main (59e0f1759) vs feat/Migrate-Decoder-And-Tests-to-NNX (4f4617640)| Metric | main 59e0f1759 | feat/Migrate-Decoder-And-Tests-to-NNX 4f4617640 | Diff (feat/Migrate-Decoder-And-Tests-to-NNX − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 8.1790 | 8.1790 | 0 |
| TFLOP/s | 92.251 | 92.220 | -0.031 |
| Tok/s | 13905.0 | 13900.4 | -4.624 |
| Avg s/step | 2.639 | 2.471 | -0.168 |
| 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: Sat Apr 25 07:17:14 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. 2026-04-25 07:17:39.733860: 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) I0425 07:17:39.937303 132585359980352 max_utils.py:273] Attempting to initialize the jax distributed system... I0425 07:17:48.978338 132585359980352 distributed.py:149] Starting JAX distributed service on [::]:8482 I0425 07:17:48.980713 132585359980352 distributed.py:172] Connecting to JAX distributed service on mt-02-synthetic-5gd3f-slice-job-0-0.mt-02-synthetic-5gd3f:8482 I0425 07:17:51.154454 132585359980352 max_utils.py:284] Jax distributed system initialized! I0425 07:17:57.271776 132585359980352 max_utils.py:800] System Information: Jax Version: 0.9.2 I0425 07:17:57.271882 132585359980352 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0425 07:17:57.271925 132585359980352 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 I0425 07:17:57.271961 132585359980352 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0425 07:17:57.964924 132585359980352 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1) I0425 07:17:57.965222 132585359980352 checkpointing.py:677] Setting up checkpoint logger... I0425 07:17:57.965283 132585359980352 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0425 07:17:57.965326 132585359980352 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0425 07:17:57.965666 132585359980352 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 0x789555d18d70>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0425 07:18:00.779971 132585359980352 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0425 07:18:00.780222 132585359980352 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 0x7881105538c0>}, handler_registry=None I0425 07:18:00.780462 132585359980352 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7881105538c0>` 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`. I0425 07:18:00.780511 132585359980352 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x788110557680>` 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`. I0425 07:18:00.780546 132585359980352 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 0x7881105538c0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7881105538c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x788110557680>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x788110557680>}). I0425 07:18:00.780868 132585359980352 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0425 07:18:00.780937 132585359980352 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7881102f1a80> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0425 07:18:01.897024 132585359980352 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_02_synthetic/checkpoints I0425 07:18:02.743060 132585359980352 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_20260425_071456/linen_xpk_main_20260425_071456_02_synthetic/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7881280af440> I0425 07:18:02.743262 132585359980352 checkpointing.py:301] Checkpoint manager created! I0425 07:18:03.680360 132585359980352 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 07:18:03.680480 132585359980352 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:18:04.064875 132585359980352 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn'). I0425 07:18:04.064978 132585359980352 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:18:04.081506 132585359980352 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch_attn', 'activation_length_attn', 'activation_embed_attn'). I0425 07:18:04.081569 132585359980352 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 07:18:04.105378 132585359980352 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 07:18:04.105452 132585359980352 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:18:04.122063 132585359980352 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 07:18:04.122138 132585359980352 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:18:04.138668 132585359980352 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_length_attn', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 07:18:04.138730 132585359980352 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:18:04.163480 132585359980352 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch_attn', 'activation_length_attn', 'activation_heads', 'activation_kv'). I0425 07:18:04.163554 132585359980352 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 07:18:04.184833 132585359980352 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0425 07:18:04.184900 132585359980352 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0425 07:18:04.395190 132585359980352 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0425 07:18:04.395302 132585359980352 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0425 07:18:05.831336 132585359980352 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0425 07:18:05.831466 132585359980352 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) I0425 07:18:05.831520 132585359980352 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) I0425 07:18:05.831578 132585359980352 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') I0425 07:18:05.831632 