main (8a17c3d19) vs feat/nnx-post-train-fixes (7f06c99ac)| Metric | main 8a17c3d19 | 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 | 92.263 | 92.143 | -0.12 |
| Tok/s | 13906.9 | 13888.8 | -18.108 |
| Avg s/step | 8.220 | 8.241 | +0.021 |
| 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: Wed Apr 22 08:40:44 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. 2026-04-22 08:41:09.325443: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0422 08:41:09.535939 135318487500608 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 08:41:18.578330 135318487500608 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 08:41:18.580645 135318487500608 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-sv-n7gip-slice-job-0-0.mt-16-zarr3-false-sv-n7gip:8482 I0422 08:41:19.799145 135318487500608 max_utils.py:284] Jax distributed system initialized! I0422 08:41:24.828911 135318487500608 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 08:41:24.829019 135318487500608 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 08:41:24.829060 135318487500608 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0422 08:41:24.829124 135318487500608 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 08:41:25.527453 135318487500608 maxtext_utils.py:1565] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 08:41:25.527732 135318487500608 checkpointing.py:677] Setting up checkpoint logger... I0422 08:41:25.527787 135318487500608 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=False I0422 08:41:25.527832 135318487500608 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 08:41:25.528192 135318487500608 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 0x7b11bb914110>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 08:41:28.532209 135318487500608 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0422 08:41:28.532449 135318487500608 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 0x7afd14361ac0>}, handler_registry=None I0422 08:41:28.532692 135318487500608 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7afd14361ac0>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0422 08:41:28.532742 135318487500608 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7afd14365be0>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0422 08:41:28.532775 135318487500608 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 0x7afd14361ac0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7afd14361ac0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7afd14365be0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7afd14365be0>}). I0422 08:41:28.533114 135318487500608 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 08:41:28.533186 135318487500608 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7afcd0679580> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 08:41:29.676495 135318487500608 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_false/checkpoints I0422 08:41:30.101228 135318487500608 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_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7afd14362b10> I0422 08:41:30.101410 135318487500608 checkpointing.py:301] Checkpoint manager created! I0422 08:41:31.023857 135318487500608 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 08:41:31.023976 135318487500608 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 08:41:31.409573 135318487500608 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 08:41:31.409671 135318487500608 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 08:41:31.426220 135318487500608 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 08:41:31.426279 135318487500608 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 08:41:31.450253 135318487500608 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 08:41:31.450326 135318487500608 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:41:31.466898 135318487500608 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 08:41:31.466960 135318487500608 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:41:31.483467 135318487500608 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 08:41:31.483532 135318487500608 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:41:31.508499 135318487500608 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 08:41:31.508568 135318487500608 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 08:41:31.529926 135318487500608 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 08:41:31.530002 135318487500608 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 08:41:31.740560 135318487500608 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0422 08:41:31.740670 135318487500608 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 08:41:33.178661 135318487500608 maxtext_utils.py:1668] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0422 08:41:33.178790 135318487500608 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 08:41:33.178844 135318487500608 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 08:41:33.178902 135318487500608 maxtext_utils.py:1668] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 08:41:33.178954 135318487500608 maxtext_utils.py:1668] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 08:41:33.178992 135318487500608 maxtext_utils.py:1668] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 08:41:33.179043 135318487500608 maxtext_utils.py:1668] 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) I0422 08:41:33.179107 135318487500608 maxtext_utils.py:1668] 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') I0422 08:41:33.179150 135318487500608 maxtext_utils.py:1668] 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) I0422 08:41:33.179188 135318487500608 maxtext_utils.py:1668] 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) I0422 08:41:33.179236 135318487500608 maxtext_utils.py:1668] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 08:41:33.179284 135318487500608 maxtext_utils.py:1668] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 08:41:33.671064 135318487500608 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 08:41:33.671172 135318487500608 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 08:41:33.686783 135318487500608 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 