2026-04-17 13:39:06.830613: 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) I0417 13:39:06.948746 125724073790592 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0417 13:39:54.371632 125724073790592 max_utils.py:800] System Information: Jax Version: 0.9.2 I0417 13:39:54.371740 125724073790592 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0417 13:39:54.371774 125724073790592 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Apr 6 2026 20:48:10 (1775533690) cl/895581894 I0417 13:39:54.371799 125724073790592 train_utils.py:335] WARNING: 'dataset_path' might be pointing your local file system I0417 13:39:54.371821 125724073790592 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0417 13:39:54.371908 125724073790592 train.py:671] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. W0417 13:39:54.463325 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:39:54.866580 125724073790592 maxtext_utils.py:1548] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0417 13:39:54.866971 125724073790592 checkpointing.py:677] Setting up checkpoint logger... I0417 13:39:54.867025 125724073790592 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0417 13:39:54.867103 125724073790592 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0417 13:39:54.867649 125724073790592 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 0x7257d3df8950>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0417 13:39:57.238583 125724073790592 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0417 13:39:57.238782 125724073790592 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 0x725157b323c0>}, handler_registry=None I0417 13:39:57.239125 125724073790592 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x725157b323c0>` 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`. I0417 13:39:57.239171 125724073790592 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x725157b32510>` 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`. I0417 13:39:57.239199 125724073790592 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 0x725157b323c0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x725157b323c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x725157b32510>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x725157b32510>}). I0417 13:39:57.239498 125724073790592 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0417 13:39:57.381645 125724073790592 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints I0417 13:39:57.381882 125724073790592 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=False, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x725157f40110> I0417 13:39:57.381976 125724073790592 checkpointing.py:301] Checkpoint manager created! W0417 13:39:57.398626 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:39:57.963788 125724073790592 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 13:39:57.963902 125724073790592 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0417 13:39:58.050346 125724073790592 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 13:39:58.050424 125724073790592 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0417 13:39:58.064670 125724073790592 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 13:39:58.064719 125724073790592 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0417 13:39:58.085474 125724073790592 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 13:39:58.085528 125724073790592 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0417 13:39:58.099816 125724073790592 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 13:39:58.099866 125724073790592 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0417 13:39:58.114144 125724073790592 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 13:39:58.114194 125724073790592 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0417 13:39:58.135830 125724073790592 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0417 13:39:58.135888 125724073790592 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0417 13:39:58.154184 125724073790592 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0417 13:39:58.154235 125724073790592 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0417 13:39:58.349770 125724073790592 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0417 13:39:58.349944 125724073790592 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. W0417 13:39:58.400298 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. [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 I0417 13:39:58.783690 125724073790592 maxtext_utils.py:1651] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0417 13:39:58.783807 125724073790592 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0417 13:39:58.783847 125724073790592 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0417 13:39:58.783890 125724073790592 maxtext_utils.py:1651] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0417 13:39:58.783928 125724073790592 maxtext_utils.py:1651] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0417 13:39:58.783957 125724073790592 maxtext_utils.py:1651] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0417 13:39:58.783995 125724073790592 maxtext_utils.py:1651] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0417 13:39:58.784033 125724073790592 maxtext_utils.py:1651] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0417 13:39:58.784116 125724073790592 maxtext_utils.py:1651] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0417 13:39:58.784145 125724073790592 maxtext_utils.py:1651] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0417 13:39:58.784182 125724073790592 maxtext_utils.py:1651] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0417 13:39:58.784216 125724073790592 maxtext_utils.py:1651] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0417 13:39:59.210533 125724073790592 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0417 13:39:59.210617 125724073790592 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0417 13:39:59.224344 125724073790592 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0417 13:39:59.224397 125724073790592 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). W0417 13:39:59.478232 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:39:59.591809 125724073790592 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. I0417 13:39:59.592302 125724073790592 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0417 13:39:59.592455 125724073790592 metric_logger.py:301] number parameters: 1.104 billion W0417 13:40:00.428253 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:40:00.523201 125724073790592 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0417 13:40:00.640325 125724073790592 checkpointing.py:776] Waited 0.1171114444732666 seconds for step 0 to finish before starting checkpointing. I0417 13:40:00.640848 125724073790592 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 13:40:00.641029 125724073790592 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0417 13:40:00.641109 125724073790592 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/0. I0417 13:40:00.816727 125724073790592 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/0 I0417 13:40:01.641756 125724073790592 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776433201385524896, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776433201385524896, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA I0417 13:40:01.642859 125724073790592 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient W0417 13:40:01.643212 125724073790592 future.py:328] [process=0][thread=MainThread][operation_id=1] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 13:40:01.741599 125616659629632 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/0/items I0417 13:40:02.317790 125724073790592 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0417 13:40:02.317895 125724073790592 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 I0417 13:40:02.335372 482799 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0417 13:40:02.335419 482799 google_auth_provider.cc:156] Using OAuth2 AuthProvider W0417 13:40:02.543780 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:40:02.555451 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:40:02.564893 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:40:02.578130 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:40:02.590893 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:40:02.600748 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:40:02.606547 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:40:02.611470 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:40:02.925794 125724073790592 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.608373s I0417 13:40:02.926485 125724073790592 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 17.626 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 700 milliseconds) (per-host) I0417 13:40:02.930722 125724073790592 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.704568s (batch_requests_ready=0.087950s, total_serialization_initiated=0.611936s, others=0.004682s) I0417 13:40:02.930832 125724073790592 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.288049s (all_items=0.000047s, per_item={'items': '0.00004697'}, temp_paths=1.288002) W0417 13:40:02.930875 125724073790592 future.py:328] [process=0][thread=MainThread][operation_id=1] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 13:40:03.707537 125616644949568 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/0/items/array_metadatas/process_0 I0417 13:40:39.930232 125622332425792 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 36.998998s (commit=36.566272s, array_metadata_write=0.432726s) I0417 13:40:39.931299 125724073790592 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 335.148 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 37 seconds) (per-host) I0417 13:40:40.170283 125724073790592 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776433201385524896, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA I0417 13:40:40.566645 125724073790592 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776433201385524896, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/0/_CHECKPOINT_METADATA I0417 13:40:40.715372 125724073790592 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0417 13:40:41.122694 125724073790592 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0417 13:40:41.122848 125724073790592 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.555825s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/0/items I0417 13:40:41.123538 125724073790592 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/0/items I0417 13:40:41.365336 125724073790592 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/0 I0417 13:40:42.066714 125724073790592 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/0`. I0417 13:40:42.067324 125724073790592 checkpointer.py:290] Finished synchronous save in 41.43 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/0 I0417 13:40:42.067523 125724073790592 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0417 13:40:42.067569 125724073790592 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=0] CheckpointManager Save Finalize is done on all hosts. I0417 13:40:42.067593 125724073790592 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=0] Finished synchronous save. I0417 13:40:42.067641 125724073790592 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776433200.6408336, 'wait_for_prev_duration_secs': 4.1961669921875e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776433200.6410668, 'checkpointer_blocking_duration_secs': 41.42630171775818, 'get_old_steps_start_time': 1776433242.0674047, 'get_old_steps_duration_secs': 3.719329833984375e-05, 'checkpoint_manager_blocking_start_time': 1776433200.640712, 'checkpoint_manager_blocking_duration_secs': 41.4269015789032} I0417 13:40:42.067736 125724073790592 checkpointing.py:410] Saved a checkpoint at step 0. I0417 13:40:42.067787 125724073790592 max_utils.py:750] Memstats: After params initialized: I0417 13:40:42.067839 125724073790592 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_0(process=0,(0,0,0,0)) I0417 13:40:42.067866 125724073790592 