XPK Start: Sun Apr 19 01:04:41 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-19 01:05:05.866717: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0419 01:05:06.072859 140227334051648 max_utils.py:273] Attempting to initialize the jax distributed system... I0419 01:05:15.113605 140227334051648 distributed.py:149] Starting JAX distributed service on [::]:8482 I0419 01:05:15.115913 140227334051648 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-true-sv-d20cg-slice-job-0-0.mt-16-zarr3-true-sv-d20cg:8482 I0419 01:05:17.072883 140227334051648 max_utils.py:284] Jax distributed system initialized! I0419 01:05:23.209570 140227334051648 max_utils.py:800] System Information: Jax Version: 0.9.2 I0419 01:05:23.209671 140227334051648 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0419 01:05:23.209711 140227334051648 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0419 01:05:23.209748 140227334051648 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0419 01:05:24.254890 140227334051648 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 01:05:24.255499 140227334051648 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 01:05:24.255677 140227334051648 checkpointing.py:688] Setting up checkpoint logger... I0419 01:05:24.255728 140227334051648 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0419 01:05:24.255773 140227334051648 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0419 01:05:24.256123 140227334051648 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 0x7f8880c501d0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0419 01:05:27.499982 140227334051648 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0419 01:05:27.500212 140227334051648 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 0x7f877484f950>}, handler_registry=None I0419 01:05:27.500457 140227334051648 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f877484f950>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0419 01:05:27.500506 140227334051648 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f877484e420>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0419 01:05:27.500542 140227334051648 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 0x7f877484f950>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f877484f950>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f877484e420>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f877484e420>}). I0419 01:05:27.500869 140227334051648 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0419 01:05:27.500939 140227334051648 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f74347d2c00> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0419 01:05:29.154170 140227334051648 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints I0419 01:05:29.170003 140227334051648 checkpoint_manager.py:929] [process=0][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f877484ebd0> I0419 01:05:29.170124 140227334051648 checkpointing.py:302] Checkpoint manager created! I0419 01:05:30.085786 140227334051648 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 01:05:30.085908 140227334051648 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 01:05:30.470465 140227334051648 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 01:05:30.470556 140227334051648 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 01:05:30.487095 140227334051648 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 01:05:30.487152 140227334051648 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 01:05:30.510792 140227334051648 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 01:05:30.510872 140227334051648 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 01:05:30.527338 140227334051648 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 01:05:30.527401 140227334051648 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 01:05:30.543957 140227334051648 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 01:05:30.544020 140227334051648 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 01:05:30.569763 140227334051648 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0419 01:05:30.569841 140227334051648 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 01:05:30.590748 140227334051648 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0419 01:05:30.590831 140227334051648 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0419 01:05:30.806747 140227334051648 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0419 01:05:30.806911 140227334051648 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0419 01:05:32.249380 140227334051648 maxtext_utils.py:1827] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0419 01:05:32.249508 140227334051648 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0419 01:05:32.249565 140227334051648 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0419 01:05:32.249626 140227334051648 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0419 01:05:32.249686 140227334051648 maxtext_utils.py:1827] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0419 01:05:32.249727 140227334051648 maxtext_utils.py:1827] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0419 01:05:32.249785 140227334051648 maxtext_utils.py:1827] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0419 01:05:32.249902 140227334051648 maxtext_utils.py:1827] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0419 01:05:32.249952 140227334051648 maxtext_utils.py:1827] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0419 01:05:32.249992 140227334051648 maxtext_utils.py:1827] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0419 01:05:32.250044 140227334051648 maxtext_utils.py:1827] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0419 01:05:32.250096 140227334051648 maxtext_utils.py:1827] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0419 01:05:32.743291 140227334051648 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 