08:41:33.686844 135318487500608 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 08:41:44.340560 135318487500608 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0422 08:41:44.341428 135318487500608 metric_logger.py:301] number parameters: 1.104 billion I0422 08:41:55.567914 135318487500608 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0422 08:41:56.597834 135318487500608 checkpointing.py:776] Waited 1.0298995971679688 seconds for step 0 to finish before starting checkpointing. I0422 08:41:56.601635 135318487500608 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 08:41:56.603039 135318487500608 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0422 08:41:56.604694 135318487500608 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_false/checkpoints/0. I0422 08:41:56.939553 135318487500608 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 08:41:56.940483 135318487500608 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 08:41:56.940545 135318487500608 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 08:41:57.218256 135318487500608 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278779s I0422 08:41:57.218428 135318487500608 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.420 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2845947742462158 s) (per-host) I0422 08:41:57.218480 135318487500608 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.284657s (batch_requests_ready=0.002628s, total_serialization_initiated=0.281960s, others=0.000070s) I0422 08:41:57.218577 135318487500608 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.289388s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=0.289370) I0422 08:41:57.219437 135318487500608 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.62 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_false/checkpoints/0. I0422 08:41:57.219701 135191494289152 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 09:01:57.219679 I0422 08:41:57.224771 135318487500608 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 08:41:57.225042 135189348034304 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 08:41:57.225185 135318487500608 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776847316.6016166, 'wait_for_prev_duration_secs': 6.270408630371094e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776847316.603078, 'checkpointer_blocking_duration_secs': 0.6167240142822266, 'get_old_steps_start_time': 1776847317.2198267, 'get_old_steps_duration_secs': 3.0279159545898438e-05, 'checkpoint_manager_blocking_start_time': 1776847316.5984056, 'checkpoint_manager_blocking_duration_secs': 0.6267430782318115} I0422 08:41:57.225348 135318487500608 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0422 08:41:57.225398 135318487500608 max_utils.py:750] Memstats: After params initialized: I0422 08:41:57.225444 135318487500608 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0422 08:41:57.225476 135318487500608 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0422 08:41:57.225502 135318487500608 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0422 08:41:57.225527 135318487500608 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0422 08:41:57.540187 135318487500608 metric_logger.py:196] completed step: 0, seconds: 11.226, TFLOP/s/device: 1.210, Tokens/s/device: 182.427, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 08:41:57.729230 135318487500608 metric_logger.py:196] completed step: 1, seconds: 1.971, TFLOP/s/device: 6.894, Tokens/s/device: 1039.194, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 08:41:58.124860 135318487500608 metric_logger.py:196] completed step: 2, seconds: 0.042, TFLOP/s/device: 327.314, Tokens/s/device: 49336.320, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0422 08:41:58.272199 135318487500608 metric_logger.py:196] completed step: 3, seconds: 0.396, TFLOP/s/device: 34.333, Tokens/s/device: 5175.023, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0422 08:41:58.276638 135318487500608 checkpointing.py:772] Waiting for step 5 to finish before checkpoint... I0422 08:41:58.566693 135318487500608 checkpointing.py:776] Waited 0.29001855850219727 seconds for step 5 to finish before starting checkpointing. I0422 08:41:58.569689 135318487500608 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 08:42:00.416050 2788 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 08:42:02.451356 135189397239552 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_false/checkpoints/0/items/array_metadatas/process_6 I0422 08:42:40.066043 135191494289152 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 36.622 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.13217067718506 s) (per-host) I0422 08:42:40.066176 135191494289152 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.846403s. I0422 08:42:46.203572 135191494289152 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 48.983783s. I0422 08:42:46.203844 135189348034304 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 08:42:46.203965 135189348034304 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 08:42:46.204015 135189348034304 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 08:42:46.206436 135189348034304 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 08:42:46.206630 135318487500608 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0422 08:42:46.206764 135318487500608 checkpoint_manager.py:1452] Waiting for previous save to complete took 47.637078 seconds. If this number is high, consider checkpointing less frequently. I0422 08:42:46.208734 135318487500608 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 5 I0422 08:42:46.211117 135318487500608 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_false/checkpoints/5. I0422 08:42:46.578204 135318487500608 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 08:42:46.578300 135318487500608 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 08:42:46.614932 135318487500608 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.037850s I0422 08:42:46.615111 135318487500608 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 37.285 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04137277603149414 s) (per-host) I0422 08:42:46.615165 135318487500608 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.041448s (batch_requests_ready=0.001880s, total_serialization_initiated=0.039486s, others=0.000082s) I0422 08:42:46.615255 135318487500608 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.047063s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.047047) I0422 08:42:46.616000 135318487500608 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.41 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_false/checkpoints/5. I0422 08:42:46.616358 135189348034304 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 09:02:46.616321 I0422 08:42:46.621889 135318487500608 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0422 08:42:46.622149 135189397239552 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 08:42:46.622315 135318487500608 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776847318.5696592, 'wait_for_prev_duration_secs': 47.63707756996155, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776847366.2087727, 'checkpointer_blocking_duration_secs': 0.4077341556549072, 'get_old_steps_start_time': 1776847366.6165307, 'get_old_steps_duration_secs': 3.170967102050781e-05, 'checkpoint_manager_blocking_start_time': 1776847318.5673585, 'checkpoint_manager_blocking_duration_secs': 48.05492091178894} I0422 08:42:46.622426 135318487500608 checkpointing.py:408] Started an asynchronous checkpoint save for step 5 I0422 08:42:46.623183 135318487500608 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.632, Tokens/s/device: 13359.600, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0422 08:42:46.629508 135318487500608 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.231, Tokens/s/device: 13901.994, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0422 08:42:51.717997 135169885312768 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_false/checkpoints/5/items/array_metadatas/process_6 I0422 08:43:09.393676 135318487500608 metric_logger.py:196] completed step: 6, seconds: 48.351, TFLOP/s/device: 0.281, Tokens/s/device: 42.357, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0422 08:43:09.540952 135318487500608 metric_logger.py:196] completed step: 7, seconds: 22.618, TFLOP/s/device: 0.601, Tokens/s/device: 90.548, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0422 08:43:09.688172 135318487500608 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.223, Tokens/s/device: 13448.733, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0422 08:43:09.834817 135318487500608 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0422 08:43:09.835445 135318487500608 checkpointing.py:776] Waited 0.0006470680236816406 seconds for step 9 to finish before starting checkpointing. I0422 08:43:09.837998 135318487500608 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 08:43:28.627410 135189348034304 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.561 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.05364274978638 s) (per-host) I0422 08:43:28.627532 135189348034304 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.011059s. I0422 08:43:34.709909 135189348034304 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 48.093421s. I0422 08:43:34.710202 135189397239552 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 08:43:34.710323 135189397239552 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 08:43:34.710382 135189397239552 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0422 08:43:34.712007 135189397239552 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0422 08:43:34.712203 135318487500608 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. W0422 08:43:34.712334 135318487500608 checkpoint_manager.py:1452] Waiting for previous save to complete took 24.874343 seconds. If this number is high, consider checkpointing less frequently. I0422 08:43:34.713825 135318487500608 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 9 I0422 08:43:34.716621 135318487500608 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_false/checkpoints/9. I0422 08:43:35.022747 135318487500608 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 08:43:35.022839 135318487500608 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 08:43:35.058349 135318487500608 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036422s I0422 08:43:35.058490 135318487500608 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 38.651 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03991055488586426 s) (per-host) I0422 08:43:35.058539 135318487500608 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039969s (batch_requests_ready=0.001771s, total_serialization_initiated=0.038134s, others=0.000064s) I0422 08:43:35.058615 135318487500608 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.044093s (all_items=0.000012s, per_item={'items': '0.00001216'}, temp_paths=0.044081) I0422 08:43:35.059262 135318487500608 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.35 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_false/checkpoints/9. I0422 08:43:35.059609 135169885312768 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 09:03:35.059570 I0422 08:43:35.061624 135318487500608 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 08:43:35.061843 135189397239552 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 08:43:35.061963 135318487500608 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776847389.837966, 'wait_for_prev_duration_secs': 24.874343156814575, 'time_between_consecutive_saves_sec': 23.63148832321167, 'checkpointer_blocking_start_time': 1776847414.713866, 'checkpointer_blocking_duration_secs': 0.3459012508392334, 'get_old_steps_start_time': 1776847415.0597866, 'get_old_steps_duration_secs': 2.3603439331054688e-05, 'checkpoint_manager_blocking_start_time': 1776847389.8356843, 'checkpoint_manager_blocking_duration_secs': 25.226248741149902} I0422 08:43:35.062069 135318487500608 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0422 08:43:35.062122 135318487500608 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 08:43:40.091585 135189369005824 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_16_zarr3_false/checkpoints/9/items/array_metadatas/process_6 I0422 08:44:17.319960 135169885312768 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.342 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.30135488510132 s) (per-host) I0422 08:44:17.320062 135169885312768 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.260338s. I0422 08:44:22.680726 135169885312768 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 47.620986s. I0422 08:44:22.681028 135189397239552 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 08:44:22.681188 135189397239552 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 08:44:22.681231 135189397239552 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 08:44:22.683763 135189397239552 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 08:44:22.683943 135318487500608 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. I0422 08:44:22.684105 135318487500608 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 08:44:22.685141 135318487500608 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.263, Tokens/s/device: 13906.903, 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: Wed Apr 22 08:44:34 UTC 2026 EXIT_CODE=0
XPK Start: Wed Apr 22 14:07:24 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-22 14:07:48.843920: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0422 14:07:49.378295 135820639934272 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 14:07:58.417292 135820639934272 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 14:07:58.419897 135820639934272 distributed.py:172] Connecting to JAX distributed service on mt-16-zarr3-false-sv-1rj9e-slice-job-0-0.mt-16-zarr3-false-sv-1rj9e:8482 I0422 14:07:59.971943 135820639934272 max_utils.py:284] Jax distributed system initialized! I0422 14:08:06.851454 135820639934272 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 14:08:06.851555 135820639934272 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 14:08:06.851621 135820639934272 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0422 14:08:06.851662 135820639934272 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 14:08:07.533423 135820639934272 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 14:08:07.534063 135820639934272 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 14:08:07.534262 135820639934272 checkpointing.py:688] Setting up checkpoint logger... I0422 14:08:07.534329 135820639934272 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=False I0422 14:08:07.534382 135820639934272 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 14:08:07.534756 135820639934272 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 0x7b867d650200>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 14:08:10.460525 135820639934272 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0422 14:08:10.460793 135820639934272 checkpoint_manager.py:708] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b73e96f5c70>}, handler_registry=None I0422 14:08:10.461070 135820639934272 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b73e96f5c70>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0422 14:08:10.461127 135820639934272 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b73e89525d0>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0422 14:08:10.461180 135820639934272 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 0x7b73e96f5c70>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b73e96f5c70>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b73e89525d0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b73e89525d0>}). I0422 14:08:10.461543 135820639934272 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 14:08:10.461638 135820639934272 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7b73597dc180> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 14:08:11.628281 135820639934272 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints I0422 14:08:11.835649 135820639934272 checkpoint_manager.py:929] [process=0][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_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7b73e8950dd0> I0422 14:08:11.835819 135820639934272 checkpointing.py:302] Checkpoint manager created! I0422 14:08:12.761007 135820639934272 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 14:08:12.761111 135820639934272 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 14:08:13.142478 135820639934272 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 14:08:13.142570 135820639934272 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 14:08:13.159174 135820639934272 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 14:08:13.159242 135820639934272 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 14:08:13.183829 135820639934272 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 14:08:13.183896 135820639934272 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 14:08:13.200555 135820639934272 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 14:08:13.200627 135820639934272 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 14:08:13.217228 135820639934272 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 14:08:13.217288 135820639934272 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 14:08:13.242289 135820639934272 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 14:08:13.242357 135820639934272 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 14:08:13.263716 135820639934272 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 14:08:13.263778 135820639934272 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 14:08:13.484481 135820639934272 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0422 14:08:13.484600 135820639934272 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 14:08:14.925665 135820639934272 maxtext_utils.py:1827] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0422 14:08:14.925791 135820639934272 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 14:08:14.925844 135820639934272 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 14:08:14.925912 135820639934272 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 14:08:14.925969 135820639934272 maxtext_utils.py:1827] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 14:08:14.926020 135820639934272 maxtext_utils.py:1827] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0422 14:08:14.926114 135820639934272 maxtext_utils.py:1827] 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) I0422 