main (b341212ce) vs main (172d0f10a)| Metric | main b341212ce | main 172d0f10a | Diff (main − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 5.9150 | 5.9150 | 0 |
| TFLOP/s | 113.290 | 113.522 | +0.232 |
| Tok/s | 17076.3 | 17111.3 | +34.956 |
| Avg s/step | 1.864 | 2.159 | +0.295 |
| Memory % | 5.09 | 5.09 | 0 |
| JAX | 0.8.1 | 0.8.1 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
2026-04-21 15:23:51.696511: 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 00:00:1776785031.706407 1515172 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776785031.709449 1515172 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776785031.719386 1515172 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776785031.719399 1515172 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776785031.719401 1515172 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776785031.719402 1515172 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-21 15:23:55.125980: 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 15:23:55.242109 136956042857600 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0421 15:24:24.779057 136956042857600 max_utils.py:800] System Information: Jax Version: 0.8.1 I0421 15:24:24.779256 136956042857600 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0421 15:24:24.779300 136956042857600 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 15:24:24.779331 136956042857600 train_utils.py:335] WARNING: 'dataset_path' might be pointing your local file system I0421 15:24:24.779354 136956042857600 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0421 15:24:24.779474 136956042857600 train.py:703] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0421 15:24:25.139702 136956042857600 maxtext_utils.py:1551] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0421 15:24:25.140313 136956042857600 checkpointing.py:677] Setting up checkpoint logger... I0421 15:24:25.140370 136956042857600 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0421 15:24:25.140465 136956042857600 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0421 15:24:25.141418 136956042857600 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 0x7c8f469aaf30>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0421 15:24:27.536988 136956042857600 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0421 15:24:27.537298 136956042857600 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 0x7c8d9fb83950>}, handler_registry=None I0421 15:24:27.537722 136956042857600 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c8d9fb83950>` 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 15:24:27.537767 136956042857600 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c8d9ef6f0e0>` 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 15:24:27.537796 136956042857600 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 0x7c8d9fb83950>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c8d9fb83950>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c8d9ef6f0e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c8d9ef6f0e0>}). I0421 15:24:27.538111 136956042857600 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0421 15:24:27.538176 136956042857600 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x7c8d9ef90fe0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0421 15:24:27.661110 136956042857600 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints I0421 15:24:27.661381 136956042857600 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_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7c8d9ef6e540> I0421 15:24:27.661472 136956042857600 checkpointing.py:301] Checkpoint manager created! I0421 15:24:28.295080 136956042857600 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0421 15:24:28.295178 136956042857600 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 15:24:28.574198 136956042857600 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 15:24:28.574284 136956042857600 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 15:24:28.586229 136956042857600 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0421 15:24:28.586276 136956042857600 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0421 15:24:28.605013 136956042857600 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 15:24:28.605086 136956042857600 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 15:24:28.616992 136956042857600 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 15:24:28.617052 136956042857600 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 15:24:28.628901 136956042857600 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 15:24:28.628950 136956042857600 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 15:24:28.647920 136956042857600 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0421 15:24:28.647983 136956042857600 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0421 15:24:28.663002 136956042857600 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0421 15:24:28.663068 136956042857600 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0421 15:24:28.810455 136956042857600 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0421 15:24:28.810555 136956042857600 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. [DECOUPLED NO-OP] gcs_storage: using stubs. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] workload_monitor: using stub. [DECOUPLED NO-OP] vertex_tensorboard: using stub. fsdp: 8 I0421 15:24:29.397882 136956042857600 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0421 15:24:29.397991 136956042857600 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0421 15:24:29.398032 136956042857600 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0421 15:24:29.398093 136956042857600 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0421 15:24:29.398134 136956042857600 maxtext_utils.py:1654] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0421 15:24:29.398162 136956042857600 maxtext_utils.py:1654] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0421 15:24:29.398200 136956042857600 maxtext_utils.py:1654] 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 15:24:29.398240 136956042857600 maxtext_utils.py:1654] 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 15:24:29.398268 136956042857600 maxtext_utils.py:1654] 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 15:24:29.398296 136956042857600 maxtext_utils.py:1654] 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 15:24:29.398330 136956042857600 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0421 15:24:29.398366 136956042857600 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0421 15:24:29.798226 136956042857600 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 15:24:29.798318 136956042857600 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0421 15:24:29.809538 136956042857600 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0421 15:24:29.809588 136956042857600 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0421 15:24:30.540656 136956042857600 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 15:24:30.541272 136956042857600 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0421 15:24:30.541415 136956042857600 metric_logger.py:301] number parameters: 1.104 billion I0421 15:24:33.227350 136956042857600 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0421 15:24:33.344177 136956042857600 checkpointing.py:776] Waited 0.11679935455322266 seconds for step 0 to finish before starting checkpointing. I0421 15:24:33.344641 136956042857600 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 15:24:33.344789 136956042857600 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0421 15:24:33.345215 136956042857600 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/0. I0421 15:24:33.418459 136956042857600 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0421 15:24:33.498671 136871860504128 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/0 I0421 15:24:33.505937 136956042857600 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 15:24:33.506052 136956042857600 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 15:24:34.264490 136871850018368 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/0/items I0421 15:24:34.317469 136956042857600 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.812004s I0421 15:24:34.317829 136956042857600 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 13.744 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 897 milliseconds) (per-host) I0421 15:24:34.317884 136956042857600 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.898015s (batch_requests_ready=0.082170s, total_serialization_initiated=0.815600s, others=0.000245s) I0421 15:24:34.317944 136956042857600 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.898932s (all_items=0.000019s, per_item={'items': '0.00001884'}, temp_paths=0.898913) I0421 15:24:34.318756 136871797589568 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 15:24:34.318861 136956042857600 async_checkpointer.py:561] Finished blocking save. Time taken: 0.974033s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/0. I0421 15:24:34.319059 136956042857600 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0421 15:24:34.319261 136956042857600 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776785073.3446279, 'wait_for_prev_duration_secs': 3.9577484130859375e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776785073.34481, 'checkpointer_blocking_duration_secs': 0.9741642475128174, 'get_old_steps_start_time': 1776785074.3189893, 'get_old_steps_duration_secs': 2.0503997802734375e-05, 'checkpoint_manager_blocking_start_time': 1776785073.3445435, 'checkpoint_manager_blocking_duration_secs': 0.9746959209442139} I0421 15:24:34.319329 136956042857600 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0421 15:24:34.319369 136956042857600 max_utils.py:750] Memstats: After params initialized: I0421 15:24:34.319449 136956042857600 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0421 15:24:34.319475 136956042857600 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0421 15:24:34.319496 136956042857600 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0421 15:24:34.319515 136956042857600 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0421 15:24:34.319533 136956042857600 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0421 15:24:34.319550 136956042857600 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0421 15:24:34.319566 136956042857600 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0421 15:24:34.319583 136956042857600 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0421 15:24:34.394199 136871818561088 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776785074138668290, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776785074138668290, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/0/_CHECKPOINT_METADATA I0421 15:24:34.394500 136871787103808 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 15:24:34.701693 136956042857600 metric_logger.py:196] completed step: 0, seconds: 2.685, TFLOP/s/device: 5.060, Tokens/s/device: 762.640, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0421 15:24:34.702705 136956042857600 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/tensorboard/' I0421 15:24:34.764745 1517204 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0421 15:24:34.764823 1517204 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0421 15:24:35.056968 136956042857600 metric_logger.py:196] completed step: 1, seconds: 1.472, TFLOP/s/device: 9.229, Tokens/s/device: 1391.151, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0421 15:24:35.077001 136956042857600 metric_logger.py:196] completed step: 2, seconds: 0.007, TFLOP/s/device: 1880.832, Tokens/s/device: 283499.446, total_weights: 16384, loss: 9.775, lm_loss: 9.775, perplexity: 17589.287 I0421 15:24:35.194440 136956042857600 metric_logger.py:196] completed step: 3, seconds: 0.369, TFLOP/s/device: 36.795, Tokens/s/device: 5546.183, total_weights: 16384, loss: 8.838, lm_loss: 8.838, perplexity: 6892.933 I0421 15:24:49.734628 136956042857600 metric_logger.py:196] completed step: 4, seconds: 0.006, TFLOP/s/device: 2346.249, Tokens/s/device: 353652.219, total_weights: 16384, loss: 7.965, lm_loss: 7.965, perplexity: 2878.644 I0421 15:24:49.737121 136871808075328 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/0/items/array_metadatas/process_0 I0421 15:24:49.758646 136956042857600 metric_logger.py:196] completed step: 5, seconds: 0.118, TFLOP/s/device: 114.736, Tokens/s/device: 17294.230, total_weights: 16384, loss: 7.233, lm_loss: 7.233, perplexity: 1383.750 I0421 15:24:49.875982 136956042857600 metric_logger.py:196] completed step: 6, seconds: 14.557, TFLOP/s/device: 0.933, Tokens/s/device: 140.689, total_weights: 16384, loss: 6.676, lm_loss: 6.676, perplexity: 793.531 I0421 15:24:49.994923 136956042857600 metric_logger.py:196] completed step: 7, seconds: 0.006, TFLOP/s/device: 2223.389, Tokens/s/device: 335133.366, total_weights: 16384, loss: 6.293, lm_loss: 6.293, perplexity: 540.860 I0421 15:24:50.114122 136956042857600 metric_logger.py:196] completed step: 8, seconds: 0.119, TFLOP/s/device: 114.595, Tokens/s/device: 17272.934, total_weights: 16384, loss: 6.055, lm_loss: 6.055, perplexity: 426.371 I0421 15:24:50.232694 136956042857600 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0421 15:24:50.233699 136956042857600 checkpointing.py:776] Waited 0.0011243820190429688 seconds for step 9 to finish before starting checkpointing. I0421 15:24:50.234137 136956042857600 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 15:25:25.161576 136871839532608 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 50.401254s (commit=49.954031s, array_metadata_write=0.447223s) I0421 15:25:25.162788 136871797589568 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 244.223 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 51 seconds) (per-host) I0421 15:25:25.162910 136871797589568 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 50.843997s. I0421 15:25:25.425235 136871797589568 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776785074138668290, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/0/_CHECKPOINT_METADATA I0421 15:25:25.600790 136871797589568 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 15:25:25.805526 136871818561088 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776785074138668290, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/0/_CHECKPOINT_METADATA I0421 15:25:26.043251 136871797589568 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 15:25:26.043484 136871797589568 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.576780s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/0/items I0421 15:25:26.044299 136871797589568 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/0/items I0421 15:25:26.298930 136871797589568 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/0 I0421 15:25:26.976717 136871797589568 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/0`. I0421 15:25:26.977469 136871797589568 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 53.632643s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/0 I0421 15:25:26.977552 136871797589568 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 52.658643s. I0421 15:25:26.977687 136871787103808 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 15:25:26.977740 136871787103808 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 15:25:26.977802 136871787103808 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0421 15:25:26.977854 136871787103808 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0421 15:25:26.977997 136956042857600 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 15:25:26.978161 136956042857600 checkpoint_manager.py:1441] Waiting for previous save to complete took 36.744093 seconds. If this number is high, consider checkpointing less frequently. I0421 15:25:26.979332 136956042857600 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0421 15:25:26.979660 136956042857600 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/9. I0421 15:25:27.159153 136956042857600 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0421 15:25:27.159260 136956042857600 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 15:25:27.160985 136871787103808 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/9 I0421 15:25:27.926234 136871726286400 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/9/items I0421 15:25:40.134824 136956042857600 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 12.976163s I0421 15:25:40.139705 136956042857600 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 966.641 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 13 seconds) (per-host) I0421 15:25:40.139778 136956042857600 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 13.072997s (batch_requests_ready=0.088705s, total_serialization_initiated=12.979568s, others=0.004724s) I0421 15:25:40.139897 136956042857600 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 13.073619s (all_items=0.000027s, per_item={'items': '0.00002742'}, temp_paths=13.073591) I0421 15:25:40.141572 136871808075328 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0421 15:25:40.141711 136956042857600 async_checkpointer.py:561] Finished blocking save. Time taken: 13.162329s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/9. I0421 15:25:40.142012 136956042857600 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0421 15:25:40.142194 136871797589568 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0421 15:25:40.142295 136956042857600 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776785090.2340279, 'wait_for_prev_duration_secs': 36.74409294128418, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776785126.9793568, 'checkpointer_blocking_duration_secs': 13.16253137588501, 'get_old_steps_start_time': 1776785140.141909, 'get_old_steps_duration_secs': 4.3392181396484375e-05, 'checkpoint_manager_blocking_start_time': 1776785090.2339573, 'checkpoint_manager_blocking_duration_secs': 49.908305406570435} I0421 15:25:40.142442 136956042857600 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0421 15:25:40.142477 136956042857600 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0421 15:25:40.888804 136871870989888 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/9/items/array_metadatas/process_0 I0421 15:26:14.780184 136871839532608 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 34.639631s (commit=34.201800s, array_metadata_write=0.437831s) I0421 15:26:14.781507 136871808075328 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 264.841 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 47 seconds) (per-host) I0421 15:26:14.781624 136871808075328 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 34.639867s. I0421 15:26:15.212251 136871808075328 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 15:26:15.662258 136871808075328 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0421 15:26:15.662466 136871808075328 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.599585s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/9/items I0421 15:26:15.663246 136871808075328 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/9/items I0421 15:26:15.917948 136871808075328 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/9 I0421 15:26:16.601546 136871808075328 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/9`. I0421 15:26:16.602327 136871808075328 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 49.622954s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260421_144657/linen_main_20260421_144657_11_optimizer_offload_false/checkpoints/9 I0421 15:26:16.602405 136871808075328 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 36.460649s. I0421 15:26:16.602537 136871797589568 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0421 15:26:16.602589 136871797589568 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0421 15:26:16.602651 136871797589568 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0421 15:26:16.602691 136871797589568 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0421 15:26:16.603143 136956042857600 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 15:26:16.603306 136956042857600 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0421 15:26:16.604545 136956042857600 metric_logger.py:196] completed step: 9, seconds: 0.120, TFLOP/s/device: 113.290, Tokens/s/device: 17076.343, total_weights: 16384, loss: 5.915, lm_loss: 5.915, perplexity: 370.653 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops
2026-04-23 01:55:46.289145: 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 00:00:1776909346.302341 2295055 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776909346.306187 2295055 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776909346.317683 2295055 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776909346.317698 2295055 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776909346.317700 2295055 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776909346.317702 2295055 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} `rope_scaling`'s factor field must be a float >= 1, got 40 `rope_scaling`'s beta_fast field must be a float, got 32 `rope_scaling`'s beta_slow field must be a float, got 1 Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} 2026-04-23 01:55:49.997354: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0423 01:55:50.132845 131446876743488 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0423 01:56:30.851004 131446876743488 max_utils.py:800] System Information: Jax Version: 0.8.1 I0423 01:56:30.851285 131446876743488 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0423 01:56:30.851341 131446876743488 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709 I0423 01:56:30.851370 131446876743488 train_utils.py:348] WARNING: 'dataset_path' might be pointing your local file system I0423 01:56:30.851392 131446876743488 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0423 01:56:30.851557 131446876743488 train.py:703] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0423 01:56:31.503234 131446876743488 maxtext_utils.py:1604] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 01:56:31.503494 131446876743488 checkpointing.py:677] Setting up checkpoint logger... I0423 01:56:31.503539 131446876743488 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0423 01:56:31.503575 131446876743488 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0423 01:56:31.504007 131446876743488 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 0x778c92daae10>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0423 01:56:33.901598 131446876743488 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0423 01:56:33.901990 131446876743488 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 0x778abdf220c0>}, handler_registry=None I0423 01:56:33.902237 131446876743488 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x778abdf220c0>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0423 01:56:33.902279 131446876743488 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x778abdf223f0>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0423 01:56:33.902308 131446876743488 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 0x778abdf220c0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x778abdf220c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x778abdf223f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x778abdf223f0>}). I0423 01:56:33.902570 131446876743488 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0423 01:56:33.902621 131446876743488 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x778abdf30680> timeout: 600 secs and primary_host=0 for async checkpoint writes I0423 01:56:34.031847 131446876743488 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints I0423 01:56:34.032101 131446876743488 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_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x778abdf20050> I0423 01:56:34.032380 131446876743488 checkpointing.py:301] Checkpoint manager created! I0423 01:56:34.424547 131446876743488 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0423 01:56:34.424648 131446876743488 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 01:56:34.704878 131446876743488 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 01:56:34.704980 131446876743488 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 01:56:34.717075 131446876743488 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 01:56:34.717124 131446876743488 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 01:56:34.736198 131446876743488 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 01:56:34.736258 131446876743488 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 01:56:34.748537 131446876743488 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 01:56:34.748589 131446876743488 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 01:56:34.760733 131446876743488 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 01:56:34.760785 131446876743488 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 01:56:34.780233 131446876743488 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0423 01:56:34.780290 131446876743488 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 01:56:34.795753 131446876743488 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0423 01:56:34.795806 131446876743488 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0423 01:56:34.946493 131446876743488 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0423 01:56:34.946595 131446876743488 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. [DECOUPLED NO-OP] gcs_storage: using stubs. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] workload_monitor: using stub. [DECOUPLED NO-OP] vertex_tensorboard: using stub. fsdp: 8 I0423 01:56:35.571729 131446876743488 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0423 01:56:35.571844 131446876743488 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0423 01:56:35.571886 131446876743488 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0423 01:56:35.571930 131446876743488 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0423 01:56:35.571969 131446876743488 maxtext_utils.py:1707] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0423 01:56:35.572002 131446876743488 maxtext_utils.py:1707] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0423 01:56:35.572058 131446876743488 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0423 01:56:35.572104 131446876743488 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: PartitionSpec('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0423 01:56:35.572132 131446876743488 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0423 01:56:35.572158 131446876743488 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0423 01:56:35.572192 131446876743488 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0423 01:56:35.572227 131446876743488 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0423 01:56:35.982933 131446876743488 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 01:56:35.983026 131446876743488 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0423 01:56:35.994913 131446876743488 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 01:56:35.994966 131446876743488 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0423 01:56:36.795854 131446876743488 max_utils.py:791] Total memory size: 3.6 GB, Output size: 1.5 GB, Temp size: 2.0 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0423 01:56:36.796408 131446876743488 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0423 01:56:36.796565 131446876743488 metric_logger.py:301] number parameters: 1.104 billion I0423 01:56:39.441504 131446876743488 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0423 01:56:39.559345 131446876743488 checkpointing.py:776] Waited 0.11783027648925781 seconds for step 0 to finish before starting checkpointing. I0423 01:56:39.559988 131446876743488 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 01:56:39.560154 131446876743488 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0423 01:56:39.560551 131446876743488 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/0. I0423 01:56:39.648159 131446876743488 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0423 01:56:39.735633 131446876743488 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 01:56:39.735735 131446876743488 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0423 01:56:39.745298 131415719020096 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/0 I0423 01:56:40.560717 131415708534336 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/0/items I0423 01:56:40.712381 131415677077056 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776909400466867886, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776909400466867886, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/0/_CHECKPOINT_METADATA I0423 01:56:41.025611 2297097 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0423 01:56:41.025671 2297097 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0423 01:56:41.187361 131446876743488 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.452211s I0423 01:56:41.193089 131446876743488 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 7.994 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host) I0423 01:56:41.193157 131446876743488 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.543878s (batch_requests_ready=0.082082s, total_serialization_initiated=1.456182s, others=0.005615s) I0423 01:56:41.193238 131446876743488 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.544499s (all_items=0.000019s, per_item={'items': '0.00001907'}, temp_paths=1.544480) I0423 01:56:41.194720 131415729505856 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 01:56:41.194831 131446876743488 async_checkpointer.py:561] Finished blocking save. Time taken: 1.634637s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/0. I0423 01:56:41.195056 131446876743488 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0423 01:56:41.195253 131414878062144 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 01:56:41.195354 131446876743488 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776909399.5599756, 'wait_for_prev_duration_secs': 4.1484832763671875e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776909399.5601761, 'checkpointer_blocking_duration_secs': 1.634791612625122, 'get_old_steps_start_time': 1776909401.194984, 'get_old_steps_duration_secs': 2.7179718017578125e-05, 'checkpoint_manager_blocking_start_time': 1776909399.559893, 'checkpoint_manager_blocking_duration_secs': 1.6354341506958008} I0423 01:56:41.195476 131446876743488 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0423 01:56:41.195515 131446876743488 max_utils.py:750] Memstats: After params initialized: I0423 01:56:41.195562 131446876743488 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0423 01:56:41.195588 131446876743488 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0423 01:56:41.195610 131446876743488 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0423 01:56:41.195628 131446876743488 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0423 01:56:41.195645 131446876743488 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0423 01:56:41.195662 131446876743488 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0423 01:56:41.195678 131446876743488 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0423 01:56:41.195695 131446876743488 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) I0423 01:56:41.558083 131446876743488 metric_logger.py:196] completed step: 0, seconds: 2.644, TFLOP/s/device: 5.138, Tokens/s/device: 774.467, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0423 01:56:41.559188 131446876743488 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/tensorboard/' I0423 01:56:41.673954 131446876743488 metric_logger.py:196] completed step: 1, seconds: 2.113, TFLOP/s/device: 6.429, Tokens/s/device: 969.031, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0423 01:56:41.793388 131446876743488 metric_logger.py:196] completed step: 2, seconds: 0.011, TFLOP/s/device: 1293.150, Tokens/s/device: 194917.674, total_weights: 16384, loss: 9.775, lm_loss: 9.775, perplexity: 17589.287 I0423 01:56:41.912375 131446876743488 metric_logger.py:196] completed step: 3, seconds: 0.115, TFLOP/s/device: 117.916, Tokens/s/device: 17773.612, total_weights: 16384, loss: 8.838, lm_loss: 8.838, perplexity: 6892.933 I0423 01:56:41.947586 131415089874496 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/0/items/array_metadatas/process_0 I0423 01:56:58.621647 131446876743488 metric_logger.py:196] completed step: 4, seconds: 0.119, TFLOP/s/device: 114.323, Tokens/s/device: 17231.950, total_weights: 16384, loss: 7.965, lm_loss: 7.965, perplexity: 2878.644 I0423 01:56:58.629383 131446876743488 metric_logger.py:196] completed step: 5, seconds: 0.119, TFLOP/s/device: 114.179, Tokens/s/device: 17210.229, total_weights: 16384, loss: 7.233, lm_loss: 7.233, perplexity: 1383.750 I0423 01:56:58.746505 131446876743488 metric_logger.py:196] completed step: 6, seconds: 16.710, TFLOP/s/device: 0.813, Tokens/s/device: 122.564, total_weights: 16384, loss: 6.676, lm_loss: 6.676, perplexity: 793.531 I0423 01:56:58.864999 131446876743488 metric_logger.py:196] completed step: 7, seconds: 0.005, TFLOP/s/device: 2489.853, Tokens/s/device: 375297.783, total_weights: 16384, loss: 6.293, lm_loss: 6.293, perplexity: 540.860 I0423 01:56:58.984033 131446876743488 metric_logger.py:196] completed step: 8, seconds: 0.118, TFLOP/s/device: 115.605, Tokens/s/device: 17425.190, total_weights: 16384, loss: 6.055, lm_loss: 6.055, perplexity: 426.371 I0423 01:56:59.101453 131446876743488 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0423 01:56:59.102457 131446876743488 checkpointing.py:776] Waited 0.0010409355163574219 seconds for step 9 to finish before starting checkpointing. I0423 01:56:59.102746 131446876743488 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 01:57:22.039113 131415079388736 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 40.845238s (commit=40.352303s, array_metadata_write=0.492936s) I0423 01:57:22.040284 131415729505856 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 298.101 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 42 seconds) (per-host) I0423 01:57:22.040397 131415729505856 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.845506s. I0423 01:57:22.297066 131415729505856 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776909400466867886, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/0/_CHECKPOINT_METADATA I0423 01:57:22.511452 131415729505856 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 01:57:22.728073 131415677077056 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776909400466867886, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/0/_CHECKPOINT_METADATA I0423 01:57:22.889558 131415729505856 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 01:57:22.889784 131415729505856 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.547374s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/0/items