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_1(process=0,(1,0,0,0)) I0417 13:40:42.067888 125724073790592 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_2(process=0,(0,1,0,0)) I0417 13:40:42.067907 125724073790592 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_3(process=0,(1,1,0,0)) I0417 13:40:42.067924 125724073790592 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_4(process=0,(0,2,0,0)) I0417 13:40:42.067941 125724073790592 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_5(process=0,(1,2,0,0)) I0417 13:40:42.067960 125724073790592 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_6(process=0,(0,3,0,0)) I0417 13:40:42.067978 125724073790592 max_utils.py:756] Using (GB) 1.59 / 31.25 (5.088000%) on TPU_7(process=0,(1,3,0,0)) W0417 13:40:42.074012 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:40:42.079543 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. W0417 13:40:42.084055 480752 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions. I0417 13:40:42.274143 125724073790592 metric_logger.py:196] completed step: 0, seconds: 0.930, TFLOP/s/device: 14.606, Tokens/s/device: 2201.552, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0417 13:40:42.274932 125724073790592 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/tensorboard/' I0417 13:40:42.389511 125724073790592 metric_logger.py:196] completed step: 1, seconds: 41.748, TFLOP/s/device: 0.325, Tokens/s/device: 49.056, total_weights: 16384, loss: 10.874, lm_loss: 10.874, perplexity: 52770.941 I0417 13:40:42.507838 125724073790592 metric_logger.py:196] completed step: 2, seconds: 0.007, TFLOP/s/device: 2047.179, Tokens/s/device: 308573.151, total_weights: 16384, loss: 10.276, lm_loss: 10.276, perplexity: 29019.973 I0417 13:40:42.626439 125724073790592 metric_logger.py:196] completed step: 3, seconds: 0.117, TFLOP/s/device: 115.873, Tokens/s/device: 17465.610, total_weights: 16384, loss: 9.282, lm_loss: 9.282, perplexity: 10739.224 I0417 13:40:42.630965 125724073790592 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0417 13:40:42.862290 125724073790592 checkpointing.py:776] Waited 0.23128962516784668 seconds for step 5 to finish before starting checkpointing. I0417 13:40:42.862671 125724073790592 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 13:40:42.862818 125724073790592 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 5 I0417 13:40:42.862874 125724073790592 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/5. I0417 13:40:43.047066 125724073790592 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/5 W0417 13:40:43.863943 125724073790592 future.py:328] [process=0][thread=MainThread][operation_id=2] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 13:40:43.949433 125622332425792 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/5/items I0417 13:40:44.496084 125724073790592 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0417 13:40:44.496187 125724073790592 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 I0417 13:40:44.656248 125724073790592 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.160529s I0417 13:40:44.656602 125724073790592 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 49.519 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 249 milliseconds) (per-host) I0417 13:40:44.659911 125724073790592 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.252555s (batch_requests_ready=0.085485s, total_serialization_initiated=0.163540s, others=0.003529s) I0417 13:40:44.659981 125724073790592 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.796378s (all_items=0.000037s, per_item={'items': '0.00003743'}, temp_paths=0.796340) W0417 13:40:44.660012 125724073790592 future.py:328] [process=0][thread=MainThread][operation_id=2] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 13:40:45.436625 125622321940032 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/5/items/array_metadatas/process_0 I0417 13:41:13.957195 125616754001472 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 29.296874s (commit=28.862451s, array_metadata_write=0.434422s) I0417 13:41:13.958623 125724073790592 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 427.623 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 29 seconds) (per-host) I0417 13:41:14.713070 125724073790592 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0417 13:41:15.150013 125724073790592 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0417 13:41:15.150272 125724073790592 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.594796s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/5/items I0417 13:41:15.150974 125724073790592 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/5/items I0417 13:41:15.396723 125724073790592 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/5 I0417 13:41:16.085979 125724073790592 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/5`. I0417 13:41:16.086655 125724073790592 checkpointer.py:290] Finished synchronous save in 33.22 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/5 I0417 13:41:16.086840 125724073790592 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0417 13:41:16.086881 125724073790592 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=5] CheckpointManager Save Finalize is done on all hosts. I0417 13:41:16.086905 125724073790592 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=5] Finished synchronous save. I0417 13:41:16.086941 125724073790592 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776433242.8626502, 'wait_for_prev_duration_secs': 4.696846008300781e-05, 'time_between_consecutive_saves_sec': 0.7950654029846191, 'checkpointer_blocking_start_time': 1776433242.8628397, 'checkpointer_blocking_duration_secs': 33.223865270614624, 'get_old_steps_start_time': 1776433276.0867376, 'get_old_steps_duration_secs': 3.552436828613281e-05, 'checkpoint_manager_blocking_start_time': 1776433242.8625798, 'checkpoint_manager_blocking_duration_secs': 33.22433948516846} I0417 13:41:16.087036 125724073790592 checkpointing.py:410] Saved a checkpoint at step 5. I0417 13:41:16.088240 125724073790592 metric_logger.py:196] completed step: 4, seconds: 0.118, TFLOP/s/device: 