14:08:14.926180 135820639934272 maxtext_utils.py:1827] 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') I0422 14:08:14.926225 135820639934272 maxtext_utils.py:1827] 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) I0422 14:08:14.926262 135820639934272 maxtext_utils.py:1827] 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) I0422 14:08:14.926312 135820639934272 maxtext_utils.py:1827] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 14:08:14.926368 135820639934272 maxtext_utils.py:1827] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 14:08:15.427598 135820639934272 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 14:08:15.427697 135820639934272 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 14:08:15.443753 135820639934272 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 14:08:15.443814 135820639934272 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 14:08:26.446223 135820639934272 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0422 14:08:26.454000 135820639934272 metric_logger.py:301] number parameters: 1.104 billion I0422 14:08:38.092798 135820639934272 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0422 14:08:38.245299 135820639934272 checkpointing.py:798] Waited 0.15247845649719238 seconds for step 0 to finish before starting checkpointing. I0422 14:08:38.247883 135820639934272 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 14:08:38.249507 135820639934272 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0422 14:08:38.251212 135820639934272 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0. I0422 14:08:38.567526 135820639934272 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 14:08:38.581732 135820639934272 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 14:08:38.581798 135820639934272 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 14:08:38.835915 135690020972288 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0 I0422 14:08:38.850947 135820639934272 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.269979s I0422 14:08:38.851794 135820639934272 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.584 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2762563228607178 s) (per-host) I0422 14:08:38.851866 135820639934272 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.276348s (batch_requests_ready=0.002060s, total_serialization_initiated=0.273524s, others=0.000764s) I0422 14:08:38.851946 135820639934272 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.280530s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.280512) I0422 14:08:38.852784 135820639934272 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.60 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0. I0422 14:08:38.853111 135690029364992 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 14:28:38.853073 I0422 14:08:39.299780 135820639934272 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 14:08:39.300184 135688410359552 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0422 14:08:39.300352 135820639934272 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776866918.24786, 'wait_for_prev_duration_secs': 7.081031799316406e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776866918.2495434, 'checkpointer_blocking_duration_secs': 0.6037154197692871, 'get_old_steps_start_time': 1776866918.853284, 'get_old_steps_duration_secs': 3.0517578125e-05, 'checkpoint_manager_blocking_start_time': 1776866918.2458227, 'checkpoint_manager_blocking_duration_secs': 1.054488182067871} I0422 14:08:39.300459 135820639934272 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0422 14:08:39.300510 135820639934272 max_utils.py:750] Memstats: After params initialized: I0422 14:08:39.300567 135820639934272 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_0(process=0,(0,0,0,0)) I0422 14:08:39.300616 135820639934272 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_1(process=0,(1,0,0,0)) I0422 14:08:39.300643 135820639934272 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_4(process=0,(0,1,0,0)) I0422 14:08:39.300668 135820639934272 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_5(process=0,(1,1,0,0)) I0422 14:08:39.646992 135820639934272 metric_logger.py:196] completed step: 0, seconds: 11.330, TFLOP/s/device: 1.199, Tokens/s/device: 180.753, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 14:08:39.650315 135820639934272 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/tensorboard/' I0422 14:08:40.097596 135820639934272 metric_logger.py:196] completed step: 1, seconds: 1.552, TFLOP/s/device: 8.753, Tokens/s/device: 1319.315, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 14:08:40.243613 135820639934272 metric_logger.py:196] completed step: 2, seconds: 0.451, TFLOP/s/device: 30.106, Tokens/s/device: 4537.891, total_weights: 65536, loss: 10.560, lm_loss: 10.560, perplexity: 38547.719 I0422 14:08:40.391058 135820639934272 metric_logger.py:196] completed step: 3, seconds: 0.009, TFLOP/s/device: 1453.791, Tokens/s/device: 219131.179, total_weights: 65536, loss: 9.980, lm_loss: 9.980, perplexity: 21590.506 I0422 14:08:40.400972 135820639934272 checkpointing.py:794] Waiting for step 5 to finish before checkpoint... I0422 14:08:40.685174 135820639934272 checkpointing.py:798] Waited 0.28406453132629395 seconds for step 5 to finish before starting checkpointing. I0422 14:08:40.688168 135820639934272 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 14:08:40.708854 135689492494080 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0/items I0422 14:08:40.887897 135686241904384 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776866920433518276, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776866920433518276, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0/_CHECKPOINT_METADATA I0422 14:08:41.853980 2915 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 14:08:43.666119 135689446344448 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0/items/array_metadatas/process_0 I0422 14:09:24.206758 135689437951744 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 42.351672s (commit=40.664134s, array_metadata_write=1.687539s) I0422 14:09:24.208947 135690029364992 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.615 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.63338565826416 s) (per-host) I0422 14:09:24.209050 135690029364992 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.355822s. I0422 14:09:24.989681 135690029364992 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776866920433518276, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0/_CHECKPOINT_METADATA I0422 14:09:26.002560 135690029364992 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000400s. I0422 14:09:26.206770 135686241904384 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776866920433518276, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0/_CHECKPOINT_METADATA I0422 14:09:26.705835 135690029364992 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.590626s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0/items I0422 14:09:26.708089 135690029364992 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0/items I0422 14:09:27.304026 135690029364992 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0 I0422 14:09:28.777734 135690029364992 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0`. I0422 14:09:28.778868 135690029364992 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 50.53 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/0 I0422 14:09:28.780609 135690029364992 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 49.927377s. I0422 14:09:28.781483 135688410359552 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 14:09:28.781607 135688410359552 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0422 14:09:28.781714 135688410359552 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 14:09:28.783437 135688410359552 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 14:09:28.783695 135820639934272 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0422 14:09:28.783875 135820639934272 checkpoint_manager.py:1452] Waiting for previous save to complete took 48.095708 seconds. If this number is high, consider checkpointing less frequently. I0422 14:09:28.785644 135820639934272 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 5 I0422 14:09:28.788026 135820639934272 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/5. I0422 14:09:29.095705 135820639934272 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 14:09:29.095839 135820639934272 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 14:09:29.130713 135820639934272 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036021s I0422 14:09:29.131642 135820639934272 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 34.589 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04459667205810547 s) (per-host) I0422 14:09:29.131743 135820639934272 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.044721s (batch_requests_ready=0.004049s, total_serialization_initiated=0.039839s, others=0.000833s) I0422 14:09:29.131870 135820639934272 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.049608s (all_items=0.000025s, per_item={'items': '0.00002503'}, temp_paths=0.049583) I0422 14:09:29.132912 135820639934272 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.35 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/5. I0422 14:09:29.133286 135689458923264 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 14:29:29.133244 I0422 14:09:29.136748 135820639934272 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=5] Starting CheckpointManager Save Finalize thread=save_finalize I0422 14:09:29.138048 135688410359552 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0422 14:09:29.138284 135820639934272 standard_logger.py:34] {'step': 5, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776866920.6881247, 'wait_for_prev_duration_secs': 48.09570789337158, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776866968.785711, 'checkpointer_blocking_duration_secs': 0.347766637802124, 'get_old_steps_start_time': 1776866969.1335084, 'get_old_steps_duration_secs': 4.1961669921875e-05, 'checkpoint_manager_blocking_start_time': 1776866920.6858702, 'checkpoint_manager_blocking_duration_secs': 48.45235991477966} I0422 14:09:29.138433 135820639934272 checkpointing.py:409] Started an asynchronous checkpoint save for step 5 I0422 14:09:29.339741 135690029364992 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/5 I0422 14:09:31.156517 135688955623168 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/5/items I0422 14:09:34.135728 135689475708672 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/5/items/array_metadatas/process_0 I0422 14:09:51.958632 135820639934272 metric_logger.py:196] completed step: 4, seconds: 0.147, TFLOP/s/device: 92.565, Tokens/s/device: 13952.474, total_weights: 65536, loss: 9.460, lm_loss: 9.460, perplexity: 12840.355 I0422 14:09:51.967701 135820639934272 metric_logger.py:196] completed step: 5, seconds: 0.148, TFLOP/s/device: 91.642, Tokens/s/device: 13813.291, total_weights: 65536, loss: 8.959, lm_loss: 8.959, perplexity: 7776.467 I0422 14:09:52.113196 135820639934272 metric_logger.py:196] completed step: 6, seconds: 71.565, TFLOP/s/device: 0.190, Tokens/s/device: 28.617, total_weights: 65536, loss: 8.469, lm_loss: 8.469, perplexity: 4762.740 I0422 14:09:52.260702 135820639934272 metric_logger.py:196] completed step: 7, seconds: 0.007, TFLOP/s/device: 1932.186, Tokens/s/device: 291240.046, total_weights: 65536, loss: 8.003, lm_loss: 8.003, perplexity: 2989.151 I0422 14:09:52.408084 135820639934272 metric_logger.py:196] completed step: 8, seconds: 0.147, TFLOP/s/device: 92.653, Tokens/s/device: 13965.604, total_weights: 65536, loss: 7.572, lm_loss: 7.572, perplexity: 1943.098 I0422 14:09:52.554646 135820639934272 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0422 14:09:52.555312 135820639934272 checkpointing.py:798] Waited 0.0006861686706542969 seconds for step 9 to finish before starting checkpointing. I0422 14:09:52.557289 135820639934272 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=5][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 14:10:13.355796 135689467315968 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.038144s (commit=39.780110s, array_metadata_write=1.258034s) I0422 14:10:13.356853 135689458923264 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 35.681 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 44.269797801971436 s) (per-host) I0422 14:10:13.356965 135689458923264 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 44.223523s. I0422 14:10:15.193848 135689458923264 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000243s. I0422 14:10:15.746500 135689458923264 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.568937s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/5/items I0422 14:10:15.748378 135689458923264 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/5/items I0422 14:10:16.502289 135689458923264 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/5 I0422 14:10:18.381555 135689458923264 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/5`. I0422 14:10:18.382350 135689458923264 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 49.60 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/5 I0422 14:10:18.383631 135689458923264 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 49.250188s. I0422 14:10:18.383762 135688410359552 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 14:10:18.383826 135688410359552 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0422 14:10:18.383894 135688410359552 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is syncing with other hosts... I0422 14:10:18.385835 135688410359552 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=5] CheckpointManager Save Finalize is done on all hosts. I0422 14:10:18.386013 135820639934272 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=5][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=5. W0422 14:10:18.386152 135820639934272 checkpoint_manager.py:1452] Waiting for previous save to complete took 25.828867 seconds. If this number is high, consider checkpointing less frequently. I0422 14:10:18.387549 135820639934272 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0422 14:10:18.389564 135820639934272 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9. I0422 14:10:18.705642 135820639934272 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 14:10:18.705738 135820639934272 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 14:10:18.731563 135820639934272 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.026884s I0422 14:10:18.734266 135820639934272 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 46.577 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03311872482299805 s) (per-host) I0422 14:10:18.734343 135820639934272 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.033215s (batch_requests_ready=0.001760s, total_serialization_initiated=0.028839s, others=0.002616s) I0422 14:10:18.734426 135820639934272 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.036998s (all_items=0.000022s, per_item={'items': '0.00002193'}, temp_paths=0.036976) I0422 14:10:18.735173 135820639934272 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.35 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9. I0422 14:10:18.735442 135689475708672 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 14:30:18.735405 I0422 14:10:18.750875 135820639934272 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 14:10:18.751200 135688410359552 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0422 14:10:18.751346 135820639934272 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776866992.5572572, 'wait_for_prev_duration_secs': 25.828866958618164, 'time_between_consecutive_saves_sec': 23.773751735687256, 'checkpointer_blocking_start_time': 1776867018.387602, 'checkpointer_blocking_duration_secs': 0.3479888439178467, 'get_old_steps_start_time': 1776867018.7356105, 'get_old_steps_duration_secs': 2.7894973754882812e-05, 'checkpoint_manager_blocking_start_time': 1776866992.5555425, 'checkpoint_manager_blocking_duration_secs': 26.195770263671875} I0422 14:10:18.751452 135820639934272 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0422 14:10:18.751496 135820639934272 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 14:10:18.966164 135689458923264 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9 I0422 14:10:21.079052 135690029364992 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9/items I0422 14:10:24.063647 135692663326464 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9/items/array_metadatas/process_0 I0422 14:11:03.999393 135691086321408 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 41.771352s (commit=40.489933s, array_metadata_write=1.281419s) I0422 14:11:04.000533 135689475708672 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.870 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.2993528842926 s) (per-host) I0422 14:11:04.000672 135689475708672 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 45.265115s. I0422 14:11:05.796289 135689475708672 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000241s. I0422 14:11:06.310630 135689475708672 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.483822s. use_zarr3=False, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9/items I0422 14:11:06.312276 135689475708672 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9/items I0422 14:11:06.895986 135689475708672 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9 I0422 14:11:08.351174 135689475708672 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9`. I0422 14:11:08.351971 135689475708672 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 49.96 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260422_123906/linen_xpk_feat_nnx_post_train_fixes_20260422_123906_16_zarr3_false/checkpoints/9 I0422 14:11:08.353545 135689475708672 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 49.617984s. I0422 14:11:08.353761 135688410359552 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 14:11:08.353873 135688410359552 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0422 14:11:08.353924 135688410359552 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 14:11:08.355459 135688410359552 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 14:11:08.355606 135820639934272 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0422 14:11:08.355706 135820639934272 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 14:11:08.356710 135820639934272 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.143, Tokens/s/device: 13888.795, 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: Wed Apr 22 14:11:20 UTC 2026 EXIT_CODE=0