XPK Start: Wed Apr 22 14:25:11 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-22 14:25:35.842503: 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) I0422 14:25:36.025049 133537724462912 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 14:25:45.067665 133537724462912 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 14:25:45.070137 133537724462912 distributed.py:172] Connecting to JAX distributed service on mt-05-fp8-os420-slice-job-0-0.mt-05-fp8-os420:8482 I0422 14:25:46.420503 133537724462912 max_utils.py:284] Jax distributed system initialized! I0422 14:25:52.646644 133537724462912 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 14:25:52.646748 133537724462912 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 14:25:52.646789 133537724462912 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 I0422 14:25:52.646824 133537724462912 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 14:25:53.348891 133537724462912 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 14:25:53.780835 133537724462912 checkpointing.py:688] Setting up checkpoint logger... I0422 14:25:53.780954 133537724462912 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 14:25:53.781000 133537724462912 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 14:25:53.781274 133537724462912 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 0x797327d8d790>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 14:25:56.605245 133537724462912 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0422 14:25:56.605483 133537724462912 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 0x795dc8682120>}, handler_registry=None I0422 14:25:56.605725 133537724462912 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x795dc8682120>` 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`. I0422 14:25:56.605772 133537724462912 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x795dc82d83b0>` 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`. I0422 14:25:56.605808 133537724462912 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 0x795dc8682120>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x795dc8682120>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x795dc82d83b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x795dc82d83b0>}). I0422 14:25:56.606150 133537724462912 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 14:25:56.606221 133537724462912 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x795dc83a8400> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 14:25:57.448528 133537724462912 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_05_fp8/checkpoints I0422 14:25:57.877359 133537724462912 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=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_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x795dc8122030> I0422 14:25:57.877531 133537724462912 checkpointing.py:302] Checkpoint manager created! I0422 14:25:59.157063 133537724462912 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0422 14:25:59.157199 133537724462912 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 14:26:01.257925 133537724462912 maxtext_utils.py:1836] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0422 14:26:01.258043 133537724462912 maxtext_utils.py:1836] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0422 14:26:01.258107 133537724462912 maxtext_utils.py:1836] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0422 14:26:01.258152 133537724462912 maxtext_utils.py:1836] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0422 14:26:01.258190 133537724462912 maxtext_utils.py:1836] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0422 14:26:01.258225 133537724462912 maxtext_utils.py:1836] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0422 14:26:01.258262 133537724462912 maxtext_utils.py:1836] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0422 14:26:01.258303 133537724462912 maxtext_utils.py:1836] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0422 14:26:01.258348 133537724462912 maxtext_utils.py:1836] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0422 14:26:01.258407 133537724462912 maxtext_utils.py:1836] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0422 14:26:01.258447 133537724462912 maxtext_utils.py:1836] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0422 14:26:01.258480 133537724462912 maxtext_utils.py:1836] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0422 14:26:01.686455 133537724462912 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 14:26:01.686556 133537724462912 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 14:26:01.692967 133537724462912 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 14:26:01.693071 133537724462912 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 14:26:01.714922 133537724462912 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 14:26:01.714991 133537724462912 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 14:26:01.731675 133537724462912 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 14:26:01.731733 133537724462912 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 14:26:01.792593 133537724462912 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 14:26:01.792680 133537724462912 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 14:26:01.809448 133537724462912 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 14:26:01.809507 133537724462912 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 14:26:01.826368 133537724462912 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 14:26:01.826427 