132585359980352 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) I0425 07:18:05.831670 132585359980352 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) I0425 07:18:05.831722 132585359980352 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) I0425 07:18:05.831773 132585359980352 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') I0425 07:18:05.831813 132585359980352 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) I0425 07:18:05.831849 132585359980352 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) I0425 07:18:05.831895 132585359980352 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0425 07:18:05.831943 132585359980352 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0425 07:18:06.324674 132585359980352 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 07:18:06.324772 132585359980352 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0425 07:18:06.340363 132585359980352 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 07:18:06.340423 132585359980352 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0425 07:18:17.394465 132585359980352 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. I0425 07:18:17.395348 132585359980352 metric_logger.py:301] number parameters: 1.104 billion I0425 07:18:29.301989 132585359980352 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0425 07:18:29.455207 132585359980352 checkpointing.py:776] Waited 0.1531982421875 seconds for step 0 to finish before starting checkpointing. I0425 07:18:29.457512 132585359980352 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 07:18:29.459188 132585359980352 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0425 07:18:29.460691 132585359980352 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_02_synthetic/checkpoints/0. I0425 07:18:29.791239 132585359980352 signaling_client.py:364] Using JaxDistributedSignalingClient I0425 07:18:29.792126 132585359980352 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 07:18:29.792184 132585359980352 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 I0425 07:18:30.070457 132585359980352 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.279293s I0425 07:18:30.070638 132585359980352 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.410 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28513145446777344 s) (per-host) I0425 07:18:30.070692 132585359980352 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.285194s (batch_requests_ready=0.002397s, total_serialization_initiated=0.282726s, others=0.000071s) I0425 07:18:30.070789 132585359980352 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.289593s (all_items=0.000017s, per_item={'items': '0.00001740'}, temp_paths=0.289575) I0425 07:18:30.071574 132585359980352 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_02_synthetic/checkpoints/0. I0425 07:18:30.071905 132457022441216 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 07:38:30.071868 I0425 07:18:30.077250 132585359980352 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0425 07:18:30.077560 132455431669504 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 07:18:30.077717 132585359980352 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777101509.4574933, 'wait_for_prev_duration_secs': 6.246566772460938e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777101509.4592268, 'checkpointer_blocking_duration_secs': 0.6128330230712891, 'get_old_steps_start_time': 1777101510.072085, 'get_old_steps_duration_secs': 4.0531158447265625e-05, 'checkpoint_manager_blocking_start_time': 1777101509.4556687, 'checkpoint_manager_blocking_duration_secs': 0.62200927734375} I0425 07:18:30.077824 132585359980352 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0425 07:18:30.077874 132585359980352 max_utils.py:750] Memstats: After params initialized: I0425 07:18:30.077927 132585359980352 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0425 07:18:30.077960 132585359980352 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0425 07:18:30.077986 132585359980352 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0425 07:18:30.078012 132585359980352 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0425 07:18:30.391735 132585359980352 metric_logger.py:196] completed step: 0, seconds: 11.907, TFLOP/s/device: 1.141, Tokens/s/device: 172.006, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0425 07:18:30.563669 132585359980352 metric_logger.py:196] completed step: 1, seconds: 1.088, TFLOP/s/device: 12.486, Tokens/s/device: 1881.995, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0425 07:18:30.977359 132585359980352 metric_logger.py:196] completed step: 2, seconds: 0.023, TFLOP/s/device: 583.439, Tokens/s/device: 87942.288, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0425 07:18:31.124693 132585359980352 metric_logger.py:196] completed step: 3, seconds: 0.415, TFLOP/s/device: 32.716, Tokens/s/device: 4931.339, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0425 07:18:31.419734 132585359980352 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.822, Tokens/s/device: 13388.246, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0425 07:18:31.425373 132585359980352 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.359, Tokens/s/device: 13921.366, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0425 07:18:32.962966 2785 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0425 07:18:34.973730 132455440062208 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_02_synthetic/checkpoints/0/items/array_metadatas/process_6 I0425 07:18:52.905250 132585359980352 metric_logger.py:196] completed step: 6, seconds: 0.295, TFLOP/s/device: 