XPK Start: Tue Apr 21 16:43:49 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-21 16:44:13.854072: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0421 16:44:14.033844 137912354039616 max_utils.py:273] Attempting to initialize the jax distributed system... I0421 16:44:23.073486 137912354039616 distributed.py:149] Starting JAX distributed service on [::]:8482 I0421 16:44:23.075888 137912354039616 distributed.py:172] Connecting to JAX distributed service on mt-05-fp8-z40cv-slice-job-0-0.mt-05-fp8-z40cv:8482 I0421 16:44:24.328381 137912354039616 max_utils.py:284] Jax distributed system initialized! I0421 16:44:30.505658 137912354039616 max_utils.py:800] System Information: Jax Version: 0.9.2 I0421 16:44:30.505760 137912354039616 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0421 16:44:30.505802 137912354039616 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0421 16:44:30.505836 137912354039616 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0421 16:44:31.193658 137912354039616 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 16:44:31.629023 137912354039616 checkpointing.py:688] Setting up checkpoint logger... I0421 16:44:31.629145 137912354039616 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0421 16:44:31.629189 137912354039616 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0421 16:44:31.629403 137912354039616 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7d6daf9ee600>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 16:44:34.512494 137912354039616 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0421 16:44:34.512751 137912354039616 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 0x7d59203ea570>}, handler_registry=None I0421 16:44:34.512992 137912354039616 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d59203ea570>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0421 16:44:34.513040 137912354039616 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d59203d93d0>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0421 16:44:34.513075 137912354039616 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 0x7d59203ea570>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d59203ea570>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d59203d93d0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d59203d93d0>}). I0421 16:44:34.513460 137912354039616 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0421 16:44:34.513531 137912354039616 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7d59107d0860> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0421 16:44:35.990569 137912354039616 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints I0421 16:44:35.993210 137912354039616 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=10), 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/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7d59203dae10> I0421 16:44:35.993324 137912354039616 checkpointing.py:302] Checkpoint manager created! I0421 16:44:37.260828 137912354039616 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0421 16:44:37.260947 137912354039616 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0421 16:44:39.787602 137912354039616 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 16:44:39.787694 137912354039616 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 16:44:39.793305 137912354039616 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 16:44:39.793362 137912354039616 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 16:44:39.810550 137912354039616 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 16:44:39.810620 137912354039616 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 16:44:39.826983 137912354039616 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 16:44:39.827039 137912354039616 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0421 16:44:39.887966 137912354039616 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 16:44:39.888051 137912354039616 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 16:44:39.904561 137912354039616 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 16:44:39.904629 137912354039616 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 16:44:39.921032 137912354039616 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 16:44:39.921089 137912354039616 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 16:44:39.953405 137912354039616 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 16:44:39.953485 137912354039616 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0421 16:44:40.014361 137912354039616 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 16:44:40.014437 137912354039616 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0421 16:44:53.878890 137912354039616 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0421 16:44:53.889624 137912354039616 metric_logger.py:301] number parameters: 1.104 billion I0421 16:45:08.272851 137912354039616 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0421 16:45:08.435790 137912354039616 