2026-04-21 16:19:35.506249: 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 00:00:1776788375.517629 1593038 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776788375.521770 1593038 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776788375.531627 1593038 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776788375.531639 1593038 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776788375.531641 1593038 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776788375.531642 1593038 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'} 2026-04-21 16:19:39.234623: 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 16:19:39.370744 134065433132160 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0421 16:20:08.557062 134065433132160 max_utils.py:800] System Information: Jax Version: 0.8.1 I0421 16:20:08.557241 134065433132160 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0421 16:20:08.557282 134065433132160 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 16:20:08.557311 134065433132160 train_utils.py:378] WARNING: 'dataset_path' might be pointing your local file system I0421 16:20:08.557439 134065433132160 train.py:806] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0421 16:20:09.188443 134065433132160 maxtext_utils.py:1732] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 16:20:09.188638 134065433132160 maxtext_utils.py:1732] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 16:20:09.188801 134065433132160 checkpointing.py:688] Setting up checkpoint logger... I0421 16:20:09.188844 134065433132160 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0421 16:20:09.188879 134065433132160 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0421 16:20:09.189325 134065433132160 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 0x79ee40db3140>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 16:20:12.039175 134065433132160 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0421 16:20:12.039367 134065433132160 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 0x79ece9d4aba0>}, handler_registry=None I0421 16:20:12.039598 134065433132160 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79ece9d4aba0>` 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 16:20:12.039639 134065433132160 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79ecea7f1190>` 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 16:20:12.039667 134065433132160 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 0x79ece9d4aba0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x79ece9d4aba0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79ecea7f1190>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x79ecea7f1190>}). I0421 16:20:12.039937 134065433132160 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0421 16:20:12.039999 134065433132160 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x79ec95bbd620> timeout: 600 secs and primary_host=0 for async checkpoint writes I0421 16:20:12.163843 134065433132160 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints I0421 16:20:12.164107 134065433132160 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=10), 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_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x79ecea7f23c0> I0421 16:20:12.164188 134065433132160 checkpointing.py:302] Checkpoint manager created! I0421 16:20:12.241403 134065433132160 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0421 16:20:12.244322 134065433132160 reader.py:262] Creating a tf.data.Dataset reading 8 files located in folders: tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0. I0421 16:20:12.297840 134065433132160 logging_logger.py:49] Constructing tf.data.Dataset __local_c4_builder for split train, from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0421 16:20:12.321094 134065433132160 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2 I0421 16:20:12.321157 134065433132160 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2 I0421 16:20:12.929288 134065433132160 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 16:20:12.929388 134065433132160 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 16:20:13.230662 134065433132160 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 16:20:13.230753 134065433132160 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 16:20:13.243486 134065433132160 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 16:20:13.243539 134065433132160 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 16:20:13.277496 134065433132160 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 16:20:13.277598 134065433132160 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:20:13.290813 134065433132160 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 16:20:13.290871 134065433132160 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:20:13.303539 134065433132160 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 16:20:13.303589 134065433132160 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:20:13.323688 134065433132160 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 16:20:13.323752 134065433132160 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:20:13.339798 134065433132160 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 16:20:13.339854 134065433132160 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0421 16:20:13.494214 134065433132160 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0421 16:20:13.494318 134065433132160 checkpointing.py:676] 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 I0421 16:20:14.110939 134065433132160 maxtext_utils.py:1835] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0421 16:20:14.111069 134065433132160 maxtext_utils.py:1835] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0421 16:20:14.111119 134065433132160 maxtext_utils.py:1835] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0421 16:20:14.111167 134065433132160 maxtext_utils.py:1835] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0421 16:20:14.111208 134065433132160 maxtext_utils.py:1835] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0421 16:20:14.111238 134065433132160 maxtext_utils.py:1835] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0421 16:20:14.111278 134065433132160 maxtext_utils.py:1835] 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 16:20:14.111319 134065433132160 maxtext_utils.py:1835] 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 16:20:14.111350 134065433132160 maxtext_utils.py:1835] 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 16:20:14.111378 134065433132160 maxtext_utils.py:1835] 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 16:20:14.111414 134065433132160 maxtext_utils.py:1835] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0421 16:20:14.111449 134065433132160 maxtext_utils.py:1835] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0421 16:20:14.792191 134065433132160 train.py:157] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 16:20:14.792280 134065433132160 train.py:157] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0421 16:20:14.804444 134065433132160 train.py:164] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 16:20:14.804496 134065433132160 train.py:164] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0421 16:20:15.539105 134065433132160 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 16:20:15.539644 134065433132160 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0421 16:20:15.539796 134065433132160 metric_logger.py:301] number parameters: 1.104 billion I0421 16:20:19.712708 134065433132160 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0421 16:20:19.834221 134065433132160 checkpointing.py:798] Waited 0.1214911937713623 seconds for step 0 to finish before starting checkpointing. I0421 16:20:19.834689 134065433132160 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 16:20:19.834847 134065433132160 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0421 16:20:19.835278 134065433132160 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/0. I0421 16:20:19.923594 134065433132160 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0421 16:20:20.009236 133974814885440 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/0 I0421 16:20:20.030975 134065433132160 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 16:20:20.031085 134065433132160 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 16:20:20.674870 134065433132160 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.644333s I0421 16:20:20.675259 134065433132160 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 16.443 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 750 milliseconds) (per-host) I0421 16:20:20.675321 134065433132160 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.750623s (batch_requests_ready=0.102443s, total_serialization_initiated=0.647894s, others=0.000286s) I0421 16:20:20.675390 134065433132160 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.751215s (all_items=0.000021s, per_item={'items': '0.00002098'}, temp_paths=0.751194) I0421 16:20:20.676065 133974747776576 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 16:20:20.676166 134065433132160 async_checkpointer.py:561] Finished blocking save. Time taken: 0.841279s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/0. I0421 16:20:20.676329 134065433132160 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0421 16:20:20.676513 133974737290816 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 16:20:20.676602 134065433132160 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776788419.8346748, 'wait_for_prev_duration_secs': 4.220008850097656e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776788419.834869, 'checkpointer_blocking_duration_secs': 0.841392993927002, 'get_old_steps_start_time': 1776788420.6762757, 'get_old_steps_duration_secs': 1.9073486328125e-05, 'checkpoint_manager_blocking_start_time': 1776788419.8345892, 'checkpoint_manager_blocking_duration_secs': 0.8419840335845947} I0421 16:20:20.676680 134065433132160 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0421 16:20:20.676721 134065433132160 max_utils.py:750] Memstats: After params initialized: I0421 16:20:20.676803 134065433132160 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0421 16:20:20.676829 134065433132160 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0421 16:20:20.676848 134065433132160 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0421 16:20:20.676866 134065433132160 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0421 16:20:20.676882 134065433132160 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0421 16:20:20.676897 134065433132160 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0421 16:20:20.676913 134065433132160 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0421 16:20:20.676928 134065433132160 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0421 16:20:20.802609 133974804399680 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/0/items I0421 16:20:21.049203 134065433132160 metric_logger.py:196] completed step: 0, seconds: 4.172, TFLOP/s/device: 3.256, Tokens/s/device: 490.851, total_weights: 13328, loss: 10.828, lm_loss: 10.828, perplexity: 50404.086 I0421 16:20:21.050168 134065433132160 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/tensorboard/' I0421 16:20:21.165215 134065433132160 metric_logger.py:196] completed step: 1, seconds: 1.334, TFLOP/s/device: 10.185, Tokens/s/device: 1535.161, total_weights: 12332, loss: 10.828, lm_loss: 10.828, perplexity: 50391.758 I0421 16:20:21.175693 133974825371200 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776788420703435717, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776788420703435717, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/0/_CHECKPOINT_METADATA I0421 16:20:21.282921 134065433132160 metric_logger.py:196] completed step: 2, seconds: 0.010, TFLOP/s/device: 1370.914, Tokens/s/device: 206639.088, total_weights: 15161, loss: 9.823, lm_loss: 9.823, perplexity: 18444.867 I0421 16:20:21.400703 134065433132160 metric_logger.py:196] completed step: 3, seconds: 0.117, TFLOP/s/device: 116.099, Tokens/s/device: 17499.637, total_weights: 13327, loss: 9.405, lm_loss: 9.405, perplexity: 12149.331 I0421 16:20:21.484922 1595411 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0421 16:20:21.484991 1595411 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0421 16:20:36.507625 134065433132160 metric_logger.py:196] completed step: 4, seconds: 0.118, TFLOP/s/device: 115.060, Tokens/s/device: 17343.145, total_weights: 11939, loss: 9.045, lm_loss: 9.045, perplexity: 8471.842 I0421 16:20:36.536431 134065433132160 metric_logger.py:196] completed step: 5, seconds: 0.118, TFLOP/s/device: 115.196, Tokens/s/device: 17363.584, total_weights: 15502, loss: 8.953, lm_loss: 8.953, perplexity: 7728.546 I0421 16:20:36.652033 134065433132160 metric_logger.py:196] completed step: 6, seconds: 15.124, TFLOP/s/device: 0.898, Tokens/s/device: 135.418, total_weights: 13864, loss: 8.775, lm_loss: 8.775, perplexity: 6468.536 I0421 16:20:36.769465 134065433132160 metric_logger.py:196] completed step: 7, seconds: 0.010, TFLOP/s/device: 1380.385, Tokens/s/device: 208066.646, total_weights: 12988, loss: 8.688, lm_loss: 8.688, perplexity: 5928.887 I0421 16:20:36.841436 133974768748096 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/0/items/array_metadatas/process_0 I0421 16:20:36.886757 134065433132160 metric_logger.py:196] completed step: 8, seconds: 0.117, TFLOP/s/device: 116.247, Tokens/s/device: 17521.945, total_weights: 13820, loss: 8.740, lm_loss: 8.740, perplexity: 6250.123 I0421 16:20:37.003880 134065433132160 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0421 16:20:37.004954 134065433132160 checkpointing.py:798] Waited 0.0011827945709228516 seconds for step 9 to finish before starting checkpointing. I0421 16:20:37.005599 134065433132160 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 16:21:18.788630 133974758262336 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 57.303255s (commit=56.849283s, array_metadata_write=0.453972s) I0421 16:21:18.789832 133974747776576 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 214.674 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 58 seconds) (per-host) I0421 16:21:18.789950 133974747776576 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 58.113732s. I0421 16:21:19.036098 133974747776576 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776788420703435717, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/0/_CHECKPOINT_METADATA I0421 16:21:19.228451 133974747776576 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 16:21:19.425801 133974825371200 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776788420703435717, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/0/_CHECKPOINT_METADATA I0421 16:21:19.636286 133974747776576 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 16:21:19.636899 133974747776576 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.554232s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/0/items I0421 16:21:19.637551 133974747776576 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/0/items I0421 16:21:19.879280 133974747776576 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/0 I0421 16:21:20.556724 133974747776576 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/0`. I0421 16:21:20.557477 133974747776576 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 60.722592s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/0 I0421 16:21:20.557567 133974747776576 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 59.881353s. I0421 16:21:20.557706 133974737290816 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 16:21:20.557760 133974737290816 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 16:21:20.557832 133974737290816 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0421 16:21:20.557883 133974737290816 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0421 16:21:20.558180 134065433132160 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. W0421 16:21:20.558345 134065433132160 checkpoint_manager.py:1441] Waiting for previous save to complete took 43.552904 seconds. If this number is high, consider checkpointing less frequently. I0421 16:21:20.559602 134065433132160 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0421 16:21:20.559960 134065433132160 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/9. I0421 16:21:20.738674 133974737290816 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/9 I0421 16:21:20.745414 134065433132160 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 16:21:20.745519 134065433132160 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 16:21:21.484476 133974621947456 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/9/items I0421 16:21:21.708210 134065433132160 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.963305s I0421 16:21:21.708671 134065433132160 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 11.633 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0421 16:21:21.708723 134065433132160 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.060898s (batch_requests_ready=0.094053s, total_serialization_initiated=0.966501s, others=0.000344s) I0421 16:21:21.708782 134065433132160 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.061448s (all_items=0.000024s, per_item={'items': '0.00002360'}, temp_paths=1.061424) I0421 16:21:21.709734 133974783428160 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 16:21:21.709842 134065433132160 async_checkpointer.py:561] Finished blocking save. Time taken: 1.150193s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/9. I0421 16:21:21.710082 134065433132160 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0421 16:21:21.710260 133974747776576 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 16:21:21.710379 134065433132160 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776788437.0054023, 'wait_for_prev_duration_secs': 43.55290389060974, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776788480.5596256, 'checkpointer_blocking_duration_secs': 1.1503241062164307, 'get_old_steps_start_time': 1776788481.709968, 'get_old_steps_duration_secs': 4.863739013671875e-05, 'checkpoint_manager_blocking_start_time': 1776788437.0052764, 'checkpoint_manager_blocking_duration_secs': 44.70507788658142} I0421 16:21:21.710504 134065433132160 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0421 16:21:21.710536 134065433132160 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 16:21:22.748054 133974737290816 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/9/items/array_metadatas/process_0 I0421 16:22:14.833898 133974793913920 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 52.862108s (commit=52.393284s, array_metadata_write=0.468825s) I0421 16:22:14.835313 133974783428160 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 233.205 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 54 seconds) (per-host) I0421 16:22:14.835430 133974783428160 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 53.125541s. I0421 16:22:15.270036 133974783428160 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 16:22:15.708320 133974783428160 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 16:22:15.708991 133974783428160 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.579503s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/9/items I0421 16:22:15.709810 133974783428160 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/9/items I0421 16:22:15.955325 133974783428160 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/9 I0421 16:22:16.647908 133974783428160 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/9`. I0421 16:22:16.648701 133974783428160 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 56.089058s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_01_base/checkpoints/9 I0421 16:22:16.648791 133974783428160 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 54.938903s. I0421 16:22:16.648998 133974747776576 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 16:22:16.649151 133974747776576 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 16:22:16.649230 133974747776576 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0421 16:22:16.649274 133974747776576 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0421 16:22:16.649447 134065433132160 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. I0421 16:22:16.649598 134065433132160 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 16:22:16.650714 134065433132160 metric_logger.py:196] completed step: 9, seconds: 0.118, TFLOP/s/device: 115.398, Tokens/s/device: 17394.111, total_weights: 12300, loss: 8.710, lm_loss: 8.710, perplexity: 6065.008 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops