main (59e0f1759) vs feat/nnx-post-train-fixes (28b1e4a16)| Metric | main 59e0f1759 | feat/nnx-post-train-fixes 28b1e4a16 | Diff (feat/nnx-post-train-fixes − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 5.3910 | 5.3910 | 0 |
| TFLOP/s | 92.261 | 92.129 | -0.132 |
| Tok/s | 13906.5 | 13886.7 | -19.802 |
| Avg s/step | 4.935 | 4.994 | +0.059 |
| 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 25 08:21:01 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-25 08:21:26.430239: 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) I0425 08:21:26.638729 135425501923136 max_utils.py:273] Attempting to initialize the jax distributed system... I0425 08:21:35.679386 135425501923136 distributed.py:149] Starting JAX distributed service on [::]:8482 I0425 08:21:35.681872 135425501923136 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-false-rs-up2qb-slice-job-0-0.mt-15-ocdbt-false-rs-up2qb:8482 I0425 08:21:36.739534 135425501923136 max_utils.py:284] Jax distributed system initialized! I0425 08:21:42.921614 135425501923136 max_utils.py:800] System Information: Jax Version: 0.9.2 I0425 08:21:42.921723 135425501923136 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0425 08:21:42.921764 135425501923136 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 I0425 08:21:42.921800 135425501923136 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0425 08:21:43.621895 135425501923136 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1) I0425 08:21:43.622196 135425501923136 checkpointing.py:677] Setting up checkpoint logger... I0425 08:21:43.622252 135425501923136 checkpointing.py:233] Creating checkpoint manager with ocdbt=False and zarr3=True I0425 08:21:43.622298 135425501923136 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0425 08:21:43.622655 135425501923136 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 0x7b2aa2f180b0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0425 08:21:44.950059 135425501923136 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0425 08:21:44.950330 135425501923136 checkpoint_manager.py:708] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b16142b4c80>}, handler_registry=None I0425 08:21:44.950569 135425501923136 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b16142b4c80>` 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`. I0425 08:21:44.950620 135425501923136 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b16142bce30>` 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`. I0425 08:21:44.950658 135425501923136 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 0x7b16142b4c80>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b16142b4c80>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b16142bce30>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b16142bce30>}). I0425 08:21:44.950983 135425501923136 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0425 08:21:44.951052 135425501923136 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7b15887a5800> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0425 08:21:46.854681 135302811219712 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1777105149312405739, 'commit_timestamp_nsecs': 1777105162536042327, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints/5/_CHECKPOINT_METADATA I0425 08:21:47.101351 135302802827008 checkpoint_manager.py:1871] Missing metrics for step 9 I0425 08:21:47.105781 135425501923136 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints I0425 08:21:47.549004 135425501923136 checkpoint_manager.py:929] [process=6][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_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7b16142b5970> I0425 08:21:47.549226 135425501923136 checkpointing.py:301] Checkpoint manager created! I0425 08:21:48.478007 135425501923136 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 08:21:48.478146 135425501923136 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 08:21:48.865483 135425501923136 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn'). I0425 08:21:48.865576 135425501923136 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 08:21:48.882128 135425501923136 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn'). I0425 08:21:48.882189 135425501923136 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 08:21:48.906154 135425501923136 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 08:21:48.906224 135425501923136 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 08:21:48.922989 135425501923136 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 08:21:48.923053 135425501923136 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 08:21:48.939547 135425501923136 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 08:21:48.939609 135425501923136 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 08:21:48.964768 135425501923136 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch_attn', 'activation_length_attn', 'activation_heads', 'activation_kv'). I0425 08:21:48.964841 135425501923136 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 08:21:48.986213 135425501923136 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0425 08:21:48.986277 135425501923136 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0425 08:21:49.196198 135425501923136 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0425 08:21:49.196309 135425501923136 checkpointing.py:581] restoring from this run's directory