main (8a17c3d19) vs feat/nnx-set-defaults-true (73213e044)| Metric | main 8a17c3d19 | feat/nnx-set-defaults-true 73213e044 | Diff (feat/nnx-set-defaults-true − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 8.1790 | 8.1790 | 0 |
| TFLOP/s | 91.809 | 92.198 | +0.389 |
| Tok/s | 13838.5 | 13897.1 | +58.596 |
| Avg s/step | 2.704 | 2.449 | -0.255 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.8.1 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Wed Apr 22 07:31:45 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 07:32:10.267660: 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 07:32:10.477755 138415253305152 max_utils.py:273] Attempting to initialize the jax distributed system... I0422 07:32:19.519728 138415253305152 distributed.py:149] Starting JAX distributed service on [::]:8482 I0422 07:32:19.522104 138415253305152 distributed.py:172] Connecting to JAX distributed service on mt-07-eval-xyj8l-slice-job-0-0.mt-07-eval-xyj8l:8482 I0422 07:32:20.851543 138415253305152 max_utils.py:284] Jax distributed system initialized! I0422 07:32:27.178910 138415253305152 max_utils.py:800] System Information: Jax Version: 0.9.2 I0422 07:32:27.179017 138415253305152 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0422 07:32:27.179058 138415253305152 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 07:32:27.179098 138415253305152 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 07:32:27.875126 138415253305152 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 07:32:27.875436 138415253305152 checkpointing.py:677] Setting up checkpoint logger... I0422 07:32:27.875495 138415253305152 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 07:32:27.875540 138415253305152 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0422 07:32:27.875878 138415253305152 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7de2af118140>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 07:32:31.189399 138415253305152 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0422 07:32:31.189644 138415253305152 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 0x7dcf6c1c5a60>}, handler_registry=None I0422 07:32:31.189884 138415253305152 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dcf6c1c5a60>` 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 07:32:31.189933 138415253305152 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dcf6c1c9490>` 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 07:32:31.189968 138415253305152 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 0x7dcf6c1c5a60>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7dcf6c1c5a60>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dcf6c1c9490>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7dcf6c1c9490>}). I0422 07:32:31.190319 138415253305152 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0422 07:32:31.190398 138415253305152 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7dce846cd760> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0422 07:32:31.914428 138415253305152 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_07_eval/checkpoints I0422 07:32:32.349873 138415253305152 checkpoint_manager.py:929] [process=6][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_07_eval/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7dcf6c1c6db0> I0422 07:32:32.350060 138415253305152 checkpointing.py:301] Checkpoint manager created! I0422 07:32:33.272232 138415253305152 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 07:32:33.272362 138415253305152 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 07:32:33.657633 138415253305152 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 07:32:33.657729 138415253305152 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 07:32:33.674393 138415253305152 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 07:32:33.674455 138415253305152 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 07:32:33.698431 138415253305152 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 07:32:33.698496 138415253305152 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 07:32:33.715211 138415253305152 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 07:32:33.715274 138415253305152 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 07:32:33.731905 138415253305152 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 07:32:33.731970 138415253305152 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 07:32:33.756933 138415253305152 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 07:32:33.757005 138415253305152 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 07:32:33.778370 138415253305152 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 07:32:33.778436 138415253305152 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 07:32:33.996763 138415253305152 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0422 07:32:33.996871 138415253305152 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 07:32:35.425643 138415253305152 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0422 07:32:35.425771 138415253305152 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) I0422 07:32:35.425824 138415253305152 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) I0422 07:32:35.425881 138415253305152 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') I0422 07:32:35.425932 138415253305152 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) I0422 07:32:35.425974 138415253305152 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) I0422 07:32:35.426028 138415253305152 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) I0422 07:32:35.426079 138415253305152 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') I0422 07:32:35.426124 138415253305152 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) I0422 07:32:35.426162 138415253305152 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) I0422 07:32:35.426209 138415253305152 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0422 07:32:35.426258 138415253305152 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0422 07:32:35.918988 138415253305152 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 07:32:35.919085 138415253305152 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 07:32:35.934738 138415253305152 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 07:32:35.934801 138415253305152 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 07:32:46.844157 138415253305152 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 07:32:46.844950 138415253305152 metric_logger.py:301] number parameters: 1.104 billion I0422 07:32:58.004917 138415253305152 