2026-04-21 16:33:36.023480: 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 00:00:1776789216.033886 1606561 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776789216.037132 1606561 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776789216.047187 1606561 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776789216.047199 1606561 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776789216.047200 1606561 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776789216.047202 1606561 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 16:33:39.710957: 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 16:33:39.846666 127378984799360 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0421 16:34:07.281920 127378984799360 max_utils.py:800] System Information: Jax Version: 0.8.1 I0421 16:34:07.282212 127378984799360 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0421 16:34:07.282263 127378984799360 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 16:34:07.282292 127378984799360 train_utils.py:378] WARNING: 'dataset_path' might be pointing your local file system I0421 16:34:07.282319 127378984799360 train_utils.py:391] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0421 16:34:07.282497 127378984799360 train.py:806] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0421 16:34:07.916852 127378984799360 maxtext_utils.py:1732] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 16:34:07.917033 127378984799360 maxtext_utils.py:1732] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 16:34:07.917208 127378984799360 checkpointing.py:688] Setting up checkpoint logger... I0421 16:34:07.917248 127378984799360 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0421 16:34:07.917283 127378984799360 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0421 16:34:07.917725 127378984799360 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 0x73d96eda3650>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 16:34:10.360604 127378984799360 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0421 16:34:10.360820 127378984799360 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 0x73d819d54470>}, handler_registry=None I0421 16:34:10.361081 127378984799360 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x73d819d54470>` 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 16:34:10.361123 127378984799360 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x73d7c898d370>` 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 16:34:10.361153 127378984799360 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 0x73d819d54470>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x73d819d54470>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x73d7c898d370>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x73d7c898d370>}). I0421 16:34:10.361439 127378984799360 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0421 16:34:10.361489 127378984799360 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x73d7c89b8ae0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0421 16:34:10.495850 127378984799360 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_06_grad_accum/checkpoints I0421 16:34:10.496109 127378984799360 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_06_grad_accum/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x73d7c898d100> I0421 16:34:10.496190 127378984799360 checkpointing.py:302] Checkpoint manager created! I0421 16:34:11.318373 127378984799360 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 16:34:11.318467 127378984799360 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 16:34:11.598489 127378984799360 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 16:34:11.598577 127378984799360 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 16:34:11.610599 127378984799360 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 16:34:11.610648 127378984799360 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 16:34:11.629298 127378984799360 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 16:34:11.629355 127378984799360 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:34:11.655308 127378984799360 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 16:34:11.655383 127378984799360 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:34:11.667485 127378984799360 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 16:34:11.667536 127378984799360 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:34:11.686783 127378984799360 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 16:34:11.686845 127378984799360 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 16:34:11.702040 127378984799360 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 16:34:11.702102 127378984799360 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0421 16:34:11.850828 127378984799360 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0421 16:34:11.850939 127378984799360 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 16:34:12.454759 127378984799360 maxtext_utils.py:1835] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0421 16:34:12.454873 127378984799360 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 16:34:12.454915 127378984799360 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 16:34:12.454960 127378984799360 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 16:34:12.455008 127378984799360 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 16:34:12.455038 127378984799360 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 16:34:12.455099 127378984799360 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 16:34:12.455140 127378984799360 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 16:34:12.455170 127378984799360 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 16:34:12.455197 127378984799360 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 