01:05:32.743388 140227334051648 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0419 01:05:32.758730 140227334051648 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 01:05:32.758792 140227334051648 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0419 01:05:43.694640 140227334051648 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0419 01:05:43.702669 140227334051648 metric_logger.py:301] number parameters: 1.104 billion I0419 01:05:55.526988 140227334051648 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0419 01:05:55.680938 140227334051648 checkpointing.py:798] Waited 0.15393304824829102 seconds for step 0 to finish before starting checkpointing. I0419 01:05:55.683322 140227334051648 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 01:05:55.685200 140227334051648 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0419 01:05:55.686669 140227334051648 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/0. I0419 01:05:56.000829 140227334051648 signaling_client.py:364] Using JaxDistributedSignalingClient I0419 01:05:56.016789 140227334051648 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 01:05:56.016898 140227334051648 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0419 01:05:56.289344 140227334051648 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.273305s I0419 01:05:56.290205 140227334051648 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.508 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.280078649520874 s) (per-host) I0419 01:05:56.290285 140227334051648 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.280180s (batch_requests_ready=0.002221s, total_serialization_initiated=0.277181s, others=0.000777s) I0419 01:05:56.290371 140227334051648 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.285150s (all_items=0.000019s, per_item={'items': '0.00001931'}, temp_paths=0.285130) I0419 01:05:56.291157 140227334051648 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/0. I0419 01:05:56.291459 140098410641152 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 01:25:56.291426 I0419 01:05:56.298972 140227334051648 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0419 01:05:56.299250 140096745486080 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 01:05:56.299405 140227334051648 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776560755.6833043, 'wait_for_prev_duration_secs': 6.794929504394531e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776560755.6852367, 'checkpointer_blocking_duration_secs': 0.6063218116760254, 'get_old_steps_start_time': 1776560756.291583, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1776560755.6814783, 'checkpoint_manager_blocking_duration_secs': 0.6178886890411377} I0419 01:05:56.299506 140227334051648 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0419 01:05:56.299555 140227334051648 max_utils.py:750] Memstats: After params initialized: I0419 01:05:56.299604 140227334051648 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0419 01:05:56.299639 140227334051648 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0419 01:05:56.299667 140227334051648 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0419 01:05:56.299692 140227334051648 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0419 01:05:56.361074 140098402248448 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/0 I0419 01:05:56.659094 140227334051648 metric_logger.py:196] completed step: 0, seconds: 11.516, TFLOP/s/device: 1.180, Tokens/s/device: 177.837, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0419 01:05:56.662609 140227334051648 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/tensorboard/' I0419 01:05:57.071640 140227334051648 metric_logger.py:196] completed step: 1, seconds: 1.130, TFLOP/s/device: 12.022, Tokens/s/device: 1812.017, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0419 01:05:57.217225 140227334051648 metric_logger.py:196] completed step: 2, seconds: 0.413, TFLOP/s/device: 32.896, Tokens/s/device: 4958.382, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0419 01:05:57.364744 140227334051648 metric_logger.py:196] completed step: 3, seconds: 0.011, TFLOP/s/device: 1225.833, Tokens/s/device: 184770.841, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0419 01:05:57.373409 140227334051648 checkpointing.py:794] Waiting for step 5 to finish before checkpoint... I0419 01:05:57.658964 140227334051648 checkpointing.py:798] Waited 0.2853844165802002 seconds for step 5 to finish before starting checkpointing. I0419 01:05:57.661874 140227334051648 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 01:05:58.790098 140098385463040 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/0/items I0419 01:05:58.937008 140095720175360 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776560758537916969, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776560758537916969, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0419 01:06:00.368228 2834 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0419 01:06:02.561267 140096762271488 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/0/items/array_metadatas/process_0 I0419 01:06:38.847524 140096753878784 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 38.481981s (commit=36.858037s, array_metadata_write=1.623943s) I0419 01:06:38.849153 140098410641152 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 36.873 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.8389630317688 s) (per-host) I0419 01:06:38.849324 140098410641152 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.557227s. I0419 01:06:39.594828 140098410641152 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776560758537916969, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0419 01:06:40.354879 140098410641152 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000413s. I0419 01:06:40.388601 140095720175360 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776560758537916969, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0419 01:06:40.989993 140098410641152 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0419 01:06:41.902108 140098410641152 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.164515s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/0/items I0419 01:06:41.904231 140098410641152 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/0/items I0419 01:06:42.866967 140098410641152 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/0 I0419 01:06:44.234719 140098410641152 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/0`. I0419 01:06:44.235838 140098410641152 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 48.55 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/0 I0419 01:06:44.237488 140098410641152 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 47.945387s. I0419 01:06:44.237715 140096745486080 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 01:06:44.237864 140096745486080 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 01:06:44.237978 140096745486080 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0419 01:06:44.240896 140096745486080 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0419 01:06:44.241189 140227334051648 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0419 01:06:44.241367 140227334051648 checkpoint_manager.py:1452] Waiting for previous save to complete took 46.579506 seconds. If this number is high, consider checkpointing less frequently. I0419 01:06:44.243319 140227334051648 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 5 I0419 01:06:44.245704 140227334051648 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/5. I0419 01:06:44.959563 140227334051648 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 01:06:44.959698 140227334051648 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0419 01:06:44.996441 140227334051648 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.037904s I0419 01:06:44.997415 140227334051648 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 32.899 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.046888113021850586 s) (per-host) I0419 01:06:44.997521 140227334051648 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.047019s (batch_requests_ready=0.004352s, total_serialization_initiated=0.041795s, others=0.000872s) I0419 01:06:44.997654 140227334051648 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.052083s (all_items=0.000025s, per_item={'items': '0.00002456'}, temp_paths=0.052058) I0419 01:06:44.998648 140227334051648 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.76 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/5. I0419 01:06:44.999049 140096745486080 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 01:26:44.999012 I0419 01:06:45.002854 140227334051648 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0419 01:06:45.003264 140096774850304 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 01:06:45.003505 140227334051648 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776560757.6618016, 'wait_for_prev_duration_secs': 46.579506397247314, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776560804.243391, 'checkpointer_blocking_duration_secs': 0.7558577060699463, 'get_old_steps_start_time': 1776560804.9992752, 'get_old_steps_duration_secs': 4.291534423828125e-05, 'checkpoint_manager_blocking_start_time': 1776560757.6596124, 'checkpoint_manager_blocking_duration_secs': 47.34383797645569} I0419 01:06:45.003655 140227334051648 checkpointing.py:409] Started an asynchronous checkpoint save for step 5 I0419 01:06:45.192841 140098410641152 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/5 I0419 01:06:46.944507 140098385463040 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/5/items I0419 01:06:51.011902 140096791635712 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/5/items/array_metadatas/process_0 I0419 01:07:07.250744 140227334051648 metric_logger.py:196] completed step: 4, seconds: 0.145, TFLOP/s/device: 93.791, Tokens/s/device: 14137.203, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0419 01:07:07.260057 140227334051648 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.382, Tokens/s/device: 13924.868, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0419 01:07:07.405592 140227334051648 metric_logger.py:196] completed step: 6, seconds: 69.885, TFLOP/s/device: 0.194, Tokens/s/device: 29.305, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0419 01:07:07.553066 140227334051648 metric_logger.py:196] completed step: 7, seconds: 0.007, TFLOP/s/device: 1913.141, Tokens/s/device: 288369.473, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0419 01:07:07.700509 140227334051648 metric_logger.py:196] completed step: 8, seconds: 0.147, TFLOP/s/device: 92.438, Tokens/s/device: 13933.300, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0419 01:07:07.847230 140227334051648 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0419 01:07:07.848069 140227334051648 checkpointing.py:798] Waited 0.0008556842803955078 seconds for step 9 to finish before starting checkpointing. I0419 01:07:07.850327 140227334051648 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 01:07:30.430776 140096783243008 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.012283s (commit=40.835832s, array_metadata_write=1.176450s) I0419 01:07:30.431913 140096745486080 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.731 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.48135733604431 s) (per-host) I0419 01:07:30.432027 140096745486080 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.432834s. I0419 01:07:32.540210 140096745486080 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0419 01:07:32.734959 140096745486080 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000241s. I0419 01:07:33.589490 140096745486080 