main (586e69205) vs feat/nnx-post-train-fixes (7f06c99ac)| Metric | main 586e69205 | feat/nnx-post-train-fixes 7f06c99ac | Diff (feat/nnx-post-train-fixes − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 7.1810 | 7.1810 | 0 |
| TFLOP/s | 92.288 | 92.314 | +0.026 |
| Tok/s | 13910.7 | 13914.6 | +3.875 |
| Avg s/step | 8.399 | 8.214 | -0.185 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.9.2 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Thu Apr 23 08:46:48 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-23 08:47:13.562407: 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) I0423 08:47:13.774648 139867416733504 max_utils.py:273] Attempting to initialize the jax distributed system... I0423 08:47:22.816728 139867416733504 distributed.py:149] Starting JAX distributed service on [::]:8482 I0423 08:47:22.819414 139867416733504 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-sv-onwex-slice-job-0-0.mt-16-zarr3-false-sv-onwex:8482 I0423 08:47:23.678599 139867416733504 max_utils.py:284] Jax distributed system initialized! I0423 08:47:29.917233 139867416733504 max_utils.py:800] System Information: Jax Version: 0.9.2 I0423 08:47:29.917340 139867416733504 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0423 08:47:29.917380 139867416733504 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 I0423 08:47:29.917414 139867416733504 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0423 08:47:30.611746 139867416733504 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 08:47:30.612029 139867416733504 checkpointing.py:677] Setting up checkpoint logger... I0423 08:47:30.612083 139867416733504 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=False I0423 08:47:30.612139 139867416733504 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0423 08:47:30.612481 139867416733504 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7f34c8deb680>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0423 08:47:34.527814 139867416733504 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0423 08:47:34.528053 139867416733504 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 0x7f20cc43ade0>}, handler_registry=None I0423 08:47:34.528307 139867416733504 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f20cc43ade0>` 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`. I0423 08:47:34.528365 139867416733504 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f20cc43d1f0>` 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`. I0423 08:47:34.528402 139867416733504 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 0x7f20cc43ade0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f20cc43ade0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f20cc43d1f0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f20cc43d1f0>}). I0423 08:47:34.528720 139867416733504 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0423 08:47:34.528790 139867416733504 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f20cc1a5940> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0423 08:47:35.668851 139867416733504 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints I0423 08:47:35.671213 139867416733504 checkpoint_manager.py:929] [process=6][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f20cc43cd70> I0423 08:47:35.671328 139867416733504 checkpointing.py:301] Checkpoint manager created! I0423 08:47:36.622150 139867416733504 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0423 08:47:36.622272 139867416733504 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 08:47:36.999774 139867416733504 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn'). I0423 08:47:36.999868 139867416733504 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 08:47:37.016295 139867416733504 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn'). I0423 08:47:37.016355 139867416733504 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 08:47:37.040025 139867416733504 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 08:47:37.040103 139867416733504 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 08:47:37.056447 139867416733504 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 08:47:37.056511 139867416733504 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 08:47:37.072727 139867416733504 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 08:47:37.072786 139867416733504 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 08:47:37.097404 139867416733504 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch_attn', 'activation_length_attn', 'activation_heads', 'activation_kv'). I0423 08:47:37.097486 139867416733504 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 08:47:37.118535 139867416733504 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0423 08:47:37.118600 139867416733504 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0423 08:47:37.328664 139867416733504 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0423 08:47:37.328772 139867416733504 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0423 08:47:38.759458 139867416733504 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0423 08:47:38.759588 139867416733504 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0423 08:47:38.759641 139867416733504 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0423 08:47:38.759699 139867416733504 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0423 08:47:38.759754 139867416733504 maxtext_utils.py:1707] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0423 08:47:38.759797 139867416733504 maxtext_utils.py:1707] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0423 08:47:38.759850 139867416733504 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0423 08:47:38.759903 139867416733504 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0423 08:47:38.759944 139867416733504 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0423 08:47:38.759980 139867416733504 maxtext_utils.py:1707] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0423 08:47:38.760027 139867416733504 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0423 08:47:38.760073 139867416733504 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0423 08:47:39.252972 139867416733504 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0423 08:47:39.253067 139867416733504 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0423 08:47:39.268737 139867416733504 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0423 08:47:39.268798 139867416733504 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0423 08:47:50.150228 139867416733504 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. I0423 08:47:50.151036 139867416733504 metric_logger.py:301] number parameters: 1.104 billion I0423 08:48:01.407902 139867416733504 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0423 08:48:01.854852 139867416733504 checkpointing.py:776] Waited 0.4469327926635742 seconds for step 0 to finish before starting checkpointing. I0423 08:48:01.857286 139867416733504 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 08:48:01.858985 139867416733504 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0423 08:48:01.860438 139867416733504 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/0. I0423 08:48:02.618947 139867416733504 signaling_client.py:364] Using JaxDistributedSignalingClient I0423 08:48:02.619827 139867416733504 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0423 08:48:02.619882 139867416733504 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 I0423 08:48:02.895859 139867416733504 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.276985s I0423 08:48:02.896037 139867416733504 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.457 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28269004821777344 s) (per-host) I0423 08:48:02.896090 139867416733504 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.282753s (batch_requests_ready=0.002489s, total_serialization_initiated=0.280192s, others=0.000072s) I0423 08:48:02.896198 139867416733504 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.287034s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.287017) I0423 08:48:02.897081 139867416733504 event_tracking.py:125] [process=6] [async] Finished blocking save in 1.04 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/0. I0423 08:48:02.897441 139739009103616 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 09:08:02.897403 I0423 08:48:02.902519 139867416733504 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0423 08:48:02.902857 139737939568384 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0423 08:48:02.903016 139867416733504 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776934081.8572667, 'wait_for_prev_duration_secs': 6.031990051269531e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776934081.8590229, 'checkpointer_blocking_duration_secs': 1.0385775566101074, 'get_old_steps_start_time': 1776934082.897627, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1776934081.8554137, 'checkpoint_manager_blocking_duration_secs': 1.0475637912750244} I0423 08:48:02.903143 139867416733504 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0423 08:48:02.903198 139867416733504 max_utils.py:750] Memstats: After params initialized: I0423 08:48:02.903248 139867416733504 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0423 08:48:02.903281 139867416733504 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0423 08:48:02.903307 139867416733504 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0423 08:48:02.903332 139867416733504 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0423 08:48:03.217172 139867416733504 metric_logger.py:196] completed step: 0, seconds: 11.257, TFLOP/s/device: 1.207, Tokens/s/device: 181.935, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0423 08:48:03.397023 139867416733504 metric_logger.py:196] completed step: 1, seconds: 1.808, TFLOP/s/device: 7.516, Tokens/s/device: 1132.912, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0423 08:48:03.790885 139867416733504 metric_logger.py:196] completed step: 2, seconds: 0.033, TFLOP/s/device: 415.432, Tokens/s/device: 62618.480, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0423 08:48:03.938125 139867416733504 metric_logger.py:196] completed step: 3, seconds: 0.394, TFLOP/s/device: 34.491, Tokens/s/device: 5198.801, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0423 08:48:03.942614 139867416733504 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0423 08:48:04.232016 139867416733504 checkpointing.py:776] Waited 0.2893691062927246 seconds for step 5 to finish before starting checkpointing. I0423 08:48:04.234923 139867416733504 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 08:48:06.293809 2806 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0423 08:48:08.527228 139738992318208 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/0/items/array_metadatas/process_6 I0423 08:48:45.126480 139739009103616 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.156 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.51309323310852 s) (per-host) I0423 08:48:45.126603 139739009103616 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.229042s. I0423 08:48:52.928188 139739009103616 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.030611s. I0423 08:48:52.928489 139737939568384 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 08:48:52.928605 139737939568384 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0423 08:48:52.928655 139737939568384 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0423 08:48:52.932559 139737939568384 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0423 08:48:52.932762 139867416733504 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. W0423 08:48:52.932897 139867416733504 checkpoint_manager.py:1452] Waiting for previous save to complete took 48.697976 seconds. If this number is high, consider checkpointing less frequently. I0423 08:48:52.934987 139867416733504 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 5 I0423 08:48:52.937536 139867416733504 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/5. I0423 08:48:53.656377 139867416733504 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0423 08:48:53.656470 139867416733504 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 I0423 08:48:53.688395 139867416733504 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.032954s I0423 08:48:53.688551 139867416733504 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 42.331 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.036440372467041016 s) (per-host) I0423 08:48:53.688602 139867416733504 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.036500s (batch_requests_ready=0.001777s, total_serialization_initiated=0.034657s, others=0.000067s) I0423 08:48:53.688688 139867416733504 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.041592s (all_items=0.000016s, per_item={'items': '0.00001574'}, temp_paths=0.041576) I0423 08:48:53.689392 139867416733504 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/5. I0423 08:48:53.689722 139737939568384 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 09:08:53.689684 I0423 08:48:53.699635 139867416733504 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0423 08:48:53.699895 139713914562304 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0423 08:48:53.700055 139867416733504 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776934084.2348936, 'wait_for_prev_duration_secs': 48.697975635528564, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776934132.935024, 'checkpointer_blocking_duration_secs': 0.7548608779907227, 'get_old_steps_start_time': 1776934133.6899087, 'get_old_steps_duration_secs': 3.0517578125e-05, 'checkpoint_manager_blocking_start_time': 1776934084.2326822, 'checkpoint_manager_blocking_duration_secs': 49.46733641624451} I0423 08:48:53.700178 139867416733504 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0423 08:48:53.700996 139867416733504 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.817, Tokens/s/device: 13387.458, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0423 08:48:53.706916 139867416733504 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.272, Tokens/s/device: 13908.225, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0423 08:48:59.657141 139738992318208 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/5/items/array_metadatas/process_6 I0423 08:49:16.843115 139867416733504 metric_logger.py:196] completed step: 6, seconds: 49.763, TFLOP/s/device: 0.273, Tokens/s/device: 41.155, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0423 08:49:16.990335 139867416733504 metric_logger.py:196] completed step: 7, seconds: 22.990, TFLOP/s/device: 0.591, Tokens/s/device: 89.081, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0423 08:49:17.137680 139867416733504 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.539, Tokens/s/device: 13496.326, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0423 08:49:17.284496 139867416733504 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0423 08:49:17.285199 139867416733504 checkpointing.py:776] Waited 0.0007252693176269531 seconds for step 9 to finish before starting checkpointing. I0423 08:49:17.287221 139867416733504 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 08:49:35.529479 139737939568384 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.720 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.87732744216919 s) (per-host) I0423 08:49:35.529607 139737939568384 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.839773s. I0423 08:49:43.877835 139737939568384 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.187987s. I0423 08:49:43.878162 139713914562304 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 08:49:43.878303 139713914562304 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0423 08:49:43.878368 139713914562304 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0423 08:49:43.880889 139713914562304 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0423 08:49:43.881073 139867416733504 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0423 08:49:43.881210 139867416733504 checkpoint_manager.py:1452] Waiting for previous save to complete took 26.593997 seconds. If this number is high, consider checkpointing less frequently. I0423 08:49:43.883128 139867416733504 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 9 I0423 08:49:43.885191 139867416733504 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/9. I0423 08:49:44.598329 139867416733504 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0423 08:49:44.598425 139867416733504 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 I0423 08:49:44.633173 139867416733504 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.035708s I0423 08:49:44.633339 139867416733504 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 39.380 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.039171695709228516 s) (per-host) I0423 08:49:44.633388 139867416733504 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039232s (batch_requests_ready=0.001633s, total_serialization_initiated=0.037534s, others=0.000065s) I0423 08:49:44.633479 139867416733504 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.043579s (all_items=0.000010s, per_item={'items': '0.00001049'}, temp_paths=0.043569) I0423 08:49:44.634237 139867416733504 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/9. I0423 08:49:44.634545 139713914562304 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 09:09:44.634508 I0423 08:49:44.636612 139867416733504 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0423 