XPK Start: Thu Apr 23 15:26:11 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. 2026-04-23 15:26:36.395465: 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) I0423 15:26:36.605559 134871240152896 max_utils.py:273] Attempting to initialize the jax distributed system... I0423 15:26:45.646678 134871240152896 distributed.py:149] Starting JAX distributed service on [::]:8482 I0423 15:26:45.649167 134871240152896 distributed.py:172] Connecting to JAX distributed service on mt-15-ocdbt-false-sv-l8gh7-slice-job-0-0.mt-15-ocdbt-false-sv-l8gh7:8482 I0423 15:26:46.800383 134871240152896 max_utils.py:284] Jax distributed system initialized! I0423 15:26:52.803948 134871240152896 max_utils.py:800] System Information: Jax Version: 0.9.2 I0423 15:26:52.804055 134871240152896 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0423 15:26:52.804096 134871240152896 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 I0423 15:26:52.804134 134871240152896 train_utils.py:391] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0423 15:26:53.485777 134871240152896 maxtext_utils.py:1771] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 15:26:53.600395 134871240152896 checkpointing.py:688] Setting up checkpoint logger... I0423 15:26:53.600520 134871240152896 checkpointing.py:234] Creating checkpoint manager with ocdbt=False and zarr3=True I0423 15:26:53.600566 134871240152896 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0423 15:26:53.600780 134871240152896 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 0x7aa98ed1d220>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0423 15:26:56.635103 134871240152896 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0423 15:26:56.635341 134871240152896 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 0x7a9600587320>}, handler_registry=None I0423 15:26:56.635602 134871240152896 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a9600587320>` 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`. I0423 15:26:56.635653 134871240152896 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a95e06ef5f0>` 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`. I0423 15:26:56.635691 134871240152896 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 0x7a9600587320>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a9600587320>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a95e06ef5f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a95e06ef5f0>}). I0423 15:26:56.636019 134871240152896 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0423 15:26:56.636092 134871240152896 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7a95e06d36a0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0423 15:26:57.738845 134871240152896 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints I0423 15:26:58.194733 134871240152896 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/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7a95e06ee090> I0423 15:26:58.194915 134871240152896 checkpointing.py:302] Checkpoint manager created! I0423 15:26:59.136880 134871240152896 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0423 15:26:59.137003 134871240152896 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0423 15:27:00.500949 134871240152896 maxtext_utils.py:1889] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0423 15:27:00.501067 134871240152896 maxtext_utils.py:1889] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0423 15:27:00.501121 134871240152896 maxtext_utils.py:1889] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0423 15:27:00.501165 134871240152896 maxtext_utils.py:1889] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0423 15:27:00.501203 134871240152896 maxtext_utils.py:1889] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0423 15:27:00.501236 134871240152896 maxtext_utils.py:1889] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0423 15:27:00.501271 134871240152896 maxtext_utils.py:1889] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0423 15:27:00.501306 134871240152896 maxtext_utils.py:1889] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0423 15:27:00.501340 134871240152896 maxtext_utils.py:1889] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0423 15:27:00.501372 134871240152896 maxtext_utils.py:1889] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0423 15:27:00.501410 134871240152896 maxtext_utils.py:1889] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0423 15:27:00.501443 134871240152896 maxtext_utils.py:1889] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0423 15:27:00.907769 134871240152896 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0423 15:27:00.907862 134871240152896 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 15:27:00.913210 134871240152896 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0423 15:27:00.913268 134871240152896 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 15:27:00.930373 134871240152896 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 15:27:00.930440 134871240152896 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 15:27:00.946531 134871240152896 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 15:27:00.946597 134871240152896 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 15:27:00.972015 134871240152896 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 15:27:00.972084 134871240152896 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 15:27:00.988326 134871240152896 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 15:27:00.988393 134871240152896 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 15:27:01.004503 134871240152896 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 15:27:01.004566 134871240152896 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 15:27:01.034957 134871240152896 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0423 15:27:01.035026 134871240152896 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 15:27:01.055412 134871240152896 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0423 15:27:01.055474 134871240152896 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0423 15:27:12.302369 134871240152896 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. I0423 15:27:12.314125 134871240152896 metric_logger.py:301] number parameters: 1.104 billion I0423 15:27:23.827420 134871240152896 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0423 15:27:23.980296 134871240152896 checkpointing.py:798] Waited 0.1528606414794922 seconds for step 0 to finish before starting checkpointing. I0423 15:27:23.982551 134871240152896 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 15:27:23.984172 134871240152896 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0423 15:27:23.985728 134871240152896 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/0. I0423 15:27:24.713660 134871240152896 signaling_client.py:364] Using JaxDistributedSignalingClient I0423 15:27:24.734406 134871240152896 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0423 15:27:24.734501 134871240152896 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 I0423 15:27:24.963669 134746134800128 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/0 I0423 15:27:25.007737 134871240152896 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.274751s I0423 15:27:25.008538 134871240152896 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.396 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28585219383239746 s) (per-host) I0423 15:27:25.008628 134871240152896 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.285962s (batch_requests_ready=0.003354s, total_serialization_initiated=0.281870s, others=0.000738s) I0423 15:27:25.008711 134871240152896 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.290960s (all_items=0.000022s, per_item={'items': '0.00002241'}, temp_paths=0.290938) I0423 15:27:25.009551 134871240152896 event_tracking.py:125] [process=0] [async] Finished blocking save in 1.03 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/0. I0423 15:27:25.009871 134746143192832 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 15:47:25.009836 I0423 15:27:25.015944 134871240152896 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0423 15:27:25.016248 134744478037760 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 15:27:25.016407 134871240152896 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776958043.9825337, 'wait_for_prev_duration_secs': 7.033348083496094e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776958043.9842107, 'checkpointer_blocking_duration_secs': 1.0258045196533203, 'get_old_steps_start_time': 1776958045.0100408, 'get_old_steps_duration_secs': 3.0279159545898438e-05, 'checkpoint_manager_blocking_start_time': 1776958043.9807236, 'checkpoint_manager_blocking_duration_secs': 1.0356409549713135} I0423 15:27:25.016520 134871240152896 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0423 15:27:25.016595 134871240152896 max_utils.py:750] Memstats: After params initialized: I0423 15:27:25.016648 134871240152896 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0423 15:27:25.016680 134871240152896 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0423 15:27:25.016707 134871240152896 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0423 15:27:25.016731 134871240152896 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0423 15:27:25.362944 134871240152896 metric_logger.py:196] completed step: 0, seconds: 11.203, TFLOP/s/device: 1.213, Tokens/s/device: 182.802, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52786.469 I0423 15:27:25.366191 134871240152896 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/tensorboard/' I0423 15:27:25.790365 134871240152896 metric_logger.py:196] completed step: 1, seconds: 1.532, TFLOP/s/device: 8.870, Tokens/s/device: 1336.921, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52786.469 I0423 15:27:25.932096 134871240152896 metric_logger.py:196] completed step: 2, seconds: 0.428, TFLOP/s/device: 31.726, Tokens/s/device: 4782.075, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38565.637 I0423 15:27:26.079715 134871240152896 metric_logger.py:196] completed step: 3, seconds: 0.013, TFLOP/s/device: 1069.516, Tokens/s/device: 161209.068, total_weights: 65536, loss: 9.979, lm_loss: 9.979, perplexity: 21564.932 I0423 15:27:26.092189 134871240152896 checkpointing.py:794] Waiting for step 5 to finish before checkpoint... I0423 15:27:26.374958 134871240152896 checkpointing.py:798] Waited 0.28273653984069824 seconds for step 5 to finish before starting checkpointing. I0423 15:27:26.377787 134871240152896 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 15:27:26.785099 134746118014720 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/0/items I0423 15:27:26.950440 134741848225536 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776958046517794873, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776958046517794873, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/0/_CHECKPOINT_METADATA I0423 15:27:27.955309 2781 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0423 15:27:29.722429 134744494823168 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/0/items/array_metadatas/process_0 I0423 15:27:38.483304 134744486430464 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 10.535370s (commit=9.476315s, array_metadata_write=1.059054s) I0423 15:27:38.484737 134746143192832 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 114.780 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 13.762009620666504 s) (per-host) I0423 15:27:38.484904 134746143192832 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 13.474920s. I0423 15:27:39.141489 134746143192832 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776958046517794873, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/0/_CHECKPOINT_METADATA W0423 15:27:39.908882 134746143192832 ocdbt_utils.py:168] [process=0][thread=async_save] Skipping merge of OCDBT checkpoints: No per-process OCDBT checkpoint subdirs found in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/0/items, I0423 15:27:39.932229 134741848225536 