16:34:12.455233 127378984799360 maxtext_utils.py:1835] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0421 16:34:12.455268 127378984799360 maxtext_utils.py:1835] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0421 16:34:12.472970 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[2048]............................................... Unknown. I0421 16:34:12.473026 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[2048]............................................... (None,). I0421 16:34:12.483970 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[2048,16,7168]....................................... Unknown. I0421 16:34:12.484019 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[2048,16,7168]....................................... ('fsdp', None, None). I0421 16:34:12.505706 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[7168,16,2048]....................................... Unknown. I0421 16:34:12.505761 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[7168,16,2048]....................................... (None, None, 'fsdp'). I0421 16:34:12.516614 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[2048,16]............................................ Unknown. I0421 16:34:12.516662 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[2048,16]............................................ (None, None). I0421 16:34:12.538298 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[2048,16,16,128]..................................... Unknown. I0421 16:34:12.538353 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[2048,16,16,128]..................................... ('fsdp', None, None, None). I0421 16:34:12.549219 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[16,16,128,2048]..................................... Unknown. I0421 16:34:12.549266 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[16,16,128,2048]..................................... (None, None, None, 'fsdp'). I0421 16:34:12.581771 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[2048,32000]......................................... Unknown. I0421 16:34:12.581832 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[2048,32000]......................................... ('fsdp', None). I0421 16:34:12.592704 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Logical: float32[32000,2048]......................................... Unknown. I0421 16:34:12.592751 127378984799360 gradient_accumulation.py:70] gradient_accumulation/inputs Physical: float32[32000,2048]......................................... (None, 'fsdp'). I0421 16:34:13.118414 127378984799360 train.py:157] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 16:34:13.118506 127378984799360 train.py:157] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0421 16:34:13.130254 127378984799360 train.py:164] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 16:34:13.130304 127378984799360 train.py:164] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0421 16:34:14.086668 127378984799360 max_utils.py:791] Total memory size: 4.5 GB, Output size: 1.5 GB, Temp size: 2.9 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0421 16:34:14.087171 127378984799360 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0421 16:34:14.087327 127378984799360 metric_logger.py:301] number parameters: 1.104 billion I0421 16:34:17.552735 127378984799360 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0421 16:34:17.980727 127378984799360 checkpointing.py:798] Waited 0.427976131439209 seconds for step 0 to finish before starting checkpointing. I0421 16:34:17.981241 127378984799360 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 16:34:17.981393 127378984799360 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0421 16:34:17.981788 127378984799360 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_06_grad_accum/checkpoints/0. I0421 16:34:18.073991 127378984799360 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0421 16:34:18.164059 127294798562880 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_06_grad_accum/checkpoints/0 I0421 16:34:18.172912 127378984799360 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 16:34:18.173011 127378984799360 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 16:34:18.936067 127378984799360 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.763552s I0421 16:34:18.936486 127378984799360 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 14.327 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 861 milliseconds) (per-host) I0421 16:34:18.936542 127378984799360 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.861451s (batch_requests_ready=0.094065s, total_serialization_initiated=0.767103s, others=0.000282s) I0421 16:34:18.936607 127378984799360 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.862043s (all_items=0.000020s, per_item={'items': '0.00001955'}, temp_paths=0.862024) I0421 16:34:18.937505 127294735648320 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 16:34:18.937652 127378984799360 async_checkpointer.py:561] Finished blocking save. Time taken: 0.956214s. 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_06_grad_accum/checkpoints/0. I0421 16:34:18.937903 127378984799360 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0421 16:34:18.938139 127378984799360 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_06_grad_accum/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776789257.9812288, 'wait_for_prev_duration_secs': 4.1484832763671875e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776789257.9814136, 'checkpointer_blocking_duration_secs': 0.9563882350921631, 'get_old_steps_start_time': 1776789258.9378264, 'get_old_steps_duration_secs': 2.86102294921875e-05, 'checkpoint_manager_blocking_start_time': 1776789257.9811418, 'checkpoint_manager_blocking_duration_secs': 0.9569668769836426} I0421 16:34:18.938224 127378984799360 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0421 16:34:18.938274 127378984799360 max_utils.py:750] Memstats: After params initialized: I0421 16:34:18.938349 127378984799360 max_utils.py:756] Using (GB) 1.6 / 31.25 (5.120000%) on TPU_0(process=0,(0,0,0,0)) I0421 16:34:18.938376 127378984799360 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0421 16:34:18.938397 127378984799360 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0421 16:34:18.938416 127378984799360 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0421 16:34:18.938435 127378984799360 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0421 16:34:18.938452 127378984799360 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0421 16:34:18.938470 127378984799360 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0421 16:34:18.938487 127378984799360 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0421 16:34:19.009013 127294788077120 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_06_grad_accum/checkpoints/0/items I0421 16:34:19.136784 127294756619840 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776789258884193419, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776789258884193419, "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_06_grad_accum/checkpoints/0/_CHECKPOINT_METADATA I0421 16:34:19.137113 127294809048640 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 16:34:19.319282 127378984799360 metric_logger.py:196] completed step: 0, seconds: 3.465, TFLOP/s/device: 15.686, Tokens/s/device: 2364.306, total_weights: 65536, loss: 10.871, lm_loss: 10.871, perplexity: 52637.648 I0421 16:34:19.320300 127378984799360 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_06_grad_accum/tensorboard/' I0421 16:34:19.703116 1608471 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0421 16:34:19.703198 1608471 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0421 16:34:19.740805 127378984799360 metric_logger.py:196] completed step: 1, seconds: 1.764, TFLOP/s/device: 30.809, Tokens/s/device: 4643.936, total_weights: 65536, loss: 10.871, lm_loss: 10.871, perplexity: 52637.648 I0421 16:34:20.158015 127378984799360 metric_logger.py:196] completed step: 2, seconds: 0.007, TFLOP/s/device: 7327.560, Tokens/s/device: 1104489.686, total_weights: 65536, loss: 10.261, lm_loss: 10.261, perplexity: 28586.895 I0421 16:34:20.465117 127294777591360 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_06_grad_accum/checkpoints/0/items/array_metadatas/process_0 I0421 16:34:20.587457 127378984799360 metric_logger.py:196] completed step: 3, seconds: 0.422, TFLOP/s/device: 128.671, Tokens/s/device: 19394.766, total_weights: 65536, loss: 9.732, lm_loss: 9.732, perplexity: 16845.295 I0421 16:34:47.548446 127378984799360 metric_logger.py:196] completed step: 4, seconds: 0.417, TFLOP/s/device: 130.301, Tokens/s/device: 19640.327, total_weights: 65536, loss: 9.274, lm_loss: 9.274, perplexity: 10654.189 I0421 16:34:47.558568 127378984799360 metric_logger.py:196] completed step: 5, seconds: 0.430, TFLOP/s/device: 126.533, Tokens/s/device: 19072.497, total_weights: 65536, loss: 8.888, lm_loss: 8.888, perplexity: 7241.146 I0421 16:34:47.981119 127378984799360 metric_logger.py:196] completed step: 6, seconds: 26.964, TFLOP/s/device: 2.016, Tokens/s/device: 303.813, total_weights: 65536, loss: 8.588, lm_loss: 8.588, perplexity: 5368.870 I0421 16:34:48.410636 127378984799360 metric_logger.py:196] completed step: 7, seconds: 0.005, TFLOP/s/device: 9977.697, Tokens/s/device: 1503947.127, total_weights: 65536, loss: 8.380, lm_loss: 8.380, perplexity: 4361.125 I0421 16:34:48.829478 127378984799360 metric_logger.py:196] completed step: 8, seconds: 0.423, TFLOP/s/device: 128.333, Tokens/s/device: 19343.748, total_weights: 65536, loss: 8.251, lm_loss: 8.251, perplexity: 3832.059 I0421 16:34:49.252698 127378984799360 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0421 16:34:49.253547 127378984799360 checkpointing.py:798] Waited 0.0008790493011474609 seconds for step 9 to finish before starting checkpointing. I0421 16:34:49.253876 127378984799360 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 16:35:04.210904 127294798562880 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 44.505888s (commit=44.002527s, array_metadata_write=0.503362s) I0421 16:35:04.212321 127294735648320 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 273.896 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 46 seconds) (per-host) I0421 16:35:04.212464 127294735648320 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.274747s. I0421 16:35:04.452589 127294735648320 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776789258884193419, '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_06_grad_accum/checkpoints/0/_CHECKPOINT_METADATA I0421 16:35:04.665593 127294735648320 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 16:35:04.862447 127294756619840 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776789258884193419, '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_06_grad_accum/checkpoints/0/_CHECKPOINT_METADATA I0421 16:35:05.095485 127294735648320 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 16:35:05.096150 127294735648320 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.591630s. 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_06_grad_accum/checkpoints/0/items I0421 16:35:05.096785 127294735648320 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_06_grad_accum/checkpoints/0/items I0421 16:35:05.345378 127294735648320 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_06_grad_accum/checkpoints/0 I0421 16:35:06.067594 127294735648320 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_06_grad_accum/checkpoints/0`. I0421 16:35:06.068366 127294735648320 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 48.086938s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_06_grad_accum/checkpoints/0 I0421 16:35:06.068457 127294735648320 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 47.130747s. I0421 16:35:06.068714 127294809048640 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 16:35:06.068843 127294809048640 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 16:35:06.068912 127294809048640 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0421 16:35:06.068967 127294809048640 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0421 16:35:06.069173 127378984799360 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 16:35:06.069315 127378984799360 checkpoint_manager.py:1441] Waiting for previous save to complete took 16.815473 seconds. If this number is high, consider checkpointing less frequently. I0421 16:35:06.070544 127378984799360 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0421 16:35:06.070890 127378984799360 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_06_grad_accum/checkpoints/9. I0421 16:35:06.255377 127378984799360 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 16:35:06.255482 127378984799360 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 16:35:06.257157 127294809048640 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_06_grad_accum/checkpoints/9 I0421 16:35:07.001302 127294634985024 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_06_grad_accum/checkpoints/9/items I0421 16:35:15.286479 127378984799360 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 9.031605s I0421 16:35:15.291104 127378984799360 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.352 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 9 seconds) (per-host) I0421 16:35:15.291185 127378984799360 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 9.127576s (batch_requests_ready=0.088096s, total_serialization_initiated=9.034956s, others=0.004524s) I0421 16:35:15.291265 127378984799360 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 9.128155s (all_items=0.000024s, per_item={'items': '0.00002360'}, temp_paths=9.128132) I0421 16:35:15.292784 127294809048640 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 16:35:15.292872 127378984799360 async_checkpointer.py:561] Finished blocking save. Time taken: 9.222277s. 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_06_grad_accum/checkpoints/9. I0421 16:35:15.293097 127378984799360 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0421 16:35:15.293420 127294697899584 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 16:35:15.293544 127378984799360 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_06_grad_accum/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776789289.2538087, 'wait_for_prev_duration_secs': 16.81547260284424, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776789306.0705688, 'checkpointer_blocking_duration_secs': 9.222411632537842, 'get_old_steps_start_time': 1776789315.292995, 'get_old_steps_duration_secs': 4.38690185546875e-05, 'checkpoint_manager_blocking_start_time': 1776789289.2537394, 'checkpoint_manager_blocking_duration_secs': 26.039777278900146} I0421 16:35:15.293662 127378984799360 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0421 16:35:15.293695 127378984799360 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 16:35:16.073303 127294777591360 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_06_grad_accum/checkpoints/9/items/array_metadatas/process_0 I0421 16:35:55.068220 127293313779264 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 39.776415s (commit=39.357261s, array_metadata_write=0.419154s) I0421 16:35:55.069672 127294809048640 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 258.389 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 48 seconds) (per-host) I0421 16:35:55.069802 127294809048640 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 39.776883s. I0421 16:35:55.516701 127294809048640 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 16:35:55.975843 127294809048640 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 16:35:55.976036 127294809048640 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.600003s. 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_06_grad_accum/checkpoints/9/items I0421 16:35:55.976814 127294809048640 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_06_grad_accum/checkpoints/9/items I0421 16:35:56.205381 127294809048640 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_06_grad_accum/checkpoints/9 I0421 16:35:56.909516 127294809048640 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_06_grad_accum/checkpoints/9`. I0421 16:35:56.910311 127294809048640 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 50.839726s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260421_161934/linen_feat_nnx_trainstate_and_training_loop_20260421_161934_06_grad_accum/checkpoints/9 I0421 16:35:56.910390 127294809048640 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 41.617472s. I0421 16:35:56.910604 127294697899584 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 16:35:56.910733 127294697899584 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 16:35:56.910807 127294697899584 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0421 16:35:56.910853 127294697899584 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0421 16:35:56.911031 127378984799360 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 16:35:56.911207 127378984799360 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 16:35:56.912828 127378984799360 metric_logger.py:196] completed step: 9, seconds: 0.430, TFLOP/s/device: 126.512, Tokens/s/device: 19069.256, total_weights: 65536, loss: 8.175, lm_loss: 8.175, perplexity: 3551.987 Per train step: Total TFLOPs: 54.35 split as 93.93% learnable weight flops and 6.07% attention flops