45.981, Tokens/s/device: 6930.743, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759 I0425 07:18:53.052491 132585359980352 metric_logger.py:196] completed step: 7, seconds: 21.333, TFLOP/s/device: 0.637, Tokens/s/device: 96.002, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0425 07:18:53.199927 132585359980352 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.268, Tokens/s/device: 13455.449, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0425 07:18:53.346583 132585359980352 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0425 07:18:53.347171 132585359980352 checkpointing.py:776] Waited 0.0006105899810791016 seconds for step 9 to finish before starting checkpointing. I0425 07:18:53.349421 132585359980352 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 07:19:06.949070 132457022441216 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 42.504 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 37.16352462768555 s) (per-host) I0425 07:19:06.949221 132457022441216 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.877203s. I0425 07:19:15.695349 132457022441216 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 45.623317s. I0425 07:19:15.695673 132455431669504 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 07:19:15.695803 132455431669504 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 07:19:15.695857 132455431669504 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0425 07:19:15.697361 132455431669504 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0425 07:19:15.697559 132585359980352 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. W0425 07:19:15.697701 132585359980352 checkpoint_manager.py:1452] Waiting for previous save to complete took 22.348280 seconds. If this number is high, consider checkpointing less frequently. I0425 07:19:15.699618 132585359980352 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 9 I0425 07:19:15.701623 132585359980352 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_02_synthetic/checkpoints/9. I0425 07:19:15.978228 132585359980352 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0425 07:19:15.978327 132585359980352 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 I0425 07:19:16.012946 132585359980352 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.035756s I0425 07:19:16.013134 132585359980352 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 39.160 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.03939199447631836 s) (per-host) I0425 07:19:16.013184 132585359980352 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039452s (batch_requests_ready=0.001932s, total_serialization_initiated=0.037451s, others=0.000069s) I0425 07:19:16.013283 132585359980352 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.043586s (all_items=0.000015s, per_item={'items': '0.00001526'}, temp_paths=0.043571) I0425 07:19:16.013988 132585359980352 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.31 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_02_synthetic/checkpoints/9. I0425 07:19:16.014299 132455431669504 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-25 07:39:16.014260 I0425 07:19:16.016323 132585359980352 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0425 07:19:16.016622 132432078558976 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 07:19:16.016759 132585359980352 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777101533.3493905, 'wait_for_prev_duration_secs': 22.34827995300293, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777101555.6996565, 'checkpointer_blocking_duration_secs': 0.3147892951965332, 'get_old_steps_start_time': 1777101556.0144696, 'get_old_steps_duration_secs': 3.0040740966796875e-05, 'checkpoint_manager_blocking_start_time': 1777101533.3473964, 'checkpoint_manager_blocking_duration_secs': 22.669329404830933} I0425 07:19:16.016923 132585359980352 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0425 07:19:16.016968 132585359980352 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 07:19:21.697988 132455440062208 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260425_071456/linen_xpk_main_20260425_071456_02_synthetic/checkpoints/9/items/array_metadatas/process_6 I0425 07:19:57.902441 132455431669504 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 37.673 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.92865991592407 s) (per-host) I0425 07:19:57.902578 132455431669504 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.888165s. I0425 07:20:06.055785 132455431669504 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.041358s. I0425 07:20:06.056069 132432078558976 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 07:20:06.056205 132432078558976 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 07:20:06.056255 132432078558976 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0425 07:20:06.057914 132432078558976 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0425 07:20:06.058083 132585359980352 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. I0425 07:20:06.058242 132585359980352 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 07:20:06.059201 132585359980352 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.251, Tokens/s/device: 13905.014, 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: Sat Apr 25 07:20:14 UTC 2026 EXIT_CODE=0
4f4617640
· feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433
· full log