step 9 I0425 08:21:49.433411 135425501923136 event_tracking.py:70] [process=6] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints/9. I0425 08:21:49.672085 135425501923136 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints/9. I0425 08:21:49.672570 135425501923136 event_tracking.py:125] [process=6] [sync] Finished blocking load in 0.24 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints/9. I0425 08:21:52.393080 1499 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0425 08:22:01.382524 135425501923136 jax_array_handlers.py:843] [process=6] /jax/orbax/read/worker/io/requested throughput: 155.539 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.159993171691895 s) (per-host) I0425 08:22:01.390771 135425501923136 base_pytree_checkpoint_handler.py:130] [process=6] /jax/checkpoint/read/gbytes_per_sec: 149.671 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.573837995529175 s) (per-host) I0425 08:22:01.392780 135425501923136 event_tracking.py:138] [process=6] [sync] Finished load in 11.96 seconds @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints/9 I0425 08:22:01.392876 135425501923136 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints', 'checkpointer_start_time': 1777105309.4333105, 'checkpointer_duration_secs': 11.959538459777832, 'checkpoint_manager_start_time': 1777105309.196564, 'checkpoint_manager_duration_secs': 12.196287155151367} fsdp: 32 I0425 08:22:01.809212 135425501923136 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0425 08:22:01.809348 135425501923136 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0425 08:22:01.809401 135425501923136 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0425 08:22:01.809460 135425501923136 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0425 08:22:01.809513 135425501923136 maxtext_utils.py:1707] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0425 08:22:01.809551 135425501923136 maxtext_utils.py:1707] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0425 08:22:01.809610 135425501923136 maxtext_utils.py:1707] 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) I0425 08:22:01.809662 135425501923136 maxtext_utils.py:1707] 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') I0425 08:22:01.809701 135425501923136 maxtext_utils.py:1707] 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) I0425 08:22:01.809735 135425501923136 maxtext_utils.py:1707] 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) I0425 08:22:01.809781 135425501923136 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0425 08:22:01.809828 135425501923136 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0425 08:22:02.293936 135425501923136 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 08:22:02.294030 135425501923136 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0425 08:22:02.309599 135425501923136 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 08:22:02.309660 135425501923136 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0425 08:22:13.278911 135425501923136 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. I0425 08:22:13.279759 135425501923136 metric_logger.py:301] number parameters: 1.104 billion I0425 08:22:24.624788 135425501923136 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0425 08:22:24.777982 135425501923136 checkpointing.py:776] Waited 0.15317320823669434 seconds for step 10 to finish before starting checkpointing. I0425 08:22:24.780636 135425501923136 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 08:22:24.782498 135425501923136 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 10 I0425 08:22:24.783962 135425501923136 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints/10. I0425 08:22:25.122514 135425501923136 signaling_client.py:364] Using JaxDistributedSignalingClient I0425 08:22:25.123513 135425501923136 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 08:22:25.123584 135425501923136 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 I0425 08:22:25.403945 135425501923136 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.281509s I0425 08:22:25.404134 135425501923136 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.402 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2855815887451172 s) (per-host) I0425 08:22:25.404191 135425501923136 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.285650s (batch_requests_ready=0.002310s, total_serialization_initiated=0.283249s, others=0.000091s) I0425 08:22:25.404290 135425501923136 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.289765s (all_items=0.000019s, per_item={'items': '0.00001907'}, temp_paths=0.289746) I0425 08:22:25.405011 135425501923136 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.62 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints/10. I0425 08:22:25.405379 135300546270976 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 08:42:25.405341 I0425 08:22:25.408070 135425501923136 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0425 08:22:25.408323 135299497707264 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 08:22:25.408471 135425501923136 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777105344.780611, 'wait_for_prev_duration_secs': 9.179115295410156e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777105344.7825575, 'checkpointer_blocking_duration_secs': 0.622978687286377, 'get_old_steps_start_time': 1777105345.4055586, 'get_old_steps_duration_secs': 