08:49:44.636898 139738992318208 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0423 08:49:44.637030 139867416733504 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776934157.2871919, 'wait_for_prev_duration_secs': 26.59399652481079, 'time_between_consecutive_saves_sec': 24.354591131210327, 'checkpointer_blocking_start_time': 1776934183.8831654, 'checkpointer_blocking_duration_secs': 0.7515277862548828, 'get_old_steps_start_time': 1776934184.6347144, 'get_old_steps_duration_secs': 3.075599670410156e-05, 'checkpoint_manager_blocking_start_time': 1776934157.2854364, 'checkpoint_manager_blocking_duration_secs': 27.351561784744263} I0423 08:49:44.637231 139867416733504 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0423 08:49:44.637292 139867416733504 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 08:49:50.234463 139735821448960 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260423_071538/linen_xpk_main_20260423_071538_16_zarr3_false/checkpoints/9/items/array_metadatas/process_6 I0423 08:50:27.980868 139713914562304 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 36.407 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.38665461540222 s) (per-host) I0423 08:50:27.980994 139713914562304 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.346336s. I0423 08:50:35.076799 139713914562304 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.442126s. I0423 08:50:35.077128 139738992318208 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 08:50:35.077267 139738992318208 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0423 08:50:35.077330 139738992318208 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0423 08:50:35.078850 139738992318208 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0423 08:50:35.079024 139867416733504 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0423 08:50:35.079187 139867416733504 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 08:50:35.080340 139867416733504 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.288, Tokens/s/device: 13910.681, total_weights: 65536, loss: 7.181, lm_loss: 7.181, perplexity: 1314.495 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Thu Apr 23 08:50:43 UTC 2026 EXIT_CODE=0
XPK Start: Thu Apr 23 14:12:54 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-23 14:13:19.504052: 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) I0423 14:13:19.716277 134304599349056 max_utils.py:273] Attempting to initialize the jax distributed system... I0423 14:13:28.756671 134304599349056 distributed.py:149] Starting JAX distributed service on [::]:8482 I0423 14:13:28.758912 134304599349056 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-sv-1d6sn-slice-job-0-0.mt-16-zarr3-false-sv-1d6sn:8482 I0423 14:13:30.025001 134304599349056 max_utils.py:284] Jax distributed system initialized! I0423 14:13:36.193501 134304599349056 max_utils.py:800] System Information: Jax Version: 0.9.2 I0423 14:13:36.193610 134304599349056 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0423 14:13:36.193651 134304599349056 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 I0423 14:13:36.193689 134304599349056 train_utils.py:391] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0423 14:13:36.891258 134304599349056 maxtext_utils.py:1771] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 14:13:36.891849 134304599349056 maxtext_utils.py:1771] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 14:13:36.892043 134304599349056 checkpointing.py:688] Setting up checkpoint logger... I0423 14:13:36.892097 134304599349056 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=False I0423 14:13:36.892141 134304599349056 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0423 14:13:36.892486 134304599349056 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7a25a611c1a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0423 14:13:40.261672 134304599349056 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0423 14:13:40.261912 134304599349056 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 0x7a111c5324b0>}, handler_registry=None I0423 14:13:40.262176 134304599349056 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a111c5324b0>` 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`. I0423 14:13:40.262227 134304599349056 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a111c5348c0>` 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`. I0423 14:13:40.262263 134304599349056 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 0x7a111c5324b0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a111c5324b0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a111c5348c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a111c5348c0>}). I0423 14:13:40.262591 134304599349056 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0423 14:13:40.262662 134304599349056 async_checkpointer.py:192] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7a10442e9e40> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0423 14:13:41.413041 134304599349056 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/linen_xpk_feat_nnx_post_train_fixes_20260423_124541_16_zarr3_false/checkpoints I0423 14:13:41.415313 134304599349056 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=5), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/linen_xpk_feat_nnx_post_train_fixes_20260423_124541_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7a111c533890> I0423 14:13:41.415427 134304599349056 checkpointing.py:302] Checkpoint manager created! I0423 14:13:42.356605 134304599349056 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0423 14:13:42.356719 134304599349056 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 14:13:42.736771 134304599349056 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 14:13:42.736864 134304599349056 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 14:13:42.753356 134304599349056 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 14:13:42.753417 134304599349056 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0423 14:13:42.777866 134304599349056 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 14:13:42.777948 134304599349056 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 14:13:42.794612 134304599349056 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 14:13:42.794684 134304599349056 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 14:13:42.811264 134304599349056 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 14:13:42.811327 134304599349056 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 14:13:42.836517 134304599349056 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0423 14:13:42.836588 134304599349056 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0423 14:13:42.857779 134304599349056 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0423 14:13:42.857846 134304599349056 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0423 14:13:43.084563 134304599349056 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0423 14:13:43.084697 134304599349056 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0423 14:13:44.520172 134304599349056 maxtext_utils.py:1880] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0423 14:13:44.520300 134304599349056 maxtext_utils.py:1880] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0423 14:13:44.520357 134304599349056 maxtext_utils.py:1880] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0423 14:13:44.520415 134304599349056 maxtext_utils.py:1880] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0423 14:13:44.520474 134304599349056 maxtext_utils.py:1880] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0423 14:13:44.520519 134304599349056 maxtext_utils.py:1880] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0423 14:13:44.520574 134304599349056 maxtext_utils.py:1880] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0423 14:13:44.520629 134304599349056 maxtext_utils.py:1880] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0423 14:13:44.520670 134304599349056 maxtext_utils.py:1880] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0423 14:13:44.520708 134304599349056 maxtext_utils.py:1880] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0423 14:13:44.520756 134304599349056 maxtext_utils.py:1880] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0423 14:13:44.520808 134304599349056 maxtext_utils.py:1880] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0423 14:13:45.020088 134304599349056 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0423 14:13:45.020183 134304599349056 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0423 14:13:45.035892 134304599349056 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0423 14:13:45.035965 134304599349056 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0423 14:13:56.034121 134304599349056 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. I0423 14:13:56.035011 134304599349056 metric_logger.py:301] number parameters: 1.104 billion I0423 14:14:07.392852 134304599349056 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0423 14:14:07.548515 134304599349056 checkpointing.py:798] Waited 0.15564346313476562 seconds for step 0 to finish before starting checkpointing. I0423 14:14:07.550871 134304599349056 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 14:14:07.552539 134304599349056 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 0 I0423 14:14:07.554007 134304599349056 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/linen_xpk_feat_nnx_post_train_fixes_20260423_124541_16_zarr3_false/checkpoints/0. I0423 14:14:08.336760 134304599349056 signaling_client.py:364] Using JaxDistributedSignalingClient I0423 14:14:08.337943 134304599349056 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0423 14:14:08.338004 134304599349056 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 I0423 14:14:08.616610 134304599349056 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.279942s I0423 14:14:08.616791 134304599349056 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 2.144 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.7195043563842773 s) (per-host) I0423 14:14:08.616846 134304599349056 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.719569s (batch_requests_ready=0.436264s, total_serialization_initiated=0.283232s, others=0.000073s) I0423 14:14:08.616954 134304599349056 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.723746s (all_items=0.000019s, per_item={'items': '0.00001907'}, temp_paths=0.723727) I0423 14:14:08.617728 134304599349056 event_tracking.py:125] [process=3] [async] Finished blocking save in 1.07 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/linen_xpk_feat_nnx_post_train_fixes_20260423_124541_16_zarr3_false/checkpoints/0. I0423 14:14:08.618072 134176044226304 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 14:34:08.618033 I0423 14:14:08.620117 134304599349056 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0423 14:14:08.620366 134174425995008 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0423 14:14:08.620488 134304599349056 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/linen_xpk_feat_nnx_post_train_fixes_20260423_124541_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776953647.5508523, 'wait_for_prev_duration_secs': 6.29425048828125e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776953647.552579, 'checkpointer_blocking_duration_secs': 1.0656490325927734, 'get_old_steps_start_time': 1776953648.6182535, 'get_old_steps_duration_secs': 3.123283386230469e-05, 'checkpoint_manager_blocking_start_time': 1776953647.5489075, 'checkpoint_manager_blocking_duration_secs': 1.0715417861938477} I0423 14:14:08.620601 134304599349056 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0423 14:14:08.620651 134304599349056 max_utils.py:750] Memstats: After params initialized: I0423 14:14:08.620704 134304599349056 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_10(process=3,(2,2,0,0)) I0423 14:14:08.620737 134304599349056 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_11(process=3,(3,2,0,0)) I0423 14:14:08.620765 134304599349056 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_14(process=3,(2,3,0,0)) I0423 14:14:08.620789 134304599349056 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_15(process=3,(3,3,0,0)) I0423 14:14:08.937261 134304599349056 metric_logger.py:196] completed step: 0, seconds: 11.358, TFLOP/s/device: 1.196, Tokens/s/device: 180.318, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0423 14:14:09.111504 134304599349056 metric_logger.py:196] completed step: 1, seconds: 1.543, TFLOP/s/device: 8.806, Tokens/s/device: 1327.369, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0423 14:14:09.541759 134304599349056 metric_logger.py:196] completed step: 2, seconds: 0.027, TFLOP/s/device: 505.775, Tokens/s/device: 76235.855, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0423 14:14:09.689065 134304599349056 metric_logger.py:196] completed step: 3, seconds: 0.430, TFLOP/s/device: 31.577, Tokens/s/device: 4759.592, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0423 14:14:09.693298 134304599349056 checkpointing.py:794] Waiting for step 5 to finish before checkpoint... I0423 14:14:09.983109 134304599349056 checkpointing.py:798] Waited 0.28978443145751953 seconds for step 5 to finish before starting checkpointing. I0423 14:14:09.985922 134304599349056 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 14:14:11.158600 2791 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0423 14:14:14.129397 134174442780416 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/linen_xpk_feat_nnx_post_train_fixes_20260423_124541_16_zarr3_false/checkpoints/0/items/array_metadatas/process_3 I0423 14:14:47.474846 134176044226304 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 39.911 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 39.57751154899597 s) (per-host) I0423 14:14:47.474987 134176044226304 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 38.856798s. I0423 14:14:58.299845 134176044226304 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 49.681640s. I0423 14:14:58.300163 134174425995008 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 14:14:58.300287 134174425995008 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0423 14:14:58.300339 134174425995008 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0423 14:14:58.301884 134174425995008 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0423 14:14:58.302044 134304599349056 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. W0423 14:14:58.302161 134304599349056 checkpoint_manager.py:1452] Waiting for previous save to complete took 48.316236 seconds. If this number is high, consider checkpointing less frequently. I0423 14:14:58.303959 134304599349056 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 5 I0423 14:14:58.306438 134304599349056 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/linen_xpk_feat_nnx_post_train_fixes_20260423_124541_16_zarr3_false/checkpoints/5. I0423 14:14:58.619950 134304599349056 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0423 14:14:58.620044 134304599349056 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 I0423 14:14:58.656356 134304599349056 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.037445s I0423 14:14:58.656509 134304599349056 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 37.711 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04090523719787598 s) (per-host) I0423 14:14:58.656560 134304599349056 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.040966s (batch_requests_ready=0.001788s, total_serialization_initiated=0.039109s, others=0.000069s) I0423 14:14:58.656643 134304599349056 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.045835s (all_items=0.000017s, per_item={'items': '0.00001669'}, temp_paths=0.045818) I0423 14:14:58.657383 134304599349056 event_tracking.py:125] [process=3] [async] Finished blocking save in 0.35 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/linen_xpk_feat_nnx_post_train_fixes_20260423_124541_16_zarr3_false/checkpoints/5. I0423 14:14:58.657720 134174425995008 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 14:34:58.657680 I0423 14:14:58.664358 134304599349056 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0423 14:14:58.664611 134175516522240 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0423 14:14:58.664778 134304599349056 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/linen_xpk_feat_nnx_post_train_fixes_20260423_124541_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776953649.9858918, 'wait_for_prev_duration_secs': 48.31623554229736, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776953698.3039973, 'checkpointer_blocking_duration_secs': 0.3538689613342285, 'get_old_steps_start_time': 1776953698.6578896, 'get_old_steps_duration_secs': 3.075599670410156e-05, 'checkpoint_manager_blocking_start_time': 1776953649.983689, 'checkpoint_manager_blocking_duration_secs': 48.68105149269104} I0423 14:14:58.664883 134304599349056 checkpointing.py:409] Started an asynchronous checkpoint save for step 5 I0423 14:14:58.665714 134304599349056 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.827, Tokens/s/device: 13389.034, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0423 14:14:58.671395 134304599349056 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.222, Tokens/s/device: 13900.767, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0423 14:15:03.808786 134174442780416 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/linen_xpk_feat_nnx_post_train_fixes_20260423_124541_16_zarr3_false/checkpoints/5/items/array_metadatas/process_3 