main (574ad3fb9) vs feat/nnx-post-train-fixes (574ad3fb9)| Metric | main 574ad3fb9 | feat/nnx-post-train-fixes 574ad3fb9 | Diff (feat/nnx-post-train-fixes − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 7.1810 | 7.1810 | 0 |
| TFLOP/s | 92.149 | 92.203 | +0.054 |
| Tok/s | 13889.6 | 13897.8 | +8.201 |
| Avg s/step | 4.754 | 4.751 | -0.003 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.9.2 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Sat Apr 18 19:33:21 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 19:33:45.955889: 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 19:33:46.136519 140126308788032 max_utils.py:273] Attempting to initialize the jax distributed system... I0418 19:33:55.175676 140126308788032 distributed.py:149] Starting JAX distributed service on [::]:8482 I0418 19:33:55.177983 140126308788032 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-false-sv-r4zd1-slice-job-0-0.mt-15-ocdbt-false-sv-r4zd1:8482 I0418 19:33:56.079634 140126308788032 max_utils.py:284] Jax distributed system initialized! I0418 19:34:02.536898 140126308788032 max_utils.py:800] System Information: Jax Version: 0.9.2 I0418 19:34:02.537000 140126308788032 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0418 19:34:02.537040 140126308788032 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 19:34:02.537084 140126308788032 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0418 19:34:03.237103 140126308788032 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0418 19:34:03.237379 140126308788032 checkpointing.py:677] Setting up checkpoint logger... I0418 19:34:03.237432 140126308788032 checkpointing.py:233] Creating checkpoint manager with ocdbt=False and zarr3=True I0418 19:34:03.237479 140126308788032 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0418 19:34:03.237810 140126308788032 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=False, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7f70fb244080>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0418 19:34:06.565551 140126308788032 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0418 19:34:06.565785 140126308788032 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 0x7f5c8811aea0>}, handler_registry=None I0418 19:34:06.566030 140126308788032 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f5c8811aea0>` 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 19:34:06.566097 140126308788032 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f5c8811c800>` 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 19:34:06.566146 140126308788032 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 0x7f5c8811aea0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f5c8811aea0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f5c8811c800>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f5c8811c800>}). I0418 19:34:06.566547 140126308788032 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0418 19:34:06.566619 140126308788032 async_checkpointer.py:192] [process=7][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f5bb8463d80> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0418 19:34:08.100911 140126308788032 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_false/checkpoints I0418 19:34:08.540988 140126308788032 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_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f5c8811a450> I0418 19:34:08.541186 140126308788032 checkpointing.py:301] Checkpoint manager created! I0418 19:34:09.492749 140126308788032 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0418 19:34:09.492861 140126308788032 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:34:09.874751 140126308788032 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 19:34:09.874843 140126308788032 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:34:09.890986 140126308788032 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 19:34:09.891045 140126308788032 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:34:09.914574 140126308788032 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 19:34:09.914643 140126308788032 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:34:09.931006 140126308788032 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 19:34:09.931066 140126308788032 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:34:09.947210 140126308788032 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 19:34:09.947269 140126308788032 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:34:09.971763 140126308788032 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0418 19:34:09.971842 140126308788032 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:34:09.993645 140126308788032 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0418 19:34:09.993708 140126308788032 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0418 19:34:10.209131 140126308788032 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0418 19:34:10.209238 140126308788032 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0418 19:34:11.647469 140126308788032 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0418 19:34:11.647593 140126308788032 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 19:34:11.647647 140126308788032 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 19:34:11.647717 140126308788032 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 19:34:11.647772 140126308788032 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 19:34:11.647811 140126308788032 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 19:34:11.647863 140126308788032 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 19:34:11.647914 140126308788032 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 19:34:11.647955 140126308788032 