3.123283386230469e-05, 'checkpoint_manager_blocking_start_time': 1777105344.7784815, 'checkpoint_manager_blocking_duration_secs': 0.6299550533294678} I0425 08:22:25.408576 135425501923136 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0425 08:22:25.408627 135425501923136 max_utils.py:750] Memstats: After params initialized: I0425 08:22:25.408674 135425501923136 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0425 08:22:25.408706 135425501923136 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0425 08:22:25.408732 135425501923136 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0425 08:22:25.408756 135425501923136 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0425 08:22:25.719407 135425501923136 metric_logger.py:196] completed step: 10, seconds: 11.345, TFLOP/s/device: 1.198, Tokens/s/device: 180.521, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0425 08:22:25.913782 135425501923136 metric_logger.py:196] completed step: 11, seconds: 1.093, TFLOP/s/device: 12.430, Tokens/s/device: 1873.550, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0425 08:22:26.061455 135425501923136 metric_logger.py:196] completed step: 12, seconds: 0.047, TFLOP/s/device: 288.486, Tokens/s/device: 43483.800, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0425 08:22:26.208889 135425501923136 metric_logger.py:196] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.717, Tokens/s/device: 13372.336, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0425 08:22:26.213389 135425501923136 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0425 08:22:26.503343 135425501923136 checkpointing.py:776] Waited 0.2899188995361328 seconds for step 15 to finish before starting checkpointing. I0425 08:22:26.506352 135425501923136 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 08:22:30.401465 135299506099968 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints/10/items/array_metadatas/process_6 I0425 08:22:35.565484 135300546270976 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 151.202 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.44689679145813 s) (per-host) I0425 08:22:35.565612 135300546270976 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.160120s. I0425 08:22:41.450679 135300546270976 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 16.045170s. I0425 08:22:41.450946 135299497707264 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 08:22:41.451060 135299497707264 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 08:22:41.451123 135299497707264 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0425 08:22:41.453016 135299497707264 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0425 08:22:41.453184 135425501923136 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0425 08:22:41.453315 135425501923136 checkpoint_manager.py:1452] Waiting for previous save to complete took 14.946961 seconds. If this number is high, consider checkpointing less frequently. I0425 08:22:41.455424 135425501923136 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 15 I0425 08:22:41.457880 135425501923136 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints/15. I0425 08:22:41.738972 135425501923136 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 08:22:41.739065 135425501923136 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 I0425 08:22:41.780711 135425501923136 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042732s I0425 08:22:41.780873 135425501923136 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 33.349 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04625558853149414 s) (per-host) I0425 08:22:41.780923 135425501923136 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.046314s (batch_requests_ready=0.001890s, total_serialization_initiated=0.044357s, others=0.000067s) I0425 08:22:41.781010 135425501923136 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.051292s (all_items=0.000013s, per_item={'items': '0.00001311'}, temp_paths=0.051279) I0425 08:22:41.781720 135425501923136 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints/15. I0425 08:22:41.782057 135299497707264 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 08:42:41.782018 I0425 08:22:41.784130 135425501923136 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0425 08:22:41.784446 135295748003584 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 08:22:41.784582 135425501923136 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777105346.506323, 'wait_for_prev_duration_secs': 14.946961402893066, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777105361.4554675, 'checkpointer_blocking_duration_secs': 0.3267514705657959, 'get_old_steps_start_time': 1777105361.782245, 'get_old_steps_duration_secs': 3.0040740966796875e-05, 'checkpoint_manager_blocking_start_time': 1777105346.5039601, 'checkpoint_manager_blocking_duration_secs': 15.280587434768677} I0425 08:22:41.784746 135425501923136 checkpointing.py:408] Started an asynchronous checkpoint save for step 15 I0425 08:22:41.785534 135425501923136 metric_logger.py:196] completed step: 14, seconds: 0.148, TFLOP/s/device: 92.076, Tokens/s/device: 13878.724, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0425 08:22:41.791905 135425501923136 metric_logger.py:196] completed step: 15, seconds: 0.148, TFLOP/s/device: 92.069, Tokens/s/device: 13877.689, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0425 08:22:46.796527 135299506099968 