2026-04-17 13:58:27.808747: 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) I0417 13:58:27.928456 139377489497216 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0417 13:59:17.859846 139377489497216 max_utils.py:800] System Information: Jax Version: 0.9.2 I0417 13:59:17.859963 139377489497216 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0417 13:59:17.859997 139377489497216 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Apr 6 2026 20:48:10 (1775533690) cl/895581894 I0417 13:59:17.860022 139377489497216 train_utils.py:335] WARNING: 'dataset_path' might be pointing your local file system I0417 13:59:17.860058 139377489497216 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0417 13:59:17.860164 139377489497216 train.py:671] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0417 13:59:17.951022 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:59:18.336382 139377489497216 maxtext_utils.py:1548] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0417 13:59:18.336672 139377489497216 checkpointing.py:677] Setting up checkpoint logger... I0417 13:59:18.336714 139377489497216 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0417 13:59:18.336762 139377489497216 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0417 13:59:18.337277 139377489497216 base_pytree_checkpoint_handler.py:411] 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 0x7ec2c37d5670>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0417 13:59:20.789028 139377489497216 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0417 13:59:20.789236 139377489497216 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 0x7ebc46b320f0>}, handler_registry=None I0417 13:59:20.789491 139377489497216 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ebc46b320f0>` 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`. I0417 13:59:20.789533 139377489497216 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ebc46b31880>` 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`. I0417 13:59:20.789562 139377489497216 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 0x7ebc46b320f0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ebc46b320f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ebc46b31880>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ebc46b31880>}). I0417 13:59:20.789852 139377489497216 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0417 13:59:20.789903 139377489497216 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7ebc46b2e200> timeout: 600 secs and primary_host=0 for async checkpoint writes I0417 13:59:20.902497 139377489497216 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints I0417 13:59:20.902728 139377489497216 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_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7ebc46f06840> I0417 13:59:20.902823 139377489497216 checkpointing.py:301] Checkpoint manager created! W0417 13:59:20.919302 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:59:21.435034 139377489497216 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 13:59:21.435153 139377489497216 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0417 13:59:21.517596 139377489497216 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 13:59:21.517690 139377489497216 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0417 13:59:21.531812 139377489497216 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 13:59:21.531861 139377489497216 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0417 13:59:21.552163 139377489497216 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 13:59:21.552217 139377489497216 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0417 13:59:21.566343 139377489497216 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 13:59:21.566393 139377489497216 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0417 13:59:21.580580 139377489497216 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 13:59:21.580629 139377489497216 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0417 13:59:21.602131 139377489497216 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0417 13:59:21.602186 139377489497216 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0417 13:59:21.620208 139377489497216 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0417 13:59:21.620259 139377489497216 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0417 13:59:21.808278 139377489497216 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0417 13:59:21.808370 139377489497216 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. W0417 13:59:21.857093 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. [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 I0417 13:59:22.234808 139377489497216 maxtext_utils.py:1651] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0417 13:59:22.234920 139377489497216 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0417 13:59:22.234970 139377489497216 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0417 13:59:22.235015 139377489497216 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0417 13:59:22.235069 139377489497216 maxtext_utils.py:1651] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0417 13:59:22.235106 139377489497216 maxtext_utils.py:1651] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0417 13:59:22.235147 139377489497216 maxtext_utils.py:1651] 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) I0417 13:59:22.235186 139377489497216 maxtext_utils.py:1651] 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') I0417 13:59:22.235214 139377489497216 maxtext_utils.py:1651] 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) I0417 13:59:22.235240 139377489497216 maxtext_utils.py:1651] 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) I0417 13:59:22.235274 139377489497216 maxtext_utils.py:1651] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0417 13:59:22.235307 139377489497216 maxtext_utils.py:1651] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0417 13:59:22.652663 139377489497216 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0417 13:59:22.652745 139377489497216 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0417 13:59:22.666303 139377489497216 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0417 13:59:22.666352 139377489497216 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). W0417 13:59:22.917142 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:59:23.037055 139377489497216 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. I0417 13:59:23.037528 139377489497216 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0417 13:59:23.037686 139377489497216 metric_logger.py:301] number parameters: 1.104 billion W0417 13:59:23.860083 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:59:23.955248 139377489497216 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0417 13:59:24.072337 139377489497216 checkpointing.py:776] Waited 0.11707234382629395 seconds for step 0 to finish before starting checkpointing. I0417 13:59:24.072763 139377489497216 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 13:59:24.072910 139377489497216 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0417 13:59:24.073280 139377489497216 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/0. I0417 13:59:24.155892 139377489497216 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0417 13:59:24.241887 139269989467712 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/0 I0417 13:59:24.273935 139377489497216 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0417 13:59:24.274020 139377489497216 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 W0417 13:59:24.487454 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:59:24.497991 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:59:24.504226 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:59:24.516368 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:59:24.531334 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:59:24.540507 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:59:24.546544 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:59:24.551855 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:59:24.748871 139377489497216 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.475343s I0417 13:59:24.749229 139377489497216 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 20.831 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 592 milliseconds) (per-host) I0417 13:59:24.749287 139377489497216 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.592530s (batch_requests_ready=0.113490s, total_serialization_initiated=0.478780s, others=0.000260s) I0417 13:59:24.749346 139377489497216 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.593033s (all_items=0.000019s, per_item={'items': '0.00001860'}, temp_paths=0.593014) I0417 13:59:24.750037 139269926553152 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0417 13:59:24.750141 139377489497216 async_checkpointer.py:561] Finished blocking save. Time taken: 0.677183s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/0. I0417 13:59:24.750313 139377489497216 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0417 13:59:24.750452 139269999953472 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 13:59:24.750530 139377489497216 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776434364.0727499, 'wait_for_prev_duration_secs': 4.0531158447265625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776434364.0729303, 'checkpointer_blocking_duration_secs': 0.6773099899291992, 'get_old_steps_start_time': 1776434364.7502582, 'get_old_steps_duration_secs': 2.0265579223632812e-05, 'checkpoint_manager_blocking_start_time': 1776434364.072669, 'checkpoint_manager_blocking_duration_secs': 0.6778318881988525} I0417 13:59:24.750609 139377489497216 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0417 13:59:24.750646 139377489497216 max_utils.py:750] Memstats: After params initialized: I0417 13:59:24.750723 139377489497216 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0417 13:59:24.750757 139377489497216 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0417 13:59:24.750779 139377489497216 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0417 13:59:24.750797 139377489497216 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0417 13:59:24.750818 139377489497216 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0417 13:59:24.750835 139377489497216 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0417 13:59:24.750852 139377489497216 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0417 13:59:24.750869 139377489497216 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0417 13:59:24.756583 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:59:24.762434 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:59:24.767188 507191 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:59:24.882216 139269978981952 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/0/items I0417 13:59:24.965216 139377489497216 metric_logger.py:196] completed step: 0, seconds: 0.917, TFLOP/s/device: 14.815, Tokens/s/device: 2233.153, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0417 13:59:24.966107 139377489497216 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/tensorboard/' I0417 13:59:25.071575 139269911873088 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776434364793365654, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776434364793365654, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0417 13:59:25.081435 139377489497216 metric_logger.py:196] completed step: 1, seconds: 1.008, TFLOP/s/device: 13.478, Tokens/s/device: 2031.555, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0417 13:59:25.200240 139377489497216 metric_logger.py:196] completed step: 2, seconds: 0.006, TFLOP/s/device: 2389.576, Tokens/s/device: 360182.905, total_weights: 16384, loss: 10.276, lm_loss: 10.276, perplexity: 29019.973 I0417 13:59:25.318345 139377489497216 metric_logger.py:196] completed step: 3, seconds: 0.117, TFLOP/s/device: 116.307, Tokens/s/device: 17531.095, total_weights: 16384, loss: 9.282, lm_loss: 9.282, perplexity: 10739.224 I0417 13:59:25.323005 139377489497216 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0417 13:59:25.372011 509774 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0417 13:59:25.372080 509774 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0417 13:59:25.554218 139377489497216 checkpointing.py:776] Waited 0.23117923736572266 seconds for step 5 to finish before starting checkpointing. I0417 13:59:25.554619 139377489497216 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 13:59:26.120721 139269947524672 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/0/items/array_metadatas/process_0 I0417 13:59:54.317453 139269937038912 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 28.948913s (commit=28.543420s, array_metadata_write=0.405493s) I0417 13:59:54.318579 139269926553152 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 418.967 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 30 seconds) (per-host) I0417 13:59:54.318677 139269926553152 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 29.568491s. I0417 13:59:54.540509 139269926553152 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776434364793365654, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0417 13:59:54.717324 139269926553152 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0417 13:59:54.946650 139269911873088 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776434364793365654, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/0/_CHECKPOINT_METADATA I0417 13:59:55.156105 139269926553152 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0417 13:59:55.156280 139269926553152 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.576571s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/0/items I0417 13:59:55.156901 139269926553152 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/0/items I0417 13:59:55.412110 139269926553152 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/0 I0417 13:59:56.093844 139269926553152 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/0`. I0417 13:59:56.094455 139269926553152 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 32.021504s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/0 I0417 13:59:56.094527 139269926553152 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 31.344343s. I0417 13:59:56.094704 139269999953472 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 13:59:56.094808 139269999953472 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 13:59:56.094873 139269999953472 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0417 13:59:56.094920 139269999953472 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0417 13:59:56.095199 139377489497216 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0417 13:59:56.095326 139377489497216 checkpoint_manager.py:1441] Waiting for previous save to complete took 30.540706 seconds. If this number is high, consider checkpointing less frequently. I0417 13:59:56.096209 139377489497216 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 5 I0417 13:59:56.096493 139377489497216 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/5. I0417 13:59:56.263404 139377489497216 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0417 13:59:56.263500 139377489497216 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 I0417 13:59:56.264611 139269999953472 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/5 I0417 13:59:56.951251 139269978981952 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/5/items I0417 14:00:03.933297 139377489497216 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 7.670596s I0417 14:00:03.937274 139377489497216 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.590 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 7 seconds) (per-host) I0417 14:00:03.937339 139377489497216 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 7.759113s (batch_requests_ready=0.080736s, total_serialization_initiated=7.674502s, others=0.003876s) I0417 14:00:03.937414 139377489497216 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 7.759679s (all_items=0.000020s, per_item={'items': '0.00002003'}, temp_paths=7.759659) I0417 14:00:03.938843 139269937038912 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0417 14:00:03.938963 139377489497216 async_checkpointer.py:561] Finished blocking save. Time taken: 7.842708s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/5. I0417 14:00:03.939118 139377489497216 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0417 14:00:03.939350 139269926553152 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 14:00:03.939427 139377489497216 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776434365.5545964, 'wait_for_prev_duration_secs': 30.540706157684326, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776434396.0962338, 'checkpointer_blocking_duration_secs': 7.842787742614746, 'get_old_steps_start_time': 1776434403.9390378, 'get_old_steps_duration_secs': 4.00543212890625e-05, 'checkpoint_manager_blocking_start_time': 1776434365.554533, 'checkpoint_manager_blocking_duration_secs': 38.38487005233765} I0417 14:00:03.939563 139377489497216 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0417 14:00:03.940179 139377489497216 metric_logger.py:196] completed step: 4, seconds: 0.119, TFLOP/s/device: 114.220, Tokens/s/device: 17216.450, total_weights: 16384, loss: 8.327, lm_loss: 8.327, perplexity: 4134.812 I0417 14:00:03.947731 139377489497216 metric_logger.py:196] completed step: 5, seconds: 0.118, TFLOP/s/device: 115.097, Tokens/s/device: 17348.728, total_weights: 16384, loss: 7.367, lm_loss: 7.367, perplexity: 1583.522 I0417 14:00:04.063797 139377489497216 metric_logger.py:196] completed step: 6, seconds: 38.623, TFLOP/s/device: 0.352, Tokens/s/device: 53.026, total_weights: 16384, loss: 6.459, lm_loss: 6.459, perplexity: 638.363 I0417 14:00:04.181612 139377489497216 metric_logger.py:196] completed step: 7, seconds: 0.006, TFLOP/s/device: 2338.979, Tokens/s/device: 352556.378, total_weights: 16384, loss: 5.622, lm_loss: 5.622, perplexity: 276.577 I0417 14:00:04.299883 139377489497216 metric_logger.py:196] completed step: 8, seconds: 0.117, TFLOP/s/device: 116.234, Tokens/s/device: 17519.997, total_weights: 16384, loss: 4.864, lm_loss: 4.864, perplexity: 129.523 I0417 14:00:04.417433 139377489497216 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0417 14:00:04.418232 139377489497216 checkpointing.py:776] Waited 0.0008141994476318359 seconds for step 9 to finish before starting checkpointing. I0417 14:00:04.418477 139377489497216 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 14:00:04.717698 139269947524672 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/5/items/array_metadatas/process_0 I0417 14:00:37.247836 139269978981952 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 33.309856s (commit=32.871755s, array_metadata_write=0.438101s) I0417 14:00:37.248957 139269937038912 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 