2026-04-21 16:53:00.241705: 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 00:00:1776790380.251737 1623601 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776790380.254854 1623601 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776790380.264619 1623601 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776790380.264629 1623601 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776790380.264631 1623601 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776790380.264632 1623601 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 16:53:03.942334: 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 16:53:04.080674 138553583582336 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0421 16:53:31.711067 138553583582336 max_utils.py:800] System Information: Jax Version: 0.8.1 I0421 16:53:31.711334 138553583582336 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0421 16:53:31.711381 138553583582336 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 16:53:31.711410 138553583582336 train_utils.py:378] WARNING: 'dataset_path' might be pointing your local file system I0421 16:53:31.711583 138553583582336 train.py:806] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0421 16:53:32.344282 138553583582336 maxtext_utils.py:1732] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 16:53:32.344468 138553583582336 maxtext_utils.py:1732] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 16:53:32.344631 138553583582336 checkpointing.py:688] Setting up checkpoint logger... I0421 16:53:32.344674 138553583582336 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0421 16:53:32.344712 138553583582336 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0421 16:53:32.345154 138553583582336 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 0x7e033b7af170>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 16:53:34.887912 138553583582336 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0421 16:53:34.888119 138553583582336 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 0x7e01e47469f0>}, handler_registry=None I0421 16:53:34.888348 138553583582336 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e01e47469f0>` 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 16:53:34.888388 138553583582336 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e019317eba0>` 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 16:53:34.888417 138553583582336 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 0x7e01e47469f0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e01e47469f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e019317eba0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e019317eba0>}). I0421 16:53:34.888685 138553583582336 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0421 16:53:34.888746 138553583582336 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7e01931a4ae0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0421 16:53:35.011926 138553583582336 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_10_shardy_false/checkpoints I0421 16:53:35.012187 138553583582336 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_10_shardy_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e019317e6f0> I0421 16:53:35.012269 138553583582336 checkpointing.py:302] Checkpoint manager created! I0421 16:53:35.090064 138553583582336 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0421 16:53:35.092943 138553583582336 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 16:53:35.147693 138553583582336 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 16:53:35.170267 138553583582336 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2 I0421 16:53:35.170332 138553583582336 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2 I0421 16:53:35.769646 138553583582336 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 16:53:35.769752 138553583582336 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 16:53:36.073713 138553583582336 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 16:53:36.073809 138553583582336 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 16:53:36.086398 138553583582336 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 16:53:36.086447 138553583582336 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 16:53:36.105336 138553583582336 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 16:53:36.105397 138553583582336 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:53:36.118072 138553583582336 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 16:53:36.118126 138553583582336 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:53:36.130718 138553583582336 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 16:53:36.130771 138553583582336 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:53:36.150618 138553583582336 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 16:53:36.150681 138553583582336 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:53:36.166486 138553583582336 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 16:53:36.166542 138553583582336 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0421 16:53:36.319937 138553583582336 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0421 16:53:36.320056 138553583582336 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 16:53:36.934750 138553583582336 maxtext_utils.py:1835] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0421 16:53:36.934868 138553583582336 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 16:53:36.934910 138553583582336 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 16:53:36.934955 138553583582336 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 16:53:36.935001 138553583582336 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 16:53:36.935030 138553583582336 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 16:53:36.935090 138553583582336 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 16:53:36.935133 138553583582336 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 16:53:36.935163 138553583582336 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 16:53:36.935190 138553583582336 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 16:53:36.935226 138553583582336 maxtext_utils.py:1835] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0421 16:53:36.935261 138553583582336 maxtext_utils.py:1835] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0421 16:53:37.608507 138553583582336 train.py:157] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 16:53:37.608597 138553583582336 train.py:157] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0421 16:53:37.620520 138553583582336 train.py:164] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 16:53:37.620574 138553583582336 train.py:164] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0421 16:53:38.363970 138553583582336 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 16:53:38.364682 138553583582336 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0421 16:53:38.364920 138553583582336 metric_logger.py:301] number parameters: 1.104 billion I0421 16:53:42.671830 138553583582336 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0421 16:53:42.793759 138553583582336 checkpointing.py:798] Waited 0.12194204330444336 seconds for step 0 to finish before starting checkpointing. I0421 16:53:42.794447 138553583582336 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 16:53:42.794648 138553583582336 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0421 16:53:42.795307 138553583582336 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_10_shardy_false/checkpoints/0. I0421 16:53:42.886702 138553583582336 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0421 16:53:42.980993 138462982309440 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_10_shardy_false/checkpoints/0 I0421 16:53:42.992264 138553583582336 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 16:53:42.992361 138553583582336 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 16:53:43.630827 138553583582336 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.639027s I0421 16:53:43.631221 138553583582336 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 16.611 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 742 milliseconds) (per-host) I0421 16:53:43.631277 138553583582336 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.743064s (batch_requests_ready=0.100155s, total_serialization_initiated=0.642622s, others=0.000287s) I0421 16:53:43.631340 138553583582336 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.744024s (all_items=0.000033s, per_item={'items': '0.00003338'}, temp_paths=0.743990) I0421 16:53:43.632118 138462911006272 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 16:53:43.632219 138553583582336 async_checkpointer.py:561] Finished blocking save. Time taken: 0.837526s. 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_10_shardy_false/checkpoints/0. I0421 16:53:43.632418 138553583582336 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0421 16:53:43.632566 138462992795200 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 16:53:43.632642 138553583582336 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_10_shardy_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776790422.7944288, 'wait_for_prev_duration_secs': 4.9591064453125e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776790422.7946706, 'checkpointer_blocking_duration_secs': 0.8376493453979492, 'get_old_steps_start_time': 1776790423.6323392, 'get_old_steps_duration_secs': 3.743171691894531e-05, 'checkpoint_manager_blocking_start_time': 1776790422.7942703, 'checkpoint_manager_blocking_duration_secs': 0.8383340835571289} I0421 16:53:43.632753 138553583582336 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0421 16:53:43.632795 138553583582336 max_utils.py:750] Memstats: After params initialized: I0421 16:53:43.632872 138553583582336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0421 16:53:43.632899 138553583582336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0421 16:53:43.632919 138553583582336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0421 16:53:43.632936 138553583582336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0421 16:53:43.632953 138553583582336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0421 16:53:43.632969 138553583582336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0421 16:53:43.632985 138553583582336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0421 16:53:43.633000 138553583582336 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0421 16:53:43.787125 138462971823680 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_10_shardy_false/checkpoints/0/items I0421 16:53:44.025839 138553583582336 metric_logger.py:196] completed step: 0, seconds: 4.306, TFLOP/s/device: 3.155, Tokens/s/device: 475.590, total_weights: 13328, loss: 10.828, lm_loss: 10.828, perplexity: 50404.086 I0421 16:53:44.026849 138553583582336 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_10_shardy_false/tensorboard/' I0421 16:53:44.141948 138553583582336 metric_logger.py:196] completed step: 1, seconds: 1.352, TFLOP/s/device: 10.051, Tokens/s/device: 1514.988, total_weights: 12332, loss: 10.828, lm_loss: 10.828, perplexity: 50391.758 I0421 16:53:44.153377 138462896326208 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776790423700655550, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776790423700655550, "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_10_shardy_false/checkpoints/0/_CHECKPOINT_METADATA I0421 16:53:44.259561 138553583582336 metric_logger.py:196] completed step: 2, seconds: 0.010, TFLOP/s/device: 1391.981, Tokens/s/device: 209814.568, total_weights: 15161, loss: 9.823, lm_loss: 9.823, perplexity: 18444.867 I0421 16:53:44.318509 1625973 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0421 16:53:44.318588 1625973 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0421 16:53:44.377277 138553583582336 metric_logger.py:196] completed step: 3, seconds: 0.117, TFLOP/s/device: 115.796, Tokens/s/device: 17454.000, total_weights: 13327, loss: 9.405, lm_loss: 9.405, perplexity: 12149.331 I0421 16:53:59.784267 138553583582336 metric_logger.py:196] completed step: 4, seconds: 0.117, TFLOP/s/device: 115.656, Tokens/s/device: 17432.903, total_weights: 11939, loss: 9.045, lm_loss: 9.045, perplexity: 8471.842 I0421 16:53:59.809912 138553583582336 metric_logger.py:196] completed step: 5, seconds: 0.118, TFLOP/s/device: 115.295, Tokens/s/device: 17378.465, total_weights: 15502, loss: 8.953, lm_loss: 8.953, perplexity: 7728.546 I0421 16:53:59.926981 138553583582336 metric_logger.py:196] completed step: 6, seconds: 15.423, TFLOP/s/device: 0.881, Tokens/s/device: 132.792, total_weights: 13864, loss: 8.775, lm_loss: 8.775, perplexity: 6468.536 I0421 16:54:00.044716 138553583582336 metric_logger.py:196] completed step: 7, seconds: 0.007, TFLOP/s/device: 2064.913, Tokens/s/device: 311246.201, total_weights: 12988, loss: 8.688, lm_loss: 8.688, perplexity: 5928.887 I0421 16:54:00.111789 138462931977792 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_10_shardy_false/checkpoints/0/items/array_metadatas/process_0 I0421 16:54:00.162537 138553583582336 metric_logger.py:196] completed step: 8, seconds: 0.119, TFLOP/s/device: 114.522, Tokens/s/device: 17262.015, total_weights: 13820, loss: 8.740, lm_loss: 8.740, perplexity: 6250.123 I0421 16:54:00.279443 138553583582336 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0421 16:54:00.280441 138553583582336 checkpointing.py:798] Waited 0.001125335693359375 seconds for step 9 to finish before starting checkpointing. I0421 16:54:00.280856 138553583582336 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 16:54:28.652073 138462921492032 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 44.330946s (commit=43.884240s, array_metadata_write=0.446706s) I0421 16:54:28.653174 138462911006272 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 276.124 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 45 seconds) (per-host) I0421 16:54:28.653243 138462911006272 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.020974s. I0421 16:54:28.934946 138462911006272 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776790423700655550, '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_10_shardy_false/checkpoints/0/_CHECKPOINT_METADATA I0421 16:54:29.133163 138462911006272 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 16:54:29.332904 138462896326208 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776790423700655550, '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_10_shardy_false/checkpoints/0/_CHECKPOINT_METADATA I0421 16:54:29.549150 138462911006272 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 16:54:29.549805 138462911006272 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.575669s. 