checkpointing.py:798] Waited 0.16292476654052734 seconds for step 0 to finish before starting checkpointing. I0421 16:45:08.438172 137912354039616 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 16:45:08.440159 137912354039616 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0421 16:45:08.442055 137912354039616 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/0. I0421 16:45:09.194734 137912354039616 signaling_client.py:364] Using JaxDistributedSignalingClient I0421 16:45:09.224657 137912354039616 jax_array_handlers.py:360] Scheduling D2H of 153 prioritized jax.Array. I0421 16:45:09.224739 137912354039616 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0421 16:45:09.464726 137782651500288 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/0 I0421 16:45:09.626646 137912354039616 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.404417s I0421 16:45:09.627459 137912354039616 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 3.631 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.42487645149230957 s) (per-host) I0421 16:45:09.627530 137912354039616 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.424966s (batch_requests_ready=0.006277s, total_serialization_initiated=0.417962s, others=0.000727s) I0421 16:45:09.627636 137912354039616 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.428916s (all_items=0.000019s, per_item={'items': '0.00001931'}, temp_paths=0.428897) I0421 16:45:09.628441 137912354039616 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.19 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/0. I0421 16:45:09.628789 137782257239808 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-21 17:05:09.628748 I0421 16:45:09.644289 137912354039616 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0421 16:45:09.644566 137782659892992 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 16:45:09.644732 137912354039616 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776789908.4381516, 'wait_for_prev_duration_secs': 6.151199340820312e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776789908.440198, 'checkpointer_blocking_duration_secs': 1.1887450218200684, 'get_old_steps_start_time': 1776789909.6289663, 'get_old_steps_duration_secs': 2.9325485229492188e-05, 'checkpoint_manager_blocking_start_time': 1776789908.4363124, 'checkpoint_manager_blocking_duration_secs': 1.2083823680877686} I0421 16:45:09.644843 137912354039616 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0421 16:45:09.644919 137912354039616 max_utils.py:750] Memstats: After params initialized: I0421 16:45:09.644969 137912354039616 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_0(process=0,(0,0,0,0)) I0421 16:45:09.645001 137912354039616 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_1(process=0,(1,0,0,0)) I0421 16:45:09.645027 137912354039616 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_4(process=0,(0,1,0,0)) I0421 16:45:09.645049 137912354039616 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_5(process=0,(1,1,0,0)) I0421 16:45:10.003492 137912354039616 metric_logger.py:196] completed step: 0, seconds: 13.945, TFLOP/s/device: 0.974, Tokens/s/device: 146.864, total_weights: 65536, loss: 10.872, lm_loss: 10.872, perplexity: 52680.742 I0421 16:45:10.006927 137912354039616 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/tensorboard/' I0421 16:45:10.459666 137912354039616 metric_logger.py:196] completed step: 1, seconds: 1.724, TFLOP/s/device: 7.880, Tokens/s/device: 1187.769, total_weights: 65536, loss: 10.872, lm_loss: 10.872, perplexity: 52680.742 I0421 16:45:10.605651 137912354039616 metric_logger.py:196] completed step: 2, seconds: 0.458, TFLOP/s/device: 29.683, Tokens/s/device: 4474.205, total_weights: 65536, loss: 10.856, lm_loss: 10.856, perplexity: 51864.762 I0421 16:45:10.762773 137912354039616 metric_logger.py:196] completed step: 3, seconds: 0.018, TFLOP/s/device: 769.504, Tokens/s/device: 115987.993, total_weights: 65536, loss: 10.824, lm_loss: 10.824, perplexity: 50225.512 I0421 16:45:11.288564 137782634714880 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/0/items I0421 16:45:11.805415 137781053069056 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776789911026652334, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776789911026652334, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/0/_CHECKPOINT_METADATA I0421 16:45:12.616143 2879 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0421 16:45:14.800376 137782274025216 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/0/items/array_metadatas/process_0 I0421 16:45:32.488340 137782265632512 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 19.864169s (commit=18.671525s, array_metadata_write=1.192645s) I0421 16:45:32.489511 137782257239808 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 67.839 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 23.28692317008972 s) (per-host) I0421 16:45:32.489589 137782257239808 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 22.860675s. I0421 16:45:33.169305 137782257239808 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776789911026652334, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/0/_CHECKPOINT_METADATA I0421 16:45:33.920450 137781053069056 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776789911026652334, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/0/_CHECKPOINT_METADATA I0421 16:45:33.934375 137782257239808 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.001081s. I0421 16:45:34.304263 137782257239808 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0421 16:45:35.559001 137912354039616 metric_logger.py:196] completed step: 4, seconds: 0.146, TFLOP/s/device: 93.144, Tokens/s/device: 14039.706, total_weights: 65536, loss: 10.793, lm_loss: 10.793, perplexity: 48689.156 I0421 16:45:35.571857 137912354039616 metric_logger.py:196] completed step: 5, seconds: 0.158, TFLOP/s/device: 86.188, Tokens/s/device: 12991.132, total_weights: 65536, loss: 10.763, lm_loss: 10.763, perplexity: 47229.012 I0421 16:45:35.723685 137912354039616 metric_logger.py:196] completed step: 6, seconds: 24.794, TFLOP/s/device: 0.548, Tokens/s/device: 82.600, total_weights: 65536, loss: 10.733, lm_loss: 10.733, perplexity: 45864.258 I0421 16:45:35.726766 137782257239808 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.430822s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/0/items I0421 16:45:35.729058 137782257239808 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/0/items I0421 16:45:35.880776 137912354039616 metric_logger.py:196] completed step: 7, seconds: 0.010, TFLOP/s/device: 1314.163, Tokens/s/device: 198084.921, total_weights: 65536, loss: 10.706, lm_loss: 10.706, perplexity: 44620.242 I0421 16:45:36.037700 137912354039616 metric_logger.py:196] completed step: 8, seconds: 0.157, TFLOP/s/device: 86.785, Tokens/s/device: 13081.247, total_weights: 65536, loss: 10.680, lm_loss: 10.680, perplexity: 43477.055 I0421 16:45:36.048183 137912354039616 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0421 16:45:36.252425 137782257239808 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/0 I0421 16:45:36.352140 137912354039616 checkpointing.py:798] Waited 0.30392932891845703 seconds for step 10 to finish before starting checkpointing. I0421 16:45:36.354405 137912354039616 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 16:45:37.981449 137782257239808 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/0`. I0421 16:45:37.982168 137782257239808 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 29.54 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/0 I0421 16:45:37.983677 137782257239808 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 28.354769s. I0421 16:45:37.983867 137782659892992 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 16:45:37.983991 137782659892992 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 16:45:37.984059 137782659892992 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0421 16:45:37.987356 137782659892992 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0421 16:45:37.987507 137912354039616 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. W0421 16:45:37.987644 137912354039616 checkpoint_manager.py:1452] Waiting for previous save to complete took 1.633240 seconds. If this number is high, consider checkpointing less frequently. I0421 16:45:37.989661 137912354039616 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 10 I0421 16:45:37.991549 137912354039616 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/10. I0421 16:45:38.296104 137912354039616 jax_array_handlers.py:360] Scheduling D2H of 153 prioritized jax.Array. I0421 16:45:38.296264 137912354039616 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0421 16:45:38.337426 137912354039616 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.043629s I0421 16:45:38.338295 137912354039616 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 25.161 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.06131577491760254 s) (per-host) I0421 16:45:38.338371 137912354039616 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.061410s (batch_requests_ready=0.005602s, total_serialization_initiated=0.054986s, others=0.000821s) I0421 16:45:38.338460 137912354039616 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.065647s (all_items=0.000017s, per_item={'items': '0.00001740'}, temp_paths=0.065630) I0421 16:45:38.339176 137912354039616 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.35 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/10. I0421 16:45:38.339512 137782659892992 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-21 17:05:38.339473 I0421 16:45:38.528321 137782257239808 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/10 I0421 16:45:38.790234 137912354039616 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0421 16:45:38.790605 137781068347136 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 16:45:38.790783 137912354039616 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776789936.354372, 'wait_for_prev_duration_secs': 1.633239984512329, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776789937.9896991, 'checkpointer_blocking_duration_secs': 0.3499279022216797, 'get_old_steps_start_time': 1776789938.3396513, 'get_old_steps_duration_secs': 3.0994415283203125e-05, 'checkpoint_manager_blocking_start_time': 1776789936.352621, 'checkpoint_manager_blocking_duration_secs': 2.4381232261657715} I0421 16:45:38.790902 137912354039616 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0421 16:45:38.791607 137912354039616 metric_logger.py:196] completed step: 9, seconds: 0.154, TFLOP/s/device: 88.237, Tokens/s/device: 13299.997, total_weights: 65536, loss: 10.656, lm_loss: 10.656, perplexity: 42467.320 I0421 16:45:38.806039 137912354039616 metric_logger.py:196] completed step: 10, seconds: 0.157, TFLOP/s/device: 86.680, Tokens/s/device: 13065.307, total_weights: 65536, loss: 10.636, lm_loss: 10.636, perplexity: 41601.941 I0421 16:45:38.956310 137912354039616 metric_logger.py:196] completed step: 11, seconds: 2.755, TFLOP/s/device: 4.932, Tokens/s/device: 743.413, total_weights: 65536, loss: 10.618, lm_loss: 10.618, perplexity: 40849.570 I0421 16:45:39.918672 137912354039616 metric_logger.py:196] completed step: 12, seconds: 0.013, TFLOP/s/device: 1056.460, Tokens/s/device: 159241.117, total_weights: 65536, loss: 10.602, lm_loss: 10.602, perplexity: 40203.926 I0421 16:45:39.930643 137912354039616 metric_logger.py:196] completed step: 13, seconds: 1.101, TFLOP/s/device: 12.341, Tokens/s/device: 1860.191, total_weights: 65536, loss: 10.588, lm_loss: 10.588, perplexity: 39652.145 I0421 16:45:40.083164 137912354039616 metric_logger.py:196] completed step: 14, seconds: 0.013, TFLOP/s/device: 1060.997, Tokens/s/device: 159925.035, total_weights: 65536, loss: 10.577, lm_loss: 10.577, perplexity: 39212.879 I0421 16:45:40.239917 137912354039616 metric_logger.py:196] completed step: 15, seconds: 0.010, TFLOP/s/device: 1322.863, Tokens/s/device: 199396.359, total_weights: 65536, loss: 10.568, lm_loss: 10.568, perplexity: 38879.223 I0421 16:45:40.396831 137912354039616 metric_logger.py:196] completed step: 16, seconds: 0.154, TFLOP/s/device: 88.194, Tokens/s/device: 13293.608, total_weights: 65536, loss: 10.562, lm_loss: 10.562, perplexity: 38619.105 I0421 16:45:40.512338 137777844840192 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/10/items I0421 16:45:40.553694 137912354039616 metric_logger.py:196] completed step: 17, seconds: 0.157, TFLOP/s/device: 86.695, Tokens/s/device: 13067.558, total_weights: 65536, loss: 10.556, lm_loss: 10.556, perplexity: 38395.816 I0421 16:45:40.710537 137912354039616 metric_logger.py:196] completed step: 18, seconds: 0.157, TFLOP/s/device: 86.697, Tokens/s/device: 13067.892, total_weights: 65536, loss: 10.552, lm_loss: 10.552, perplexity: 38236.652 I0421 16:45:40.869219 137912354039616 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0421 16:45:40.870481 137912354039616 checkpointing.py:798] Waited 0.0012738704681396484 seconds for step 19 to finish before starting checkpointing. I0421 16:45:40.872769 137912354039616 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 16:45:43.457007 137782274025216 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/10/items/array_metadatas/process_0 I0421 16:46:02.796718 137782265632512 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 21.125673s (commit=19.847866s, array_metadata_write=1.277807s) I0421 16:46:02.797991 137782659892992 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 64.425 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 24.520984649658203 s) (per-host) I0421 16:46:02.798120 137782659892992 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 24.457938s. I0421 16:46:04.217011 137782659892992 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.001032s. I0421 16:46:04.431556 137782659892992 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0421 16:46:05.531136 137782659892992 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.918803s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/10/items I0421 16:46:05.533001 137782659892992 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/10/items I0421 16:46:06.053134 137782659892992 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/10 I0421 16:46:07.435261 137782659892992 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/10`. I0421 16:46:07.436094 137782659892992 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 29.45 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/10 I0421 16:46:07.437467 137782659892992 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 29.097282s. I0421 16:46:07.437613 137781068347136 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 16:46:07.437689 137781068347136 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 16:46:07.437756 137781068347136 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0421 16:46:07.440501 137781068347136 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0421 16:46:07.440664 137912354039616 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0421 16:46:07.440788 137912354039616 checkpoint_manager.py:1452] Waiting for previous save to complete took 26.568024 seconds. If this number is high, consider checkpointing less frequently. I0421 16:46:07.442377 137912354039616 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 19 I0421 16:46:07.444378 137912354039616 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/19. I0421 16:46:08.168443 137912354039616 jax_array_handlers.py:360] Scheduling D2H of 153 prioritized jax.Array. I0421 16:46:08.168540 137912354039616 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0421 16:46:08.209304 137912354039616 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.043185s I0421 16:46:08.210023 137912354039616 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 25.391 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.060758352279663086 s) (per-host) I0421 16:46:08.210090 137912354039616 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.060844s (batch_requests_ready=0.005734s, total_serialization_initiated=0.054448s, others=0.000663s) I0421 16:46:08.210171 137912354039616 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.064747s (all_items=0.000012s, per_item={'items': '0.00001216'}, temp_paths=0.064735) I0421 16:46:08.210883 137912354039616 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.77 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/19. I0421 16:46:08.211159 137782266169088 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-21 17:06:08.211123 I0421 16:46:08.213145 137912354039616 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0421 16:46:08.213437 137781068347136 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 16:46:08.213573 137912354039616 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776789940.872739, 'wait_for_prev_duration_secs': 26.568024158477783, 'time_between_consecutive_saves_sec': 2.8853371143341064, 'checkpointer_blocking_start_time': 1776789967.4424171, 'checkpointer_blocking_duration_secs': 0.7688755989074707, 'get_old_steps_start_time': 1776789968.211312, 'get_old_steps_duration_secs': 2.5510787963867188e-05, 'checkpoint_manager_blocking_start_time': 1776789940.8710093, 'checkpoint_manager_blocking_duration_secs': 27.342531204223633} I0421 16:46:08.213766 137912354039616 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0421 16:46:08.213810 137912354039616 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 16:46:08.395439 137777282791168 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/19 I0421 16:46:10.212873 137782257239808 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/19/items I0421 16:46:13.911762 137782668285696 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/19/items/array_metadatas/process_0 I0421 16:46:32.667631 137782274561792 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 20.985107s (commit=19.605150s, array_metadata_write=1.379957s) I0421 16:46:32.668848 137782266169088 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 64.429 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 24.5195369720459 s) (per-host) I0421 16:46:32.668988 137782266169088 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 24.457717s. I0421 16:46:34.129058 137782266169088 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000937s. I0421 16:46:34.290556 137782266169088 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0421 16:46:35.379241 137782266169088 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.904730s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/19/items I0421 16:46:35.382326 137782266169088 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/19/items I0421 16:46:35.901203 137782266169088 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/19 I0421 16:46:37.288517 137782266169088 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/19`. I0421 16:46:37.289224 137782266169088 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 29.85 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900/nnx_xpk_feat_nnx_trainstate_and_training_loop_20260421_161900_05_fp8/checkpoints/19 I0421 16:46:37.290843 137782266169088 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 29.079570s. I0421 16:46:37.291038 137781068347136 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 16:46:37.291174 137781068347136 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 16:46:37.291227 137781068347136 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0421 16:46:37.294086 137781068347136 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0421 16:46:37.294270 137912354039616 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0421 16:46:37.294453 137912354039616 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 16:46:37.295390 137912354039616 metric_logger.py:196] completed step: 19, seconds: 0.157, TFLOP/s/device: 86.406, Tokens/s/device: 13024.096, total_weights: 65536, loss: 10.549, lm_loss: 10.549, perplexity: 38125.074 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Tue Apr 21 16:46:45 UTC 2026 EXIT_CODE=0