation_id=3] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 14:22:47.791670 136850878494464 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/9/items/array_metadatas/process_0 I0417 14:23:26.194483 136850853328640 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 40.686651s (commit=39.404115s, array_metadata_write=1.282537s) I0417 14:23:26.195533 136979417962304 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 38.761 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 40.75260519981384 s) (per-host) I0417 14:23:28.574013 136979417962304 event_tracking.py:125] [process=0] [sync] Finished blocking save in 48.12 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/9. I0417 14:23:29.656632 136979417962304 array_metadata_store.py:411] [process=0][thread=MainThread] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000242s. I0417 14:23:30.000865 136979417962304 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0417 14:23:30.942878 136979417962304 base_pytree_checkpoint_handler.py:1406] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.368537s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/9/items I0417 14:23:30.944579 136979417962304 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/9/items I0417 14:23:31.503812 136979417962304 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/9 I0417 14:23:32.964047 136979417962304 atomicity.py:847] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/9`. I0417 14:23:32.966567 136979417962304 event_tracking.py:138] [process=0] [sync] Finished save in 52.51 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints/9 I0417 14:23:32.969482 136979417962304 checkpoint_manager.py:2137] [process=0][thread=MainThread][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0417 14:23:32.970888 136979417962304 checkpoint_manager.py:2146] [process=0][thread=MainThread][step=9] CheckpointManager Save Finalize is done on all hosts. I0417 14:23:32.970937 136979417962304 checkpoint_manager.py:1581] [process=0][thread=MainThread][step=9] Finished synchronous save. I0417 14:23:32.970988 136979417962304 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260417_125726/linen_xpk_main_20260417_125726_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776435760.4529707, 'wait_for_prev_duration_secs': 5.817413330078125e-05, 'time_between_consecutive_saves_sec': 23.234905004501343, 'checkpointer_blocking_start_time': 1776435760.4549727, 'checkpointer_blocking_duration_secs': 52.511696338653564, 'get_old_steps_start_time': 1776435812.9666858, 'get_old_steps_duration_secs': 2.4318695068359375e-05, 'checkpoint_manager_blocking_start_time': 1776435760.4509385, 'checkpoint_manager_blocking_duration_secs': 52.52002429962158} I0417 14:23:32.971074 136979417962304 checkpointing.py:410] Saved a checkpoint at step 9. I0417 14:23:32.971106 136979417962304 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 14:23:32.971136 136979417962304 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 14:23:32.971732 136979417962304 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.180, Tokens/s/device: 13894.448, total_weights: 65536, loss: 7.181, lm_loss: 7.181, perplexity: 1314.495 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Fri Apr 17 14:23:44 UTC 2026 EXIT_CODE=0