MaxView

‹ 14_async_ckpt_false_resumeCase: 14_async_ckpt_false_save15_ocdbt_false_resume ›

Metrics: Linen vs NNX  ·  main

MetricLinen  147168c3aNNX  1b5e4eb5aDiff (NNX − Linen)
Parameters1.104 billion
Final loss7.1810
TFLOP/s91.925
Tok/s13855.9
Avg s/step14.427
Memory %1.38
JAX0.9.20.9.2

Diff = NNX value − Linen value. Green = NNX improved. Red = NNX regressed.

Linen  ·  147168c3a  ·  main_20260417_235819  ·  full log
XPK Start: Sat Apr 18 01:12:53 UTC 2026
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
2026-04-18 01:13:17.146526: 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)
I0418 01:13:17.679888 139892134672192 max_utils.py:273] Attempting to initialize the jax distributed system...
I0418 01:13:26.720099 139892134672192 distributed.py:149] Starting JAX distributed service on [::]:8482
I0418 01:13:26.722523 139892134672192 distributed.py:172] Connecting to JAX distributed service on mt-14-async-ckpt-false--mofsv-slice-job-0-0.mt-14-async-ckpt-false--mofsv:8482
I0418 01:13:27.626196 139892134672192 max_utils.py:284] Jax distributed system initialized!
I0418 01:13:32.869967 139892134672192 max_utils.py:800] System Information: Jax Version: 0.9.2
I0418 01:13:32.870066 139892134672192 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0418 01:13:32.870107 139892134672192 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
I0418 01:13:32.870145 139892134672192 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
I0418 01:13:33.554522 139892134672192 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0418 01:13:33.554793 139892134672192 checkpointing.py:677] Setting up checkpoint logger...
I0418 01:13:33.554867 139892134672192 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True
I0418 01:13:33.554914 139892134672192 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None
I0418 01:13:33.555259 139892134672192 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 0x7f3a752501a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0418 01:13:36.477112 139892134672192 checkpointing.py:265] Enabling policy for fixed interval checkpointing.
I0418 01:13:36.477357 139892134672192 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 0x7f26f022a9f0>}, handler_registry=None
I0418 01:13:36.477597 139892134672192 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f26f022a9f0>` 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`.
I0418 01:13:36.477646 139892134672192 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f26f0235580>` 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`.
I0418 01:13:36.477680 139892134672192 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 0x7f26f022a9f0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f26f022a9f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f26f0235580>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f26f0235580>}).
I0418 01:13:36.478020 139892134672192 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34
I0418 01:13:38.013316 139892134672192 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints
I0418 01:13:38.764526 139892134672192 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_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f26f022bc20>
I0418 01:13:38.764699 139892134672192 checkpointing.py:301] Checkpoint manager created!
I0418 01:13:39.676131 139892134672192 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0418 01:13:39.676244 139892134672192 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0418 01:13:40.059473 139892134672192 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0418 01:13:40.059559 139892134672192 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0418 01:13:40.075745 139892134672192 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0418 01:13:40.075801 139892134672192 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None).
I0418 01:13:40.099209 139892134672192 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0418 01:13:40.099273 139892134672192 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0418 01:13:40.115553 139892134672192 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0418 01:13:40.115614 139892134672192 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0418 01:13:40.131817 139892134672192 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0418 01:13:40.131888 139892134672192 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0418 01:13:40.157120 139892134672192 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0418 01:13:40.157191 139892134672192 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None).
I0418 01:13:40.178095 139892134672192 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0418 01:13:40.178159 139892134672192 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None).
I0418 01:13:40.393706 139892134672192 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint
I0418 01:13:40.393811 139892134672192 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint.