115.556, Tokens/s/device: 17417.928, total_weights: 16384, loss: 8.327, lm_loss: 8.327, perplexity: 4134.812 I0417 13:41:16.099014 125724073790592 metric_logger.py:196] completed step: 5, seconds: 0.118, TFLOP/s/device: 114.956, Tokens/s/device: 17327.445, total_weights: 16384, loss: 7.367, lm_loss: 7.367, perplexity: 1583.522 I0417 13:41:16.211726 125724073790592 metric_logger.py:196] completed step: 6, seconds: 33.463, TFLOP/s/device: 0.406, Tokens/s/device: 61.201, total_weights: 16384, loss: 6.459, lm_loss: 6.459, perplexity: 638.363 I0417 13:41:16.329963 125724073790592 metric_logger.py:196] completed step: 7, seconds: 0.009, TFLOP/s/device: 1595.857, Tokens/s/device: 240544.985, total_weights: 16384, loss: 5.622, lm_loss: 5.622, perplexity: 276.577 I0417 13:41:16.448141 125724073790592 metric_logger.py:196] completed step: 8, seconds: 0.114, TFLOP/s/device: 119.698, Tokens/s/device: 18042.145, total_weights: 16384, loss: 4.864, lm_loss: 4.864, perplexity: 129.523 I0417 13:41:16.565759 125724073790592 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0417 13:41:16.566557 125724073790592 checkpointing.py:776] Waited 0.0008208751678466797 seconds for step 9 to finish before starting checkpointing. I0417 13:41:16.566805 125724073790592 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 13:41:16.566946 125724073790592 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 9 I0417 13:41:16.566998 125724073790592 checkpointer.py:241] [process=0] Started saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/9. I0417 13:41:16.739139 125724073790592 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/9 W0417 13:41:17.602861 125724073790592 future.py:328] [process=0][thread=MainThread][operation_id=3] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 13:41:17.684300 125622332425792 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/9/items I0417 13:41:18.250531 125724073790592 jax_array_handlers.py:347] Scheduling D2H of 39 prioritized jax.Array. I0417 13:41:18.250636 125724073790592 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 I0417 13:41:18.419724 125724073790592 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.169524s I0417 13:41:18.420007 125724073790592 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 48.296 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 255 milliseconds) (per-host) I0417 13:41:18.420077 125724073790592 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.255623s (batch_requests_ready=0.082863s, total_serialization_initiated=0.172537s, others=0.000223s) I0417 13:41:18.420137 125724073790592 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.817511s (all_items=0.000029s, per_item={'items': '0.00002861'}, temp_paths=0.817483) W0417 13:41:18.420167 125724073790592 future.py:328] [process=0][thread=MainThread][operation_id=3] _SignalingThread.join() waiting for signals ([]) blocking the main thread will slow down blocking save times. This is likely due to main thread calling result() on a CommitFuture. I0417 13:41:19.216918 125616754001472 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/9/items/array_metadatas/process_0 I0417 13:42:07.030754 125616743515712 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 48.610378s (commit=48.173873s, array_metadata_write=0.436505s) I0417 13:42:07.031813 125724073790592 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 258.594 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 48 seconds) (per-host) I0417 13:42:07.827305 125724073790592 array_metadata_store.py:367] [process=0][thread=MainThread] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0417 13:42:08.277869 125724073790592 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0417 13:42:08.278068 125724073790592 base_pytree_checkpoint_handler.py:1342] [process=0][thread=MainThread] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.585468s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/9/items I0417 13:42:08.278700 125724073790592 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/9/items I0417 13:42:08.537964 125724073790592 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/9 I0417 13:42:09.221427 125724073790592 atomicity.py:794] [process=0][thread=MainThread] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/9`. I0417 13:42:09.222093 125724073790592 checkpointer.py:290] Finished synchronous save in 52.66 seconds to gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints/9 I0417 13:42:09.222219 125724073790592 checkpoint_manager.py:2103] [process=0][thread=MainThread][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0417 13:42:09.222254 125724073790592 checkpoint_manager.py:2112] [process=0][thread=MainThread][step=9] CheckpointManager Save Finalize is done on all hosts. I0417 13:42:09.222277 125724073790592 checkpoint_manager.py:1570] [process=0][thread=MainThread][step=9] Finished synchronous save. I0417 13:42:09.222313 125724073790592 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260417_125240/linen_main_20260417_125240_14_async_ckpt_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': True, 'wait_for_prev_start_time': 1776433276.566783, 'wait_for_prev_duration_secs': 4.506111145019531e-05, 'time_between_consecutive_saves_sec': 0.479886531829834, 'checkpointer_blocking_start_time': 1776433276.5669658, 'checkpointer_blocking_duration_secs': 52.65517544746399, 'get_old_steps_start_time': 1776433329.2221582, 'get_old_steps_duration_secs': 1.8358230590820312e-05, 'checkpoint_manager_blocking_start_time': 1776433276.5667498, 'checkpoint_manager_blocking_duration_secs': 52.65554189682007} I0417 13:42:09.222412 125724073790592 checkpointing.py:410] Saved a checkpoint at step 9. I0417 13:42:09.222434 125724073790592 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 13:42:09.222455 125724073790592 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 13:42:09.223104 125724073790592 metric_logger.py:196] completed step: 9, seconds: 0.118, TFLOP/s/device: 114.765, Tokens/s/device: 17298.612, total_weights: 16384, loss: 4.189, lm_loss: 4.189, perplexity: 65.979 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops