2026-04-21 16:08:46.638356: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered WARNING: All log messages before absl::InitializeLog() is called are written to STDERR E0000 00:00:1776787726.649218 1551961 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776787726.652392 1551961 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776787726.663264 1551961 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776787726.663275 1551961 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776787726.663277 1551961 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776787726.663278 1551961 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-21 16:08:50.028075: 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 16:08:50.144751 132411908254848 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. ~/maxtext/maxtext_venv/lib/python3.12/site-packages/jax/_src/xla_bridge.py:204: UserWarning: TPU backend initialization is taking more than 60.0 seconds. Did you run your code on all TPU hosts? See https://docs.jax.dev/en/latest/multi_process.html for more information. warnings.warn( I0421 16:09:53.080490 132411908254848 max_utils.py:800] System Information: Jax Version: 0.8.1 I0421 16:09:53.080672 132411908254848 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0421 16:09:53.080713 132411908254848 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709 I0421 16:09:53.080741 132411908254848 train_utils.py:335] WARNING: 'dataset_path' might be pointing your local file system I0421 16:09:53.080763 132411908254848 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0421 16:09:53.080863 132411908254848 train.py:703] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0421 16:09:53.439562 132411908254848 maxtext_utils.py:1551] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 16:09:53.440026 132411908254848 checkpointing.py:677] Setting up checkpoint logger... I0421 16:09:53.440089 132411908254848 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=False I0421 16:09:53.440179 132411908254848 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0421 16:09:53.441022 132411908254848 base_pytree_checkpoint_handler.py:411] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x786d431af170>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 16:09:55.096856 132411908254848 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0421 16:09:55.097131 132411908254848 checkpoint_manager.py:702] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x786b9a377380>}, handler_registry=None I0421 16:09:55.097521 132411908254848 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x786b9a377380>` 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 16:09:55.097565 132411908254848 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x786bec1505c0>` 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 16:09:55.097607 132411908254848 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 0x786b9a377380>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x786b9a377380>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x786bec1505c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x786bec1505c0>}). I0421 16:09:55.097901 132411908254848 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0421 16:09:55.097963 132411908254848 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x786b9a3ad300> timeout: 600 secs and primary_host=0 for async checkpoint writes I0421 16:09:55.535646 132328269547072 checkpoint.py:228] Read Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776787540199141590, 'commit_timestamp_nsecs': 1776787583599797018, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/0/_CHECKPOINT_METADATA I0421 16:09:55.626330 132328269547072 checkpoint_manager.py:1847] Missing metrics for step 0 I0421 16:09:55.657443 132411908254848 checkpoint_manager.py:1788] Found 3 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints I0421 16:09:55.657950 132411908254848 checkpoint_manager.py:921] [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_hns=False, 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), root_directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x786b9a73bc20> I0421 16:09:55.658067 132411908254848 checkpointing.py:301] Checkpoint manager created! I0421 16:09:56.294203 132411908254848 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 16:09:56.294289 132411908254848 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 16:09:56.589661 132411908254848 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 16:09:56.589740 132411908254848 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 16:09:56.601637 132411908254848 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 16:09:56.601687 132411908254848 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 16:09:56.620363 132411908254848 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 16:09:56.620417 132411908254848 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:09:56.632444 132411908254848 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 16:09:56.632498 132411908254848 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:09:56.644332 132411908254848 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0421 16:09:56.644379 132411908254848 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:09:56.663413 132411908254848 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 16:09:56.663465 132411908254848 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:09:56.678572 132411908254848 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 16:09:56.678624 132411908254848 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0421 16:09:56.826363 132411908254848 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0421 16:09:56.826455 132411908254848 checkpointing.py:581] restoring from this run's directory step 9 I0421 16:09:57.003161 132411908254848 checkpointer.py:304] Restoring checkpoint from gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/9. I0421 16:09:57.530622 1553496 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0421 16:09:57.530689 1553496 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0421 16:10:04.228175 132411908254848 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/read/gbytes_per_sec: 1.792 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host) I0421 16:10:04.228450 132411908254848 checkpointer.py:318] Finished restoring checkpoint in 7.31 seconds from gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/9. I0421 16:10:04.228526 132411908254848 standard_logger.py:34] {'step': 9, 'event_type': 'restore', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints', 'checkpointer_start_time': 1776787796.9179003, 'checkpointer_duration_secs': 7.310598134994507, 'checkpoint_manager_start_time': 1776787796.8266597, 'checkpoint_manager_duration_secs': 7.40184211730957} [DECOUPLED NO-OP] gcs_storage: using stubs. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] workload_monitor: using stub. [DECOUPLED NO-OP] vertex_tensorboard: using stub. fsdp: 8 I0421 16:10:04.553091 132411908254848 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0421 16:10:04.553199 132411908254848 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0421 16:10:04.553241 132411908254848 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0421 16:10:04.553284 132411908254848 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0421 16:10:04.553322 132411908254848 maxtext_utils.py:1654] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0421 16:10:04.553350 132411908254848 maxtext_utils.py:1654] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0421 16:10:04.553390 132411908254848 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0421 16:10:04.553427 132411908254848 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: PartitionSpec('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0421 16:10:04.553456 132411908254848 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0421 16:10:04.553482 132411908254848 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0421 16:10:04.553516 132411908254848 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0421 16:10:04.553551 132411908254848 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0421 16:10:04.951210 132411908254848 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 16:10:04.951291 132411908254848 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0421 16:10:04.962534 132411908254848 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 16:10:04.962583 132411908254848 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0421 16:10:05.586605 132411908254848 max_utils.py:791] Total memory size: 3.6 GB, Output size: 1.5 GB, Temp size: 2.0 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0421 16:10:05.590229 132411908254848 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0421 16:10:05.590425 132411908254848 metric_logger.py:301] number parameters: 1.104 billion I0421 16:10:07.623090 132411908254848 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0421 16:10:07.740020 132411908254848 checkpointing.py:776] Waited 0.11691665649414062 seconds for step 10 to finish before starting checkpointing. I0421 16:10:07.740451 132411908254848 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 16:10:07.740621 132411908254848 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 10 I0421 16:10:07.740930 132411908254848 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/10. I0421 16:10:07.835032 132411908254848 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0421 16:10:07.913587 132327592166976 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/10 I0421 16:10:07.933537 132411908254848 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 16:10:07.933623 132411908254848 replica_slices.py:410] 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 16:10:08.619133 132327581681216 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/10/items I0421 16:10:08.776476 132327550223936 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776787808532229772, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776787808532229772, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/10/_CHECKPOINT_METADATA I0421 16:10:09.924329 132411908254848 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.991183s I0421 16:10:09.924889 132411908254848 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.909 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 2 seconds) (per-host) I0421 16:10:09.924947 132411908254848 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 2.088640s (batch_requests_ready=0.094243s, total_serialization_initiated=1.993943s, others=0.000455s) I0421 16:10:09.925014 132411908254848 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.089534s (all_items=0.000021s, per_item={'items': '0.00002098'}, temp_paths=2.089513) I0421 16:10:09.929805 132327602652736 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 16:10:09.929901 132411908254848 async_checkpointer.py:561] Finished blocking save. Time taken: 2.189240s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/10. I0421 16:10:09.930103 132411908254848 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0421 16:10:09.930263 132327571195456 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 16:10:09.930357 132411908254848 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776787807.7404382, 'wait_for_prev_duration_secs': 3.790855407714844e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776787807.7406418, 'checkpointer_blocking_duration_secs': 2.189371109008789, 'get_old_steps_start_time': 1776787809.9300258, 'get_old_steps_duration_secs': 3.886222839355469e-05, 'checkpoint_manager_blocking_start_time': 1776787807.7403572, 'checkpoint_manager_blocking_duration_secs': 2.189974069595337} I0421 16:10:09.930463 132411908254848 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0421 16:10:09.930499 132411908254848 max_utils.py:750] Memstats: After params initialized: I0421 16:10:09.930545 132411908254848 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0421 16:10:09.930573 132411908254848 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0421 16:10:09.930593 132411908254848 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0421 16:10:09.930617 132411908254848 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0421 16:10:09.930634 132411908254848 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0421 16:10:09.930652 132411908254848 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0421 16:10:09.930669 132411908254848 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0421 16:10:09.930687 132411908254848 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0421 16:10:10.184108 132411908254848 metric_logger.py:196] completed step: 10, seconds: 2.032, TFLOP/s/device: 6.686, Tokens/s/device: 1007.848, total_weights: 16384, loss: 3.602, lm_loss: 3.602, perplexity: 36.686 I0421 16:10:10.301196 132411908254848 metric_logger.py:196] completed step: 11, seconds: 2.559, TFLOP/s/device: 5.309, Tokens/s/device: 800.212, total_weights: 16384, loss: 3.101, lm_loss: 3.101, perplexity: 22.211 I0421 16:10:10.420575 132411908254848 metric_logger.py:196] completed step: 12, seconds: 0.005, TFLOP/s/device: 2520.803, Tokens/s/device: 379962.894, total_weights: 16384, loss: 2.684, lm_loss: 2.684, perplexity: 14.644 I0421 16:10:10.539530 