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776958046517794873, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/0/_CHECKPOINT_METADATA I0423 15:27:40.730317 134746143192832 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000807s. I0423 15:27:40.730621 134746143192832 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.466804s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/0/items I0423 15:27:40.732486 134746143192832 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/0/items I0423 15:27:41.267301 134746143192832 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/0 I0423 15:27:42.657246 134746143192832 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/0`. I0423 15:27:42.658391 134746143192832 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 18.67 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/0 I0423 15:27:42.660065 134746143192832 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 17.650078s. I0423 15:27:42.660344 134744478037760 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 15:27:42.660536 134744478037760 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 15:27:42.660659 134744478037760 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0423 15:27:42.663439 134744478037760 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0423 15:27:42.663719 134871240152896 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. W0423 15:27:42.663831 134871240152896 checkpoint_manager.py:1452] Waiting for previous save to complete took 16.286061 seconds. If this number is high, consider checkpointing less frequently. I0423 15:27:42.665920 134871240152896 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 5 I0423 15:27:42.668601 134871240152896 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/5. I0423 15:27:42.974898 134871240152896 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0423 15:27:42.975031 134871240152896 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 I0423 15:27:43.012214 134871240152896 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.040305s I0423 15:27:43.013291 134871240152896 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 25.513 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.060461997985839844 s) (per-host) I0423 15:27:43.013410 134871240152896 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.060606s (batch_requests_ready=0.006684s, total_serialization_initiated=0.052940s, others=0.000982s) I0423 15:27:43.013553 134871240152896 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.065550s (all_items=0.000026s, per_item={'items': '0.00002551'}, temp_paths=0.065524) I0423 15:27:43.014527 134871240152896 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.35 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/5. I0423 15:27:43.014995 134744478037760 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 15:47:43.014952 I0423 15:27:43.017600 134871240152896 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0423 15:27:43.017998 134743950358272 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 15:27:43.018236 134871240152896 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776958046.3777447, 'wait_for_prev_duration_secs': 16.286060571670532, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776958062.666011, 'checkpointer_blocking_duration_secs': 0.34917712211608887, 'get_old_steps_start_time': 1776958063.0152152, 'get_old_steps_duration_secs': 4.38690185546875e-05, 'checkpoint_manager_blocking_start_time': 1776958046.375447, 'checkpoint_manager_blocking_duration_secs': 16.642733812332153} I0423 15:27:43.018389 134871240152896 checkpointing.py:409] Started an asynchronous checkpoint save for step 5 I0423 15:27:43.201662 134746143192832 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/5 I0423 15:27:45.027515 134745052665600 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/5/items I0423 15:27:48.359316 134743967143680 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/5/items/array_metadatas/process_0 I0423 15:27:56.749846 134743958750976 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 10.600781s (commit=9.069212s, array_metadata_write=1.531569s) I0423 15:27:56.750930 134744478037760 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 114.480 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 13.798088073730469 s) (per-host) I0423 15:27:56.751046 134744478037760 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 13.735898s. W0423 15:27:58.401957 134744478037760 ocdbt_utils.py:168] [process=0][thread=async_save] Skipping merge of OCDBT checkpoints: No per-process OCDBT checkpoint subdirs found in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/5/items, I0423 15:27:59.249726 134744478037760 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000462s. I0423 15:27:59.249949 134744478037760 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.673948s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/5/items I0423 15:27:59.251521 134744478037760 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/5/items I0423 15:28:00.155965 134744478037760 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/5 I0423 15:28:01.575319 134744478037760 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/5`. I0423 15:28:01.576173 134744478037760 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 18.91 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/5 I0423 15:28:01.577791 134744478037760 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 18.562642s. I0423 15:28:01.578028 134743950358272 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 15:28:01.578170 134743950358272 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 15:28:01.578258 134743950358272 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0423 15:28:01.580676 134743950358272 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0423 15:28:06.074356 134871240152896 metric_logger.py:196] completed step: 4, seconds: 0.144, TFLOP/s/device: 94.674, Tokens/s/device: 14270.285, total_weights: 65536, loss: 9.457, lm_loss: 9.457, perplexity: 12802.436 I0423 15:28:06.089471 134871240152896 