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0422 07:32:58.158346 138415253305152 checkpointing.py:776] Waited 0.15338993072509766 seconds for step 0 to finish before starting checkpointing. I0422 07:32:58.160617 138415253305152 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 07:32:58.162344 138415253305152 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0422 07:32:58.163715 138415253305152 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_07_eval/checkpoints/0. I0422 07:32:58.903980 138415253305152 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 07:32:58.905026 138415253305152 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 07:32:58.905087 138415253305152 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 07:32:59.182047 138415253305152 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278141s I0422 07:32:59.182229 138415253305152 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.438 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2836735248565674 s) (per-host) I0422 07:32:59.182278 138415253305152 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.283733s (batch_requests_ready=0.002196s, total_serialization_initiated=0.281470s, others=0.000067s) I0422 07:32:59.182403 138415253305152 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.287849s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.287831) I0422 07:32:59.183234 138415253305152 event_tracking.py:125] [process=6] [async] Finished blocking save in 1.02 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_07_eval/checkpoints/0. I0422 07:32:59.183514 138286602475264 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 07:52:59.183491 I0422 07:32:59.190648 138415253305152 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 07:32:59.190928 138280670586624 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 07:32:59.191081 138415253305152 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776843178.1605985, 'wait_for_prev_duration_secs': 6.771087646484375e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776843178.1623833, 'checkpointer_blocking_duration_secs': 1.0212352275848389, 'get_old_steps_start_time': 1776843179.183645, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1776843178.1588352, 'checkpoint_manager_blocking_duration_secs': 1.0322089195251465} I0422 07:32:59.191179 138415253305152 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0422 07:32:59.191225 138415253305152 max_utils.py:750] Memstats: After params initialized: I0422 07:32:59.191272 138415253305152 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0422 07:32:59.191321 138415253305152 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0422 07:32:59.191348 138415253305152 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0422 07:32:59.191372 138415253305152 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0422 07:32:59.505611 138415253305152 metric_logger.py:196] completed step: 0, seconds: 11.160, TFLOP/s/device: 1.218, Tokens/s/device: 183.515, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 07:32:59.676992 138415253305152 metric_logger.py:196] completed step: 1, seconds: 1.499, TFLOP/s/device: 9.063, Tokens/s/device: 1366.149, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0422 07:33:00.087598 138415253305152 metric_logger.py:196] completed step: 2, seconds: 0.024, TFLOP/s/device: 555.189, Tokens/s/device: 83684.060, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0422 07:33:00.234599 138415253305152 metric_logger.py:196] completed step: 3, seconds: 0.411, TFLOP/s/device: 33.079, Tokens/s/device: 4986.038, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0422 07:33:00.529904 138415253305152 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 89.064, Tokens/s/device: 13424.755, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0422 07:33:00.535792 138415253305152 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.377, Tokens/s/device: 13924.016, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0422 07:33:02.674101 2818 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 07:33:04.664684 138283354941184 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_07_eval/checkpoints/0/items/array_metadatas/process_6 I0422 07:33:22.186776 138415253305152 metric_logger.py:196] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.915, Tokens/s/device: 6920.766, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759 I0422 07:33:22.334272 138415253305152 metric_logger.py:196] completed step: 7, seconds: 21.504, TFLOP/s/device: 0.632, Tokens/s/device: 95.236, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0422 07:33:22.481721 138415253305152 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.370, Tokens/s/device: 13470.760, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0422 07:33:22.628390 138415253305152 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0422 07:33:22.628957 138415253305152 checkpointing.py:776] Waited 0.0005886554718017578 seconds for step 9 to finish before starting checkpointing. I0422 07:33:22.631186 138415253305152 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 07:33:36.908568 138286602475264 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 41.557 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 38.00997447967529 s) (per-host) I0422 07:33:36.908689 138286602475264 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.725101s. I0422 07:33:42.632206 138286602475264 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 43.448601s. I0422 07:33:42.632553 138280670586624 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 07:33:42.632666 138280670586624 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 07:33:42.632715 138280670586624 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 07:33:42.635015 138280670586624 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 07:33:42.635229 138415253305152 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 07:33:42.635384 138415253305152 checkpoint_manager.py:1452] Waiting for previous save to complete took 20.004197 seconds. If this number is high, consider checkpointing less frequently. I0422 07:33:42.637204 138415253305152 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 9 I0422 07:33:42.639219 138415253305152 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_07_eval/checkpoints/9. I0422 07:33:42.928536 138415253305152 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0422 07:33:42.928633 138415253305152 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 07:33:42.961774 138415253305152 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.034407s