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 19:34:11.647992 140126308788032 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 19:34:11.648039 140126308788032 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0418 19:34:11.648102 140126308788032 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0418 19:34:12.143023 140126308788032 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 19:34:12.143135 140126308788032 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0418 19:34:12.158481 140126308788032 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 19:34:12.158541 140126308788032 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0418 19:34:22.990097 140126308788032 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 19:34:22.990978 140126308788032 metric_logger.py:301] number parameters: 1.104 billion I0418 19:34:34.414645 140126308788032 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0418 19:34:34.568638 140126308788032 checkpointing.py:776] Waited 0.15397024154663086 seconds for step 0 to finish before starting checkpointing. I0418 19:34:34.571367 140126308788032 checkpoint_manager.py:2009] [process=7][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 19:34:34.573066 140126308788032 checkpoint_manager.py:1512] [process=7] Saving checkpoint at step 0 I0418 19:34:34.574655 140126308788032 event_tracking.py:70] [process=7] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_false/checkpoints/0. I0418 19:34:35.765806 140126308788032 signaling_client.py:364] Using JaxDistributedSignalingClient I0418 19:34:35.766843 140126308788032 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 19:34:35.766906 140126308788032 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 19:34:36.038364 140126308788032 base_pytree_checkpoint_handler.py:154] [process=7][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.272634s I0418 19:34:36.038545 140126308788032 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/blocking_gbytes_per_sec: 5.546 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2781527042388916 s) (per-host) I0418 19:34:36.038595 140126308788032 base_pytree_checkpoint_handler.py:768] [process=7][thread=MainThread] Initiated Pytree async_save. Time taken: 0.278212s (batch_requests_ready=0.002219s, total_serialization_initiated=0.275925s, others=0.000068s) I0418 19:34:36.038686 140126308788032 composite_checkpoint_handler.py:715] [process=7][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.282803s (all_items=0.000017s, per_item={'items': '0.00001740'}, temp_paths=0.282785) I0418 19:34:36.039513 140126308788032 event_tracking.py:125] [process=7] [async] Finished blocking save in 1.47 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_false/checkpoints/0. I0418 19:34:36.039838 139996118398720 async_checkpointer.py:76] [process=7][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 19:54:36.039799 I0418 19:34:36.051358 140126308788032 checkpoint_manager.py:1560] [process=7][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0418 19:34:36.051639 139993949525760 async_checkpointer.py:280] [process=7][thread=save_finalize] Waiting for background save thread=async_save. I0418 19:34:36.051794 140126308788032 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776540874.5713487, 'wait_for_prev_duration_secs': 6.270408630371094e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776540874.573115, 'checkpointer_blocking_duration_secs': 1.466883659362793, 'get_old_steps_start_time': 1776540876.0400257, 'get_old_steps_duration_secs': 3.5762786865234375e-05, 'checkpoint_manager_blocking_start_time': 1776540874.5692194, 'checkpoint_manager_blocking_duration_secs': 1.4825351238250732} I0418 19:34:36.051903 140126308788032 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0418 19:34:36.051952 140126308788032 max_utils.py:750] Memstats: After params initialized: I0418 19:34:36.052002 140126308788032 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_26(process=7,(2,6,0,0)) I0418 19:34:36.052035 140126308788032 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_27(process=7,(3,6,0,0)) I0418 19:34:36.052067 140126308788032 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_30(process=7,(2,7,0,0)) I0418 19:34:36.052107 140126308788032 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_31(process=7,(3,7,0,0)) I0418 19:34:36.367942 140126308788032 metric_logger.py:196] completed step: 0, seconds: 11.424, TFLOP/s/device: 1.189, Tokens/s/device: 179.279, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0418 19:34:36.539875 140126308788032 metric_logger.py:196] completed step: 1, seconds: 1.952, TFLOP/s/device: 6.961, Tokens/s/device: 1049.306, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0418 19:34:36.944589 140126308788032 metric_logger.py:196] completed step: 2, seconds: 0.025, TFLOP/s/device: 548.532, Tokens/s/device: 82680.662, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0418 19:34:37.092037 140126308788032 metric_logger.py:196] completed step: 3, seconds: 0.405, TFLOP/s/device: 33.565, Tokens/s/device: 5059.301, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0418 19:34:37.096585 140126308788032 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0418 19:34:37.386711 140126308788032 checkpointing.py:776] Waited 0.29009366035461426 seconds for step 5 to finish before starting checkpointing. I0418 19:34:37.389573 140126308788032 checkpoint_manager.py:2020] [process=7][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 19:34:39.021648 2938 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0418 19:34:42.385246 139993987340032 array_metadata_store.py:203] [process=7][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_false/checkpoints/0/items/array_metadatas/process_7 I0418 19:34:46.018138 139996118398720 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/gbytes_per_sec: 153.991 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.257695198059082 s) (per-host) I0418 19:34:46.018249 139996118398720 async_checkpointer.py:90] [process=7][thread=async_save] 3 Handler Commit operations completed. Time taken: 9.978296s. I0418 19:34:53.237900 139996118398720 async_checkpointer.py:160] [process=7][thread=async_save] Background save thread done. Time taken: 17.197931s. I0418 19:34:53.238123 139993949525760 async_checkpointer.py:288] [process=7][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 19:34:53.238185 139993949525760 async_checkpointer.py:298] [process=7][thread=save_finalize] No errors found in background save thread=async_save. I0418 19:34:53.238252 139993949525760 checkpoint_manager.py:2137] [process=7][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0418 19:34:53.239844 139993949525760 checkpoint_manager.py:2146] [process=7][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0418 19:34:53.240010 140126308788032 checkpoint_manager.py:2032] [process=7][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0418 19:34:53.240160 140126308788032 checkpoint_manager.py:1452] Waiting for previous save to complete took 15.850587 seconds. If this number is high, consider checkpointing less frequently. I0418 19:34:53.242214 140126308788032 checkpoint_manager.py:1512] [process=7] Saving checkpoint at step 5 I0418 19:34:53.244679 140126308788032 event_tracking.py:70] [process=7] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_false/checkpoints/5. I0418 19:34:54.383461 140126308788032 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 19:34:54.383554 140126308788032 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 19:34:54.418418 140126308788032 base_pytree_checkpoint_handler.py:154] [process=7][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.035967s I0418 19:34:54.418588 140126308788032 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/blocking_gbytes_per_sec: 39.094 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.039458513259887695 s) (per-host) I0418 19:34:54.418641 140126308788032 base_pytree_checkpoint_handler.py:768] [process=7][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039520s (batch_requests_ready=0.001745s, total_serialization_initiated=0.037708s, others=0.000068s) I0418 19:34:54.418730 140126308788032 composite_checkpoint_handler.py:715] [process=7][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.044526s (all_items=0.000015s, per_item={'items': '0.00001550'}, temp_paths=0.044511) I0418 19:34:54.419397 140126308788032 event_tracking.py:125] [process=7] [async] Finished blocking save in 1.18 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_false/checkpoints/5. I0418 19:34:54.419757 139996101613312 async_checkpointer.py:76] [process=7][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 19:54:54.419716 I0418 19:34:54.428587 140126308788032 checkpoint_manager.py:1560] [process=7][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0418 19:34:54.428892 139993949525760 async_checkpointer.py:280] [process=7][thread=save_finalize] Waiting for background save thread=async_save. I0418 19:34:54.429052 140126308788032 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776540877.3895442, 'wait_for_prev_duration_secs': 15.850587368011475, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776540893.2422533, 'checkpointer_blocking_duration_secs': 1.1776604652404785, 'get_old_steps_start_time': 1776540894.4199364, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776540877.3873465, 'checkpoint_manager_blocking_duration_secs': 17.041672229766846} I0418 19:34:54.429180 140126308788032 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0418 19:34:54.430090 140126308788032 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.705, Tokens/s/device: 13370.590, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0418 19:34:54.436083 140126308788032 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.227, Tokens/s/device: 13901.522, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0418 19:34:59.831129 139993987340032 array_metadata_store.py:203] [process=7][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_false/checkpoints/5/items/array_metadatas/process_7 I0418 19:35:05.138834 139996101613312 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/gbytes_per_sec: 146.807 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.759648323059082 s) (per-host) I0418 19:35:05.138949 139996101613312 async_checkpointer.py:90] [process=7][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.719071s. I0418 19:35:14.977532 139996101613312 async_checkpointer.py:160] [process=7][thread=async_save] Background save thread done. Time taken: 20.557637s. I0418 19:35:14.977770 139993949525760 async_checkpointer.py:288] [process=7][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 19:35:14.977852 139993949525760 async_checkpointer.py:298] [process=7][thread=save_finalize] No errors found in background save thread=async_save. I0418 19:35:14.977901 139993949525760 checkpoint_manager.py:2137] [process=7][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0418 19:35:14.979835 139993949525760 checkpoint_manager.py:2146] [process=7][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0418 19:35:17.051611 140126308788032 metric_logger.py:196] completed step: 6, seconds: 17.338, TFLOP/s/device: 0.784, Tokens/s/device: 118.122, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0418 19:35:17.199275 140126308788032 metric_logger.py:196] completed step: 7, seconds: 22.469, TFLOP/s/device: 0.605, Tokens/s/device: 91.146, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0418 19:35:17.346816 140126308788032 