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_10_shardy_false/checkpoints/0/items I0421 16:54:29.550592 138462911006272 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_10_shardy_false/checkpoints/0/items I0421 16:54:29.799144 138462911006272 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_10_shardy_false/checkpoints/0 I0421 16:54:30.492649 138462911006272 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_10_shardy_false/checkpoints/0`. I0421 16:54:30.493419 138462911006272 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 47.698727s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_10_shardy_false/checkpoints/0 I0421 16:54:30.493507 138462911006272 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 46.861241s. I0421 16:54:30.493634 138462992795200 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 16:54:30.493693 138462992795200 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 16:54:30.493763 138462992795200 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0421 16:54:30.493814 138462992795200 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0421 16:54:30.494175 138553583582336 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 16:54:30.494325 138553583582336 checkpoint_manager.py:1441] Waiting for previous save to complete took 30.213506 seconds. If this number is high, consider checkpointing less frequently. I0421 16:54:30.495534 138553583582336 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0421 16:54:30.495882 138553583582336 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_10_shardy_false/checkpoints/9. I0421 16:54:30.683584 138462992795200 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_10_shardy_false/checkpoints/9 I0421 16:54:30.704358 138553583582336 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 16:54:30.704462 138553583582336 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 16:54:31.467938 138462797760064 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_10_shardy_false/checkpoints/9/items I0421 16:54:35.602881 138553583582336 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 4.898973s I0421 16:54:35.607312 138553583582336 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.462 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 5 seconds) (per-host) I0421 16:54:35.607389 138553583582336 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 5.013547s (batch_requests_ready=0.107119s, total_serialization_initiated=4.902132s, others=0.004297s) I0421 16:54:35.607468 138553583582336 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 5.014140s (all_items=0.000026s, per_item={'items': '0.00002575'}, temp_paths=5.014114) I0421 16:54:35.609094 138462992795200 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 16:54:35.609180 138553583582336 async_checkpointer.py:561] Finished blocking save. Time taken: 5.113598s. 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_10_shardy_false/checkpoints/9. I0421 16:54:35.609391 138553583582336 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0421 16:54:35.609575 138462911006272 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 16:54:35.609681 138553583582336 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_10_shardy_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776790440.2807865, 'wait_for_prev_duration_secs': 30.213505744934082, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776790470.4955578, 'checkpointer_blocking_duration_secs': 5.113730430603027, 'get_old_steps_start_time': 1776790475.6093025, 'get_old_steps_duration_secs': 4.482269287109375e-05, 'checkpoint_manager_blocking_start_time': 1776790440.2807057, 'checkpoint_manager_blocking_duration_secs': 35.32894730567932} I0421 16:54:35.609797 138553583582336 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0421 16:54:35.609835 138553583582336 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 16:54:36.404440 138462982309440 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_10_shardy_false/checkpoints/9/items/array_metadatas/process_0 I0421 16:55:13.431316 138462797760064 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 37.823177s (commit=37.340302s, array_metadata_write=0.482876s) I0421 16:55:13.432648 138462992795200 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 294.985 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 42 seconds) (per-host) I0421 16:55:13.432714 138462992795200 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.823484s. I0421 16:55:13.885696 138462992795200 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 16:55:14.347754 138462992795200 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 16:55:14.348459 138462992795200 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.629546s. 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_10_shardy_false/checkpoints/9/items I0421 16:55:14.349263 138462992795200 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_10_shardy_false/checkpoints/9/items I0421 16:55:14.608383 138462992795200 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_10_shardy_false/checkpoints/9 I0421 16:55:15.312906 138462992795200 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_10_shardy_false/checkpoints/9`. I0421 16:55:15.313671 138462992795200 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 44.818093s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_10_shardy_false/checkpoints/9 I0421 16:55:15.313755 138462992795200 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 39.704531s. I0421 16:55:15.314031 138462911006272 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 16:55:15.314174 138462911006272 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 16:55:15.314282 138462911006272 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0421 16:55:15.314336 138462911006272 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0421 16:55:15.314495 138553583582336 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 16:55:15.314657 138553583582336 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 16:55:15.319881 138553583582336 metric_logger.py:196] completed step: 9, seconds: 0.119, TFLOP/s/device: 114.141, Tokens/s/device: 17204.590, 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