133537724462912 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 14:26:01.858769 133537724462912 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 14:26:01.858841 133537724462912 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 14:26:01.918715 133537724462912 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 14:26:01.918795 133537724462912 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 14:26:15.971796 133537724462912 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. I0422 14:26:15.975692 133537724462912 metric_logger.py:301] number parameters: 1.104 billion I0422 14:26:29.796051 133537724462912 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0422 14:26:29.956433 133537724462912 checkpointing.py:798] Waited 0.16037583351135254 seconds for step 0 to finish before starting checkpointing. I0422 14:26:29.959032 133537724462912 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 14:26:29.960901 133537724462912 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0422 14:26:29.962285 133537724462912 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_05_fp8/checkpoints/0. I0422 14:26:30.546191 133537724462912 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 14:26:30.547266 133537724462912 jax_array_handlers.py:360] Scheduling D2H of 153 prioritized jax.Array. I0422 14:26:30.547335 133537724462912 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 I0422 14:26:30.961229 133537724462912 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.416732s I0422 14:26:30.961405 133537724462912 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 3.557 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.4336719512939453 s) (per-host) I0422 14:26:30.961457 133537724462912 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.433735s (batch_requests_ready=0.005615s, total_serialization_initiated=0.428047s, others=0.000072s) I0422 14:26:30.961582 133537724462912 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.437951s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=0.437933) I0422 14:26:30.962385 133537724462912 event_tracking.py:125] [process=6] [async] Finished blocking save in 1.00 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_05_fp8/checkpoints/0. I0422 14:26:30.962658 133408485029632 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 14:46:30.962633 I0422 14:26:30.970968 133537724462912 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 14:26:30.971241 133406918702848 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 14:26:30.971403 133537724462912 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776867989.9590144, 'wait_for_prev_duration_secs': 6.127357482910156e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776867989.9609401, 'checkpointer_blocking_duration_secs': 1.001833438873291, 'get_old_steps_start_time': 1776867990.9627998, 'get_old_steps_duration_secs': 3.0517578125e-05, 'checkpoint_manager_blocking_start_time': 1776867989.9569054, 'checkpoint_manager_blocking_duration_secs': 1.0144603252410889} I0422 14:26:30.971517 133537724462912 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0422 14:26:30.971598 133537724462912 max_utils.py:750] Memstats: After params initialized: I0422 14:26:30.971648 133537724462912 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_24(process=6,(0,6,0,0)) I0422 14:26:30.971681 133537724462912 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_25(process=6,(1,6,0,0)) I0422 14:26:30.971709 133537724462912 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_28(process=6,(0,7,0,0)) I0422 14:26:30.971732 133537724462912 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_29(process=6,(1,7,0,0)) I0422 14:26:31.303387 133537724462912 metric_logger.py:196] completed step: 0, seconds: 13.818, TFLOP/s/device: 0.983, Tokens/s/device: 148.210, total_weights: 65536, loss: 10.872, lm_loss: 10.872, perplexity: 52680.742 I0422 14:26:31.466867 133537724462912 metric_logger.py:196] completed step: 1, seconds: 1.504, TFLOP/s/device: 9.031, Tokens/s/device: 1361.270, total_weights: 65536, loss: 10.872, lm_loss: 10.872, perplexity: 52680.742 I0422 14:26:31.932430 133537724462912 metric_logger.py:196] completed step: 2, seconds: 0.018, TFLOP/s/device: 756.269, Tokens/s/device: 113993.098, total_weights: 65536, loss: 10.856, lm_loss: 10.856, perplexity: 51864.762 I0422 14:26:32.089334 133537724462912 metric_logger.py:196] completed step: 3, seconds: 0.435, TFLOP/s/device: 31.204, Tokens/s/device: 4703.451, total_weights: 65536, loss: 10.824, lm_loss: 10.824, perplexity: 50225.512 I0422 14:26:32.404244 133537724462912 metric_logger.py:196] completed step: 4, seconds: 0.187, TFLOP/s/device: 72.840, Tokens/s/device: 10979.232, total_weights: 65536, loss: 10.793, lm_loss: 10.793, perplexity: 48689.156 I0422 14:26:32.414822 133537724462912 metric_logger.py:196] completed step: 5, seconds: 0.157, TFLOP/s/device: 86.729, Tokens/s/device: 13072.813, total_weights: 65536, loss: 10.763, lm_loss: 10.763, perplexity: 47229.012 I0422 14:26:34.254345 2840 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 14:26:36.284121 133407971706624 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_05_fp8/checkpoints/0/items/array_metadatas/process_6 I0422 14:26:51.608150 133408485029632 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 74.940 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 21.08036494255066 s) (per-host) I0422 14:26:51.608278 133408485029632 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 20.645540s. I0422 14:26:57.388587 133537724462912 metric_logger.py:196] completed step: 6, seconds: 0.316, TFLOP/s/device: 42.984, Tokens/s/device: 6478.983, total_weights: 65536, loss: 10.733, lm_loss: 10.733, perplexity: 45864.258 I0422 14:26:57.545694 133537724462912 metric_logger.py:196] completed step: 7, seconds: 24.820, TFLOP/s/device: 0.547, Tokens/s/device: 82.513, total_weights: 65536, loss: 10.706, lm_loss: 10.706, perplexity: 44620.242 I0422 14:26:57.702398 133537724462912 metric_logger.py:196] completed step: 8, seconds: 0.163, TFLOP/s/device: 83.199, Tokens/s/device: 12540.721, total_weights: 65536, loss: 10.680, lm_loss: 10.680, perplexity: 43477.055 I0422 14:26:57.711045 133537724462912 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0422 14:26:58.016659 133537724462912 checkpointing.py:798] Waited 0.3055891990661621 seconds for step 10 to finish before starting checkpointing. I0422 14:26:58.019039 133537724462912 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 14:26:59.192013 133408485029632 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 28.229257s. I0422 14:26:59.192361 133406918702848 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 14:26:59.192485 133406918702848 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 14:26:59.192534 133406918702848 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 14:26:59.194730 133406918702848 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 14:26:59.194886 133537724462912 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. W0422 14:26:59.195030 133537724462912 checkpoint_manager.py:1452] Waiting for previous save to complete took 1.175988 seconds. If this number is high, consider checkpointing less frequently. I0422 14:26:59.197047 133537724462912 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 10 I0422 14:26:59.199145 133537724462912 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_05_fp8/checkpoints/10. I0422 14:27:00.053573 133537724462912 jax_array_handlers.py:360] Scheduling D2H of 153 prioritized jax.Array. I0422 14:27:00.053676 133537724462912 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 I0422 14:27:00.536792 133537724462912 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.485768s I0422 14:27:00.536956 133537724462912 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 3.084 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5002062320709229 s) (per-host) I0422 14:27:00.537005 133537724462912 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.500265s (batch_requests_ready=0.005175s, total_serialization_initiated=0.495023s, others=0.000067s) I0422 14:27:00.537139 133537724462912 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.504789s (all_items=0.000014s, per_item={'items': '0.00001431'}, temp_paths=0.504775) I0422 14:27:00.537873 133537724462912 event_tracking.py:125] [process=6] [async] Finished blocking save in 1.34 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_05_fp8/checkpoints/10. I0422 14:27:00.538231 133406918702848 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 14:47:00.538195 I0422 14:27:00.554987 133537724462912 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0422 14:27:00.555301 133407971706624 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 14:27:00.555477 133537724462912 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776868018.0190096, 'wait_for_prev_duration_secs': 1.175987958908081, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776868019.1970856, 'checkpointer_blocking_duration_secs': 1.3412885665893555, 'get_old_steps_start_time': 1776868020.5383995, 'get_old_steps_duration_secs': 3.1948089599609375e-05, 'checkpoint_manager_blocking_start_time': 1776868018.0171068, 'checkpoint_manager_blocking_duration_secs': 2.538325309753418} I0422 14:27:00.555622 133537724462912 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0422 14:27:00.556268 133537724462912 metric_logger.py:196] completed step: 9, seconds: 0.157, TFLOP/s/device: 86.596, Tokens/s/device: 13052.650, total_weights: 65536, loss: 10.656, lm_loss: 10.656, perplexity: 42467.320 I0422 14:27:00.567853 133537724462912 metric_logger.py:196] completed step: 10, seconds: 0.157, TFLOP/s/device: 86.628, Tokens/s/device: 13057.560, total_weights: 65536, loss: 10.636, lm_loss: 10.636, perplexity: 41601.941 I0422 14:27:00.721494 133537724462912 metric_logger.py:196] completed step: 11, seconds: 2.854, TFLOP/s/device: 4.761, Tokens/s/device: 717.648, total_weights: 65536, loss: 10.618, lm_loss: 10.618, perplexity: 40849.570 I0422 14:27:00.878216 133537724462912 metric_logger.py:196] completed step: 12, seconds: 0.011, TFLOP/s/device: 1289.713, Tokens/s/device: 194399.620, total_weights: 65536, loss: 10.602, lm_loss: 10.602, perplexity: 40203.926 I0422 14:27:01.447257 133537724462912 metric_logger.py:196] completed step: 13, seconds: 0.155, TFLOP/s/device: 87.663, Tokens/s/device: 13213.500, total_weights: 65536, loss: 10.588, lm_loss: 10.588, perplexity: 39652.145 I0422 14:27:01.604084 133537724462912 metric_logger.py:196] completed step: 14, seconds: 0.563, TFLOP/s/device: 24.153, Tokens/s/device: 3640.630, total_weights: 65536, loss: 10.577, lm_loss: 10.577, perplexity: 39212.879 I0422 14:27:01.760972 133537724462912 metric_logger.py:196] completed step: 15, seconds: 0.163, TFLOP/s/device: 83.389, Tokens/s/device: 12569.352, total_weights: 65536, loss: 10.568, lm_loss: 10.568, perplexity: 38879.223 I0422 14:27:01.917848 133537724462912 metric_logger.py:196] completed step: 16, seconds: 0.157, TFLOP/s/device: 86.644, Tokens/s/device: 13059.975, total_weights: 65536, loss: 10.562, lm_loss: 10.562, perplexity: 38619.105 I0422 14:27:02.074726 133537724462912 metric_logger.py:196] completed step: 17, seconds: 0.157, TFLOP/s/device: 86.696, Tokens/s/device: 13067.808, total_weights: 65536, loss: 10.556, lm_loss: 10.556, perplexity: 38395.816 I0422 14:27:02.231436 133537724462912 metric_logger.py:196] completed step: 18, seconds: 0.157, TFLOP/s/device: 86.546, Tokens/s/device: 13045.168, total_weights: 65536, loss: 10.552, lm_loss: 10.552, perplexity: 38236.652 I0422 14:27:02.390198 133537724462912 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0422 14:27:02.391519 133537724462912 checkpointing.py:798] Waited 0.0013310909271240234 seconds for step 19 to finish before starting checkpointing. I0422 14:27:02.393596 133537724462912 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 14:27:06.137948 133405274519296 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_05_fp8/checkpoints/10/items/array_metadatas/process_6 I0422 14:27:22.029452 133406918702848 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 71.831 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 21.99266266822815 s) (per-host) I0422 14:27:22.029584 133406918702848 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 21.491238s. I0422 14:27:30.125230 133406918702848 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 29.586869s. I0422 14:27:30.125528 133407971706624 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 14:27:30.125642 133407971706624 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 14:27:30.125689 133407971706624 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0422 14:27:30.128111 133407971706624 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0422 14:27:30.128288 133537724462912 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0422 14:27:30.128415 133537724462912 checkpoint_manager.py:1452] Waiting for previous save to complete took 27.734830 seconds. If this number is high, consider checkpointing less frequently. I0422 14:27:30.129950 133537724462912 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 19 I0422 14:27:30.131932 133537724462912 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_05_fp8/checkpoints/19. I0422 14:27:30.429214 133537724462912 jax_array_handlers.py:360] Scheduling D2H of 153 prioritized jax.Array. I0422 14:27:30.429369 133537724462912 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 I0422 14:27:30.464653 133537724462912 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.038024s I0422 14:27:30.464782 133537724462912 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 29.416 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.052445173263549805 s) (per-host) I0422 14:27:30.464829 133537724462912 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.052501s (batch_requests_ready=0.005153s, total_serialization_initiated=0.047288s, others=0.000061s) I0422 14:27:30.464930 133537724462912 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.056564s (all_items=0.000011s, per_item={'items': '0.00001121'}, temp_paths=0.056553) I0422 14:27:30.465650 133537724462912 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_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_05_fp8/checkpoints/19. I0422 14:27:30.465972 133407971706624 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 14:47:30.465935 I0422 14:27:30.923498 133537724462912 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0422 14:27:30.923923 133405299697408 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 14:27:30.924069 133537724462912 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776868022.3935604, 'wait_for_prev_duration_secs': 27.734830379486084, 'time_between_consecutive_saves_sec': 3.198791265487671, 'checkpointer_blocking_start_time': 1776868050.1299908, 'checkpointer_blocking_duration_secs': 0.3361341953277588, 'get_old_steps_start_time': 1776868050.466144, 'get_old_steps_duration_secs': 2.6464462280273438e-05, 'checkpoint_manager_blocking_start_time': 1776868022.3918593, 'checkpoint_manager_blocking_duration_secs': 28.53217625617981} I0422 14:27:30.924261 133537724462912 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0422 14:27:30.924305 133537724462912 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 14:27:35.629341 133407992444672 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/nnx_xpk_feat_nnx_post_train_fixes_20260422_123906_05_fp8/checkpoints/19/items/array_metadatas/process_6 I0422 14:27:51.524433 133407971706624 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 74.827 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 21.11204957962036 s) (per-host) I0422 14:27:51.524554 133407971706624 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 21.058472s. I0422 14:27:59.160149 133407971706624 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 28.694049s. I0422 14:27:59.160463 133405299697408 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 14:27:59.160585 133405299697408 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 14:27:59.160628 133405299697408 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0422 14:27:59.162512 133405299697408 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0422 14:27:59.162690 133537724462912 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0422 14:27:59.162868 133537724462912 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 14:27:59.163799 133537724462912 metric_logger.py:196] completed step: 19, seconds: 0.157, TFLOP/s/device: 86.563, Tokens/s/device: 13047.744, total_weights: 65536, loss: 10.549, lm_loss: 10.549, perplexity: 38125.074 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Wed Apr 22 14:28:11 UTC 2026 EXIT_CODE=0