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.133, Tokens/s/device: 13435.147, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0418 19:35:17.493656 140126308788032 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0418 19:35:17.494357 140126308788032 checkpointing.py:776] Waited 0.0007228851318359375 seconds for step 9 to finish before starting checkpointing. I0418 19:35:17.496384 140126308788032 checkpoint_manager.py:2009] [process=7][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 19:35:17.497919 140126308788032 checkpoint_manager.py:1512] [process=7] Saving checkpoint at step 9 I0418 19:35:17.499397 140126308788032 event_tracking.py:70] [process=7] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_false/checkpoints/9. I0418 19:35:18.193331 140126308788032 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 19:35:18.193425 140126308788032 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 19:35:18.228974 140126308788032 base_pytree_checkpoint_handler.py:154] [process=7][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036634s I0418 19:35:18.229150 140126308788032 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/blocking_gbytes_per_sec: 38.398 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04017353057861328 s) (per-host) I0418 19:35:18.229206 140126308788032 base_pytree_checkpoint_handler.py:768] [process=7][thread=MainThread] Initiated Pytree async_save. Time taken: 0.040240s (batch_requests_ready=0.001666s, total_serialization_initiated=0.038502s, others=0.000072s) I0418 19:35:18.229292 140126308788032 composite_checkpoint_handler.py:715] [process=7][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.044581s (all_items=0.000010s, per_item={'items': '0.00001025'}, temp_paths=0.044571) I0418 19:35:18.230005 140126308788032 event_tracking.py:125] [process=7] [async] Finished blocking save in 0.73 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_false/checkpoints/9. I0418 19:35:18.230312 139993949525760 async_checkpointer.py:76] [process=7][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 19:55:18.230279 I0418 19:35:18.232301 140126308788032 checkpoint_manager.py:1560] [process=7][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0418 19:35:18.232495 139975102293760 async_checkpointer.py:280] [process=7][thread=save_finalize] Waiting for background save thread=async_save. I0418 19:35:18.232603 140126308788032 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776540917.4963527, 'wait_for_prev_duration_secs': 7.462501525878906e-05, 'time_between_consecutive_saves_sec': 2.5164783000946045, 'checkpointer_blocking_start_time': 1776540917.4979563, 'checkpointer_blocking_duration_secs': 0.7324936389923096, 'get_old_steps_start_time': 1776540918.230468, 'get_old_steps_duration_secs': 2.4080276489257812e-05, 'checkpoint_manager_blocking_start_time': 1776540917.4945877, 'checkpoint_manager_blocking_duration_secs': 0.737983226776123} I0418 19:35:18.232717 140126308788032 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0418 19:35:18.232762 140126308788032 checkpoint_manager.py:2020] [process=7][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 19:35:25.297228 139995021027072 array_metadata_store.py:203] [process=7][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_15_ocdbt_false/checkpoints/9/items/array_metadatas/process_7 I0418 19:35:30.172160 139993949525760 base_pytree_checkpoint_handler.py:130] [process=7] /jax/orbax/write/gbytes_per_sec: 131.819 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 11.983110189437866 s) (per-host) I0418 19:35:30.172283 139993949525760 async_checkpointer.py:90] [process=7][thread=async_save] 3 Handler Commit operations completed. Time taken: 11.941862s. I0418 19:35:36.751338 139993949525760 async_checkpointer.py:160] [process=7][thread=async_save] Background save thread done. Time taken: 18.520900s. I0418 19:35:36.751614 139975102293760 async_checkpointer.py:288] [process=7][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 19:35:36.751740 139975102293760 async_checkpointer.py:298] [process=7][thread=save_finalize] No errors found in background save thread=async_save. I0418 19:35:36.751783 139975102293760 checkpoint_manager.py:2137] [process=7][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0418 19:35:36.753197 139975102293760 checkpoint_manager.py:2146] [process=7][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0418 19:35:36.753380 140126308788032 checkpoint_manager.py:2032] [process=7][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0418 19:35:36.753524 140126308788032 checkpoint_manager.py:2009] [process=7][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 19:35:36.754500 140126308788032 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.149, Tokens/s/device: 13889.642, 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 19:35:46 UTC 2026 EXIT_CODE=0
XPK Start: Sun Apr 19 00:57:35 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-19 00:57:59.679580: 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) I0419 00:57:59.857407 134949517694784 max_utils.py:273] Attempting to initialize the jax distributed system... I0419 00:58:08.901949 134949517694784 distributed.py:149] Starting JAX distributed service on [::]:8482 I0419 00:58:08.904222 134949517694784 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-false-sv-2sp0g-slice-job-0-0.mt-15-ocdbt-false-sv-2sp0g:8482 I0419 00:58:10.731034 134949517694784 max_utils.py:284] Jax distributed system initialized! I0419 00:58:16.612323 134949517694784 max_utils.py:800] System Information: Jax Version: 0.9.2 I0419 00:58:16.612426 134949517694784 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0419 00:58:16.612467 134949517694784 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 I0419 00:58:16.612500 134949517694784 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0419 00:58:17.663208 134949517694784 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 00:58:17.663803 134949517694784 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 00:58:17.663997 134949517694784 checkpointing.py:688] Setting up checkpoint logger... I0419 00:58:17.664048 134949517694784 checkpointing.py:234] Creating checkpoint manager with ocdbt=False and zarr3=True I0419 00:58:17.664092 134949517694784 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0419 00:58:17.664433 134949517694784 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=False, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7abbaa44f770>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0419 00:58:20.061737 134949517694784 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0419 00:58:20.062000 134949517694784 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 0x7aba9e101d60>}, handler_registry=None I0419 00:58:20.062306 134949517694784 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7aba9e101d60>` 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`. I0419 00:58:20.062375 134949517694784 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7aba9e0e7d40>` 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`. I0419 00:58:20.062427 134949517694784 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 0x7aba9e101d60>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7aba9e101d60>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7aba9e0e7d40>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7aba9e0e7d40>}). I0419 00:58:20.062779 134949517694784 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0419 00:58:20.062876 134949517694784 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7aba9e2a37e0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0419 00:58:22.208914 134949517694784 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints I0419 00:58:22.669331 134949517694784 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=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_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7aba9e1014c0> I0419 00:58:22.669509 134949517694784 checkpointing.py:302] Checkpoint manager created! I0419 00:58:23.591548 134949517694784 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 00:58:23.591652 134949517694784 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 00:58:23.968352 134949517694784 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 00:58:23.968439 134949517694784 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 00:58:23.984923 134949517694784 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 00:58:23.984985 134949517694784 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 00:58:24.008637 134949517694784 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 00:58:24.008702 134949517694784 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 00:58:24.025252 134949517694784 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 00:58:24.025315 134949517694784 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 00:58:24.041784 134949517694784 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 00:58:24.041858 134949517694784 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 00:58:24.066291 134949517694784 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0419 00:58:24.066365 134949517694784 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 00:58:24.087789 134949517694784 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0419 00:58:24.087868 134949517694784 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0419 00:58:24.305244 134949517694784 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0419 00:58:24.305350 134949517694784 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0419 00:58:25.738404 134949517694784 maxtext_utils.py:1827] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0419 00:58:25.738530 134949517694784 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) I0419 00:58:25.738584 134949517694784 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) I0419 00:58:25.738641 134949517694784 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') I0419 00:58:25.738694 134949517694784 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) I0419 00:58:25.738733 134949517694784 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) I0419 00:58:25.738790 134949517694784 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) I0419 00:58:25.738861 134949517694784 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') I0419 00:58:25.738907 134949517694784 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) I0419 00:58:25.738945 134949517694784 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) I0419 00:58:25.738995 134949517694784 maxtext_utils.py:1827] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0419 00:58:25.739044 134949517694784 maxtext_utils.py:1827] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0419 00:58:26.234442 134949517694784 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 00:58:26.234534 134949517694784 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0419 00:58:26.250020 134949517694784 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 00:58:26.250084 134949517694784 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0419 00:58:37.138800 134949517694784 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. I0419 00:58:37.146800 134949517694784 metric_logger.py:301] number parameters: 1.104 billion I0419 00:58:48.716819 134949517694784 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0419 00:58:48.870590 134949517694784 checkpointing.py:798] Waited 0.15376687049865723 seconds for step 0 to finish before starting checkpointing. I0419 00:58:48.872973 134949517694784 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 00:58:48.874798 134949517694784 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0419 00:58:48.876203 134949517694784 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/0. I0419 00:58:49.614762 134949517694784 signaling_client.py:364] Using JaxDistributedSignalingClient I0419 00:58:49.629659 134949517694784 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 00:58:49.629727 134949517694784 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 I0419 00:58:49.892871 134824358557440 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/0 I0419 00:58:49.901582 134949517694784 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.272728s I0419 00:58:49.902449 134949517694784 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.526 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2791295051574707 s) (per-host) I0419 00:58:49.902529 134949517694784 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.279233s (batch_requests_ready=0.002144s, total_serialization_initiated=0.276300s, others=0.000790s) I0419 00:58:49.902614 134949517694784 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.283600s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.283584) I0419 00:58:49.903400 134949517694784 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.03 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/0. I0419 00:58:49.903722 134824366950144 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 01:18:49.903686 I0419 00:58:49.919328 134949517694784 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0419 00:58:49.919595 134822773122816 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 00:58:49.919728 134949517694784 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776560328.8729548, 'wait_for_prev_duration_secs': 6.079673767089844e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776560328.8748507, 'checkpointer_blocking_duration_secs': 1.0290193557739258, 'get_old_steps_start_time': 1776560329.9038942, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776560328.8710012, 'checkpoint_manager_blocking_duration_secs': 1.0486855506896973} I0419 00:58:49.919845 134949517694784 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0419 00:58:49.919896 134949517694784 max_utils.py:750] Memstats: After params initialized: I0419 00:58:49.919946 134949517694784 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0419 00:58:49.919977 134949517694784 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0419 00:58:49.920003 134949517694784 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0419 00:58:49.920026 134949517694784 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0419 00:58:50.261734 134949517694784 metric_logger.py:196] completed step: 0, seconds: 11.274, TFLOP/s/device: 1.205, Tokens/s/device: 181.655, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0419 00:58:50.265224 134949517694784 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/tensorboard/' I0419 00:58:50.655513 134949517694784 metric_logger.py:196] completed step: 1, seconds: 1.543, TFLOP/s/device: 8.806, Tokens/s/device: 1327.275, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0419 00:58:50.801574 134949517694784 metric_logger.py:196] completed step: 2, seconds: 0.395, TFLOP/s/device: 34.438, Tokens/s/device: 5190.868, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0419 00:58:50.948798 134949517694784 metric_logger.py:196] completed step: 3, seconds: 0.009, TFLOP/s/device: 1438.857, Tokens/s/device: 216880.229, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0419 00:58:50.957973 134949517694784 checkpointing.py:794] Waiting for step 5 to finish before checkpoint... I0419 00:58:51.242945 134949517694784 checkpointing.py:798] Waited 0.2848086357116699 seconds for step 5 to finish before starting checkpointing. I0419 00:58:51.246031 134949517694784 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 00:58:52.160751 134824341772032 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/0/items I0419 00:58:52.295835 134821664995072 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776560331906638142, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776560331906638142, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/0/_CHECKPOINT_METADATA I0419 00:58:53.271539 2786 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0419 00:58:55.470389 134823263901440 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/0/items/array_metadatas/process_0 I0419 00:59:03.221271 134822781515520 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 9.958386s (commit=8.370821s, array_metadata_write=1.587565s) I0419 00:59:03.222720 134824366950144 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 116.152 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 13.599363327026367 s) (per-host) I0419 00:59:03.222901 134824366950144 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 13.319063s. I0419 00:59:03.875118 134824366950144 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776560331906638142, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/0/_CHECKPOINT_METADATA I0419 00:59:04.570631 134821664995072 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776560331906638142, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/0/_CHECKPOINT_METADATA W0419 00:59:04.623563 134824366950144 ocdbt_utils.py:168] [process=0][thread=async_save] Skipping merge of OCDBT checkpoints: No per-process OCDBT checkpoint subdirs found in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/0/items, I0419 00:59:05.048574 134824366950144 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000394s. I0419 00:59:05.048837 134824366950144 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.050305s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/0/items I0419 00:59:05.050675 134824366950144 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/0/items I0419 00:59:05.546473 134824366950144 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/0 I0419 00:59:07.899292 134824366950144 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/0`. I0419 00:59:07.900476 134824366950144 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 19.03 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/0 I0419 00:59:07.902109 134824366950144 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 17.998269s. I0419 00:59:07.902403 134822773122816 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 00:59:07.902575 134822773122816 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 00:59:07.902682 134822773122816 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0419 00:59:07.904331 134822773122816 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0419 00:59:07.904637 134949517694784 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0419 00:59:07.904800 134949517694784 checkpoint_manager.py:1452] Waiting for previous save to complete took 16.658787 seconds. If this number is high, consider checkpointing less frequently. I0419 00:59:07.906577 134949517694784 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 5 I0419 00:59:07.909051 134949517694784 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/5. I0419 00:59:08.621691 134949517694784 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 00:59:08.621842 134949517694784 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 I0419 00:59:08.658678 134949517694784 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.038014s I0419 00:59:08.659704 134949517694784 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 32.860 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04694318771362305 s) (per-host) I0419 00:59:08.659820 134949517694784 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.047070s (batch_requests_ready=0.004336s, total_serialization_initiated=0.041826s, others=0.000908s) I0419 00:59:08.659951 134949517694784 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.052700s (all_items=0.000031s, per_item={'items': '0.00003123'}, temp_paths=0.052669) I0419 00:59:08.660930 134949517694784 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/5. I0419 00:59:08.661334 134822773122816 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 01:19:08.661297 I0419 00:59:08.663786 134949517694784 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0419 00:59:08.664137 134821170902784 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 00:59:08.664327 134949517694784 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776560331.2459831, 'wait_for_prev_duration_secs': 16.65878677368164, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776560347.9066486, 'checkpointer_blocking_duration_secs': 0.7548830509185791, 'get_old_steps_start_time': 1776560348.661559, 'get_old_steps_duration_secs': 4.0531158447265625e-05, 'checkpoint_manager_blocking_start_time': 1776560331.2436998, 'checkpoint_manager_blocking_duration_secs': 17.420575618743896} I0419 00:59:08.664472 134949517694784 checkpointing.py:409] Started an asynchronous checkpoint save for step 5 I0419 00:59:08.856075 134824366950144 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/5 I0419 00:59:10.592877 134823263901440 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/5/items I0419 00:59:14.942654 134822756337408 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/5/items/array_metadatas/process_0 I0419 00:59:22.150303 134822739552000 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 10.049508s (commit=8.525941s, array_metadata_write=1.523566s) I0419 00:59:22.152345 134822773122816 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 116.665 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 13.539545774459839 s) (per-host) I0419 00:59:22.152472 134822773122816 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 13.490981s. W0419 00:59:23.548065 134822773122816 ocdbt_utils.py:168] [process=0][thread=async_save] Skipping merge of OCDBT checkpoints: No per-process OCDBT checkpoint subdirs found in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/5/items, I0419 00:59:24.827802 134822773122816 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000214s. I0419 00:59:24.827996 134822773122816 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.910502s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/5/items I0419 00:59:24.829467 134822773122816 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/5/items I0419 00:59:25.331514 134822773122816 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/5 I0419 00:59:26.740560 134822773122816 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/5`. I0419 00:59:26.741390 134822773122816 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 18.83 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/5 I0419 00:59:26.742925 134822773122816 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 18.081411s. I0419 00:59:26.743126 134821170902784 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 00:59:26.743258 134821170902784 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 00:59:26.743329 134821170902784 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0419 00:59:26.745120 134821170902784 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0419 00:59:31.169474 134949517694784 metric_logger.py:196] completed step: 4, seconds: 0.148, TFLOP/s/device: 91.684, Tokens/s/device: 13819.536, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0419 00:59:31.178945 134949517694784 metric_logger.py:196] completed step: 5, seconds: 0.146, TFLOP/s/device: 93.173, Tokens/s/device: 14044.038, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0419 00:59:31.324583 134949517694784 metric_logger.py:196] completed step: 6, seconds: 40.219, TFLOP/s/device: 0.338, Tokens/s/device: 50.921, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0419 00:59:31.472019 134949517694784 metric_logger.py:196] completed step: 7, seconds: 0.007, TFLOP/s/device: 1856.672, Tokens/s/device: 279857.885, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0419 00:59:31.619611 134949517694784 metric_logger.py:196] completed step: 8, seconds: 0.147, TFLOP/s/device: 92.668, Tokens/s/device: 13967.985, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0419 00:59:31.766340 134949517694784 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0419 00:59:31.767456 134949517694784 checkpointing.py:798] Waited 0.0010988712310791016 seconds for step 9 to finish before starting checkpointing. I0419 00:59:31.769603 134949517694784 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 00:59:31.771614 134949517694784 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0419 00:59:31.772900 134949517694784 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/9. I0419 00:59:32.488021 134949517694784 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 00:59:32.488126 134949517694784 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 I0419 00:59:32.513440 134949517694784 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026124s I0419 00:59:32.514369 134949517694784 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 50.099 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.030790328979492188 s) (per-host) I0419 00:59:32.514441 134949517694784 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.030879s (batch_requests_ready=0.001928s, total_serialization_initiated=0.028103s, others=0.000848s) I0419 00:59:32.514517 134949517694784 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.034887s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.034871) I0419 00:59:32.515291 134949517694784 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.74 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/9. I0419 00:59:32.515583 134822747944704 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 01:19:32.515540 I0419 00:59:32.525564 134949517694784 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0419 00:59:32.525832 134821170902784 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 00:59:32.525982 134949517694784 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776560371.7695682, 'wait_for_prev_duration_secs': 7.510185241699219e-05, 'time_between_consecutive_saves_sec': 5.024408340454102, 'checkpointer_blocking_start_time': 1776560371.7716546, 'checkpointer_blocking_duration_secs': 0.7440695762634277, 'get_old_steps_start_time': 1776560372.515745, 'get_old_steps_duration_secs': 2.9325485229492188e-05, 'checkpoint_manager_blocking_start_time': 1776560371.7678676, 'checkpoint_manager_blocking_duration_secs': 0.7580811977386475} I0419 00:59:32.526086 134949517694784 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0419 00:59:32.526138 134949517694784 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 00:59:32.727349 134822773122816 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/9 I0419 00:59:34.871970 134824366950144 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/9/items I0419 00:59:37.811001 134823263901440 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/9/items/array_metadatas/process_0 I0419 00:59:45.460187 134822756337408 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 9.514541s (commit=8.465264s, array_metadata_write=1.049277s) I0419 00:59:45.461257 134822747944704 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 121.717 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 12.977648258209229 s) (per-host) I0419 00:59:45.461385 134822747944704 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 12.945687s. W0419 00:59:47.276328 134822747944704 ocdbt_utils.py:168] [process=0][thread=async_save] Skipping merge of OCDBT checkpoints: No per-process OCDBT checkpoint subdirs found in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/9/items, I0419 00:59:47.702054 134822747944704 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000267s. I0419 00:59:47.702362 134822747944704 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.084694s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/9/items I0419 00:59:47.703727 134822747944704 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/9/items I0419 00:59:48.207323 134822747944704 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/9 I0419 00:59:50.262267 134822747944704 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/9`. I0419 00:59:50.262995 134822747944704 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 18.49 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_15_ocdbt_false/checkpoints/9 I0419 00:59:50.264458 134822747944704 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 17.748756s. I0419 00:59:50.264667 134821170902784 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 00:59:50.264790 134821170902784 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 00:59:50.264862 134821170902784 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0419 00:59:50.266690 134821170902784 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0419 00:59:50.266884 134949517694784 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0419 00:59:50.267038 134949517694784 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 00:59:50.268042 134949517694784 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.203, Tokens/s/device: 13897.843, 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: Sun Apr 19 01:00:01 UTC 2026 EXIT_CODE=0