fsdp: 32
I0418 01:13:41.806800 139892134672192 maxtext_utils.py:1654]  params/params/decoder/decoder_norm/scale
    Shape:     float32[2048]
    Logical:   P('norm',)
    Physical:  (None,)
I0418 01:13:41.806983 139892134672192 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wi_0/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0418 01:13:41.807036 139892134672192 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wi_1/kernel
    Shape:     float32[2048,16,7168]
    Logical:   P('embed', 'layers', 'mlp')
    Physical:  ('fsdp', None, None)
I0418 01:13:41.807095 139892134672192 maxtext_utils.py:1654]  params/params/decoder/layers/mlp/wo/kernel
    Shape:     float32[7168,16,2048]
    Logical:   P('mlp', 'layers', 'embed')
    Physical:  (None, None, 'fsdp')

I0418 01:13:41.807146 139892134672192 maxtext_utils.py:1654]  params/params/decoder/layers/post_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0418 01:13:41.807189 139892134672192 maxtext_utils.py:1654]  params/params/decoder/layers/pre_self_attention_layer_norm/scale
    Shape:     float32[2048,16]
    Logical:   P('norm', 'layers')
    Physical:  (None, None)
I0418 01:13:41.807244 139892134672192 maxtext_utils.py:1654]  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)
I0418 01:13:41.807299 139892134672192 maxtext_utils.py:1654]  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')
I0418 01:13:41.807339 139892134672192 maxtext_utils.py:1654]  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)
I0418 01:13:41.807375 139892134672192 maxtext_utils.py:1654]  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)
I0418 01:13:41.807422 139892134672192 maxtext_utils.py:1654]  params/params/decoder/logits_dense/kernel
    Shape:     float32[2048,32000]
    Logical:   P('embed_vocab', 'vocab')
    Physical:  ('fsdp', None)
I0418 01:13:41.807467 139892134672192 maxtext_utils.py:1654]  params/params/token_embedder/embedding
    Shape:     float32[32000,2048]
    Logical:   P('vocab', 'embed_vocab')
    Physical:  (None, 'fsdp')
I0418 01:13:42.293334 139892134672192 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0418 01:13:42.293428 139892134672192 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None).
I0418 01:13:42.308717 139892134672192 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length').
I0418 01:13:42.308780 139892134672192 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None).
I0418 01:13:53.232693 139892134672192 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.
I0418 01:13:53.240236 139892134672192 metric_logger.py:301] number parameters: 1.104 billion
I0418 01:14:04.771299 139892134672192 checkpointing.py:772] Waiting for step 0 to finish before checkpoint...
I0418 01:14:04.924991 139892134672192 checkpointing.py:776] Waited 0.1536717414855957 seconds for step 0 to finish before starting checkpointing.
I0418 01:14:04.927474 139892134672192 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0418 01:14:04.929087 139892134672192 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0
I0418 01:14:04.930530 139892134672192 event_tracking.py:70] [process=0] [sync] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/0.
I0418 01:14:06.350510 139892134672192 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/0
I0418 01:14:09.086816 139892134672192 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776474848679528123, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776474848679528123, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA
I0418 01:14:09.090126 139892134672192 signaling_client.py:364] Using JaxDistributedSignalingClient
I0418 01:14:09.091361 139892134672192 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.
I0418 01:14:09.349027 139764258821888 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/0/items
I0418 01:14:10.888854 139892134672192 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0418 01:14:10.888956 139892134672192 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
I0418 01:14:10.921122    2684 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0418 01:14:11.160161 139892134672192 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.272228s
I0418 01:14:11.167744 139892134672192 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.400 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2856712341308594 s) (per-host)
I0418 01:14:11.167860 139892134672192 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.285794s (batch_requests_ready=0.002447s, total_serialization_initiated=0.275837s, others=0.007510s)
I0418 01:14:11.167967 139892134672192 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.077936s (all_items=0.000034s, per_item={'items': '0.00003386'}, temp_paths=2.077903)
I0418 01:14:11.168017 139892134672192 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.
I0418 01:14:13.622792 139763689076480 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_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/0/items/array_metadatas/process_0
I0418 01:14:48.360344 139763172501248 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 37.192645s (commit=36.006756s, array_metadata_write=1.185889s)
I0418 01:14:48.362352 139892134672192 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 42.145 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 37.480257511138916 s) (per-host)
I0418 01:14:49.351307 139892134672192 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776474848679528123, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA
I0418 01:14:50.080536 139892134672192 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776474848679528123, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA
I0418 01:14:50.082507 139892134672192 event_tracking.py:125] [process=0] [sync] Finished blocking save in 45.15 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/0.
I0418 01:14:50.937278 139892134672192 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389).
I0418 01:14:51.073011 139892134672192 array_metadata_store.py:411] [process=0][thread=MainThread] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000268s.
I0418 01:14:52.027554 139892134672192 base_pytree_checkpoint_handler.py:1406] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.944771s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/0/items
I0418 01:14:52.030458 139892134672192 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/0/items
I0418 01:14:52.968792 139892134672192 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/0
I0418 01:14:54.362540 139892134672192 atomicity.py:847] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/0`.