I0423 14:15:21.165872 134304599349056 metric_logger.py:196] completed step: 6, seconds: 48.977, TFLOP/s/device: 0.277, Tokens/s/device: 41.816, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0423 14:15:21.313189 134304599349056 metric_logger.py:196] completed step: 7, seconds: 22.348, TFLOP/s/device: 0.608, Tokens/s/device: 91.640, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0423 14:15:21.460478 134304599349056 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.228, Tokens/s/device: 13449.440, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0423 14:15:21.607329 134304599349056 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0423 14:15:21.608020 134304599349056 checkpointing.py:798] Waited 0.0007104873657226562 seconds for step 9 to finish before starting checkpointing. I0423 14:15:21.609887 134304599349056 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 14:15:41.789442 134174425995008 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 36.587 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.17380619049072 s) (per-host) I0423 14:15:41.789555 134174425995008 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.131722s. I0423 14:15:48.993262 134174425995008 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 50.335414s. I0423 14:15:48.993576 134175516522240 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 14:15:48.993708 134175516522240 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0423 14:15:48.993772 134175516522240 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0423 14:15:48.995972 134175516522240 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0423 14:15:48.996184 134304599349056 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0423 14:15:48.996314 134304599349056 checkpoint_manager.py:1452] Waiting for previous save to complete took 27.386435 seconds. If this number is high, consider checkpointing less frequently. I0423 14:15:48.997797 134304599349056 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 9 I0423 14:15:48.999837 134304599349056 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/linen_xpk_feat_nnx_post_train_fixes_20260423_124541_16_zarr3_false/checkpoints/9. I0423 14:15:49.285303 134304599349056 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0423 14:15:49.285397 134304599349056 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 I0423 14:15:49.318594 134304599349056 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.034202s I0423 14:15:49.318745 134304599349056 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 40.944 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.037674903869628906 s) (per-host) I0423 14:15:49.318804 134304599349056 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.037743s (batch_requests_ready=0.001654s, total_serialization_initiated=0.036016s, others=0.000072s) I0423 14:15:49.318888 134304599349056 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.041833s (all_items=0.000011s, per_item={'items': '0.00001097'}, temp_paths=0.041822) I0423 14:15:49.319591 134304599349056 event_tracking.py:125] [process=3] [async] Finished blocking save in 0.32 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/linen_xpk_feat_nnx_post_train_fixes_20260423_124541_16_zarr3_false/checkpoints/9. I0423 14:15:49.319884 134174442780416 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-23 14:35:49.319851 I0423 14:15:49.328049 134304599349056 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0423 14:15:49.328284 134175516522240 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0423 14:15:49.328407 134304599349056 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/linen_xpk_feat_nnx_post_train_fixes_20260423_124541_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776953721.6098557, 'wait_for_prev_duration_secs': 27.38643503189087, 'time_between_consecutive_saves_sec': 23.307918310165405, 'checkpointer_blocking_start_time': 1776953748.9978392, 'checkpointer_blocking_duration_secs': 0.32219576835632324, 'get_old_steps_start_time': 1776953749.320053, 'get_old_steps_duration_secs': 2.3603439331054688e-05, 'checkpoint_manager_blocking_start_time': 1776953721.6082463, 'checkpoint_manager_blocking_duration_secs': 27.720129251480103} I0423 14:15:49.328525 134304599349056 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0423 14:15:49.328572 134304599349056 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 14:15:54.735002 134169602549504 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260423_124541/linen_xpk_feat_nnx_post_train_fixes_20260423_124541_16_zarr3_false/checkpoints/9/items/array_metadatas/process_3 I0423 14:16:32.230070 134174442780416 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 36.778 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.94895315170288 s) (per-host) I0423 14:16:32.230199 134174442780416 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.910193s. I0423 14:16:39.588521 134174442780416 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 50.268500s. I0423 14:16:39.588733 134175516522240 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 14:16:39.588788 134175516522240 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0423 14:16:39.588830 134175516522240 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0423 14:16:39.590230 134175516522240 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0423 14:16:39.590412 134304599349056 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. I0423 14:16:39.590578 134304599349056 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 14:16:39.591469 134304599349056 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.314, Tokens/s/device: 13914.556, total_weights: 65536, loss: 7.181, lm_loss: 7.181, perplexity: 1314.495 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Thu Apr 23 14:16:50 UTC 2026 EXIT_CODE=0