XPK Start: Fri Apr 17 00:51:29 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-17 00:51:53.693552: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0417 00:51:53.870933 135128853358400 max_utils.py:273] Attempting to initialize the jax distributed system... I0417 00:52:02.912159 135128853358400 distributed.py:149] Starting JAX distributed service on [::]:8482 I0417 00:52:02.914484 135128853358400 distributed.py:172] Connecting to JAX distributed service on mt-05-fp8-1mix0-slice-job-0-0.mt-05-fp8-1mix0:8482 I0417 00:52:03.749042 135128853358400 max_utils.py:284] Jax distributed system initialized! I0417 00:52:08.860002 135128853358400 max_utils.py:800] System Information: Jax Version: 0.9.2 I0417 00:52:08.860105 135128853358400 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0417 00:52:08.860145 135128853358400 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 I0417 00:52:08.860179 135128853358400 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0417 00:52:09.537497 135128853358400 maxtext_utils.py:1687] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0417 00:52:09.959382 135128853358400 checkpointing.py:688] Setting up checkpoint logger... I0417 00:52:09.959507 135128853358400 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0417 00:52:09.959553 135128853358400 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0417 00:52:09.959782 135128853358400 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 0x7ae5947170e0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0417 00:52:13.809269 135128853358400 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0417 00:52:13.809505 135128853358400 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 0x7ad1b869a510>}, handler_registry=None I0417 00:52:13.809754 135128853358400 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ad1b869a510>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0417 00:52:13.809802 135128853358400 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ad1504c5970>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0417 00:52:13.809851 135128853358400 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 0x7ad1b869a510>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ad1b869a510>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ad1504c5970>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ad1504c5970>}). I0417 00:52:13.810178 135128853358400 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0417 00:52:13.810247 135128853358400 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7ad1b86f1b20> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0417 00:52:15.973223 135128853358400 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_05_fp8/checkpoints I0417 00:52:16.013273 135128853358400 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_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7ad1b869a570> I0417 00:52:16.013417 135128853358400 checkpointing.py:302] Checkpoint manager created! I0417 00:52:17.274070 135128853358400 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0417 00:52:17.274193 135128853358400 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0417 00:52:19.332782 135128853358400 maxtext_utils.py:1805] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0417 00:52:19.332910 135128853358400 maxtext_utils.py:1805] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0417 00:52:19.332962 135128853358400 maxtext_utils.py:1805] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0417 00:52:19.333003 135128853358400 maxtext_utils.py:1805] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0417 00:52:19.333040 135128853358400 maxtext_utils.py:1805] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0417 00:52:19.333074 135128853358400 maxtext_utils.py:1805] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0417 00:52:19.333109 135128853358400 maxtext_utils.py:1805] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 00:52:19.333144 135128853358400 maxtext_utils.py:1805] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0417 00:52:19.333176 135128853358400 maxtext_utils.py:1805] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 00:52:19.333208 135128853358400 maxtext_utils.py:1805] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0417 00:52:19.333239 135128853358400 maxtext_utils.py:1805] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0417 00:52:19.333270 135128853358400 maxtext_utils.py:1805] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0417 00:52:19.756198 135128853358400 nnx_decoders.py:465] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 00:52:19.756295 135128853358400 nnx_decoders.py:465] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 00:52:19.761511 135128853358400 nnx_decoders.py:465] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0417 00:52:19.761569 135128853358400 nnx_decoders.py:465] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 00:52:19.778005 135128853358400 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 00:52:19.778064 135128853358400 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 00:52:19.793853 135128853358400 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0417 00:52:19.793909 135128853358400 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0417 00:52:19.853070 135128853358400 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 00:52:19.853164 135128853358400 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 00:52:19.868932 135128853358400 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 00:52:19.868990 135128853358400 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 00:52:19.884777 135128853358400 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0417 00:52:19.884859 135128853358400 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 00:52:19.914118 135128853358400 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0417 00:52:19.914188 135128853358400 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0417 00:52:19.973954 135128853358400 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0417 00:52:19.974032 135128853358400 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0417 00:52:33.873207 135128853358400 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. I0417 00:52:33.877171 135128853358400 metric_logger.py:289] number parameters: 1.104 billion I0417 00:52:47.577975 135128853358400 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0417 00:52:47.740694 135128853358400 checkpointing.py:798] Waited 0.16270184516906738 seconds for step 0 to finish before starting checkpointing. I0417 00:52:47.743193 135128853358400 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 00:52:47.745469 135128853358400 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0417 00:52:47.747159 135128853358400 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_05_fp8/checkpoints/0. I0417 00:52:48.098081 135128853358400 signaling_client.py:364] Using JaxDistributedSignalingClient I0417 00:52:48.099134 135128853358400 jax_array_handlers.py:360] Scheduling D2H of 153 prioritized jax.Array. I0417 00:52:48.099199 135128853358400 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 I0417 00:52:48.520979 135128853358400 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.424548s I0417 00:52:48.521165 135128853358400 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 3.493 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.4417135715484619 s) (per-host) I0417 00:52:48.521232 135128853358400 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.441787s (batch_requests_ready=0.006043s, total_serialization_initiated=0.435661s, others=0.000083s) I0417 00:52:48.521392 135128853358400 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.445943s (all_items=0.000019s, per_item={'items': '0.00001907'}, temp_paths=0.445924) I0417 00:52:48.522210 135128853358400 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.78 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_05_fp8/checkpoints/0. I0417 00:52:48.522559 135000242456320 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 01:12:48.522520 I0417 00:52:48.530420 135128853358400 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0417 00:52:48.530755 134999739197184 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0417 00:52:48.530949 135128853358400 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776387167.7431734, 'wait_for_prev_duration_secs': 6.222724914550781e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776387167.7455082, 'checkpointer_blocking_duration_secs': 0.7772085666656494, 'get_old_steps_start_time': 1776387168.5227427, 'get_old_steps_duration_secs': 2.956390380859375e-05, 'checkpoint_manager_blocking_start_time': 1776387167.7411895, 'checkpoint_manager_blocking_duration_secs': 0.7897205352783203} I0417 00:52:48.531070 135128853358400 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0417 00:52:48.531150 135128853358400 max_utils.py:750] Memstats: After params initialized: I0417 00:52:48.531200 135128853358400 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_24(process=6,(0,6,0,0)) I0417 00:52:48.531234 135128853358400 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_25(process=6,(1,6,0,0)) I0417 00:52:48.531262 135128853358400 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_28(process=6,(0,7,0,0)) I0417 00:52:48.531287 135128853358400 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_29(process=6,(1,7,0,0)) I0417 00:52:48.849510 135128853358400 metric_logger.py:185] completed step: 0, seconds: 13.699, TFLOP/s/device: 0.992, Tokens/s/device: 149.504, total_weights: 65536, loss: 10.872 I0417 00:52:49.022458 135128853358400 metric_logger.py:185] completed step: 1, seconds: 1.269, TFLOP/s/device: 10.708, Tokens/s/device: 1613.996, total_weights: 65536, loss: 10.872 I0417 00:52:49.475811 135128853358400 metric_logger.py:185] completed step: 2, seconds: 0.020, TFLOP/s/device: 680.343, Tokens/s/device: 102548.696, total_weights: 65536, loss: 10.856 I0417 00:52:49.632550 135128853358400 metric_logger.py:185] completed step: 3, seconds: 0.431, TFLOP/s/device: 31.549, Tokens/s/device: 4755.458, total_weights: 65536, loss: 10.824 I0417 00:52:49.947661 135128853358400 metric_logger.py:185] completed step: 4, seconds: 0.187, TFLOP/s/device: 72.830, Tokens/s/device: 10977.702, total_weights: 65536, loss: 10.793 I0417 00:52:49.957676 135128853358400 metric_logger.py:185] completed step: 5, seconds: 0.156, TFLOP/s/device: 87.007, Tokens/s/device: 13114.670, total_weights: 65536, loss: 10.763 I0417 00:52:51.429970 2810 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0417 00:52:53.444642 134999747589888 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_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_05_fp8/checkpoints/0/items/array_metadatas/process_6 I0417 00:53:06.249900 135000242456320 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 86.941 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 18.170410633087158 s) (per-host) I0417 00:53:06.250031 135000242456320 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 17.727358s. I0417 00:53:14.285493 135128853358400 metric_logger.py:185] completed step: 6, seconds: 0.316, TFLOP/s/device: 42.979, Tokens/s/device: 6478.327, total_weights: 65536, loss: 10.733 I0417 00:53:14.442410 135128853358400 metric_logger.py:185] completed step: 7, seconds: 24.174, TFLOP/s/device: 0.562, Tokens/s/device: 84.718, total_weights: 65536, loss: 10.706 I0417 00:53:14.599224 135128853358400 metric_logger.py:185] completed step: 8, seconds: 0.163, TFLOP/s/device: 83.463, Tokens/s/device: 12580.471, total_weights: 65536, loss: 10.680 I0417 00:53:14.608164 135128853358400 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0417 00:53:14.914070 135128853358400 checkpointing.py:798] Waited 0.3058900833129883 seconds for step 10 to finish before starting checkpointing. I0417 00:53:14.916353 135128853358400 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 00:53:15.909343 135000242456320 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 27.386655s. I0417 00:53:15.909609 134999739197184 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 00:53:15.909746 134999739197184 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0417 00:53:15.909800 134999739197184 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0417 00:53:15.911372 134999739197184 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0417 00:53:15.911538 135128853358400 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. I0417 00:53:15.913103 135128853358400 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 10 I0417 00:53:15.915009 135128853358400 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_05_fp8/checkpoints/10. I0417 00:53:16.633342 135128853358400 jax_array_handlers.py:360] Scheduling D2H of 153 prioritized jax.Array. I0417 00:53:16.633456 135128853358400 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 I0417 00:53:16.684360 135128853358400 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.053498s I0417 00:53:16.684527 135128853358400 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 22.612 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.06822633743286133 s) (per-host) I0417 00:53:16.684585 135128853358400 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.068293s (batch_requests_ready=0.005673s, total_serialization_initiated=0.062545s, others=0.000075s) I0417 00:53:16.684706 135128853358400 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.072645s (all_items=0.000015s, per_item={'items': '0.00001526'}, temp_paths=0.072629) I0417 00:53:16.685392 135128853358400 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.77 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_05_fp8/checkpoints/10. I0417 00:53:16.685743 134999739197184 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 01:13:16.685705 I0417 00:53:16.687681 135128853358400 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0417 00:53:16.687988 135000228452096 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0417 00:53:16.688121 135128853358400 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776387194.916322, 'wait_for_prev_duration_secs': 0.9953162670135498, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776387195.913142, 'checkpointer_blocking_duration_secs': 0.7727127075195312, 'get_old_steps_start_time': 1776387196.6858807, 'get_old_steps_duration_secs': 2.9087066650390625e-05, 'checkpoint_manager_blocking_start_time': 1776387194.9145145, 'checkpoint_manager_blocking_duration_secs': 1.7735731601715088} I0417 00:53:16.688292 135128853358400 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0417 00:53:16.688719 135128853358400 metric_logger.py:185] completed step: 9, seconds: 0.157, TFLOP/s/device: 86.602, Tokens/s/device: 13053.566, total_weights: 65536, loss: 10.656 I0417 00:53:16.699853 135128853358400 metric_logger.py:185] completed step: 10, seconds: 0.157, TFLOP/s/device: 86.380, Tokens/s/device: 13020.204, total_weights: 65536, loss: 10.636 I0417 00:53:16.852256 135128853358400 metric_logger.py:185] completed step: 11, seconds: 2.089, TFLOP/s/device: 6.504, Tokens/s/device: 980.288, total_weights: 65536, loss: 10.618 I0417 00:53:17.009032 135128853358400 metric_logger.py:185] completed step: 12, seconds: 0.010, TFLOP/s/device: 1399.148, Tokens/s/device: 210894.861, total_weights: 65536, loss: 10.602 I0417 00:53:17.166132 135128853358400 metric_logger.py:185] completed step: 13, seconds: 0.154, TFLOP/s/device: 88.380, Tokens/s/device: 13321.625, total_weights: 65536, loss: 10.588 I0417 00:53:17.322961 135128853358400 metric_logger.py:185] completed step: 14, seconds: 0.157, TFLOP/s/device: 86.642, Tokens/s/device: 13059.642, total_weights: 65536, loss: 10.577 I0417 00:53:17.479779 135128853358400 metric_logger.py:185] completed step: 15, seconds: 0.157, TFLOP/s/device: 86.570, Tokens/s/device: 13048.825, total_weights: 65536, loss: 10.568 I0417 00:53:17.636534 135128853358400 metric_logger.py:185] completed step: 16, seconds: 0.157, TFLOP/s/device: 86.650, Tokens/s/device: 13060.808, total_weights: 65536, loss: 10.562 I0417 00:53:18.196661 135128853358400 metric_logger.py:185] completed step: 17, seconds: 0.157, TFLOP/s/device: 86.579, Tokens/s/device: 13050.155, total_weights: 65536, loss: 10.556 I0417 00:53:18.353276 135128853358400 metric_logger.py:185] completed step: 18, seconds: 0.554, TFLOP/s/device: 24.521, Tokens/s/device: 3696.004, total_weights: 65536, loss: 10.552 I0417 00:53:18.512566 135128853358400 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0417 00:53:18.513833 135128853358400 checkpointing.py:798] Waited 0.0012671947479248047 seconds for step 19 to finish before starting checkpointing. I0417 00:53:18.515882 135128853358400 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 00:53:22.286731 134996509579008 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_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_05_fp8/checkpoints/10/items/array_metadatas/process_6 I0417 00:53:37.796706 134999739197184 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 74.586 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 21.180366039276123 s) (per-host) I0417 00:53:37.796867 134999739197184 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 21.110455s. I0417 00:53:45.914453 134999739197184 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 29.228038s. I0417 00:53:45.914735 135000228452096 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 00:53:45.914928 135000228452096 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0417 00:53:45.914983 135000228452096 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0417 00:53:45.916686 135000228452096 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0417 00:53:45.916890 135128853358400 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. W0417 00:53:45.917025 135128853358400 checkpoint_manager.py:1452] Waiting for previous save to complete took 27.401138 seconds. If this number is high, consider checkpointing less frequently. I0417 00:53:45.918959 135128853358400 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 19 I0417 00:53:45.920951 135128853358400 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_05_fp8/checkpoints/19. I0417 00:53:46.234434 135128853358400 jax_array_handlers.py:360] Scheduling D2H of 153 prioritized jax.Array. I0417 00:53:46.234604 135128853358400 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 I0417 00:53:46.286194 135128853358400 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.054231s I0417 00:53:46.286355 135128853358400 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 22.432 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.06877422332763672 s) (per-host) I0417 00:53:46.286403 135128853358400 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.068831s (batch_requests_ready=0.005607s, total_serialization_initiated=0.063162s, others=0.000062s) I0417 00:53:46.286517 135128853358400 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.072882s (all_items=0.000011s, per_item={'items': '0.00001073'}, temp_paths=0.072872) I0417 00:53:46.287217 135128853358400 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.37 seconds. Continuing save @ gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_05_fp8/checkpoints/19. I0417 00:53:46.287558 135000228452096 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-17 01:13:46.287521 I0417 00:53:46.289540 135128853358400 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0417 00:53:46.289857 134998623491840 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0417 00:53:46.290023 135128853358400 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_post_train_fixes_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776387198.5158548, 'wait_for_prev_duration_secs': 27.401138305664062, 'time_between_consecutive_saves_sec': 2.6044392585754395, 'checkpointer_blocking_start_time': 1776387225.9189985, 'checkpointer_blocking_duration_secs': 0.3687162399291992, 'get_old_steps_start_time': 1776387226.2877324, 'get_old_steps_duration_secs': 2.384185791015625e-05, 'checkpoint_manager_blocking_start_time': 1776387198.5141392, 'checkpoint_manager_blocking_duration_secs': 27.775851726531982} I0417 00:53:46.290140 135128853358400 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0417 00:53:46.290183 135128853358400 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0417 00:53:52.320964 134998120191744 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_20260417_003418/nnx_xpk_feat_nnx_post_train_fixes_20260417_003418_05_fp8/checkpoints/19/items/array_metadatas/process_6 I0417 00:54:07.374444 135000228452096 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 74.669 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 21.156837701797485 s) (per-host) I0417 00:54:07.374550 135000228452096 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 21.086873s. I0417 00:54:13.980917 135000228452096 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 27.693223s. I0417 00:54:13.981190 134998623491840 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0417 00:54:13.981310 134998623491840 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0417 00:54:13.981353 134998623491840 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0417 00:54:13.982896 134998623491840 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0417 00:54:13.983064 135128853358400 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. I0417 00:54:13.983247 135128853358400 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0417 00:54:13.983932 135128853358400 metric_logger.py:185] completed step: 19, seconds: 0.163, TFLOP/s/device: 83.457, Tokens/s/device: 12579.544, total_weights: 65536, loss: 10.549 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Fri Apr 17 00:54:25 UTC 2026 EXIT_CODE=0