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.908245s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/5/items I0419 01:07:33.591232 140096745486080 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/5/items I0419 01:07:34.139298 140096745486080 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/5 I0419 01:07:36.001355 140096745486080 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/5`. I0419 01:07:36.002270 140096745486080 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 51.76 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/5 I0419 01:07:36.003851 140096745486080 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 51.004656s. I0419 01:07:36.004072 140096774850304 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 01:07:36.004190 140096774850304 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 01:07:36.004257 140096774850304 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0419 01:07:36.006939 140096774850304 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0419 01:07:36.007121 140227334051648 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0419 01:07:36.007254 140227334051648 checkpoint_manager.py:1452] Waiting for previous save to complete took 28.156933 seconds. If this number is high, consider checkpointing less frequently. I0419 01:07:36.008702 140227334051648 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0419 01:07:36.010723 140227334051648 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/9. I0419 01:07:36.730041 140227334051648 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 01:07:36.730134 140227334051648 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0419 01:07:36.756200 140227334051648 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.027138s I0419 01:07:36.756942 140227334051648 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 48.948 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.031514883041381836 s) (per-host) I0419 01:07:36.757004 140227334051648 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.031595s (batch_requests_ready=0.001861s, total_serialization_initiated=0.029048s, others=0.000686s) I0419 01:07:36.757075 140227334051648 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.035435s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.035419) I0419 01:07:36.757796 140227334051648 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/9. I0419 01:07:36.758072 140096774850304 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 01:27:36.758049 I0419 01:07:36.772689 140227334051648 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0419 01:07:36.772972 140097873770240 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 01:07:36.773128 140227334051648 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776560827.8502913, 'wait_for_prev_duration_secs': 28.156932830810547, 'time_between_consecutive_saves_sec': 23.60932445526123, 'checkpointer_blocking_start_time': 1776560856.0087428, 'checkpointer_blocking_duration_secs': 0.7494275569915771, 'get_old_steps_start_time': 1776560856.7581966, 'get_old_steps_duration_secs': 2.7418136596679688e-05, 'checkpoint_manager_blocking_start_time': 1776560827.8483756, 'checkpoint_manager_blocking_duration_secs': 28.924721717834473} I0419 01:07:36.773246 140227334051648 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0419 01:07:36.773288 140227334051648 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 01:07:36.979389 140096745486080 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/9 I0419 01:07:38.721424 140094077499136 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/9/items I0419 01:07:41.795342 140098410641152 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/9/items/array_metadatas/process_0 I0419 01:08:22.152867 140098385463040 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.305246s (commit=41.116753s, array_metadata_write=1.188492s) I0419 01:08:22.153999 140096774850304 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.771 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.42856502532959 s) (per-host) I0419 01:08:22.154070 140096774850304 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.395921s. I0419 01:08:24.457993 140096774850304 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0419 01:08:24.460262 140096774850304 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000241s. I0419 01:08:25.076657 140096774850304 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.643493s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/9/items I0419 01:08:25.078485 140096774850304 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/9/items I0419 01:08:25.625138 140096774850304 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/9 I0419 01:08:27.074297 140096774850304 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/9`. I0419 01:08:27.075177 140096774850304 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 51.07 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_16_zarr3_true/checkpoints/9 I0419 01:08:27.076660 140096774850304 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 50.318508s. I0419 01:08:27.076793 140097873770240 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 01:08:27.076869 140097873770240 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 01:08:27.076927 140097873770240 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0419 01:08:27.078766 140097873770240 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0419 01:08:27.078923 140227334051648 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0419 01:08:27.079013 140227334051648 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 01:08:27.079990 140227334051648 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.300, Tokens/s/device: 13912.476, total_weights: 65536, loss: 7.181, lm_loss: 7.181, perplexity: 1314.495 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sun Apr 19 01:08:37 UTC 2026 EXIT_CODE=0