I0422 07:33:42.961946 138415253305152 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 40.105 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03846335411071777 s) (per-host) I0422 07:33:42.961992 138415253305152 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.038520s (batch_requests_ready=0.002416s, total_serialization_initiated=0.036039s, others=0.000065s) I0422 07:33:42.962080 138415253305152 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.042571s (all_items=0.000016s, per_item={'items': '0.00001597'}, temp_paths=0.042555) I0422 07:33:42.962774 138415253305152 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_07_eval/checkpoints/9. I0422 07:33:42.963057 138280670586624 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-22 07:53:42.963025 I0422 07:33:42.968592 138415253305152 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 07:33:42.968898 138261624248064 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 07:33:42.969068 138415253305152 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260422_071422/linen_xpk_main_20260422_071422_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776843202.631156, 'wait_for_prev_duration_secs': 20.004197359085083, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776843222.6372442, 'checkpointer_blocking_duration_secs': 0.32596254348754883, 'get_old_steps_start_time': 1776843222.9632292, 'get_old_steps_duration_secs': 3.0994415283203125e-05, 'checkpoint_manager_blocking_start_time': 1776843202.6291914, 'checkpoint_manager_blocking_duration_secs': 20.33984112739563} I0422 07:33:42.969182 138415253305152 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0422 07:33:42.969228 138415253305152 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 07:33:48.541206 138283354941184 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_07_eval/checkpoints/9/items/array_metadatas/process_6 I0422 07:34:25.755522 138280670586624 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 36.879 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.83200120925903 s) (per-host) I0422 07:34:25.755644 138280670586624 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.792472s. I0422 07:34:32.672493 138280670586624 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 49.709307s. I0422 07:34:32.672698 138261624248064 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 07:34:32.672762 138261624248064 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 07:34:32.672810 138261624248064 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 07:34:32.674293 138261624248064 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 07:34:32.674472 138415253305152 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 07:34:32.674618 138415253305152 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 07:34:32.675671 138415253305152 metric_logger.py:196] completed step: 9, seconds: 0.148, TFLOP/s/device: 91.809, Tokens/s/device: 13838.492, total_weights: 65536, loss: 8.179, lm_loss: 8.179, perplexity: 3564.635 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Wed Apr 22 07:34:45 UTC 2026 EXIT_CODE=0
XPK Start: Wed Apr 22 16:07:01 UTC 2026 2026-04-22 16:07:05.247142: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered WARNING: All log messages before absl::InitializeLog() is called are written to STDERR E0000 00:00:1776874025.259382 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776874025.263222 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776874025.274777 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776874025.274797 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776874025.274800 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776874025.274802 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-22 16:07:24.467765: 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 16:07:24.995177 140452400543552 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-22 16:07:34,036:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0422 16:07:34.036431 140452400543552 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-22 16:07:34,038:jax._src.distributed:157: Connecting to JAX distributed service on mt-07-eval-jhbrp-slice-job-0-0.mt-07-eval-jhbrp:8482 I0422 16:07:34.038734 140452400543552 distributed.py:157] Connecting to JAX distributed service on mt-07-eval-jhbrp-slice-job-0-0.mt-07-eval-jhbrp:8482 I0422 16:07:35.659046 140452400543552 max_utils.py:284] Jax distributed system initialized! I0422 16:07:42.843963 140452400543552 max_utils.py:800] System Information: Jax Version: 0.8.1 I0422 16:07:42.844071 140452400543552 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0422 16:07:42.844133 140452400543552 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709 I0422 16:07:42.844171 140452400543552 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 16:07:43.470810 140452400543552 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 16:07:43.471410 140452400543552 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 16:07:43.471595 140452400543552 checkpointing.py:688] Setting up checkpoint logger... I0422 16:07:43.471645 140452400543552 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 16:07:43.471687 140452400543552 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0422 16:07:43.472038 140452400543552 base_pytree_checkpoint_handler.py:415] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7fbd1e179070>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 16:07:46.518899 140452400543552 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0422 16:07:46.519263 140452400543552 checkpoint_manager.py:709] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7fa85c485cd0>}, handler_registry=None I0422 16:07:46.519497 140452400543552 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7fa85c485cd0>` 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 16:07:46.519551 140452400543552 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fbbdafc29c0>` 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 16:07:46.519587 140452400543552 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 0x7fa85c485cd0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7fa85c485cd0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fbbdafc29c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fbbdafc29c0>}). I0422 16:07:46.520092 140452400543552 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0422 16:07:46.520187 140452400543552 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7fa854708180> timeout: 600 secs and primary_host=0 for async checkpoint writes I0422 16:07:47.690013 140452400543552 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_07_eval/checkpoints I0422 16:07:48.162492 140452400543552 checkpoint_manager.py:929] [process=6][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_07_eval/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7fbbda313830> I0422 16:07:48.162676 140452400543552 checkpointing.py:302] Checkpoint manager created! I0422 16:07:49.062664 140452400543552 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 16:07:49.062784 140452400543552 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 16:07:49.431341 140452400543552 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 16:07:49.431434 140452400543552 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 16:07:49.447468 140452400543552 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 16:07:49.447530 140452400543552 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 16:07:49.470443 140452400543552 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 16:07:49.470513 140452400543552 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 16:07:49.486447 140452400543552 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 16:07:49.486511 140452400543552 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 16:07:49.502452 140452400543552 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 16:07:49.502511 140452400543552 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 16:07:49.527558 140452400543552 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 16:07:49.527634 140452400543552 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 16:07:49.547314 140452400543552 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 16:07:49.547381 140452400543552 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 16:07:49.741545 140452400543552 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0422 16:07:49.741659 140452400543552 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 16:07:51.185690 140452400543552 maxtext_utils.py:1740] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0422 16:07:51.185833 140452400543552 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 16:07:51.185890 140452400543552 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 16:07:51.185948 140452400543552 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 16:07:51.186001 140452400543552 maxtext_utils.py:1740] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0422 16:07:51.186040 140452400543552 maxtext_utils.py:1740] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0422 16:07:51.186105 140452400543552 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 16:07:51.186192 140452400543552 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: PartitionSpec('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0422 16:07:51.186255 140452400543552 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0422 16:07:51.186308 140452400543552 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 16:07:51.186362 140452400543552 maxtext_utils.py:1740] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0422 16:07:51.186412 140452400543552 maxtext_utils.py:1740] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0422 16:07:51.687599 140452400543552 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 16:07:51.687691 140452400543552 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 16:07:51.702834 140452400543552 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 16:07:51.702893 140452400543552 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 16:08:02.505564 140452400543552 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 16:08:02.506516 140452400543552 metric_logger.py:289] number parameters: 1.104 billion I0422 16:08:13.813291 140452400543552 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0422 16:08:13.966794 140452400543552 checkpointing.py:798] Waited 0.15348529815673828 seconds for step 0 to finish before starting checkpointing. I0422 16:08:13.969292 140452400543552 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 16:08:13.970961 140452400543552 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 0 I0422 16:08:13.973808 140452400543552 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_07_eval/checkpoints/0. I0422 16:08:14.954713 140452400543552 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 16:08:14.955760 140452400543552 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0422 16:08:14.955821 140452400543552 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 16:08:15.211047 140452400543552 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.256418s I0422 16:08:15.211240 140452400543552 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.996 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5149004459381104 s) (per-host) I0422 16:08:15.211292 140452400543552 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.514963s (batch_requests_ready=0.254862s, total_serialization_initiated=0.260031s, others=0.000071s) I0422 16:08:15.211381 140452400543552 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.518903s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=0.518885) I0422 16:08:15.212417 140329172846336 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0422 16:08:15.212563 140452400543552 async_checkpointer.py:561] Finished blocking save. Time taken: 1.241529s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_07_eval/checkpoints/0. I0422 16:08:15.455322 140452400543552 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 16:08:15.455691 140324877879040 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 16:08:15.455856 140452400543552 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776874093.9692733, 'wait_for_prev_duration_secs': 6.461143493652344e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776874093.9710011, 'checkpointer_blocking_duration_secs': 1.2417562007904053, 'get_old_steps_start_time': 1776874095.2127771, 'get_old_steps_duration_secs': 2.5987625122070312e-05, 'checkpoint_manager_blocking_start_time': 1776874093.967305, 'checkpoint_manager_blocking_duration_secs': 1.4885101318359375} I0422 16:08:15.455966 140452400543552 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0422 16:08:15.456017 140452400543552 max_utils.py:750] Memstats: After params initialized: I0422 16:08:15.456074 140452400543552 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0422 16:08:15.456125 140452400543552 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0422 16:08:15.456154 140452400543552 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0422 16:08:15.456179 140452400543552 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0422 16:08:15.721937 140452400543552 metric_logger.py:185] completed step: 0, seconds: 11.307, TFLOP/s/device: 1.202, Tokens/s/device: 181.132, total_weights: 65536, loss: 10.874 I0422 16:08:15.894828 140452400543552 metric_logger.py:185] completed step: 1, seconds: 1.907, TFLOP/s/device: 7.124, Tokens/s/device: 1073.750, total_weights: 65536, loss: 10.874 I0422 16:08:16.308300 140452400543552 metric_logger.py:185] completed step: 2, seconds: 0.026, TFLOP/s/device: 530.809, Tokens/s/device: 80009.376, total_weights: 65536, loss: 10.263 I0422 16:08:16.455704 140452400543552 metric_logger.py:185] completed step: 3, seconds: 0.414, TFLOP/s/device: 32.815, Tokens/s/device: 4946.227, total_weights: 65536, loss: 9.734 I0422 16:08:16.750874 140452400543552 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.973, Tokens/s/device: 13410.953, total_weights: 65536, loss: 9.278 I0422 16:08:16.756676 140452400543552 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.256, Tokens/s/device: 13905.769, total_weights: 65536, loss: 8.892 I0422 16:08:35.698647 140452400543552 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.957, Tokens/s/device: 6927.157, total_weights: 65536, loss: 8.592 I0422 16:08:35.846031 140452400543552 metric_logger.py:185] completed step: 7, seconds: 18.795, TFLOP/s/device: 0.723, Tokens/s/device: 108.963, total_weights: 65536, loss: 8.384 I0422 16:08:35.993565 140452400543552 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.311, Tokens/s/device: 13461.905, total_weights: 65536, loss: 8.255 I0422 16:08:36.140595 140452400543552 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0422 16:08:36.141275 140452400543552 checkpointing.py:798] Waited 0.0007007122039794922 seconds for step 9 to finish before starting checkpointing. I0422 16:08:36.143232 140452400543552 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 16:08:36.829828 2604 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 16:08:38.813840 140329181239040 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_07_eval/checkpoints/0/items/array_metadatas/process_6 I0422 16:09:11.109998 140329172846336 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 28.000 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 56.413634300231934 s) (per-host) I0422 16:09:11.110118 140329172846336 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 55.897469s. I0422 16:09:18.908664 140329172846336 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 63.695998s. I0422 16:09:18.909000 140324877879040 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 16:09:18.909153 140324877879040 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 16:09:18.909207 140324877879040 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 16:09:18.913368 140324877879040 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 16:09:18.913553 140452400543552 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0422 16:09:18.913703 140452400543552 checkpoint_manager.py:1458] Waiting for previous save to complete took 42.770475 seconds. If this number is high, consider checkpointing less frequently. I0422 16:09:18.915404 140452400543552 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 9 I0422 16:09:18.918737 140452400543552 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_07_eval/checkpoints/9. I0422 16:09:19.943794 140452400543552 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0422 16:09:19.943895 140452400543552 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 16:09:19.994657 140452400543552 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.051955s I0422 16:09:19.994811 140452400543552 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.966 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.31064915657043457 s) (per-host) I0422 16:09:19.994863 140452400543552 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.310710s (batch_requests_ready=0.256836s, total_serialization_initiated=0.053805s, others=0.000070s) I0422 16:09:19.994947 140452400543552 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.314637s (all_items=0.000015s, per_item={'items': '0.00001526'}, temp_paths=0.314622) I0422 16:09:19.995841 140325951620864 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0422 16:09:19.995970 140452400543552 async_checkpointer.py:561] Finished blocking save. Time taken: 1.080493s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_07_eval/checkpoints/9. I0422 16:09:20.037994 140452400543552 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0422 16:09:20.038312 140324877879040 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 16:09:20.038468 140452400543552 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_07_eval/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776874116.1431963, 'wait_for_prev_duration_secs': 42.770474910736084, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776874158.915444, 'checkpointer_blocking_duration_secs': 1.0807197093963623, 'get_old_steps_start_time': 1776874159.9961808, 'get_old_steps_duration_secs': 2.7894973754882812e-05, 'checkpoint_manager_blocking_start_time': 1776874116.1415007, 'checkpoint_manager_blocking_duration_secs': 43.896931171417236} I0422 16:09:20.038577 140452400543552 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0422 16:09:20.038622 140452400543552 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 16:09:26.284010 140329181239040 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_07_eval/checkpoints/9/items/array_metadatas/process_6 I0422 16:10:02.220300 140325951620864 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 37.135 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.53609752655029 s) (per-host) I0422 16:10:02.220431 140325951620864 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.224382s. I0422 16:10:18.428611 140325951620864 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 58.432548s. I0422 16:10:18.428890 140324877879040 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 16:10:18.429017 140324877879040 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 16:10:18.429064 140324877879040 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0422 16:10:18.430692 140324877879040 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0422 16:10:18.430945 140452400543552 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0422 16:10:18.431089 140452400543552 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 16:10:18.431821 140452400543552 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.198, Tokens/s/device: 13897.088, total_weights: 65536, loss: 8.179 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Wed Apr 22 16:10:31 UTC 2026 EXIT_CODE=0