XPK Start: Sat Apr 25 18:06:43 UTC 2026 2026-04-25 18:06:47.576529: 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 00:00:1777140407.589603 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1777140407.593319 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1777140407.604923 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777140407.604945 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777140407.604948 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777140407.604950 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-25 18:07:06.741380: 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) I0425 18:07:07.257920 136344791017280 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-25 18:07:16,298:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0425 18:07:16.298218 136344791017280 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-25 18:07:16,300:jax._src.distributed:157: Connecting to JAX distributed service on mt-02-synthetic-skz04-slice-job-0-0.mt-02-synthetic-skz04:8482 I0425 18:07:16.300481 136344791017280 distributed.py:157] Connecting to JAX distributed service on mt-02-synthetic-skz04-slice-job-0-0.mt-02-synthetic-skz04:8482 I0425 18:07:18.112329 136344791017280 max_utils.py:284] Jax distributed system initialized! I0425 18:07:24.953702 136344791017280 max_utils.py:800] System Information: Jax Version: 0.8.1 I0425 18:07:24.953808 136344791017280 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0425 18:07:24.953849 136344791017280 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 I0425 18:07:24.953884 136344791017280 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0425 18:07:25.568810 136344791017280 maxtext_utils.py:1517] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 18:07:25.569099 136344791017280 checkpointing.py:677] Setting up checkpoint logger... I0425 18:07:25.569153 136344791017280 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0425 18:07:25.569201 136344791017280 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0425 18:07:25.569541 136344791017280 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 0x7c00ba7710d0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0425 18:07:28.473160 136344791017280 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0425 18:07:28.473508 136344791017280 checkpoint_manager.py:709] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bee41e6e300>}, handler_registry=None I0425 18:07:28.473767 136344791017280 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bee41e6e300>` 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`. I0425 18:07:28.473823 136344791017280 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bee40d9c1d0>` 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`. I0425 18:07:28.473859 136344791017280 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 0x7bee41e6e300>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7bee41e6e300>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bee40d9c1d0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7bee40d9c1d0>}). I0425 18:07:28.474364 136344791017280 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0425 18:07:28.474444 136344791017280 async_checkpointer.py:177] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7bee40676840> timeout: 600 secs and primary_host=0 for async checkpoint writes I0425 18:07:29.941306 136344791017280 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_02_synthetic/checkpoints I0425 18:07:30.375767 136344791017280 checkpoint_manager.py:929] [process=5][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_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_02_synthetic/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7bee40d97c80> I0425 18:07:30.375940 136344791017280 checkpointing.py:301] Checkpoint manager created! I0425 18:07:31.264419 136344791017280 nnx_wrappers.py:442] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 18:07:31.264522 136344791017280 nnx_wrappers.py:442] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 18:07:31.631009 136344791017280 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 18:07:31.631102 136344791017280 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 18:07:31.647080 136344791017280 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 18:07:31.647143 136344791017280 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 18:07:31.670588 136344791017280 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 18:07:31.670671 136344791017280 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 18:07:31.686599 136344791017280 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 18:07:31.686672 136344791017280 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 18:07:31.702684 136344791017280 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 18:07:31.702744 136344791017280 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 18:07:31.726742 136344791017280 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0425 18:07:31.726810 136344791017280 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 18:07:31.746437 136344791017280 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0425 18:07:31.746504 136344791017280 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0425 18:07:31.935268 136344791017280 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0425 18:07:31.935376 136344791017280 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0425 18:07:33.373117 136344791017280 maxtext_utils.py:1620] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0425 18:07:33.373252 136344791017280 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0425 18:07:33.373307 136344791017280 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0425 18:07:33.373368 136344791017280 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0425 18:07:33.373422 136344791017280 maxtext_utils.py:1620] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0425 18:07:33.373463 136344791017280 maxtext_utils.py:1620] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0425 18:07:33.373516 136344791017280 maxtext_utils.py:1620] 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) I0425 18:07:33.373567 136344791017280 maxtext_utils.py:1620] 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') I0425 18:07:33.373607 136344791017280 maxtext_utils.py:1620] 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) I0425 18:07:33.373645 136344791017280 maxtext_utils.py:1620] 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) I0425 18:07:33.373732 136344791017280 maxtext_utils.py:1620] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0425 18:07:33.373788 136344791017280 maxtext_utils.py:1620] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0425 18:07:33.874881 136344791017280 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 18:07:33.874989 136344791017280 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0425 18:07:33.889925 136344791017280 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 18:07:33.889983 136344791017280 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0425 18:07:44.726111 136344791017280 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. I0425 18:07:44.727014 136344791017280 metric_logger.py:289] number parameters: 1.104 billion I0425 18:07:55.821769 136344791017280 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0425 18:07:55.974497 136344791017280 checkpointing.py:776] Waited 0.1527118682861328 seconds for step 0 to finish before starting checkpointing. I0425 18:07:55.977285 136344791017280 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 18:07:55.979009 136344791017280 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 0 I0425 18:07:55.981777 136344791017280 async_checkpointer.py:452] [process=5] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_02_synthetic/checkpoints/0. I0425 18:07:56.983342 136344791017280 signaling_client.py:364] Using JaxDistributedSignalingClient I0425 18:07:56.984354 136344791017280 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0425 18:07:56.984414 136344791017280 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 I0425 18:07:57.239962 136344791017280 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.256699s I0425 18:07:57.240128 136344791017280 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.000 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.5141503810882568 s) (per-host) I0425 18:07:57.240178 136344791017280 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.514210s (batch_requests_ready=0.253915s, total_serialization_initiated=0.260228s, others=0.000067s) I0425 18:07:57.240273 136344791017280 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.518443s (all_items=0.000021s, per_item={'items': '0.00002074'}, temp_paths=0.518422) I0425 18:07:57.241254 136221632214784 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started. I0425 18:07:57.241424 136344791017280 async_checkpointer.py:561] Finished blocking save. Time taken: 1.262343s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_02_synthetic/checkpoints/0. I0425 18:07:57.711913 136344791017280 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0425 18:07:57.712291 136217337247488 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0425 18:07:57.712456 136344791017280 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777140475.977267, 'wait_for_prev_duration_secs': 6.365776062011719e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777140475.9790478, 'checkpointer_blocking_duration_secs': 1.2624869346618652, 'get_old_steps_start_time': 1777140477.2415557, 'get_old_steps_duration_secs': 2.8133392333984375e-05, 'checkpoint_manager_blocking_start_time': 1777140475.9750881, 'checkpoint_manager_blocking_duration_secs': 1.7373247146606445} I0425 18:07:57.712563 136344791017280 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0425 18:07:57.712615 136344791017280 max_utils.py:750] Memstats: After params initialized: I0425 18:07:57.712684 136344791017280 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0)) I0425 18:07:57.712720 136344791017280 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0)) I0425 18:07:57.712748 136344791017280 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0)) I0425 18:07:57.712772 136344791017280 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0)) I0425 18:07:57.978628 136344791017280 metric_logger.py:185] completed step: 0, seconds: 11.095, TFLOP/s/device: 1.225, Tokens/s/device: 184.594, total_weights: 65536, loss: 10.874 I0425 18:07:58.163463 136344791017280 metric_logger.py:185] completed step: 1, seconds: 2.156, TFLOP/s/device: 6.303, Tokens/s/device: 950.089, total_weights: 65536, loss: 10.874 I0425 18:07:58.576117 136344791017280 metric_logger.py:185] completed step: 2, seconds: 0.038, TFLOP/s/device: 360.382, Tokens/s/device: 54320.726, total_weights: 65536, loss: 10.263 I0425 18:07:58.723513 136344791017280 metric_logger.py:185] completed step: 3, seconds: 0.413, TFLOP/s/device: 32.917, Tokens/s/device: 4961.553, total_weights: 65536, loss: 9.734 I0425 18:07:59.018904 136344791017280 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.663, Tokens/s/device: 13364.221, total_weights: 65536, loss: 9.278 I0425 18:07:59.024893 136344791017280 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.200, Tokens/s/device: 13897.371, total_weights: 65536, loss: 8.892 I0425 18:08:17.907284 136344791017280 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.947, Tokens/s/device: 6925.564, total_weights: 65536, loss: 8.592 I0425 18:08:18.054594 136344791017280 metric_logger.py:185] completed step: 7, seconds: 18.736, TFLOP/s/device: 0.725, Tokens/s/device: 109.310, total_weights: 65536, loss: 8.384 I0425 18:08:18.202149 136344791017280 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.151, Tokens/s/device: 13437.791, total_weights: 65536, loss: 8.255 I0425 18:08:18.349189 136344791017280 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0425 18:08:18.349865 136344791017280 checkpointing.py:776] Waited 0.0006988048553466797 seconds for step 9 to finish before starting checkpointing. I0425 18:08:18.351902 136344791017280 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 18:08:19.198728 2587 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0425 18:08:21.134267 136221640607488 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_02_synthetic/checkpoints/0/items/array_metadatas/process_5 I0425 18:08:52.750821 136221632214784 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 28.195 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 56.024800300598145 s) (per-host) I0425 18:08:52.750944 136221632214784 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 55.509577s. I0425 18:09:01.762427 136221632214784 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 64.521043s. I0425 18:09:01.762753 136217337247488 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 18:09:01.762887 136217337247488 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0425 18:09:01.762940 136217337247488 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0425 18:09:01.764400 136217337247488 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0425 18:09:01.764569 136344791017280 checkpoint_manager.py:2036] [process=5][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0425 18:09:01.764725 136344791017280 checkpoint_manager.py:1458] Waiting for previous save to complete took 43.412828 seconds. If this number is high, consider checkpointing less frequently. I0425 18:09:01.766928 136344791017280 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 9 I0425 18:09:01.770368 136344791017280 async_checkpointer.py:452] [process=5] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_02_synthetic/checkpoints/9. I0425 18:09:02.322569 136344791017280 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0425 18:09:02.322681 136344791017280 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 I0425 18:09:02.371320 136344791017280 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.049898s I0425 18:09:02.371486 136344791017280 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.965 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3106973171234131 s) (per-host) I0425 18:09:02.371537 136344791017280 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.310756s (batch_requests_ready=0.258954s, total_serialization_initiated=0.051735s, others=0.000067s) I0425 18:09:02.371627 136344791017280 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.314898s (all_items=0.000016s, per_item={'items': '0.00001621'}, temp_paths=0.314881) I0425 18:09:02.372575 136217899296512 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started. I0425 18:09:02.372713 136344791017280 async_checkpointer.py:561] Finished blocking save. Time taken: 0.605706s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_02_synthetic/checkpoints/9. I0425 18:09:02.388878 136344791017280 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0425 18:09:02.389140 136217337247488 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0425 18:09:02.389280 136344791017280 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_02_synthetic/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777140498.351874, 'wait_for_prev_duration_secs': 43.41282773017883, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777140541.7669685, 'checkpointer_blocking_duration_secs': 0.6059246063232422, 'get_old_steps_start_time': 1777140542.3729115, 'get_old_steps_duration_secs': 2.765655517578125e-05, 'checkpoint_manager_blocking_start_time': 1777140498.350094, 'checkpoint_manager_blocking_duration_secs': 44.03914976119995} I0425 18:09:02.389383 136344791017280 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0425 18:09:02.389427 136344791017280 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 18:09:07.953867 136221640607488 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260425_180433_02_synthetic/checkpoints/9/items/array_metadatas/process_5 I0425 18:09:45.335543 136217899296512 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 36.502 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.274707078933716 s) (per-host) I0425 18:09:45.335682 136217899296512 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.962870s. I0425 18:09:53.201330 136217899296512 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 50.828503s. I0425 18:09:53.201648 136217337247488 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 18:09:53.201787 136217337247488 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0425 18:09:53.201837 136217337247488 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0425 18:09:53.203213 136217337247488 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0425 18:09:53.203344 136344791017280 checkpoint_manager.py:2036] [process=5][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0425 18:09:53.203438 136344791017280 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 18:09:53.204068 136344791017280 metric_logger.py:185] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.220, Tokens/s/device: 13900.390, 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: Sat Apr 25 18:10:03 UTC 2026 EXIT_CODE=0