132411908254848 metric_logger.py:196] completed step: 13, seconds: 0.118, TFLOP/s/device: 115.192, Tokens/s/device: 17362.995, total_weights: 16384, loss: 2.348, lm_loss: 2.348, perplexity: 10.461 I0421 16:10:10.544337 132411908254848 checkpointing.py:772] Waiting for step 15 to finish before checkpoint... I0421 16:10:10.692290 132327560709696 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/10/items/array_metadatas/process_0 I0421 16:10:10.776747 132411908254848 checkpointing.py:776] Waited 0.2323770523071289 seconds for step 15 to finish before starting checkpointing. I0421 16:10:10.777158 132411908254848 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 16:10:46.983179 132327581681216 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 37.054144s (commit=36.597314s, array_metadata_write=0.456830s) I0421 16:10:46.984323 132327602652736 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 322.795 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 39 seconds) (per-host) I0421 16:10:46.984429 132327602652736 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.054484s. I0421 16:10:47.433027 132327602652736 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0421 16:10:47.599755 132327550223936 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776787808532229772, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/10/_CHECKPOINT_METADATA I0421 16:10:48.009424 132327602652736 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.743485s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/10/items I0421 16:10:48.010125 132327602652736 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/10/items I0421 16:10:48.252249 132327602652736 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/10 I0421 16:10:48.944989 132327602652736 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/10`. I0421 16:10:48.945635 132327602652736 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 41.204981s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/10 I0421 16:10:48.945712 132327602652736 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 39.015768s. I0421 16:10:48.945896 132327571195456 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 16:10:48.946015 132327571195456 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 16:10:48.946096 132327571195456 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0421 16:10:48.946145 132327571195456 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0421 16:10:48.946999 132411908254848 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0421 16:10:48.947708 132411908254848 checkpoint_manager.py:1441] Waiting for previous save to complete took 38.170545 seconds. If this number is high, consider checkpointing less frequently. I0421 16:10:48.948867 132411908254848 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 15 I0421 16:10:48.949161 132411908254848 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/15. I0421 16:10:49.117020 132411908254848 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 16:10:49.117128 132411908254848 replica_slices.py:410] 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 16:10:49.139332 132327571195456 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/15 I0421 16:10:49.895395 132327539738176 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/15/items I0421 16:10:51.058475 132411908254848 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.941863s I0421 16:10:51.058938 132411908254848 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 6.094 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 2 seconds) (per-host) I0421 16:10:51.058987 132411908254848 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 2.025202s (batch_requests_ready=0.079776s, total_serialization_initiated=1.945061s, others=0.000365s) I0421 16:10:51.059062 132411908254848 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.025690s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=2.025673) I0421 16:10:51.063163 132327581681216 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 16:10:51.063293 132411908254848 async_checkpointer.py:561] Finished blocking save. Time taken: 2.114379s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/15. I0421 16:10:51.063430 132411908254848 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=15] Starting CheckpointManager Save Finalize thread=save_finalize I0421 16:10:51.063611 132328191952448 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 16:10:51.063696 132411908254848 standard_logger.py:34] {'step': 15, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776787810.7771342, 'wait_for_prev_duration_secs': 38.17054533958435, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776787848.9488928, 'checkpointer_blocking_duration_secs': 2.114454507827759, 'get_old_steps_start_time': 1776787851.0633671, 'get_old_steps_duration_secs': 2.47955322265625e-05, 'checkpoint_manager_blocking_start_time': 1776787810.7770402, 'checkpoint_manager_blocking_duration_secs': 40.2866313457489} I0421 16:10:51.063790 132411908254848 checkpointing.py:408] Started an asynchronous checkpoint save for step 15 I0421 16:10:51.064452 132411908254848 metric_logger.py:196] completed step: 14, seconds: 0.120, TFLOP/s/device: 113.348, Tokens/s/device: 17085.033, total_weights: 16384, loss: 2.083, lm_loss: 2.083, perplexity: 8.030 I0421 16:10:51.071926 132411908254848 metric_logger.py:196] completed step: 15, seconds: 0.119, TFLOP/s/device: 114.575, Tokens/s/device: 17270.021, total_weights: 16384, loss: 1.880, lm_loss: 1.880, perplexity: 6.551 I0421 16:10:51.188736 132411908254848 metric_logger.py:196] completed step: 16, seconds: 40.526, TFLOP/s/device: 0.335, Tokens/s/device: 50.536, total_weights: 16384, loss: 1.726, lm_loss: 1.726, perplexity: 5.616 I0421 16:10:51.307743 132411908254848 metric_logger.py:196] completed step: 17, seconds: 0.006, TFLOP/s/device: 2400.553, Tokens/s/device: 361837.456, total_weights: 16384, loss: 1.610, lm_loss: 1.610, perplexity: 5.002 I0421 16:10:51.426768 132411908254848 metric_logger.py:196] completed step: 18, seconds: 0.118, TFLOP/s/device: 115.338, Tokens/s/device: 17384.956, total_weights: 16384, loss: 1.521, lm_loss: 1.521, perplexity: 4.577 I0421 16:10:51.544829 132411908254848 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0421 16:10:51.545642 132411908254848 checkpointing.py:776] Waited 0.0008256435394287109 seconds for step 19 to finish before starting checkpointing. I0421 16:10:51.545882 132411908254848 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=15][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 16:10:51.839437 132328238089792 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/15/items/array_metadatas/process_0 I0421 16:11:37.044972 132327539738176 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 45.982431s (commit=45.542416s, array_metadata_write=0.440015s) I0421 16:11:37.046173 132327581681216 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 263.199 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 48 seconds) (per-host) I0421 16:11:37.046280 132327581681216 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.983024s. I0421 16:11:37.464407 132327581681216 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0421 16:11:37.987053 132327581681216 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.669098s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/15/items I0421 16:11:37.987740 132327581681216 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/15/items I0421 16:11:38.222106 132327581681216 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/15 I0421 16:11:38.902802 132327581681216 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/15`. I0421 16:11:38.903488 132327581681216 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 49.954580s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/15 I0421 16:11:38.903565 132327581681216 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 47.840311s. I0421 16:11:38.903691 132328191952448 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 16:11:38.903743 132328191952448 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 16:11:38.903800 132328191952448 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is syncing with other hosts... I0421 16:11:38.903841 132328191952448 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=15] CheckpointManager Save Finalize is done on all hosts. I0421 16:11:38.903964 132411908254848 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=15][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=15. W0421 16:11:38.904081 132411908254848 checkpoint_manager.py:1441] Waiting for previous save to complete took 47.358207 seconds. If this number is high, consider checkpointing less frequently. I0421 16:11:38.904293 132411908254848 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 19 I0421 16:11:38.904569 132411908254848 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/19. I0421 16:11:39.073516 132328191952448 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/19 I0421 16:11:39.086574 132411908254848 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 16:11:39.086670 132411908254848 replica_slices.py:410] 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 16:11:39.785730 132328223409728 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/19/items I0421 16:11:42.258681 132411908254848 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 3.172495s I0421 16:11:42.259243 132411908254848 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.781 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 3 seconds) (per-host) I0421 16:11:42.259297 132411908254848 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 3.263973s (batch_requests_ready=0.087126s, total_serialization_initiated=3.176390s, others=0.000457s) I0421 16:11:42.259364 132411908254848 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 3.264458s (all_items=0.000015s, per_item={'items': '0.00001550'}, temp_paths=3.264442) I0421 16:11:42.263813 132327539738176 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 16:11:42.263939 132411908254848 async_checkpointer.py:561] Finished blocking save. Time taken: 3.359603s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/19. I0421 16:11:42.264187 132411908254848 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0421 16:11:42.264404 132328170980928 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 16:11:42.264510 132411908254848 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776787851.545858, 'wait_for_prev_duration_secs': 47.35820746421814, 'time_between_consecutive_saves_sec': 2.5996992588043213, 'checkpointer_blocking_start_time': 1776787898.904316, 'checkpointer_blocking_duration_secs': 3.3598005771636963, 'get_old_steps_start_time': 1776787902.264129, 'get_old_steps_duration_secs': 2.09808349609375e-05, 'checkpoint_manager_blocking_start_time': 1776787851.545826, 'checkpoint_manager_blocking_duration_secs': 50.71866035461426} I0421 16:11:42.264631 132411908254848 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0421 16:11:42.264664 132411908254848 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 16:11:43.046148 132328238089792 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/19/items/array_metadatas/process_0 I0421 16:12:19.576486 132328223409728 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 37.313233s (commit=36.889583s, array_metadata_write=0.423650s) I0421 16:12:19.577642 132327539738176 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 311.387 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 40 seconds) (per-host) I0421 16:12:19.577749 132327539738176 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.313766s. I0421 16:12:20.028906 132327539738176 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0421 16:12:20.510949 132327539738176 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.647790s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/19/items I0421 16:12:20.511672 132327539738176 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/19/items I0421 16:12:20.763955 132327539738176 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/19 I0421 16:12:21.553253 132327539738176 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/19`. I0421 16:12:21.553930 132327539738176 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 42.649601s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_16_zarr3_false/checkpoints/19 I0421 16:12:21.553996 132327539738176 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 39.290014s. I0421 16:12:21.554175 132328170980928 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 16:12:21.554276 132328170980928 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 16:12:21.554326 132328170980928 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0421 16:12:21.554369 132328170980928 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0421 16:12:21.554505 132411908254848 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0421 16:12:21.554649 132411908254848 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 16:12:21.555638 132411908254848 metric_logger.py:196] completed step: 19, seconds: 0.119, TFLOP/s/device: 114.229, Tokens/s/device: 17217.897, total_weights: 16384, loss: 1.449, lm_loss: 1.449, perplexity: 4.260 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops