XPK Start: Sat Apr 18 22:23:15 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-18 22:23:39.307631: 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) I0418 22:23:39.487503 139731399882560 max_utils.py:273] Attempting to initialize the jax distributed system... I0418 22:23:48.528583 139731399882560 distributed.py:149] Starting JAX distributed service on [::]:8482 I0418 22:23:48.531029 139731399882560 distributed.py:172] Connecting to JAX distributed service on mt-04-int8-hekfb-slice-job-0-0.mt-04-int8-hekfb:8482 I0418 22:23:50.443337 139731399882560 max_utils.py:284] Jax distributed system initialized! I0418 22:23:56.599589 139731399882560 max_utils.py:800] System Information: Jax Version: 0.9.2 I0418 22:23:56.599690 139731399882560 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0418 22:23:56.599730 139731399882560 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 I0418 22:23:56.599765 139731399882560 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0418 22:23:57.304933 139731399882560 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0418 22:23:57.868635 139731399882560 checkpointing.py:688] Setting up checkpoint logger... I0418 22:23:57.868754 139731399882560 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0418 22:23:57.868799 139731399882560 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0418 22:23:57.869014 139731399882560 base_pytree_checkpoint_handler.py:441] 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 0x7f1508d38200>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0418 22:24:01.189159 139731399882560 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0418 22:24:01.189360 139731399882560 checkpoint_manager.py:708] [process=3][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7eff9045d940>}, handler_registry=None I0418 22:24:01.189601 139731399882560 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7eff9045d940>` 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`. I0418 22:24:01.189651 139731399882560 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f0090576de0>` 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`. I0418 22:24:01.189686 139731399882560 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 0x7eff9045d940>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7eff9045d940>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f0090576de0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f0090576de0>}). I0418 22:24:01.190022 139731399882560 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0418 22:24:01.190092 139731399882560 async_checkpointer.py:192] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f00905189a0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0418 22:24:02.650464 139731399882560 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152/nnx_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152_04_int8/checkpoints I0418 22:24:03.089482 139731399882560 checkpoint_manager.py:929] [process=3][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=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, lightweight_initialize=False), root_directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152/nnx_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152_04_int8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f00905f37a0> I0418 22:24:03.089653 139731399882560 checkpointing.py:302] Checkpoint manager created! I0418 22:24:04.465710 139731399882560 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0418 22:24:04.465828 139731399882560 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0418 22:24:06.770022 139731399882560 maxtext_utils.py:1836] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0418 22:24:06.770147 139731399882560 maxtext_utils.py:1836] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0418 22:24:06.770201 139731399882560 maxtext_utils.py:1836] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0418 22:24:06.770287 139731399882560 maxtext_utils.py:1836] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0418 22:24:06.770335 139731399882560 maxtext_utils.py:1836] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0418 22:24:06.770369 139731399882560 maxtext_utils.py:1836] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0418 22:24:06.770403 139731399882560 maxtext_utils.py:1836] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0418 22:24:06.770439 139731399882560 maxtext_utils.py:1836] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0418 22:24:06.770472 139731399882560 maxtext_utils.py:1836] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0418 22:24:06.770504 139731399882560 maxtext_utils.py:1836] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0418 22:24:06.770536 139731399882560 maxtext_utils.py:1836] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0418 22:24:06.770567 139731399882560 maxtext_utils.py:1836] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0418 22:24:07.179726 139731399882560 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0418 22:24:07.179825 139731399882560 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 22:24:07.185414 139731399882560 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0418 22:24:07.185468 139731399882560 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 22:24:07.201694 139731399882560 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 22:24:07.201759 139731399882560 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 22:24:07.217448 139731399882560 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 22:24:07.217507 139731399882560 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 22:24:07.297454 139731399882560 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 22:24:07.297544 139731399882560 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 22:24:07.313345 139731399882560 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 22:24:07.313404 139731399882560 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 22:24:07.329102 139731399882560 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 22:24:07.329163 139731399882560 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 22:24:07.359070 139731399882560 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0418 22:24:07.359145 139731399882560 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 22:24:07.844851 139731399882560 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0418 22:24:07.844943 139731399882560 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0418 22:24:21.472935 139731399882560 max_utils.py:791] Total memory size: 1.8 GB, Output size: 0.4 GB, Temp size: 1.4 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0418 22:24:21.476411 139731399882560 metric_logger.py:301] number parameters: 1.104 billion I0418 22:24:35.803180 139731399882560 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0418 22:24:36.044021 139731399882560 checkpointing.py:798] Waited 0.24082589149475098 seconds for step 0 to finish before starting checkpointing. I0418 22:24:36.046639 139731399882560 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 22:24:36.048630 139731399882560 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 0 I0418 22:24:36.050214 139731399882560 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152/nnx_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152_04_int8/checkpoints/0. I0418 22:24:36.804735 139731399882560 signaling_client.py:364] Using JaxDistributedSignalingClient I0418 22:24:36.805697 139731399882560 jax_array_handlers.py:360] Scheduling D2H of 111 prioritized jax.Array. I0418 22:24:36.805756 139731399882560 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 I0418 22:24:37.085754 139731399882560 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.282635s I0418 22:24:37.085930 139731399882560 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 5.190 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.29720139503479004 s) (per-host) I0418 22:24:37.085983 139731399882560 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.297265s (batch_requests_ready=0.004624s, total_serialization_initiated=0.292569s, others=0.000072s) I0418 22:24:37.086098 139731399882560 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.301885s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.301867) I0418 22:24:37.086891 139731399882560 event_tracking.py:125] [process=3] [async] Finished blocking save in 1.04 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152/nnx_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152_04_int8/checkpoints/0. I0418 22:24:37.087268 139598199428864 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 22:44:37.087218 I0418 22:24:37.106895 139731399882560 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0418 22:24:37.107183 139598224606976 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0418 22:24:37.107361 139731399882560 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152/nnx_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776551076.04662, 'wait_for_prev_duration_secs': 6.198883056640625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776551076.0486703, 'checkpointer_blocking_duration_secs': 1.0387585163116455, 'get_old_steps_start_time': 1776551077.0874527, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776551076.0444767, 'checkpoint_manager_blocking_duration_secs': 1.062847375869751} I0418 22:24:37.107493 139731399882560 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0418 22:24:37.107576 139731399882560 max_utils.py:750] Memstats: After params initialized: I0418 22:24:37.107627 139731399882560 max_utils.py:756] Using (GB) 0.44 / 31.25 (1.408000%) on TPU_10(process=3,(2,2,0,0)) I0418 22:24:37.107661 139731399882560 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_11(process=3,(3,2,0,0)) I0418 22:24:37.107689 139731399882560 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_14(process=3,(2,3,0,0)) I0418 22:24:37.107715 139731399882560 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_15(process=3,(3,3,0,0)) I0418 22:24:37.437445 139731399882560 metric_logger.py:196] completed step: 0, seconds: 14.325, TFLOP/s/device: 0.948, Tokens/s/device: 142.968, total_weights: 65536, loss: 10.872, lm_loss: 10.872, perplexity: 52668.902 I0418 22:24:37.687101 139731399882560 metric_logger.py:196] completed step: 1, seconds: 1.632, TFLOP/s/device: 8.326, Tokens/s/device: 1255.048, total_weights: 65536, loss: 10.872, lm_loss: 10.872, perplexity: 52668.902 I0418 22:24:38.133258 139731399882560 metric_logger.py:196] completed step: 2, seconds: 0.023, TFLOP/s/device: 582.489, Tokens/s/device: 87799.023, total_weights: 65536, loss: 10.260, lm_loss: 10.260, perplexity: 28557.658 I0418 22:24:38.362734 139731399882560 metric_logger.py:196] completed step: 3, seconds: 0.425, TFLOP/s/device: 31.961, Tokens/s/device: 4817.452, total_weights: 65536, loss: 9.728, lm_loss: 9.728, perplexity: 16789.145 I0418 22:24:38.822539 139731399882560 metric_logger.py:196] completed step: 4, seconds: 0.257, TFLOP/s/device: 52.911, Tokens/s/device: 7975.326, total_weights: 65536, loss: 9.271, lm_loss: 9.271, perplexity: 10625.563 I0418 22:24:38.831073 139731399882560 metric_logger.py:196] completed step: 5, seconds: 0.229, TFLOP/s/device: 59.209, Tokens/s/device: 8924.564, total_weights: 65536, loss: 8.885, lm_loss: 8.885, perplexity: 7225.917 I0418 22:24:41.600633 2918 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0418 22:24:44.486366 139598207821568 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 111 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152/nnx_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152_04_int8/checkpoints/0/items/array_metadatas/process_3 I0418 22:25:07.574345 139731399882560 metric_logger.py:196] completed step: 6, seconds: 0.460, TFLOP/s/device: 29.538, Tokens/s/device: 4452.251, total_weights: 65536, loss: 8.586, lm_loss: 8.586, perplexity: 5354.806 I0418 22:25:07.803818 139731399882560 metric_logger.py:196] completed step: 7, seconds: 28.516, TFLOP/s/device: 0.476, Tokens/s/device: 71.819, total_weights: 65536, loss: 8.378, lm_loss: 8.378, perplexity: 4350.756 I0418 22:25:08.033365 139731399882560 metric_logger.py:196] completed step: 8, seconds: 0.235, TFLOP/s/device: 57.702, Tokens/s/device: 8697.462, total_weights: 65536, loss: 8.249, lm_loss: 8.249, perplexity: 3824.619 I0418 22:25:08.264459 139731399882560 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0418 22:25:08.265419 139731399882560 checkpointing.py:798] Waited 0.0009770393371582031 seconds for step 9 to finish before starting checkpointing. I0418 22:25:08.267512 139731399882560 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 22:25:11.326264 139598199428864 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 45.736 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 34.53749227523804 s) (per-host) I0418 22:25:11.326372 139598199428864 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 34.238982s. I0418 22:25:21.605541 139598199428864 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 44.518134s. I0418 22:25:21.605824 139598224606976 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 22:25:21.605952 139598224606976 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0418 22:25:21.606002 139598224606976 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0418 22:25:21.611212 139598224606976 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0418 22:25:21.611384 139731399882560 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0418 22:25:21.611527 139731399882560 checkpoint_manager.py:1452] Waiting for previous save to complete took 13.344012 seconds. If this number is high, consider checkpointing less frequently. I0418 22:25:21.613274 139731399882560 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 9 I0418 22:25:21.615602 139731399882560 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152/nnx_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152_04_int8/checkpoints/9. I0418 22:25:22.322028 139731399882560 jax_array_handlers.py:360] Scheduling D2H of 111 prioritized jax.Array. I0418 22:25:22.322126 139731399882560 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 I0418 22:25:22.367750 139731399882560 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.048209s I0418 22:25:22.367924 139731399882560 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 25.667 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.06009984016418457 s) (per-host) I0418 22:25:22.367976 139731399882560 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.060163s (batch_requests_ready=0.004230s, total_serialization_initiated=0.055862s, others=0.000072s) I0418 22:25:22.368100 139731399882560 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.064314s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.064298) I0418 22:25:22.368845 139731399882560 event_tracking.py:125] [process=3] [async] Finished blocking save in 0.76 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152/nnx_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152_04_int8/checkpoints/9. I0418 22:25:22.369164 139598224606976 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 22:45:22.369125 I0418 22:25:22.371145 139731399882560 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0418 22:25:22.371398 139598207821568 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0418 22:25:22.371529 139731399882560 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152/nnx_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152_04_int8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776551108.2674837, 'wait_for_prev_duration_secs': 13.344011783599854, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776551121.6133115, 'checkpointer_blocking_duration_secs': 0.7559678554534912, 'get_old_steps_start_time': 1776551122.369302, 'get_old_steps_duration_secs': 3.170967102050781e-05, 'checkpoint_manager_blocking_start_time': 1776551108.265696, 'checkpoint_manager_blocking_duration_secs': 14.105797529220581} I0418 22:25:22.371659 139731399882560 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0418 22:25:22.371702 139731399882560 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 22:25:27.868790 139598216214272 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 111 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152/nnx_xpk_feat_nnx_linen_converter_and_sharding_tools_20260418_204152_04_int8/checkpoints/9/items/array_metadatas/process_3 I0418 22:26:04.064660 139598224606976 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 37.828 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.75679922103882 s) (per-host) I0418 22:26:04.064786 139598224606976 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.694976s. I0418 22:26:15.312825 139598224606976 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 52.942999s. I0418 22:26:15.313127 139598207821568 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 22:26:15.313274 139598207821568 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0418 22:26:15.313322 139598207821568 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0418 22:26:15.315650 139598207821568 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0418 22:26:15.316817 139731399882560 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0418 22:26:15.316999 139731399882560 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 22:26:15.317953 139731399882560 metric_logger.py:196] completed step: 9, seconds: 0.229, TFLOP/s/device: 59.229, Tokens/s/device: 8927.637, total_weights: 65536, loss: 8.174, lm_loss: 8.174, perplexity: 3546.104 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 18 22:26:25 UTC 2026 EXIT_CODE=0