I0418 01:14:54.365204 139892134672192 event_tracking.py:138] [process=0] [sync] Finished save in 49.44 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/0
I0418 01:14:54.367985 139892134672192 checkpoint_manager.py:2137] [process=0][thread=MainThread][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0418 01:14:54.369628 139892134672192 checkpoint_manager.py:2146] [process=0][thread=MainThread][step=0] CheckpointManager Save Finalize is done on all hosts.
I0418 01:14:54.369678 139892134672192 checkpoint_manager.py:1581] [process=0][thread=MainThread][step=0] Finished synchronous save.
I0418 01:14:54.369729 139892134672192 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776474844.927454, 'wait_for_prev_duration_secs': 6.961822509765625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776474844.9291277, 'checkpointer_blocking_duration_secs': 49.43618607521057, 'get_old_steps_start_time': 1776474894.3653343, 'get_old_steps_duration_secs': 2.4080276489257812e-05, 'checkpoint_manager_blocking_start_time': 1776474844.9254067, 'checkpoint_manager_blocking_duration_secs': 49.44429612159729}
I0418 01:14:54.369817 139892134672192 checkpointing.py:410] Saved a checkpoint at step 0.
I0418 01:14:54.369874 139892134672192 max_utils.py:750] 
Memstats: After params initialized:
I0418 01:14:54.369923 139892134672192 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0))
I0418 01:14:54.369955 139892134672192 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0))
I0418 01:14:54.369980 139892134672192 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0))
I0418 01:14:54.370004 139892134672192 max_utils.py:756] 	Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0))
I0418 01:14:54.708857 139892134672192 metric_logger.py:196] completed step: 0, seconds: 11.245, TFLOP/s/device: 1.208, Tokens/s/device: 182.122, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875
I0418 01:14:54.712148 139892134672192 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/tensorboard/'
I0418 01:14:55.114496 139892134672192 metric_logger.py:196] completed step: 1, seconds: 49.930, TFLOP/s/device: 0.272, Tokens/s/device: 41.017, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875
I0418 01:14:55.260668 139892134672192 metric_logger.py:196] completed step: 2, seconds: 0.412, TFLOP/s/device: 32.982, Tokens/s/device: 4971.369, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719
I0418 01:14:55.408144 139892134672192 metric_logger.py:196] completed step: 3, seconds: 0.008, TFLOP/s/device: 1609.277, Tokens/s/device: 242567.808, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506
I0418 01:14:55.416704 139892134672192 checkpointing.py:772] Waiting for step 5 to finish before checkpoint...
I0418 01:14:55.702507 139892134672192 checkpointing.py:776] Waited 0.2856626510620117 seconds for step 5 to finish before starting checkpointing.
I0418 01:14:55.705373 139892134672192 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0418 01:14:55.707188 139892134672192 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 5
I0418 01:14:55.708866 139892134672192 event_tracking.py:70] [process=0] [sync] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/5.
I0418 01:14:57.098808 139892134672192 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/5
I0418 01:15:00.282851 139892134672192 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.
I0418 01:15:00.539922 139763193472768 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/5/items
I0418 01:15:02.095467 139892134672192 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0418 01:15:02.095613 139892134672192 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
I0418 01:15:02.180233 139892134672192 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.086052s
I0418 01:15:02.190147 139892134672192 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 14.821 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.1040809154510498 s) (per-host)
I0418 01:15:02.190283 139892134672192 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.104236s (batch_requests_ready=0.004537s, total_serialization_initiated=0.089891s, others=0.009808s)
I0418 01:15:02.190454 139892134672192 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.908776s (all_items=0.000041s, per_item={'items': '0.00004101'}, temp_paths=1.908735)
I0418 01:15:02.190562 139892134672192 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.
I0418 01:15:03.948534 139763185080064 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_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/5/items/array_metadatas/process_0
I0418 01:15:44.380953 139763201865472 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.188171s (commit=40.528088s, array_metadata_write=1.660083s)
I0418 01:15:44.382393 139892134672192 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 37.346 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.296300649642944 s) (per-host)
I0418 01:15:45.924729 139892134672192 event_tracking.py:125] [process=0] [sync] Finished blocking save in 50.22 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/5.
I0418 01:15:46.899038 139892134672192 array_metadata_store.py:411] [process=0][thread=MainThread] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000458s.
I0418 01:15:46.899802 139892134672192 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389).
I0418 01:15:47.849817 139892134672192 base_pytree_checkpoint_handler.py:1406] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.924608s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/5/items
I0418 01:15:47.851825 139892134672192 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/5/items
I0418 01:15:48.352930 139892134672192 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/5
I0418 01:15:50.589543 139892134672192 atomicity.py:847] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/5`.
I0418 01:15:50.592912 139892134672192 event_tracking.py:138] [process=0] [sync] Finished save in 54.89 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/5
I0418 01:15:50.595698 139892134672192 checkpoint_manager.py:2137] [process=0][thread=MainThread][step=5] CheckpointManager Save Finalize is syncing with other hosts...
I0418 01:15:50.597509 139892134672192 checkpoint_manager.py:2146] [process=0][thread=MainThread][step=5] CheckpointManager Save Finalize is done on all hosts.
I0418 01:15:50.597591 139892134672192 checkpoint_manager.py:1581] [process=0][thread=MainThread][step=5] Finished synchronous save.
I0418 01:15:50.597690 139892134672192 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776474895.7053406, 'wait_for_prev_duration_secs': 0.00010156631469726562, 'time_between_consecutive_saves_sec': 1.3356738090515137, 'checkpointer_blocking_start_time': 1776474895.707253, 'checkpointer_blocking_duration_secs': 54.88581728935242, 'get_old_steps_start_time': 1776474950.593093, 'get_old_steps_duration_secs': 4.172325134277344e-05, 'checkpoint_manager_blocking_start_time': 1776474895.7031553, 'checkpoint_manager_blocking_duration_secs': 54.89448952674866}
I0418 01:15:50.597832 139892134672192 checkpointing.py:410] Saved a checkpoint at step 5.
I0418 01:16:14.302516 139892134672192 metric_logger.py:196] completed step: 4, seconds: 0.147, TFLOP/s/device: 92.262, Tokens/s/device: 13906.714, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355
I0418 01:16:14.312000 139892134672192 metric_logger.py:196] completed step: 5, seconds: 0.148, TFLOP/s/device: 92.109, Tokens/s/device: 13883.710, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467
I0418 01:16:14.457410 139892134672192 metric_logger.py:196] completed step: 6, seconds: 78.894, TFLOP/s/device: 0.172, Tokens/s/device: 25.959, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740
I0418 01:16:14.605097 139892134672192 metric_logger.py:196] completed step: 7, seconds: 0.008, TFLOP/s/device: 1626.227, Tokens/s/device: 245122.681, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151
I0418 01:16:14.753212 139892134672192 metric_logger.py:196] completed step: 8, seconds: 0.145, TFLOP/s/device: 93.643, Tokens/s/device: 14114.890, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098
I0418 01:16:14.899536 139892134672192 checkpointing.py:772] Waiting for step 9 to finish before checkpoint...
I0418 01:16:14.900252 139892134672192 checkpointing.py:776] Waited 0.0007381439208984375 seconds for step 9 to finish before starting checkpointing.
I0418 01:16:14.902334 139892134672192 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0418 01:16:14.904499 139892134672192 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9
I0418 01:16:14.905972 139892134672192 event_tracking.py:70] [process=0] [sync] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/9.
I0418 01:16:15.912465 139892134672192 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/9
I0418 01:16:18.294400 139892134672192 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.
I0418 01:16:18.978032 139764807988992 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/9/items
I0418 01:16:20.967822 139892134672192 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array.
I0418 01:16:20.967938 139892134672192 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
I0418 01:16:21.011972 139892134672192 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.045079s
I0418 01:16:21.017775 139892134672192 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 28.117 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.05486297607421875 s) (per-host)
I0418 01:16:21.017867 139892134672192 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.054976s (batch_requests_ready=0.001913s, total_serialization_initiated=0.047303s, others=0.005759s)
I0418 01:16:21.017968 139892134672192 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.724605s (all_items=0.000030s, per_item={'items': '0.00002980'}, temp_paths=2.724575)
I0418 01:16:21.018013 139892134672192 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.
I0418 01:16:23.646548 139764791203584 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_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/9/items/array_metadatas/process_0
I0418 01:17:01.867288 139764267214592 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 40.849550s (commit=38.693199s, array_metadata_write=2.156350s)
I0418 01:17:01.868366 139892134672192 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 38.616 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 40.90542769432068 s) (per-host)
I0418 01:17:03.288582 139892134672192 event_tracking.py:125] [process=0] [sync] Finished blocking save in 48.38 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/9.
I0418 01:17:04.114312 139892134672192 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389).
I0418 01:17:04.272084 139892134672192 array_metadata_store.py:411] [process=0][thread=MainThread] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000268s.
I0418 01:17:05.207273 139892134672192 base_pytree_checkpoint_handler.py:1406] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.918355s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/9/items
I0418 01:17:05.208897 139892134672192 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/9/items
I0418 01:17:05.710458 139892134672192 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/9
I0418 01:17:07.114685 139892134672192 atomicity.py:847] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/9`.
I0418 01:17:07.117402 139892134672192 event_tracking.py:138] [process=0] [sync] Finished save in 52.21 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints/9
I0418 01:17:07.120365 139892134672192 checkpoint_manager.py:2137] [process=0][thread=MainThread][step=9] CheckpointManager Save Finalize is syncing with other hosts...
I0418 01:17:07.121732 139892134672192 checkpoint_manager.py:2146] [process=0][thread=MainThread][step=9] CheckpointManager Save Finalize is done on all hosts.
I0418 01:17:07.121780 139892134672192 checkpoint_manager.py:1581] [process=0][thread=MainThread][step=9] Finished synchronous save.
I0418 01:17:07.121830 139892134672192 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776474974.902315, 'wait_for_prev_duration_secs': 5.936622619628906e-05, 'time_between_consecutive_saves_sec': 24.304738759994507, 'checkpointer_blocking_start_time': 1776474974.9045374, 'checkpointer_blocking_duration_secs': 52.21296834945679, 'get_old_steps_start_time': 1776475027.1175206, 'get_old_steps_duration_secs': 2.4318695068359375e-05, 'checkpoint_manager_blocking_start_time': 1776474974.9004776, 'checkpoint_manager_blocking_duration_secs': 52.221328020095825}
I0418 01:17:07.121936 139892134672192 checkpointing.py:410] Saved a checkpoint at step 9.
I0418 01:17:07.121969 139892134672192 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0418 01:17:07.122000 139892134672192 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0418 01:17:07.122685 139892134672192 metric_logger.py:196] completed step: 9, seconds: 0.148, TFLOP/s/device: 91.925, Tokens/s/device: 13855.907, 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: Sat Apr 18 01:17:18 UTC 2026
EXIT_CODE=0
NNX  ·  1b5e4eb5a  ·  main_20260417_235819  ·  full log
XPK Start: Sat Apr 18 02:34:06 UTC 2026
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
2026-04-18 02:34:30.062110: 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)
I0418 02:34:30.601386 135160607303488 max_utils.py:273] Attempting to initialize the jax distributed system...
I0418 02:34:39.643109 135160607303488 distributed.py:149] Starting JAX distributed service on [::]:8482
I0418 02:34:39.645497 135160607303488 distributed.py:172] Connecting to JAX distributed service on mt-14-async-ckpt-false--95c0b-slice-job-0-0.mt-14-async-ckpt-false--95c0b:8482
I0418 02:34:40.764426 135160607303488 max_utils.py:284] Jax distributed system initialized!
I0418 02:34:46.811644 135160607303488 max_utils.py:800] System Information: Jax Version: 0.9.2
I0418 02:34:46.811747 135160607303488 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0418 02:34:46.811787 135160607303488 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
I0418 02:34:46.811820 135160607303488 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing.
Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "/deps/src/maxtext/trainers/pre_train/train.py", line 744, in <module>
    app.run(main)
  File "/usr/local/lib/python3.12/site-packages/absl/app.py", line 367, in run
    _run_main(main, args)
  File "/usr/local/lib/python3.12/site-packages/absl/app.py", line 312, in _run_main
    sys.exit(main(argv))
             ^^^^^^^^^^
  File "/deps/src/maxtext/trainers/pre_train/train.py", line 740, in main
    train_func()
  File "/deps/src/maxtext/trainers/pre_train/train.py", line 730, in train_func
    run(config, recorder, diagnostic_config)
  File "/deps/src/maxtext/trainers/pre_train/train.py", line 709, in run
    train_loop(config, recorder)
  File "/deps/src/maxtext/trainers/pre_train/train.py", line 536, in train_loop
    ) = train_utils.setup_train_loop(config, recorder)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/deps/src/maxtext/utils/train_utils.py", line 217, in setup_train_loop
    raise NotImplementedError("Pure NNX support has not been implemented yet.")
NotImplementedError: Pure NNX support has not been implemented yet.
XPK End: Sat Apr 18 02:34:57 UTC 2026
EXIT_CODE=1