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.569, Tokens/s/device: 13952.950, total_weights: 65536, loss: 8.955, lm_loss: 8.955, perplexity: 7748.778 I0423 15:28:06.229945 134871240152896 metric_logger.py:196] completed step: 6, seconds: 39.997, TFLOP/s/device: 0.340, Tokens/s/device: 51.204, total_weights: 65536, loss: 8.465, lm_loss: 8.465, perplexity: 4743.984 I0423 15:28:06.377402 134871240152896 metric_logger.py:196] completed step: 7, seconds: 0.009, TFLOP/s/device: 1506.501, Tokens/s/device: 227076.173, total_weights: 65536, loss: 7.998, lm_loss: 7.998, perplexity: 2976.266 I0423 15:28:06.525297 134871240152896 metric_logger.py:196] completed step: 8, seconds: 0.142, TFLOP/s/device: 95.600, Tokens/s/device: 14409.850, total_weights: 65536, loss: 7.567, lm_loss: 7.567, perplexity: 1933.479 I0423 15:28:06.673267 134871240152896 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0423 15:28:06.674005 134871240152896 checkpointing.py:798] Waited 0.0007500648498535156 seconds for step 9 to finish before starting checkpointing. I0423 15:28:06.675970 134871240152896 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 15:28:06.677630 134871240152896 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0423 15:28:06.679172 134871240152896 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/9. I0423 15:28:06.994863 134871240152896 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0423 15:28:06.994959 134871240152896 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 I0423 15:28:07.021550 134871240152896 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.028137s I0423 15:28:07.022336 134871240152896 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 41.819 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.036887407302856445 s) (per-host) I0423 15:28:07.022409 134871240152896 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.036978s (batch_requests_ready=0.002794s, total_serialization_initiated=0.033450s, others=0.000734s) I0423 15:28:07.022483 134871240152896 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.040901s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.040884) I0423 15:28:07.023192 134871240152896 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.35 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/9. I0423 15:28:07.023521 134743950358272 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 15:48:07.023484 I0423 15:28:07.029790 134871240152896 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0423 15:28:07.030085 134743434393344 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 15:28:07.030251 134871240152896 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776958086.675937, 'wait_for_prev_duration_secs': 7.295608520507812e-05, 'time_between_consecutive_saves_sec': 5.09521222114563, 'checkpointer_blocking_start_time': 1776958086.6776702, 'checkpointer_blocking_duration_secs': 0.34596753120422363, 'get_old_steps_start_time': 1776958087.0236566, 'get_old_steps_duration_secs': 2.7179718017578125e-05, 'checkpoint_manager_blocking_start_time': 1776958086.6742492, 'checkpoint_manager_blocking_duration_secs': 0.3559684753417969} I0423 15:28:07.030363 134871240152896 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0423 15:28:07.030409 134871240152896 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 15:28:07.229142 134744478037760 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/9 I0423 15:28:09.003139 134745052690176 atomicity.py:140] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/9/items I0423 15:28:12.755495 134743967143680 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/9/items/array_metadatas/process_0 I0423 15:28:20.899429 134743958750976 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 10.359052s (commit=8.991855s, array_metadata_write=1.367197s) I0423 15:28:20.900596 134743950358272 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 113.517 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 13.915092706680298 s) (per-host) I0423 15:28:20.900704 134743950358272 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 13.876539s. W0423 15:28:22.339724 134743950358272 ocdbt_utils.py:168] [process=0][thread=async_save] Skipping merge of OCDBT checkpoints: No per-process OCDBT checkpoint subdirs found in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/9/items, I0423 15:28:22.789837 134743950358272 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000448s. I0423 15:28:22.790067 134743950358272 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.094699s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/9/items I0423 15:28:22.791471 134743950358272 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/9/items I0423 15:28:23.568504 134743950358272 atomicity.py:666] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/9 I0423 15:28:24.905032 134743950358272 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/9`. I0423 15:28:24.905871 134743950358272 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 18.23 seconds @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/nnx_xpk_feat_nnx_post_train_fixes_20260423_124541_15_ocdbt_false/checkpoints/9 I0423 15:28:24.907540 134743950358272 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 17.883372s. I0423 15:28:24.907752 134743434393344 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 15:28:24.907867 134743434393344 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 15:28:24.907918 134743434393344 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0423 15:28:24.910278 134743434393344 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0423 15:28:24.910415 134871240152896 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. I0423 15:28:24.910521 134871240152896 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 15:28:24.911595 134871240152896 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.194, Tokens/s/device: 13896.428, total_weights: 65536, loss: 7.176, lm_loss: 7.176, perplexity: 1307.366 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Thu Apr 23 15:28:36 UTC 2026 EXIT_CODE=0