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints/15/items/array_metadatas/process_6 I0425 08:22:51.678791 135299497707264 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 158.847 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 9.944111824035645 s) (per-host) I0425 08:22:51.678917 135299497707264 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 9.896730s. I0425 08:22:57.061857 135299497707264 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 15.279653s. I0425 08:22:57.062164 135295748003584 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 08:22:57.062309 135295748003584 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 08:22:57.062361 135295748003584 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0425 08:22:57.063895 135295748003584 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0425 08:23:02.477665 135425501923136 metric_logger.py:196] completed step: 16, seconds: 15.577, TFLOP/s/device: 0.872, Tokens/s/device: 131.476, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0425 08:23:02.624975 135425501923136 metric_logger.py:196] completed step: 17, seconds: 20.540, TFLOP/s/device: 0.662, Tokens/s/device: 99.710, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0425 08:23:02.772286 135425501923136 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.190, Tokens/s/device: 13443.613, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0425 08:23:02.919049 135425501923136 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0425 08:23:02.919783 135425501923136 checkpointing.py:776] Waited 0.0007543563842773438 seconds for step 19 to finish before starting checkpointing. I0425 08:23:02.921813 135425501923136 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 08:23:02.923544 135425501923136 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 19 I0425 08:23:02.924838 135425501923136 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints/19. I0425 08:23:03.247838 135425501923136 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 08:23:03.247933 135425501923136 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 I0425 08:23:03.289537 135425501923136 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042638s I0425 08:23:03.289695 135425501923136 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 33.389 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04620003700256348 s) (per-host) I0425 08:23:03.289745 135425501923136 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.046259s (batch_requests_ready=0.001767s, total_serialization_initiated=0.044429s, others=0.000063s) I0425 08:23:03.289834 135425501923136 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.050433s (all_items=0.000012s, per_item={'items': '0.00001168'}, temp_paths=0.050421) I0425 08:23:03.290517 135425501923136 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.37 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints/19. I0425 08:23:03.290842 135295748003584 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 08:43:03.290804 I0425 08:23:03.292738 135425501923136 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0425 08:23:03.292951 135295194347264 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 08:23:03.293072 135425501923136 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777105382.921781, 'wait_for_prev_duration_secs': 7.200241088867188e-05, 'time_between_consecutive_saves_sec': 5.85784649848938, 'checkpointer_blocking_start_time': 1777105382.9235842, 'checkpointer_blocking_duration_secs': 0.36739611625671387, 'get_old_steps_start_time': 1777105383.2909966, 'get_old_steps_duration_secs': 2.5033950805664062e-05, 'checkpoint_manager_blocking_start_time': 1777105382.9200153, 'checkpoint_manager_blocking_duration_secs': 0.37302708625793457} I0425 08:23:03.293204 135425501923136 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0425 08:23:03.293252 135425501923136 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 08:23:08.379390 135299506099968 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_15_ocdbt_false/checkpoints/19/items/array_metadatas/process_6 I0425 08:23:13.118463 135295748003584 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 159.960 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 9.874923467636108 s) (per-host) I0425 08:23:13.118605 135295748003584 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 9.827653s. I0425 08:23:18.204630 135295748003584 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 14.913662s. I0425 08:23:18.204918 135295194347264 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 08:23:18.205042 135295194347264 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 08:23:18.205085 135295194347264 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0425 08:23:18.208024 135295194347264 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0425 08:23:18.208233 135425501923136 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0425 08:23:18.208385 135425501923136 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 08:23:18.209401 135425501923136 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.261, Tokens/s/device: 13906.525, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 25 08:23:31 UTC 2026 EXIT_CODE=0
XPK Start: Sat Apr 25 13:17:01 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-25 13:17:26.236603: 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) I0425 13:17:26.449681 133651906737984 max_utils.py:273] Attempting to initialize the jax distributed system... I0425 13:17:35.490240 133651906737984 distributed.py:149] Starting JAX distributed service on [::]:8482 I0425 13:17:35.492476 133651906737984 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-false-rs-q0hke-slice-job-0-0.mt-15-ocdbt-false-rs-q0hke:8482 I0425 13:17:36.614210 133651906737984 max_utils.py:284] Jax distributed system initialized! I0425 13:17:42.812537 133651906737984 max_utils.py:800] System Information: Jax Version: 0.9.2 I0425 13:17:42.812644 133651906737984 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0425 13:17:42.812685 133651906737984 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 I0425 13:17:42.812723 133651906737984 train_utils.py:391] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0425 13:17:43.505254 133651906737984 maxtext_utils.py:1771] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 13:17:43.505840 133651906737984 maxtext_utils.py:1771] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 13:17:43.506020 133651906737984 checkpointing.py:688] Setting up checkpoint logger... I0425 13:17:43.506068 133651906737984 checkpointing.py:234] Creating checkpoint manager with ocdbt=False and zarr3=True I0425 13:17:43.506124 133651906737984 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0425 13:17:43.506468 133651906737984 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 0x798dae363800>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0425 13:17:44.799742 133651906737984 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0425 13:17:44.799973 133651906737984 checkpoint_manager.py:708] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7979901dbb30>}, handler_registry=None I0425 13:17:44.800228 133651906737984 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7979901dbb30>` 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`. I0425 13:17:44.800287 133651906737984 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7979901dcb90>` 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`. I0425 13:17:44.800324 133651906737984 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 0x7979901dbb30>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7979901dbb30>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7979901dcb90>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7979901dcb90>}). I0425 13:17:44.800652 133651906737984 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0425 13:17:44.800720 133651906737984 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7979606f1d00> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0425 13:17:46.851550 133530952636160 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1777122933097640792, 'commit_timestamp_nsecs': 1777122945668268408, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints/9/_CHECKPOINT_METADATA I0425 13:17:47.095195 133530533230336 checkpoint_manager.py:1871] Missing metrics for step 0 I0425 13:17:47.160403 133651906737984 checkpoint_manager.py:1812] Found 3 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints I0425 13:17:47.182009 133651906737984 checkpoint_manager.py:929] [process=6][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_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x798c7e8a9040> I0425 13:17:47.182158 133651906737984 checkpointing.py:302] Checkpoint manager created! I0425 13:17:48.138411 133651906737984 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 13:17:48.138521 133651906737984 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 13:17:48.522426 133651906737984 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn'). I0425 13:17:48.522524 133651906737984 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 13:17:48.539034 133651906737984 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn'). I0425 13:17:48.539106 133651906737984 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 13:17:48.562905 133651906737984 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 13:17:48.562974 133651906737984 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 13:17:48.579551 133651906737984 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 13:17:48.579613 133651906737984 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 13:17:48.596231 133651906737984 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 13:17:48.596291 133651906737984 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 13:17:48.621203 133651906737984 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch_attn', 'activation_length_attn', 'activation_heads', 'activation_kv'). I0425 13:17:48.621276 133651906737984 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 13:17:48.642217 133651906737984 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0425 13:17:48.642285 133651906737984 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0425 13:17:48.863702 133651906737984 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0425 13:17:48.863809 133651906737984 checkpointing.py:582] restoring from this run's directory step 9 I0425 13:17:49.109238 133651906737984 event_tracking.py:70] [process=6] [sync] Started load checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints/9. I0425 13:17:49.351906 133651906737984 checkpointer.py:307] Restoring checkpoint from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints/9. I0425 13:17:49.352356 133651906737984 event_tracking.py:125] [process=6] [sync] Finished blocking load in 0.24 seconds. Continuing load @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints/9. I0425 13:17:51.936682 1482 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0425 13:18:01.075702 133651906737984 jax_array_handlers.py:843] [process=6] /jax/orbax/read/worker/io/requested throughput: 154.925 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.200215578079224 s) (per-host) I0425 13:18:01.082646 133651906737984 base_pytree_checkpoint_handler.py:130] [process=6] /jax/checkpoint/read/gbytes_per_sec: 149.326 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.59824824333191 s) (per-host) I0425 13:18:01.330401 133651906737984 event_tracking.py:138] [process=6] [sync] Finished load in 12.22 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints/9 I0425 13:18:01.330560 133651906737984 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints', 'checkpointer_start_time': 1777123069.109139, 'checkpointer_duration_secs': 12.22139024734497, 'checkpoint_manager_start_time': 1777123068.8640647, 'checkpoint_manager_duration_secs': 12.466466665267944} fsdp: 32 I0425 13:18:01.750653 133651906737984 maxtext_utils.py:1880] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0425 13:18:01.750794 133651906737984 maxtext_utils.py:1880] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0425 13:18:01.750852 133651906737984 maxtext_utils.py:1880] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0425 13:18:01.750912 133651906737984 maxtext_utils.py:1880] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0425 13:18:01.750963 133651906737984 maxtext_utils.py:1880] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0425 13:18:01.751000 133651906737984 maxtext_utils.py:1880] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0425 13:18:01.751052 133651906737984 maxtext_utils.py:1880] 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) I0425 13:18:01.751127 133651906737984 maxtext_utils.py:1880] 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') I0425 13:18:01.751173 133651906737984 maxtext_utils.py:1880] 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) I0425 13:18:01.751210 133651906737984 maxtext_utils.py:1880] 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) I0425 13:18:01.751262 133651906737984 maxtext_utils.py:1880] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0425 13:18:01.751310 133651906737984 maxtext_utils.py:1880] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0425 13:18:02.244474 133651906737984 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 13:18:02.244570 133651906737984 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0425 13:18:02.260174 133651906737984 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 13:18:02.260235 133651906737984 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0425 13:18:13.284296 133651906737984 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. I0425 13:18:13.285130 133651906737984 metric_logger.py:301] number parameters: 1.104 billion I0425 13:18:24.605849 133651906737984 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0425 13:18:24.937233 133651906737984 checkpointing.py:798] Waited 0.33136439323425293 seconds for step 10 to finish before starting checkpointing. I0425 13:18:24.939688 133651906737984 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 13:18:24.941556 133651906737984 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 10 I0425 13:18:24.942959 133651906737984 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints/10. I0425 13:18:25.726395 133651906737984 signaling_client.py:364] Using JaxDistributedSignalingClient I0425 13:18:25.727354 133651906737984 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 13:18:25.727413 133651906737984 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 I0425 13:18:26.021729 133651906737984 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.295410s I0425 13:18:26.021900 133651906737984 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.155 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2992582321166992 s) (per-host) I0425 13:18:26.021950 133651906737984 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.299319s (batch_requests_ready=0.002148s, total_serialization_initiated=0.297101s, others=0.000070s) I0425 13:18:26.022041 133651906737984 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.303601s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.303584) I0425 13:18:26.022834 133651906737984 event_tracking.py:125] [process=6] [async] Finished blocking save in 1.08 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints/10. I0425 13:18:26.023145 133527714658048 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 13:38:26.023112 I0425 13:18:26.025044 133651906737984 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0425 13:18:26.025345 133527261648640 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 13:18:26.025471 133651906737984 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777123104.9396691, 'wait_for_prev_duration_secs': 6.103515625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777123104.941593, 'checkpointer_blocking_duration_secs': 1.081705093383789, 'get_old_steps_start_time': 1777123106.0233207, 'get_old_steps_duration_secs': 3.314018249511719e-05, 'checkpoint_manager_blocking_start_time': 1777123104.9377172, 'checkpoint_manager_blocking_duration_secs': 1.0877201557159424} I0425 13:18:26.025633 133651906737984 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0425 13:18:26.025682 133651906737984 max_utils.py:750] Memstats: After params initialized: I0425 13:18:26.025731 133651906737984 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0425 13:18:26.025763 133651906737984 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0425 13:18:26.025790 133651906737984 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0425 13:18:26.025814 133651906737984 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0425 13:18:26.332724 133651906737984 metric_logger.py:196] completed step: 10, seconds: 11.321, TFLOP/s/device: 1.200, Tokens/s/device: 180.909, total_weights: 65536, loss: 6.833, lm_loss: 6.833, perplexity: 927.869 I0425 13:18:26.523251 133651906737984 metric_logger.py:196] completed step: 11, seconds: 1.725, TFLOP/s/device: 7.875, Tokens/s/device: 1187.009, total_weights: 65536, loss: 6.527, lm_loss: 6.527, perplexity: 683.659 I0425 13:18:26.670973 133651906737984 metric_logger.py:196] completed step: 12, seconds: 0.043, TFLOP/s/device: 313.986, Tokens/s/device: 47327.433, total_weights: 65536, loss: 6.265, lm_loss: 6.265, perplexity: 526.068 I0425 13:18:26.818179 133651906737984 metric_logger.py:196] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.686, Tokens/s/device: 13367.710, total_weights: 65536, loss: 6.045, lm_loss: 6.045, perplexity: 422.001 I0425 13:18:26.822572 133651906737984 checkpointing.py:794] Waiting for step 15 to finish before checkpoint... I0425 13:18:27.112011 133651906737984 checkpointing.py:798] Waited 0.28940701484680176 seconds for step 15 to finish before starting checkpointing. I0425 13:18:27.115233 133651906737984 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 13:18:30.904435 133527270041344 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints/10/items/array_metadatas/process_6 I0425 13:18:36.635162 133527714658048 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 144.751 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.912498474121094 s) (per-host) I0425 13:18:36.635270 133527714658048 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.612013s. I0425 13:18:41.550190 133527714658048 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 15.526918s. I0425 13:18:41.550452 133527261648640 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 13:18:41.550581 133527261648640 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 13:18:41.550629 133527261648640 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0425 13:18:41.552372 133527261648640 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0425 13:18:41.552535 133651906737984 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0425 13:18:41.552675 133651906737984 checkpoint_manager.py:1452] Waiting for previous save to complete took 14.437443 seconds. If this number is high, consider checkpointing less frequently. I0425 13:18:41.554844 133651906737984 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 15 I0425 13:18:41.557353 133651906737984 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints/15. I0425 13:18:41.840962 133651906737984 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 13:18:41.841052 133651906737984 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 I0425 13:18:41.889777 133651906737984 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.049955s I0425 13:18:41.889950 133651906737984 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 28.900 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.05337667465209961 s) (per-host) I0425 13:18:41.889999 133651906737984 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.053434s (batch_requests_ready=0.001767s, total_serialization_initiated=0.051602s, others=0.000065s) I0425 13:18:41.890088 133651906737984 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.058284s (all_items=0.000013s, per_item={'items': '0.00001264'}, temp_paths=0.058271) I0425 13:18:41.890799 133651906737984 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.34 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints/15. I0425 13:18:41.891108 133527261648640 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 13:38:41.891060 I0425 13:18:41.893024 133651906737984 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0425 13:18:41.893319 133522857641728 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 13:18:41.893479 133651906737984 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777123107.1152034, 'wait_for_prev_duration_secs': 14.437443017959595, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777123121.5548851, 'checkpointer_blocking_duration_secs': 0.3363649845123291, 'get_old_steps_start_time': 1777123121.8912735, 'get_old_steps_duration_secs': 3.2901763916015625e-05, 'checkpoint_manager_blocking_start_time': 1777123107.1126764, 'checkpoint_manager_blocking_duration_secs': 14.78076982498169} I0425 13:18:41.893589 133651906737984 checkpointing.py:409] Started an asynchronous checkpoint save for step 15 I0425 13:18:41.894494 133651906737984 metric_logger.py:196] completed step: 14, seconds: 0.148, TFLOP/s/device: 91.983, Tokens/s/device: 13864.724, total_weights: 65536, loss: 5.864, lm_loss: 5.864, perplexity: 352.193 I0425 13:18:41.900473 133651906737984 metric_logger.py:196] completed step: 15, seconds: 0.147, TFLOP/s/device: 92.384, Tokens/s/device: 13925.152, total_weights: 65536, loss: 5.719, lm_loss: 5.719, perplexity: 304.661 I0425 13:18:46.925970 133527270041344 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints/15/items/array_metadatas/process_6 I0425 13:18:51.648130 133527261648640 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 160.995 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 9.811486959457397 s) (per-host) I0425 13:18:51.648263 133527261648640 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 9.757038s. I0425 13:18:58.269927 133527261648640 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 16.378688s. I0425 13:18:58.270250 133522857641728 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 13:18:58.270370 133522857641728 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 13:18:58.270417 133522857641728 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0425 13:18:58.271877 133522857641728 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0425 13:19:03.075332 133651906737984 metric_logger.py:196] completed step: 16, seconds: 15.076, TFLOP/s/device: 0.901, Tokens/s/device: 135.841, total_weights: 65536, loss: 5.605, lm_loss: 5.605, perplexity: 271.872 I0425 13:19:03.222887 133651906737984 metric_logger.py:196] completed step: 17, seconds: 21.029, TFLOP/s/device: 0.646, Tokens/s/device: 97.391, total_weights: 65536, loss: 5.517, lm_loss: 5.517, perplexity: 248.975 I0425 13:19:03.370424 133651906737984 metric_logger.py:196] completed step: 18, seconds: 0.152, TFLOP/s/device: 89.133, Tokens/s/device: 13435.058, total_weights: 65536, loss: 5.448, lm_loss: 5.448, perplexity: 232.333 I0425 13:19:03.517159 133651906737984 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0425 13:19:03.517840 133651906737984 checkpointing.py:798] Waited 0.0007472038269042969 seconds for step 19 to finish before starting checkpointing. I0425 13:19:03.519940 133651906737984 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 13:19:03.521485 133651906737984 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 19 I0425 13:19:03.522956 133651906737984 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints/19. I0425 13:19:03.844882 133651906737984 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 13:19:03.844973 133651906737984 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 I0425 13:19:03.888745 133651906737984 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044734s I0425 13:19:03.888909 133651906737984 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 32.033 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.048156023025512695 s) (per-host) I0425 13:19:03.888959 133651906737984 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.048215s (batch_requests_ready=0.001635s, total_serialization_initiated=0.046516s, others=0.000063s) I0425 13:19:03.889064 133651906737984 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.052324s (all_items=0.000011s, per_item={'items': '0.00001121'}, temp_paths=0.052312) I0425 13:19:03.889761 133651906737984 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.37 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints/19. I0425 13:19:03.890124 133527270041344 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 13:39:03.890068 I0425 13:19:03.892057 133651906737984 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0425 13:19:03.892308 133522857641728 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 13:19:03.892429 133651906737984 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777123143.5199082, 'wait_for_prev_duration_secs': 7.367134094238281e-05, 'time_between_consecutive_saves_sec': 5.24799108505249, 'checkpointer_blocking_start_time': 1777123143.5215232, 'checkpointer_blocking_duration_secs': 0.36874818801879883, 'get_old_steps_start_time': 1777123143.8902895, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1777123143.518064, 'checkpoint_manager_blocking_duration_secs': 0.3743324279785156} I0425 13:19:03.892537 133651906737984 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0425 13:19:03.892584 133651906737984 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 13:19:09.116397 133527660115712 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/linen_xpk_feat_nnx_post_train_fixes_20260425_121355_15_ocdbt_false/checkpoints/19/items/array_metadatas/process_6 I0425 13:19:13.941496 133527270041344 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 156.385 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 10.100703954696655 s) (per-host) I0425 13:19:13.941621 133527270041344 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 10.051378s. I0425 13:19:18.865684 133527270041344 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 14.975425s. I0425 13:19:18.865957 133522857641728 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 13:19:18.866072 133522857641728 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 13:19:18.866127 133522857641728 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0425 13:19:18.867702 133522857641728 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0425 13:19:18.867898 133651906737984 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0425 13:19:18.868042 133651906737984 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 13:19:18.868991 133651906737984 metric_logger.py:196] completed step: 19, seconds: 0.147, TFLOP/s/device: 92.129, Tokens/s/device: 13886.723, total_weights: 65536, loss: 5.391, lm_loss: 5.391, perplexity: 219.349 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 25 13:19:31 UTC 2026 EXIT_CODE=0