I0423 01:57:22.890564 131415729505856 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/0/items I0423 01:57:23.138962 131415729505856 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/0 I0423 01:57:23.849431 131415729505856 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/0`. I0423 01:57:23.850200 131415729505856 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 44.290013s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/0 I0423 01:57:23.850275 131415729505856 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 42.655387s. I0423 01:57:23.850496 131414878062144 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 01:57:23.850622 131414878062144 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 01:57:23.850687 131414878062144 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0423 01:57:23.850735 131414878062144 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0423 01:57:23.850899 131446876743488 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0423 01:57:23.851053 131446876743488 checkpoint_manager.py:1441] Waiting for previous save to complete took 24.748315 seconds. If this number is high, consider checkpointing less frequently. I0423 01:57:23.852127 131446876743488 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0423 01:57:23.852434 131446876743488 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/9. I0423 01:57:24.026695 131446876743488 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0423 01:57:24.026797 131446876743488 replica_slices.py:410] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0423 01:57:24.032093 131415729505856 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/9 I0423 01:57:24.734935 131415089874496 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/9/items I0423 01:57:29.486341 131446876743488 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 5.460113s I0423 01:57:29.491083 131446876743488 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.222 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 5 seconds) (per-host) I0423 01:57:29.491157 131446876743488 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 5.553343s (batch_requests_ready=0.085493s, total_serialization_initiated=5.463230s, others=0.004620s) I0423 01:57:29.491234 131446876743488 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 5.553919s (all_items=0.000019s, per_item={'items': '0.00001884'}, temp_paths=5.553900) I0423 01:57:29.493002 131414878062144 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 01:57:29.493099 131446876743488 async_checkpointer.py:561] Finished blocking save. Time taken: 5.640926s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/9. I0423 01:57:29.493299 131446876743488 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0423 01:57:29.493441 131415272326720 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 01:57:29.493511 131446876743488 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776909419.1027038, 'wait_for_prev_duration_secs': 24.74831509590149, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776909443.8521526, 'checkpointer_blocking_duration_secs': 5.6410582065582275, 'get_old_steps_start_time': 1776909449.493228, 'get_old_steps_duration_secs': 3.24249267578125e-05, 'checkpoint_manager_blocking_start_time': 1776909419.102665, 'checkpoint_manager_blocking_duration_secs': 30.3908212184906} I0423 01:57:29.493624 131446876743488 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0423 01:57:29.493659 131446876743488 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 01:57:30.225884 131415687562816 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/9/items/array_metadatas/process_0 I0423 01:58:04.755298 131415729505856 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 35.263160s (commit=34.783731s, array_metadata_write=0.479429s) I0423 01:58:04.756362 131414878062144 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 309.584 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 40 seconds) (per-host) I0423 01:58:04.756420 131414878062144 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 35.263275s. I0423 01:58:05.238172 131414878062144 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 01:58:05.644512 131414878062144 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 01:58:05.648384 131414878062144 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.571366s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/9/items I0423 01:58:05.649187 131414878062144 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/9/items I0423 01:58:05.896651 131414878062144 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/9 I0423 01:58:06.589482 131414878062144 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/9`. I0423 01:58:06.590229 131414878062144 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 42.738063s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_11_optimizer_offload_false/checkpoints/9 I0423 01:58:06.590306 131414878062144 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 37.097164s. I0423 01:58:06.590430 131415272326720 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 01:58:06.590478 131415272326720 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 01:58:06.590535 131415272326720 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0423 01:58:06.590571 131415272326720 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0423 01:58:06.590756 131446876743488 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0423 01:58:06.590913 131446876743488 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 01:58:06.592026 131446876743488 metric_logger.py:196] completed step: 9, seconds: 0.120, TFLOP/s/device: 113.522, Tokens/s/device: 17111.299, total_weights: 16384, loss: 5.915, lm_loss: 5.915, perplexity: 370.653 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops