XPK Start: Sat Apr 25 14:54:42 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-25 14:55:07.154780: 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) I0425 14:55:07.365833 134503122814784 max_utils.py:273] Attempting to initialize the jax distributed system... I0425 14:55:16.405778 134503122814784 distributed.py:149] Starting JAX distributed service on [::]:8482 I0425 14:55:16.408496 134503122814784 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-sv-flixy-slice-job-0-0.mt-16-zarr3-false-sv-flixy:8482 I0425 14:55:17.396283 134503122814784 max_utils.py:284] Jax distributed system initialized! I0425 14:55:22.402793 134503122814784 max_utils.py:800] System Information: Jax Version: 0.9.2 I0425 14:55:22.402898 134503122814784 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0425 14:55:22.402938 134503122814784 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0425 14:55:22.402972 134503122814784 train_utils.py:391] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0425 14:55:23.101270 134503122814784 maxtext_utils.py:1771] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 14:55:23.214608 134503122814784 checkpointing.py:688] Setting up checkpoint logger... I0425 14:55:23.214730 134503122814784 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=False I0425 14:55:23.214777 134503122814784 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0425 14:55:23.214983 134503122814784 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7a53d77ef890>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0425 14:55:26.107438 134503122814784 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0425 14:55:26.107675 134503122814784 checkpoint_manager.py:708] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a40441aeab0>}, handler_registry=None I0425 14:55:26.107913 134503122814784 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a40441aeab0>` 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`. I0425 14:55:26.107969 134503122814784 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a400c52c5f0>` 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`. I0425 14:55:26.108018 134503122814784 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 0x7a40441aeab0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a40441aeab0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a400c52c5f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a400c52c5f0>}). I0425 14:55:26.108442 134503122814784 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0425 14:55:26.108532 134503122814784 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7a400c54f560> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0425 14:55:26.904645 134503122814784 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/nnx_xpk_feat_nnx_post_train_fixes_20260425_121355_16_zarr3_false/checkpoints I0425 14:55:27.133538 134503122814784 checkpoint_manager.py:929] [process=6][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_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=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, lightweight_initialize=False), root_directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/nnx_xpk_feat_nnx_post_train_fixes_20260425_121355_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7a52af995730> I0425 14:55:27.133712 134503122814784 checkpointing.py:302] Checkpoint manager created! I0425 14:55:28.111265 134503122814784 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0425 14:55:28.111386 134503122814784 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0425 14:55:29.574529 134503122814784 maxtext_utils.py:1889] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0425 14:55:29.574650 134503122814784 maxtext_utils.py:1889] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0425 14:55:29.574702 134503122814784 maxtext_utils.py:1889] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0425 14:55:29.574747 134503122814784 maxtext_utils.py:1889] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0425 14:55:29.574783 134503122814784 maxtext_utils.py:1889] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0425 14:55:29.574818 134503122814784 maxtext_utils.py:1889] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0425 14:55:29.574854 134503122814784 maxtext_utils.py:1889] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0425 14:55:29.574889 134503122814784 maxtext_utils.py:1889] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0425 14:55:29.574924 134503122814784 maxtext_utils.py:1889] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0425 14:55:29.574957 134503122814784 maxtext_utils.py:1889] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0425 14:55:29.574990 134503122814784 maxtext_utils.py:1889] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0425 14:55:29.575022 134503122814784 maxtext_utils.py:1889] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0425 14:55:29.973079 134503122814784 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 14:55:29.973250 134503122814784 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 14:55:29.978616 134503122814784 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 14:55:29.978678 134503122814784 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 14:55:29.995296 134503122814784 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn'). I0425 14:55:29.995362 134503122814784 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 14:55:30.011196 134503122814784 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn'). I0425 14:55:30.011254 134503122814784 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 14:55:30.036456 134503122814784 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 14:55:30.036528 134503122814784 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 14:55:30.052580 134503122814784 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 14:55:30.052642 134503122814784 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 14:55:30.068543 134503122814784 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 14:55:30.068605 134503122814784 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 14:55:30.098880 134503122814784 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch_attn', 'activation_length_attn', 'activation_heads', 'activation_kv'). I0425 14:55:30.098954 134503122814784 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 14:55:30.118993 134503122814784 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0425 14:55:30.119054 134503122814784 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0425 14:55:41.684681 134503122814784 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0425 14:55:41.686947 134503122814784 metric_logger.py:301] number parameters: 1.104 billion I0425 14:55:53.035346 134503122814784 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0425 14:55:53.187344 134503122814784 checkpointing.py:798] Waited 0.15198254585266113 seconds for step 0 to finish before starting checkpointing. I0425 14:55:53.189767 134503122814784 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 14:55:53.191644 134503122814784 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0425 14:55:53.193046 134503122814784 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/nnx_xpk_feat_nnx_post_train_fixes_20260425_121355_16_zarr3_false/checkpoints/0. I0425 14:55:53.482259 134503122814784 signaling_client.py:364] Using JaxDistributedSignalingClient I0425 14:55:53.483184 134503122814784 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0425 14:55:53.483239 134503122814784 replica_slices.py:424] 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 I0425 14:55:53.760539 134503122814784 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.279048s I0425 14:55:53.760718 134503122814784 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.354 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28813862800598145 s) (per-host) I0425 14:55:53.760771 134503122814784 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.288203s (batch_requests_ready=0.003049s, total_serialization_initiated=0.285081s, others=0.000073s) I0425 14:55:53.760884 134503122814784 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.292389s (all_items=0.000028s, per_item={'items': '0.00002766'}, temp_paths=0.292362) I0425 14:55:53.761696 134503122814784 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.57 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/nnx_xpk_feat_nnx_post_train_fixes_20260425_121355_16_zarr3_false/checkpoints/0. I0425 14:55:53.762042 134377544644352 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 15:15:53.762003 I0425 14:55:53.765912 134503122814784 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0425 14:55:53.766231 134375952365312 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 14:55:53.766393 134503122814784 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/nnx_xpk_feat_nnx_post_train_fixes_20260425_121355_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777128953.1897492, 'wait_for_prev_duration_secs': 6.031990051269531e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777128953.1916833, 'checkpointer_blocking_duration_secs': 0.5704715251922607, 'get_old_steps_start_time': 1777128953.7621794, 'get_old_steps_duration_secs': 3.147125244140625e-05, 'checkpoint_manager_blocking_start_time': 1777128953.1877344, 'checkpoint_manager_blocking_duration_secs': 0.5786218643188477} I0425 14:55:53.766501 134503122814784 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0425 14:55:53.766569 134503122814784 max_utils.py:750] Memstats: After params initialized: I0425 14:55:53.766617 134503122814784 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0425 14:55:53.766648 134503122814784 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0425 14:55:53.766674 134503122814784 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0425 14:55:53.766697 134503122814784 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0425 14:55:54.082056 134503122814784 metric_logger.py:196] completed step: 0, seconds: 11.347, TFLOP/s/device: 1.197, Tokens/s/device: 180.483, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52786.469 I0425 14:55:54.252663 134503122814784 metric_logger.py:196] completed step: 1, seconds: 1.045, TFLOP/s/device: 13.003, Tokens/s/device: 1959.946, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52786.469 I0425 14:55:54.702550 134503122814784 metric_logger.py:196] completed step: 2, seconds: 0.024, TFLOP/s/device: 554.939, Tokens/s/device: 83646.463, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38565.637 I0425 14:55:54.850163 134503122814784 metric_logger.py:196] completed step: 3, seconds: 0.440, TFLOP/s/device: 30.910, Tokens/s/device: 4659.109, total_weights: 65536, loss: 9.979, lm_loss: 9.979, perplexity: 21564.932 I0425 14:55:54.855912 134503122814784 checkpointing.py:794] Waiting for step 5 to finish before checkpoint... I0425 14:55:55.145042 134503122814784 checkpointing.py:798] Waited 0.28910136222839355 seconds for step 5 to finish before starting checkpointing. I0425 14:55:55.147745 134503122814784 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 14:55:56.817905 2757 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0425 14:55:58.828957 134376465860352 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/nnx_xpk_feat_nnx_post_train_fixes_20260425_121355_16_zarr3_false/checkpoints/0/items/array_metadatas/process_6 I0425 14:56:36.065006 134377544644352 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.086 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.59238576889038 s) (per-host) I0425 14:56:36.065153 134377544644352 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.302482s. I0425 14:56:41.948337 134377544644352 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 48.185647s. I0425 14:56:41.948613 134375952365312 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 14:56:41.948734 134375952365312 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 14:56:41.948786 134375952365312 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0425 14:56:41.951162 134375952365312 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0425 14:56:41.951343 134503122814784 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0425 14:56:41.951474 134503122814784 checkpoint_manager.py:1452] Waiting for previous save to complete took 46.803730 seconds. If this number is high, consider checkpointing less frequently. I0425 14:56:41.953635 134503122814784 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 5 I0425 14:56:41.956196 134503122814784 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/nnx_xpk_feat_nnx_post_train_fixes_20260425_121355_16_zarr3_false/checkpoints/5. I0425 14:56:42.659754 134503122814784 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0425 14:56:42.659852 134503122814784 replica_slices.py:424] 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 I0425 14:56:42.692518 134503122814784 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.034498s I0425 14:56:42.692693 134503122814784 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 37.314 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04134011268615723 s) (per-host) I0425 14:56:42.692744 134503122814784 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.041401s (batch_requests_ready=0.002515s, total_serialization_initiated=0.038819s, others=0.000067s) I0425 14:56:42.692836 134503122814784 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.046404s (all_items=0.000014s, per_item={'items': '0.00001431'}, temp_paths=0.046390) I0425 14:56:42.693470 134503122814784 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.74 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/nnx_xpk_feat_nnx_post_train_fixes_20260425_121355_16_zarr3_false/checkpoints/5. I0425 14:56:42.693793 134376465860352 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 15:16:42.693756 I0425 14:56:42.717556 134503122814784 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0425 14:56:42.717839 134375952365312 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 14:56:42.717982 134503122814784 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/nnx_xpk_feat_nnx_post_train_fixes_20260425_121355_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777128955.147715, 'wait_for_prev_duration_secs': 46.80373001098633, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777129001.953676, 'checkpointer_blocking_duration_secs': 0.7402632236480713, 'get_old_steps_start_time': 1777129002.6939623, 'get_old_steps_duration_secs': 2.8848648071289062e-05, 'checkpoint_manager_blocking_start_time': 1777128955.145432, 'checkpoint_manager_blocking_duration_secs': 47.57251334190369} I0425 14:56:42.718111 134503122814784 checkpointing.py:409] Started an asynchronous checkpoint save for step 5 I0425 14:56:42.719196 134503122814784 metric_logger.py:196] completed step: 4, seconds: 0.165, TFLOP/s/device: 82.571, Tokens/s/device: 12446.065, total_weights: 65536, loss: 9.457, lm_loss: 9.457, perplexity: 12802.436 I0425 14:56:42.727068 134503122814784 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.449, Tokens/s/device: 13934.911, total_weights: 65536, loss: 8.955, lm_loss: 8.955, perplexity: 7748.778 I0425 14:56:48.538161 134349647226624 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/nnx_xpk_feat_nnx_post_train_fixes_20260425_121355_16_zarr3_false/checkpoints/5/items/array_metadatas/process_6 I0425 14:57:06.004179 134503122814784 metric_logger.py:196] completed step: 6, seconds: 47.869, TFLOP/s/device: 0.284, Tokens/s/device: 42.783, total_weights: 65536, loss: 8.465, lm_loss: 8.465, perplexity: 4743.984 I0425 14:57:06.151960 134503122814784 metric_logger.py:196] completed step: 7, seconds: 23.132, TFLOP/s/device: 0.587, Tokens/s/device: 88.535, total_weights: 65536, loss: 7.998, lm_loss: 7.998, perplexity: 2976.266 I0425 14:57:06.299460 134503122814784 metric_logger.py:196] completed step: 8, seconds: 0.153, TFLOP/s/device: 88.856, Tokens/s/device: 13393.324, total_weights: 65536, loss: 7.567, lm_loss: 7.567, perplexity: 1933.479 I0425 14:57:06.447610 134503122814784 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0425 14:57:06.448360 134503122814784 checkpointing.py:798] Waited 0.0007655620574951172 seconds for step 9 to finish before starting checkpointing. I0425 14:57:06.450251 134503122814784 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 14:57:25.210919 134376465860352 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.115 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.559526681900024 s) (per-host) I0425 14:57:25.211043 134376465860352 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.517136s. I0425 14:57:31.596702 134376465860352 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 48.902780s. I0425 14:57:31.596987 134375952365312 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 14:57:31.597126 134375952365312 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 14:57:31.597175 134375952365312 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0425 14:57:31.604548 134375952365312 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0425 14:57:31.604724 134503122814784 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0425 14:57:31.604850 134503122814784 checkpoint_manager.py:1452] Waiting for previous save to complete took 25.154603 seconds. If this number is high, consider checkpointing less frequently. I0425 14:57:31.606525 134503122814784 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 9 I0425 14:57:31.608659 134503122814784 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/nnx_xpk_feat_nnx_post_train_fixes_20260425_121355_16_zarr3_false/checkpoints/9. I0425 14:57:31.903982 134503122814784 jax_array_handlers.py:360] Scheduling D2H of 69 prioritized jax.Array. I0425 14:57:31.904076 134503122814784 replica_slices.py:424] 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 I0425 14:57:31.940618 134503122814784 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.038219s I0425 14:57:31.940772 134503122814784 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 34.250 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04503893852233887 s) (per-host) I0425 14:57:31.940820 134503122814784 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.045097s (batch_requests_ready=0.002685s, total_serialization_initiated=0.042349s, others=0.000063s) I0425 14:57:31.940909 134503122814784 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.049217s (all_items=0.000011s, per_item={'items': '0.00001144'}, temp_paths=0.049206) I0425 14:57:31.941621 134503122814784 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.34 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/nnx_xpk_feat_nnx_post_train_fixes_20260425_121355_16_zarr3_false/checkpoints/9. I0425 14:57:31.941949 134375952365312 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 15:17:31.941912 I0425 14:57:31.944046 134503122814784 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0425 14:57:31.944276 134349647226624 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 14:57:31.944373 134503122814784 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/nnx_xpk_feat_nnx_post_train_fixes_20260425_121355_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777129026.4502242, 'wait_for_prev_duration_secs': 25.154603242874146, 'time_between_consecutive_saves_sec': 24.499019622802734, 'checkpointer_blocking_start_time': 1777129051.6065648, 'checkpointer_blocking_duration_secs': 0.33555030822753906, 'get_old_steps_start_time': 1777129051.9421372, 'get_old_steps_duration_secs': 2.9087066650390625e-05, 'checkpoint_manager_blocking_start_time': 1777129026.4486287, 'checkpoint_manager_blocking_duration_secs': 25.49571418762207} I0425 14:57:31.944525 134503122814784 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0425 14:57:31.944570 134503122814784 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 14:57:37.024600 134373772859136 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 69 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260425_121355/nnx_xpk_feat_nnx_post_train_fixes_20260425_121355_16_zarr3_false/checkpoints/9/items/array_metadatas/process_6 I0425 14:58:15.099051 134375952365312 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 36.562 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.20327806472778 s) (per-host) I0425 14:58:15.099194 134375952365312 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.157133s. I0425 14:58:19.714046 134375952365312 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 47.771969s. I0425 14:58:19.714308 134349647226624 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 14:58:19.714368 134349647226624 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 14:58:19.714405 134349647226624 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0425 14:58:19.717031 134349647226624 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0425 14:58:19.717206 134503122814784 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0425 14:58:19.717401 134503122814784 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 14:58:19.718357 134503122814784 metric_logger.py:196] completed step: 9, seconds: 0.148, TFLOP/s/device: 91.744, Tokens/s/device: 13828.588, total_weights: 65536, loss: 7.176, lm_loss: 7.176, perplexity: 1307.366 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 25 14:58:32 UTC 2026 EXIT_CODE=0