2026-04-23 01:41:03.331098: 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 00:00:1776908463.344196 2279988 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776908463.348040 2279988 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776908463.359540 2279988 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776908463.359556 2279988 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776908463.359558 2279988 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776908463.359560 2279988 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} `rope_scaling`'s factor field must be a float >= 1, got 40 `rope_scaling`'s beta_fast field must be a float, got 32 `rope_scaling`'s beta_slow field must be a float, got 1 Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} 2026-04-23 01:41:07.050082: 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 01:41:07.185827 133893495822144 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0423 01:41:37.235748 133893495822144 max_utils.py:800] System Information: Jax Version: 0.8.1 I0423 01:41:37.236014 133893495822144 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0423 01:41:37.236086 133893495822144 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 I0423 01:41:37.236116 133893495822144 train_utils.py:348] WARNING: 'dataset_path' might be pointing your local file system I0423 01:41:37.236139 133893495822144 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0423 01:41:37.236309 133893495822144 train.py:703] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0423 01:41:37.891810 133893495822144 maxtext_utils.py:1604] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 01:41:37.892093 133893495822144 checkpointing.py:677] Setting up checkpoint logger... I0423 01:41:37.892139 133893495822144 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0423 01:41:37.892175 133893495822144 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0423 01:41:37.892600 133893495822144 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 0x79c63dfc88c0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0423 01:41:40.258450 133893495822144 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0423 01:41:40.258837 133893495822144 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 0x79c4dddf77d0>}, handler_registry=None I0423 01:41:40.259093 133893495822144 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79c4dddf77d0>` 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 01:41:40.259139 133893495822144 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79c46cba9eb0>` 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 01:41:40.259170 133893495822144 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 0x79c4dddf77d0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79c4dddf77d0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79c46cba9eb0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79c46cba9eb0>}). I0423 01:41:40.259444 133893495822144 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0423 01:41:40.259497 133893495822144 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x79c46cbb8860> timeout: 600 secs and primary_host=0 for async checkpoint writes I0423 01:41:40.387694 133893495822144 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints I0423 01:41:40.387926 133893495822144 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_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x79c4dd301340> I0423 01:41:40.388229 133893495822144 checkpointing.py:301] Checkpoint manager created! I0423 01:41:40.770621 133893495822144 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0423 01:41:40.770723 133893495822144 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 01:41:41.053064 133893495822144 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 01:41:41.053153 133893495822144 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 01:41:41.065410 133893495822144 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 01:41:41.065460 133893495822144 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 01:41:41.084679 133893495822144 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 01:41:41.084739 133893495822144 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 01:41:41.097094 133893495822144 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 01:41:41.097148 133893495822144 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 01:41:41.109354 133893495822144 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 01:41:41.109406 133893495822144 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 01:41:41.128908 133893495822144 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0423 01:41:41.128973 133893495822144 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 01:41:41.144401 133893495822144 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0423 01:41:41.144455 133893495822144 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0423 01:41:41.294656 133893495822144 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0423 01:41:41.294757 133893495822144 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. [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 I0423 01:41:41.908822 133893495822144 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0423 01:41:41.908934 133893495822144 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0423 01:41:41.908983 133893495822144 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0423 01:41:41.909029 133893495822144 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0423 01:41:41.909090 133893495822144 maxtext_utils.py:1707] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0423 01:41:41.909123 133893495822144 maxtext_utils.py:1707] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0423 01:41:41.909163 133893495822144 maxtext_utils.py:1707] 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) I0423 01:41:41.909203 133893495822144 maxtext_utils.py:1707] 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') I0423 01:41:41.909231 133893495822144 maxtext_utils.py:1707] 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) I0423 01:41:41.909256 133893495822144 maxtext_utils.py:1707] 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) I0423 01:41:41.909290 133893495822144 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0423 01:41:41.909325 133893495822144 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0423 01:41:42.315375 133893495822144 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 01:41:42.315464 133893495822144 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0423 01:41:42.327493 133893495822144 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 01:41:42.327542 133893495822144 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0423 01:41:43.118733 133893495822144 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. I0423 01:41:43.119298 133893495822144 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0423 01:41:43.119448 133893495822144 metric_logger.py:301] number parameters: 1.104 billion I0423 01:41:45.478631 133893495822144 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0423 01:41:45.596505 133893495822144 checkpointing.py:776] Waited 0.11784768104553223 seconds for step 0 to finish before starting checkpointing. I0423 01:41:45.597164 133893495822144 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 01:41:45.597318 133893495822144 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0423 01:41:45.597726 133893495822144 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/0. I0423 01:41:45.682129 133893495822144 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0423 01:41:45.768426 133893495822144 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 01:41:45.768523 133893495822144 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 I0423 01:41:45.775746 133862451578432 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/0 I0423 01:41:46.498833 133862441092672 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/0/items I0423 01:41:46.661892 133862409635392 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776908506406899177, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776908506406899177, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/0/_CHECKPOINT_METADATA I0423 01:41:47.031312 2282000 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0423 01:41:47.031384 2282000 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0423 01:41:47.238280 133893495822144 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.470389s I0423 01:41:47.244569 133893495822144 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 7.906 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0423 01:41:47.244638 133893495822144 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.561115s (batch_requests_ready=0.080664s, total_serialization_initiated=1.474284s, others=0.006166s) I0423 01:41:47.244726 133893495822144 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.562056s (all_items=0.000021s, per_item={'items': '0.00002098'}, temp_paths=1.562035) I0423 01:41:47.246056 133861600134720 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 01:41:47.246150 133893495822144 async_checkpointer.py:561] Finished blocking save. Time taken: 1.648791s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/0. I0423 01:41:47.246366 133893495822144 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0423 01:41:47.246579 133862600476224 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 01:41:47.246692 133893495822144 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776908505.5971508, 'wait_for_prev_duration_secs': 4.38690185546875e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776908505.5973399, 'checkpointer_blocking_duration_secs': 1.6489291191101074, 'get_old_steps_start_time': 1776908507.2462938, 'get_old_steps_duration_secs': 3.2901763916015625e-05, 'checkpoint_manager_blocking_start_time': 1776908505.597067, 'checkpoint_manager_blocking_duration_secs': 1.6495997905731201} I0423 01:41:47.246831 133893495822144 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0423 01:41:47.246888 133893495822144 max_utils.py:750] Memstats: After params initialized: I0423 01:41:47.246935 133893495822144 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0423 01:41:47.246962 133893495822144 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0423 01:41:47.246981 133893495822144 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0423 01:41:47.246999 133893495822144 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0423 01:41:47.247016 133893495822144 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0423 01:41:47.247034 133893495822144 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0423 01:41:47.247064 133893495822144 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0423 01:41:47.247081 133893495822144 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0423 01:41:47.487395 133893495822144 metric_logger.py:196] completed step: 0, seconds: 2.359, TFLOP/s/device: 5.761, Tokens/s/device: 868.288, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0423 01:41:47.488378 133893495822144 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/tensorboard/' I0423 01:41:47.604325 133893495822144 metric_logger.py:196] completed step: 1, seconds: 2.007, TFLOP/s/device: 6.771, Tokens/s/device: 1020.645, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0423 01:41:47.723058 133893495822144 metric_logger.py:196] completed step: 2, seconds: 0.006, TFLOP/s/device: 2100.345, Tokens/s/device: 316586.799, total_weights: 16384, loss: 10.276, lm_loss: 10.276, perplexity: 29019.973 I0423 01:41:47.841852 133893495822144 metric_logger.py:196] completed step: 3, seconds: 0.119, TFLOP/s/device: 114.637, Tokens/s/device: 17279.347, total_weights: 16384, loss: 9.282, lm_loss: 9.282, perplexity: 10739.224 I0423 01:41:47.847294 133893495822144 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0423 01:41:47.993718 133861811947072 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/0/items/array_metadatas/process_0 I0423 01:41:48.079662 133893495822144 checkpointing.py:776] Waited 0.23232769966125488 seconds for step 5 to finish before starting checkpointing. I0423 01:41:48.080697 133893495822144 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 01:42:18.316807 133861801461312 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 31.071367s (commit=30.608862s, array_metadata_write=0.462504s) I0423 01:42:18.317871 133861600134720 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 387.224 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 32 seconds) (per-host) I0423 01:42:18.317924 133861600134720 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 31.071729s. I0423 01:42:18.591466 133861600134720 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776908506406899177, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/0/_CHECKPOINT_METADATA I0423 01:42:18.812528 133861600134720 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 01:42:18.995387 133862409635392 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776908506406899177, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/0/_CHECKPOINT_METADATA I0423 01:42:19.181736 133861600134720 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 01:42:19.181926 133861600134720 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.546466s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/0/items I0423 01:42:19.182725 133861600134720 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/0/items I0423 01:42:19.418079 133861600134720 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/0 I0423 01:42:20.135183 133861600134720 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/0`. I0423 01:42:20.135926 133861600134720 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 34.538574s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/0 I0423 01:42:20.136004 133861600134720 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 32.889810s. I0423 01:42:20.136204 133862600476224 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 01:42:20.136328 133862600476224 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 01:42:20.136390 133862600476224 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0423 01:42:20.136440 133862600476224 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0423 01:42:20.136519 133893495822144 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. W0423 01:42:20.136584 133893495822144 checkpoint_manager.py:1441] Waiting for previous save to complete took 32.055893 seconds. If this number is high, consider checkpointing less frequently. I0423 01:42:20.137759 133893495822144 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 5 I0423 01:42:20.138076 133893495822144 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/5. I0423 01:42:20.298137 133861600134720 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/5 I0423 01:42:20.304651 133893495822144 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 01:42:20.304759 133893495822144 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 I0423 01:42:20.994328 133861811947072 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/5/items I0423 01:42:21.361884 133893495822144 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.057627s I0423 01:42:21.362365 133893495822144 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 10.716 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0423 01:42:21.362416 133893495822144 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.151674s (batch_requests_ready=0.090392s, total_serialization_initiated=1.060940s, others=0.000342s) I0423 01:42:21.362474 133893495822144 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.152238s (all_items=0.000016s, per_item={'items': '0.00001574'}, temp_paths=1.152222) I0423 01:42:21.363525 133861994399296 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 01:42:21.363635 133893495822144 async_checkpointer.py:561] Finished blocking save. Time taken: 1.225833s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/5. I0423 01:42:21.363820 133893495822144 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0423 01:42:21.364008 133862600476224 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 01:42:21.364146 133893495822144 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776908508.08067, 'wait_for_prev_duration_secs': 32.05589294433594, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776908540.1377838, 'checkpointer_blocking_duration_secs': 1.2259562015533447, 'get_old_steps_start_time': 1776908541.363754, 'get_old_steps_duration_secs': 2.7179718017578125e-05, 'checkpoint_manager_blocking_start_time': 1776908508.080604, 'checkpoint_manager_blocking_duration_secs': 33.28351306915283} I0423 01:42:21.364297 133893495822144 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0423 01:42:21.364991 133893495822144 metric_logger.py:196] completed step: 4, seconds: 0.118, TFLOP/s/device: 114.932, Tokens/s/device: 17323.780, total_weights: 16384, loss: 8.327, lm_loss: 8.327, perplexity: 4134.812 I0423 01:42:21.373360 133893495822144 metric_logger.py:196] completed step: 5, seconds: 0.119, TFLOP/s/device: 114.509, Tokens/s/device: 17259.978, total_weights: 16384, loss: 7.367, lm_loss: 7.367, perplexity: 1583.522 I0423 01:42:21.490070 133893495822144 metric_logger.py:196] completed step: 6, seconds: 33.524, TFLOP/s/device: 0.405, Tokens/s/device: 61.090, total_weights: 16384, loss: 6.459, lm_loss: 6.459, perplexity: 638.363 I0423 01:42:21.608885 133893495822144 metric_logger.py:196] completed step: 7, seconds: 0.006, TFLOP/s/device: 2178.821, Tokens/s/device: 328415.651, total_weights: 16384, loss: 5.622, lm_loss: 5.622, perplexity: 276.577 I0423 01:42:21.727771 133893495822144 metric_logger.py:196] completed step: 8, seconds: 0.118, TFLOP/s/device: 115.285, Tokens/s/device: 17376.991, total_weights: 16384, loss: 4.864, lm_loss: 4.864, perplexity: 129.523 I0423 01:42:21.846369 133893495822144 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0423 01:42:21.847298 133893495822144 checkpointing.py:776] Waited 0.0009562969207763672 seconds for step 9 to finish before starting checkpointing. I0423 01:42:21.847563 133893495822144 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 01:42:22.185463 133862204114496 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/5/items/array_metadatas/process_0 I0423 01:42:51.709611 133861801461312 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 30.272830s (commit=29.819718s, array_metadata_write=0.453112s) I0423 01:42:51.710758 133861994399296 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 401.168 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 31 seconds) (per-host) I0423 01:42:51.710874 133861994399296 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 30.347194s. I0423 01:42:52.202313 133861994399296 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 01:42:52.571067 133861994399296 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 01:42:52.571300 133861994399296 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.538569s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/5/items I0423 01:42:52.572034 133861994399296 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/5/items I0423 01:42:52.790696 133861994399296 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/5 I0423 01:42:53.475718 133861994399296 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/5`. I0423 01:42:53.476476 133861994399296 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 33.338681s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/5 I0423 01:42:53.476556 133861994399296 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 32.112878s. I0423 01:42:53.476757 133862600476224 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 01:42:53.476889 133862600476224 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 01:42:53.476945 133862600476224 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0423 01:42:53.476984 133862600476224 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0423 01:42:53.477182 133893495822144 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. W0423 01:42:53.477312 133893495822144 checkpoint_manager.py:1441] Waiting for previous save to complete took 31.629757 seconds. If this number is high, consider checkpointing less frequently. I0423 01:42:53.477608 133893495822144 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0423 01:42:53.477921 133893495822144 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/9. I0423 01:42:53.634954 133861994399296 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/9 I0423 01:42:53.641701 133893495822144 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 01:42:53.641812 133893495822144 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 I0423 01:42:54.347597 133861610620480 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/9/items I0423 01:42:55.638174 133893495822144 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.996858s I0423 01:42:55.642208 133893495822144 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.907 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 2 seconds) (per-host) I0423 01:42:55.642266 133893495822144 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 2.089406s (batch_requests_ready=0.085330s, total_serialization_initiated=2.000176s, others=0.003900s) I0423 01:42:55.642334 133893495822144 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.089964s (all_items=0.000015s, per_item={'items': '0.00001454'}, temp_paths=2.089950) I0423 01:42:55.644282 133862600476224 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 01:42:55.644370 133893495822144 async_checkpointer.py:561] Finished blocking save. Time taken: 2.166720s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/9. I0423 01:42:55.644536 133893495822144 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0423 01:42:55.644695 133861994399296 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 01:42:55.644811 133893495822144 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776908541.8475392, 'wait_for_prev_duration_secs': 31.629757404327393, 'time_between_consecutive_saves_sec': 1.7110841274261475, 'checkpointer_blocking_start_time': 1776908573.4776325, 'checkpointer_blocking_duration_secs': 2.166843891143799, 'get_old_steps_start_time': 1776908575.6444862, 'get_old_steps_duration_secs': 1.7881393432617188e-05, 'checkpoint_manager_blocking_start_time': 1776908541.8475063, 'checkpoint_manager_blocking_duration_secs': 33.797279834747314} I0423 01:42:55.644923 133893495822144 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0423 01:42:55.644955 133893495822144 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 01:42:56.400078 133862250251840 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/9/items/array_metadatas/process_0 I0423 01:43:35.624419 133861610620480 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 39.981428s (commit=39.518160s, array_metadata_write=0.463268s) I0423 01:43:35.625638 133862600476224 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 300.356 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 42 seconds) (per-host) I0423 01:43:35.625754 133862600476224 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 39.981339s. I0423 01:43:36.175859 133862600476224 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 01:43:36.561180 133862600476224 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 01:43:36.561412 133862600476224 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.564152s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/9/items I0423 01:43:36.562196 133862600476224 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/9/items I0423 01:43:36.805813 133862600476224 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/9 I0423 01:43:37.529221 133862600476224 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/9`. I0423 01:43:37.530025 133862600476224 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 44.052382s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_08_checkpoint_async_true/checkpoints/9 I0423 01:43:37.530115 133862600476224 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 41.885702s. I0423 01:43:37.530295 133861994399296 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 01:43:37.530410 133861994399296 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 01:43:37.530469 133861994399296 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0423 01:43:37.530522 133861994399296 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0423 01:43:37.530707 133893495822144 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. I0423 01:43:37.530877 133893495822144 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 01:43:37.531980 133893495822144 metric_logger.py:196] completed step: 9, seconds: 0.121, TFLOP/s/device: 112.502, Tokens/s/device: 16957.573, 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