307.684 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 41 seconds) (per-host) I0417 14:00:37.249083 139269937038912 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 33.310154s. I0417 14:00:37.644030 139269937038912 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0417 14:00:38.112592 139269937038912 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0417 14:00:38.113258 139269937038912 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.613850s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/5/items I0417 14:00:38.113889 139269937038912 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/5/items I0417 14:00:38.361807 139269937038912 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/5 I0417 14:00:39.068163 139269937038912 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/5`. I0417 14:00:39.068792 139269937038912 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 42.972545s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/5 I0417 14:00:39.068872 139269937038912 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 35.129945s. I0417 14:00:39.068995 139269926553152 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 14:00:39.069054 139269926553152 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 14:00:39.069119 139269926553152 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0417 14:00:39.069157 139269926553152 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0417 14:00:39.069956 139377489497216 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0417 14:00:39.070726 139377489497216 checkpoint_manager.py:1441] Waiting for previous save to complete took 34.652248 seconds. If this number is high, consider checkpointing less frequently. I0417 14:00:39.071264 139377489497216 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0417 14:00:39.071541 139377489497216 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/9. I0417 14:00:39.240995 139377489497216 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0417 14:00:39.241149 139269926553152 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/9 I0417 14:00:39.241226 139377489497216 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 I0417 14:00:39.952875 139269876221504 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/9/items I0417 14:00:46.063320 139377489497216 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 6.823051s I0417 14:00:46.067577 139377489497216 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.785 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host) I0417 14:00:46.067633 139377489497216 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 6.914496s (batch_requests_ready=0.083940s, total_serialization_initiated=6.826407s, others=0.004150s) I0417 14:00:46.067710 139377489497216 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 6.915004s (all_items=0.000017s, per_item={'items': '0.00001669'}, temp_paths=6.914987) I0417 14:00:46.069168 139269926553152 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0417 14:00:46.069248 139377489497216 async_checkpointer.py:561] Finished blocking save. Time taken: 6.997939s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/9. I0417 14:00:46.069427 139377489497216 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0417 14:00:46.069581 139269937038912 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0417 14:00:46.069683 139377489497216 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776434404.418453, 'wait_for_prev_duration_secs': 34.65224838256836, 'time_between_consecutive_saves_sec': 8.323518514633179, 'checkpointer_blocking_start_time': 1776434439.0712895, 'checkpointer_blocking_duration_secs': 6.9980645179748535, 'get_old_steps_start_time': 1776434446.0693681, 'get_old_steps_duration_secs': 2.1457672119140625e-05, 'checkpoint_manager_blocking_start_time': 1776434404.4184237, 'checkpoint_manager_blocking_duration_secs': 41.65123701095581} I0417 14:00:46.069805 139377489497216 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0417 14:00:46.069838 139377489497216 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 14:00:46.856932 139269999953472 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/9/items/array_metadatas/process_0 I0417 14:01:18.497883 139269876221504 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 32.429555s (commit=31.916373s, array_metadata_write=0.513182s) I0417 14:01:18.498954 139269926553152 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 321.172 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 39 seconds) (per-host) I0417 14:01:18.499007 139269926553152 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 32.429716s. I0417 14:01:18.955786 139269926553152 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0417 14:01:19.407726 139269926553152 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0417 14:01:19.407892 139269926553152 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.619230s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/9/items I0417 14:01:19.408528 139269926553152 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/9/items I0417 14:01:19.663076 139269926553152 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/9 I0417 14:01:20.452817 139269926553152 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/9`. I0417 14:01:20.453444 139269926553152 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 41.382142s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_16_zarr3_true/checkpoints/9 I0417 14:01:20.453510 139269926553152 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 34.384220s. I0417 14:01:20.453685 139269937038912 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 14:01:20.453795 139269937038912 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0417 14:01:20.453851 139269937038912 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0417 14:01:20.453891 139269937038912 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0417 14:01:20.454155 139377489497216 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0417 14:01:20.454684 139377489497216 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 14:01:20.455708 139377489497216 metric_logger.py:196] completed step: 9, seconds: 0.118, TFLOP/s/device: 115.521, Tokens/s/device: 17412.597, total_weights: 16384, loss: 4.189, lm_loss: 4.189, perplexity: 65.979 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops