feat/nnx-set-defaults-true| Metric | Linen 574ad3fb9 | NNX 574ad3fb9 | Diff (NNX − Linen) |
|---|---|---|---|
| Parameters | 1.105 billion | 1.104 billion | — |
| Final loss | 6.1670 | 10.5490 | +4.382 |
| TFLOP/s | 85.814 | 87.557 | +1.743 |
| Tok/s | 12934.8 | 13197.5 | +262.728 |
| Avg s/step | 3.033 | 2.859 | -0.174 |
| Memory % | 1.44 | 1.44 | 0 |
| JAX | 0.8.1 | 0.8.1 | — |
Diff = NNX value − Linen value. Green = NNX improved. Red = NNX regressed.
XPK Start: Sun Apr 19 03:28:41 UTC 2026 2026-04-19 03:28:45.491696: 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 00:00:1776569325.504082 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776569325.507939 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776569325.519163 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776569325.519182 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776569325.519184 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776569325.519186 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-19 03:29:04.513340: 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) I0419 03:29:05.038861 132804013868864 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-19 03:29:14,078:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0419 03:29:14.078495 132804013868864 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-19 03:29:14,080:jax._src.distributed:157: Connecting to JAX distributed service on mt-05-fp8-qr5hf-slice-job-0-0.mt-05-fp8-qr5hf:8482 I0419 03:29:14.080758 132804013868864 distributed.py:157] Connecting to JAX distributed service on mt-05-fp8-qr5hf-slice-job-0-0.mt-05-fp8-qr5hf:8482 I0419 03:29:16.021970 132804013868864 max_utils.py:284] Jax distributed system initialized! I0419 03:29:21.800526 132804013868864 max_utils.py:800] System Information: Jax Version: 0.8.1 I0419 03:29:21.800637 132804013868864 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0419 03:29:21.800689 132804013868864 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 I0419 03:29:21.800724 132804013868864 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0419 03:29:22.436388 132804013868864 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 03:29:22.436983 132804013868864 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 03:29:22.437167 132804013868864 checkpointing.py:688] Setting up checkpoint logger... I0419 03:29:22.437218 132804013868864 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0419 03:29:22.437273 132804013868864 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0419 03:29:22.437626 132804013868864 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 0x78c851375190>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0419 03:29:27.662836 132804013868864 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0419 03:29:27.663180 132804013868864 checkpoint_manager.py:709] [process=3][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78b5fe32d910>}, handler_registry=None I0419 03:29:27.663426 132804013868864 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78b5fe32d910>` 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`. I0419 03:29:27.663476 132804013868864 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78b5fe333470>` 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`. I0419 03:29:27.663511 132804013868864 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 0x78b5fe32d910>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x78b5fe32d910>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78b5fe333470>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x78b5fe333470>}). I0419 03:29:27.664010 132804013868864 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0419 03:29:27.664089 132804013868864 async_checkpointer.py:177] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x78b5fc4ec180> timeout: 600 secs and primary_host=0 for async checkpoint writes I0419 03:29:28.891437 132804013868864 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints I0419 03:29:29.315036 132804013868864 checkpoint_manager.py:929] [process=3][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=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_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x78b5ff348fe0> I0419 03:29:29.315209 132804013868864 checkpointing.py:302] Checkpoint manager created! I0419 03:29:30.400036 132804013868864 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 03:29:30.400145 132804013868864 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 03:29:30.774746 132804013868864 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 03:29:30.774847 132804013868864 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 03:29:30.790740 132804013868864 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 03:29:30.790801 132804013868864 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 03:29:30.850616 132804013868864 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 03:29:30.850705 132804013868864 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 03:29:30.866776 132804013868864 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 03:29:30.866838 132804013868864 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 03:29:30.882643 132804013868864 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 03:29:30.882706 132804013868864 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 03:29:30.906581 132804013868864 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0419 03:29:30.906651 132804013868864 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 03:29:30.970188 132804013868864 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0419 03:29:30.970297 132804013868864 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0419 03:29:31.434576 132804013868864 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0419 03:29:31.434694 132804013868864 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0419 03:29:33.509917 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510045 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.510097 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510138 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.510175 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510209 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.510262 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510296 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.510327 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510357 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.510387 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510417 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.510449 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510479 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.510509 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510538 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.510569 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510636 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.510675 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510706 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.510735 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510763 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.510790 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510818 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.510851 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510878 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.510905 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510932 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.510959 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.510986 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.511013 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.511040 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.511067 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.511096 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.511122 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.511149 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.511175 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.511202 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.511228 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.511267 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.511294 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0419 03:29:33.511320 132804013868864 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0419 03:29:33.511371 132804013868864 maxtext_utils.py:1740] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0419 03:29:33.511426 132804013868864 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) I0419 03:29:33.511465 132804013868864 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) I0419 03:29:33.511514 132804013868864 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') I0419 03:29:33.511560 132804013868864 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) I0419 03:29:33.511594 132804013868864 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) I0419 03:29:33.511643 132804013868864 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) I0419 03:29:33.511691 132804013868864 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') I0419 03:29:33.511727 132804013868864 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) I0419 03:29:33.511761 132804013868864 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) I0419 03:29:33.511807 132804013868864 maxtext_utils.py:1740] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0419 03:29:33.511856 132804013868864 maxtext_utils.py:1740] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0419 03:29:35.302157 132804013868864 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 03:29:35.302274 132804013868864 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0419 03:29:35.318097 132804013868864 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 03:29:35.318159 132804013868864 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0419 03:29:48.626638 132804013868864 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. I0419 03:29:48.627705 132804013868864 metric_logger.py:289] number parameters: 1.105 billion I0419 03:30:03.411557 132804013868864 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0419 03:30:03.578064 132804013868864 checkpointing.py:798] Waited 0.16648554801940918 seconds for step 0 to finish before starting checkpointing. I0419 03:30:03.580872 132804013868864 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 03:30:03.582833 132804013868864 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 0 I0419 03:30:03.585977 132804013868864 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0. I0419 03:30:05.878881 132804013868864 signaling_client.py:364] Using JaxDistributedSignalingClient I0419 03:30:05.880062 132804013868864 jax_array_handlers.py:358] Scheduling D2H of 81 prioritized jax.Array. I0419 03:30:05.880181 132804013868864 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 I0419 03:30:06.241383 132804013868864 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.362617s I0419 03:30:06.241558 132804013868864 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.034 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 1.492323637008667 s) (per-host) I0419 03:30:06.241608 132804013868864 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 1.492384s (batch_requests_ready=1.124292s, total_serialization_initiated=0.368023s, others=0.000069s) I0419 03:30:06.241712 132804013868864 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.496563s (all_items=0.000018s, per_item={'items': '0.00001836'}, temp_paths=1.496545) I0419 03:30:06.242690 132683652904704 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started. I0419 03:30:06.242864 132804013868864 async_checkpointer.py:561] Finished blocking save. Time taken: 2.659959s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0. I0419 03:30:06.244798 132804013868864 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0419 03:30:06.245103 132676707153664 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0419 03:30:06.245228 132804013868864 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776569403.5808542, 'wait_for_prev_duration_secs': 6.508827209472656e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776569403.582873, 'checkpointer_blocking_duration_secs': 2.660097122192383, 'get_old_steps_start_time': 1776569406.2429917, 'get_old_steps_duration_secs': 2.5033950805664062e-05, 'checkpoint_manager_blocking_start_time': 1776569403.5788174, 'checkpoint_manager_blocking_duration_secs': 2.6663763523101807} I0419 03:30:06.245419 132804013868864 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0419 03:30:06.245483 132804013868864 max_utils.py:750] Memstats: After params initialized: I0419 03:30:06.245536 132804013868864 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_10(process=3,(2,2,0,0)) I0419 03:30:06.245569 132804013868864 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_11(process=3,(3,2,0,0)) I0419 03:30:06.245596 132804013868864 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_14(process=3,(2,3,0,0)) I0419 03:30:06.245621 132804013868864 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_15(process=3,(3,3,0,0)) I0419 03:30:06.517832 132804013868864 metric_logger.py:185] completed step: 0, seconds: 14.784, TFLOP/s/device: 0.919, Tokens/s/device: 138.531, total_weights: 65536, loss: 10.874 I0419 03:30:06.701108 132804013868864 metric_logger.py:185] completed step: 1, seconds: 3.104, TFLOP/s/device: 4.377, Tokens/s/device: 659.724, total_weights: 65536, loss: 10.874 I0419 03:30:07.100787 132804013868864 metric_logger.py:185] completed step: 2, seconds: 0.026, TFLOP/s/device: 524.114, Tokens/s/device: 79000.154, total_weights: 65536, loss: 10.816 I0419 03:30:07.259378 132804013868864 metric_logger.py:185] completed step: 3, seconds: 0.400, TFLOP/s/device: 33.940, Tokens/s/device: 5115.843, total_weights: 65536, loss: 10.431 I0419 03:30:07.576983 132804013868864 metric_logger.py:185] completed step: 4, seconds: 0.163, TFLOP/s/device: 83.103, Tokens/s/device: 12526.147, total_weights: 65536, loss: 9.992 I0419 03:30:07.583022 132804013868864 metric_logger.py:185] completed step: 5, seconds: 0.158, TFLOP/s/device: 85.789, Tokens/s/device: 12931.089, total_weights: 65536, loss: 9.549 I0419 03:30:29.348296 132804013868864 metric_logger.py:185] completed step: 6, seconds: 0.318, TFLOP/s/device: 42.729, Tokens/s/device: 6440.515, total_weights: 65536, loss: 9.111 I0419 03:30:29.506648 132804013868864 metric_logger.py:185] completed step: 7, seconds: 21.608, TFLOP/s/device: 0.629, Tokens/s/device: 94.780, total_weights: 65536, loss: 8.685 I0419 03:30:29.665006 132804013868864 metric_logger.py:185] completed step: 8, seconds: 0.163, TFLOP/s/device: 83.286, Tokens/s/device: 12553.712, total_weights: 65536, loss: 8.281 I0419 03:30:29.669931 132804013868864 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0419 03:30:29.982008 132804013868864 checkpointing.py:798] Waited 0.3120417594909668 seconds for step 10 to finish before starting checkpointing. I0419 03:30:29.984776 132804013868864 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 03:30:32.084456 2599 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0419 03:30:35.482623 132683661297408 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0/items/array_metadatas/process_3 I0419 03:30:47.159254 132683652904704 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 37.250 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.409966707229614 s) (per-host) I0419 03:30:47.159376 132683652904704 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.916577s. I0419 03:30:58.432429 132683652904704 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 52.189614s. I0419 03:30:58.432710 132676707153664 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 03:30:58.432831 132676707153664 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0419 03:30:58.432882 132676707153664 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0419 03:30:58.434521 132676707153664 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0419 03:30:58.434684 132804013868864 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0419 03:30:58.434809 132804013868864 checkpoint_manager.py:1458] Waiting for previous save to complete took 28.450038 seconds. If this number is high, consider checkpointing less frequently. I0419 03:30:58.436935 132804013868864 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 10 I0419 03:30:58.440712 132804013868864 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/10. I0419 03:30:59.408547 132804013868864 jax_array_handlers.py:358] Scheduling D2H of 81 prioritized jax.Array. I0419 03:30:59.408650 132804013868864 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 I0419 03:30:59.463340 132804013868864 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.056064s I0419 03:30:59.463507 132804013868864 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.016 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3075563907623291 s) (per-host) I0419 03:30:59.463561 132804013868864 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.307620s (batch_requests_ready=0.248044s, total_serialization_initiated=0.059506s, others=0.000070s) I0419 03:30:59.463663 132804013868864 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.311890s (all_items=0.000015s, per_item={'items': '0.00001454'}, temp_paths=0.311876) I0419 03:30:59.464555 132676116055808 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started. I0419 03:30:59.464726 132804013868864 async_checkpointer.py:561] Finished blocking save. Time taken: 1.027724s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/10. I0419 03:30:59.923879 132804013868864 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0419 03:30:59.924306 132676707153664 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0419 03:30:59.924482 132804013868864 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776569429.9847474, 'wait_for_prev_duration_secs': 28.45003843307495, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776569458.4369738, 'checkpointer_blocking_duration_secs': 1.0278599262237549, 'get_old_steps_start_time': 1776569459.4648533, 'get_old_steps_duration_secs': 2.5272369384765625e-05, 'checkpoint_manager_blocking_start_time': 1776569429.9829006, 'checkpoint_manager_blocking_duration_secs': 29.94154405593872} I0419 03:30:59.924599 132804013868864 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0419 03:30:59.925092 132804013868864 metric_logger.py:185] completed step: 9, seconds: 0.158, TFLOP/s/device: 85.842, Tokens/s/device: 12939.014, total_weights: 65536, loss: 7.908 I0419 03:30:59.932056 132804013868864 metric_logger.py:185] completed step: 10, seconds: 0.158, TFLOP/s/device: 85.842, Tokens/s/device: 12939.095, total_weights: 65536, loss: 7.568 I0419 03:31:00.090253 132804013868864 metric_logger.py:185] completed step: 11, seconds: 30.260, TFLOP/s/device: 0.449, Tokens/s/device: 67.679, total_weights: 65536, loss: 7.288 I0419 03:31:00.248556 132804013868864 metric_logger.py:185] completed step: 12, seconds: 0.006, TFLOP/s/device: 2365.859, Tokens/s/device: 356608.045, total_weights: 65536, loss: 7.032 I0419 03:31:00.406909 132804013868864 metric_logger.py:185] completed step: 13, seconds: 0.159, TFLOP/s/device: 85.253, Tokens/s/device: 12850.196, total_weights: 65536, loss: 6.815 I0419 03:31:00.565067 132804013868864 metric_logger.py:185] completed step: 14, seconds: 0.158, TFLOP/s/device: 85.807, Tokens/s/device: 12933.702, total_weights: 65536, loss: 6.635 I0419 03:31:00.723374 132804013868864 metric_logger.py:185] completed step: 15, seconds: 0.158, TFLOP/s/device: 85.891, Tokens/s/device: 12946.457, total_weights: 65536, loss: 6.492 I0419 03:31:00.881525 132804013868864 metric_logger.py:185] completed step: 16, seconds: 0.158, TFLOP/s/device: 85.864, Tokens/s/device: 12942.366, total_weights: 65536, loss: 6.380 I0419 03:31:01.039776 132804013868864 metric_logger.py:185] completed step: 17, seconds: 0.158, TFLOP/s/device: 85.791, Tokens/s/device: 12931.334, total_weights: 65536, loss: 6.293 I0419 03:31:01.197989 132804013868864 metric_logger.py:185] completed step: 18, seconds: 0.158, TFLOP/s/device: 85.959, Tokens/s/device: 12956.613, total_weights: 65536, loss: 6.223 I0419 03:31:01.355904 132804013868864 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0419 03:31:01.356901 132804013868864 checkpointing.py:798] Waited 0.0010094642639160156 seconds for step 19 to finish before starting checkpointing. I0419 03:31:01.359063 132804013868864 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 03:31:06.333789 132677202061056 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/10/items/array_metadatas/process_3 I0419 03:31:43.614849 132676116055808 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 35.533 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 44.458847999572754 s) (per-host) I0419 03:31:43.614967 132676116055808 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 44.150301s. I0419 03:31:51.296245 132676116055808 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 51.831549s. I0419 03:31:51.296547 132676707153664 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 03:31:51.296679 132676707153664 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0419 03:31:51.296740 132676707153664 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0419 03:31:51.298746 132676707153664 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0419 03:31:51.298959 132804013868864 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0419 03:31:51.299083 132804013868864 checkpoint_manager.py:1458] Waiting for previous save to complete took 49.940031 seconds. If this number is high, consider checkpointing less frequently. I0419 03:31:51.300709 132804013868864 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 19 I0419 03:31:51.304231 132804013868864 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/19. I0419 03:31:52.717336 132804013868864 jax_array_handlers.py:358] Scheduling D2H of 81 prioritized jax.Array. I0419 03:31:52.717440 132804013868864 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 I0419 03:31:52.774052 132804013868864 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.057766s I0419 03:31:52.774222 132804013868864 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.965 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3107290267944336 s) (per-host) I0419 03:31:52.774287 132804013868864 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.310805s (batch_requests_ready=0.249518s, total_serialization_initiated=0.061206s, others=0.000081s) I0419 03:31:52.774392 132804013868864 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.315069s (all_items=0.000011s, per_item={'items': '0.00001121'}, temp_paths=0.315058) I0419 03:31:52.775279 132677202061056 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started. I0419 03:31:52.775436 132804013868864 async_checkpointer.py:561] Finished blocking save. Time taken: 1.474657s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/19. I0419 03:31:53.226847 132804013868864 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0419 03:31:53.227207 132676707153664 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0419 03:31:53.227391 132804013868864 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776569461.3590364, 'wait_for_prev_duration_secs': 49.940030574798584, 'time_between_consecutive_saves_sec': 2.92447566986084, 'checkpointer_blocking_start_time': 1776569511.3007493, 'checkpointer_blocking_duration_secs': 1.4747951030731201, 'get_old_steps_start_time': 1776569512.7755585, 'get_old_steps_duration_secs': 2.1219253540039062e-05, 'checkpoint_manager_blocking_start_time': 1776569461.3571763, 'checkpoint_manager_blocking_duration_secs': 51.87018179893494} I0419 03:31:53.227525 132804013868864 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0419 03:31:53.227568 132804013868864 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 03:31:59.536285 132675088148224 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/linen_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/19/items/array_metadatas/process_3 I0419 03:32:36.018118 132677202061056 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 36.271 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.55458426475525 s) (per-host) I0419 03:32:36.018250 132677202061056 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.242849s. I0419 03:32:44.229851 132677202061056 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 51.454446s. I0419 03:32:44.230164 132676707153664 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 03:32:44.230300 132676707153664 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0419 03:32:44.230342 132676707153664 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0419 03:32:44.231553 132676707153664 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0419 03:32:44.231738 132804013868864 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0419 03:32:44.231889 132804013868864 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 03:32:44.232497 132804013868864 metric_logger.py:185] completed step: 19, seconds: 0.158, TFLOP/s/device: 85.814, Tokens/s/device: 12934.764, total_weights: 65536, loss: 6.167 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sun Apr 19 03:32:56 UTC 2026 EXIT_CODE=0
XPK Start: Sun Apr 19 05:06:58 UTC 2026 2026-04-19 05:07:02.526011: 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 00:00:1776575222.538626 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776575222.542287 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776575222.553511 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776575222.553531 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776575222.553533 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776575222.553535 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-19 05:07:21.757933: 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) I0419 05:07:22.285897 138138457016128 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-19 05:07:31,327:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0419 05:07:31.327407 138138457016128 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-19 05:07:31,329:jax._src.distributed:157: Connecting to JAX distributed service on mt-05-fp8-ex07a-slice-job-0-0.mt-05-fp8-ex07a:8482 I0419 05:07:31.329629 138138457016128 distributed.py:157] Connecting to JAX distributed service on mt-05-fp8-ex07a-slice-job-0-0.mt-05-fp8-ex07a:8482 I0419 05:07:32.829673 138138457016128 max_utils.py:284] Jax distributed system initialized! I0419 05:07:39.748353 138138457016128 max_utils.py:800] System Information: Jax Version: 0.8.1 I0419 05:07:39.748463 138138457016128 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0419 05:07:39.748505 138138457016128 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 I0419 05:07:39.748540 138138457016128 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0419 05:07:40.358528 138138457016128 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 05:07:40.816933 138138457016128 checkpointing.py:688] Setting up checkpoint logger... I0419 05:07:40.817054 138138457016128 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0419 05:07:40.817098 138138457016128 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0419 05:07:40.817332 138138457016128 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 0x7da256d751f0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0419 05:07:43.682329 138138457016128 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0419 05:07:43.682673 138138457016128 checkpoint_manager.py:709] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d8d807df080>}, handler_registry=None I0419 05:07:43.682925 138138457016128 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d8d807df080>` 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`. I0419 05:07:43.682978 138138457016128 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d8d8069d880>` 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`. I0419 05:07:43.683016 138138457016128 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 0x7d8d807df080>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d8d807df080>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d8d8069d880>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d8d8069d880>}). I0419 05:07:43.683408 138138457016128 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0419 05:07:43.683488 138138457016128 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7d8d807a6200> timeout: 600 secs and primary_host=0 for async checkpoint writes I0419 05:07:45.294183 138138457016128 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints I0419 05:07:46.131776 138138457016128 checkpoint_manager.py:929] [process=0][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7d8d8069ea80> I0419 05:07:46.131989 138138457016128 checkpointing.py:302] Checkpoint manager created! I0419 05:07:47.462959 138138457016128 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0419 05:07:47.463088 138138457016128 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0419 05:07:49.710279 138138457016128 maxtext_utils.py:1749] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0419 05:07:49.710407 138138457016128 maxtext_utils.py:1749] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0419 05:07:49.710463 138138457016128 maxtext_utils.py:1749] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0419 05:07:49.710508 138138457016128 maxtext_utils.py:1749] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0419 05:07:49.710544 138138457016128 maxtext_utils.py:1749] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0419 05:07:49.710579 138138457016128 maxtext_utils.py:1749] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0419 05:07:49.710618 138138457016128 maxtext_utils.py:1749] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0419 05:07:49.710655 138138457016128 maxtext_utils.py:1749] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0419 05:07:49.710690 138138457016128 maxtext_utils.py:1749] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0419 05:07:49.710725 138138457016128 maxtext_utils.py:1749] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0419 05:07:49.710758 138138457016128 maxtext_utils.py:1749] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0419 05:07:49.710791 138138457016128 maxtext_utils.py:1749] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0419 05:07:50.122897 138138457016128 nnx_decoders.py:616] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 05:07:50.123053 138138457016128 nnx_decoders.py:616] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 05:07:50.128548 138138457016128 nnx_decoders.py:616] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 05:07:50.128644 138138457016128 nnx_decoders.py:616] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 05:07:50.286028 138138457016128 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 05:07:50.286163 138138457016128 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 05:07:50.446172 138138457016128 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 05:07:50.446308 138138457016128 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 05:07:50.648885 138138457016128 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 05:07:50.649028 138138457016128 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 05:07:50.805072 138138457016128 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 05:07:50.805247 138138457016128 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 05:07:50.959685 138138457016128 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 05:07:50.959836 138138457016128 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 05:07:51.124677 138138457016128 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0419 05:07:51.125108 138138457016128 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 05:07:51.205130 138138457016128 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0419 05:07:51.205257 138138457016128 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0419 05:08:04.876285 138138457016128 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. I0419 05:08:04.885488 138138457016128 metric_logger.py:289] number parameters: 1.104 billion I0419 05:08:18.692481 138138457016128 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0419 05:08:18.850985 138138457016128 checkpointing.py:798] Waited 0.15848755836486816 seconds for step 0 to finish before starting checkpointing. I0419 05:08:18.853617 138138457016128 checkpoint_manager.py:2013] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 05:08:18.855564 138138457016128 checkpoint_manager.py:1518] [process=0] Saving checkpoint at step 0 I0419 05:08:18.858386 138138457016128 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0. I0419 05:08:20.020441 138138457016128 signaling_client.py:364] Using JaxDistributedSignalingClient I0419 05:08:20.277944 138011408832256 atomicity.py:137] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0 I0419 05:08:20.300338 138138457016128 jax_array_handlers.py:358] Scheduling D2H of 153 prioritized jax.Array. I0419 05:08:20.300447 138138457016128 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 I0419 05:08:20.662925 138138457016128 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.366268s I0419 05:08:20.663711 138138457016128 base_pytree_checkpoint_handler.py:129] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.430 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.6348934173583984 s) (per-host) I0419 05:08:20.663772 138138457016128 base_pytree_checkpoint_handler.py:737] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.634971s (batch_requests_ready=0.253623s, total_serialization_initiated=0.380652s, others=0.000696s) I0419 05:08:20.663870 138138457016128 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.638792s (all_items=0.000021s, per_item={'items': '0.00002050'}, temp_paths=0.638772) I0419 05:08:20.664865 138010897139456 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0419 05:08:20.665026 138138457016128 async_checkpointer.py:561] Finished blocking save. Time taken: 1.809394s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0. I0419 05:08:20.724249 138138457016128 checkpoint_manager.py:1566] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0419 05:08:20.724587 138011417224960 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 05:08:20.724752 138138457016128 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776575298.8535995, 'wait_for_prev_duration_secs': 6.222724914550781e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776575298.855603, 'checkpointer_blocking_duration_secs': 1.8095307350158691, 'get_old_steps_start_time': 1776575300.6651552, 'get_old_steps_duration_secs': 2.5272369384765625e-05, 'checkpoint_manager_blocking_start_time': 1776575298.851499, 'checkpoint_manager_blocking_duration_secs': 1.8732123374938965} I0419 05:08:20.724882 138138457016128 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0419 05:08:20.724969 138138457016128 max_utils.py:750] Memstats: After params initialized: I0419 05:08:20.725023 138138457016128 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_0(process=0,(0,0,0,0)) I0419 05:08:20.725057 138138457016128 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_1(process=0,(1,0,0,0)) I0419 05:08:20.725086 138138457016128 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_4(process=0,(0,1,0,0)) I0419 05:08:20.725113 138138457016128 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_5(process=0,(1,1,0,0)) I0419 05:08:21.032645 138138457016128 metric_logger.py:185] completed step: 0, seconds: 13.453, TFLOP/s/device: 1.010, Tokens/s/device: 152.230, total_weights: 65536, loss: 10.872 I0419 05:08:21.035563 138138457016128 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/tensorboard/' I0419 05:08:21.469023 138138457016128 metric_logger.py:185] completed step: 1, seconds: 2.334, TFLOP/s/device: 5.822, Tokens/s/device: 877.554, total_weights: 65536, loss: 10.872 I0419 05:08:21.613243 138138457016128 metric_logger.py:185] completed step: 2, seconds: 0.438, TFLOP/s/device: 31.003, Tokens/s/device: 4673.089, total_weights: 65536, loss: 10.856 I0419 05:08:21.769853 138138457016128 metric_logger.py:185] completed step: 3, seconds: 0.017, TFLOP/s/device: 789.078, Tokens/s/device: 118938.382, total_weights: 65536, loss: 10.824 I0419 05:08:22.027452 138010913924864 atomicity.py:137] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0/items I0419 05:08:42.520747 138010879579904 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776575301767977715, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776575301767977715, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0/_CHECKPOINT_METADATA I0419 05:08:42.522595 138138457016128 metric_logger.py:185] completed step: 4, seconds: 0.144, TFLOP/s/device: 94.106, Tokens/s/device: 14184.692, total_weights: 65536, loss: 10.793 I0419 05:08:42.534327 138138457016128 metric_logger.py:185] completed step: 5, seconds: 0.159, TFLOP/s/device: 85.366, Tokens/s/device: 12867.312, total_weights: 65536, loss: 10.763 I0419 05:08:42.684535 138138457016128 metric_logger.py:185] completed step: 6, seconds: 20.748, TFLOP/s/device: 0.655, Tokens/s/device: 98.711, total_weights: 65536, loss: 10.733 I0419 05:08:42.839801 138138457016128 metric_logger.py:185] completed step: 7, seconds: 0.010, TFLOP/s/device: 1357.491, Tokens/s/device: 204615.846, total_weights: 65536, loss: 10.706 I0419 05:08:42.994938 138138457016128 metric_logger.py:185] completed step: 8, seconds: 0.151, TFLOP/s/device: 89.701, Tokens/s/device: 13520.651, total_weights: 65536, loss: 10.680 I0419 05:08:43.005646 138138457016128 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0419 05:08:43.305999 138138457016128 checkpointing.py:798] Waited 0.3003253936767578 seconds for step 10 to finish before starting checkpointing. I0419 05:08:43.308393 138138457016128 checkpoint_manager.py:2024] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 05:08:44.818455 2715 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0419 05:08:47.376611 138015183714048 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0/items/array_metadatas/process_0 I0419 05:09:05.395865 138015175321344 base_pytree_checkpoint_handler.py:1232] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 20.579664s (commit=18.971025s, array_metadata_write=1.608638s) I0419 05:09:05.397109 138010897139456 base_pytree_checkpoint_handler.py:129] [process=0] /jax/checkpoint/write/gbytes_per_sec: 34.821 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 45.36825728416443 s) (per-host) I0419 05:09:05.397233 138010897139456 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 44.732253s. I0419 05:09:06.039019 138010897139456 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776575301767977715, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0/_CHECKPOINT_METADATA I0419 05:09:06.798925 138010879579904 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776575301767977715, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0/_CHECKPOINT_METADATA I0419 05:09:07.160401 138010897139456 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000886s. I0419 05:09:07.401176 138010897139456 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0419 05:09:08.299761 138010897139456 base_pytree_checkpoint_handler.py:1356] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.134205s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0/items I0419 05:09:08.301595 138010897139456 atomicity.py:608] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0/items I0419 05:09:08.698788 138010897139456 atomicity.py:608] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0 I0419 05:09:09.980790 138010897139456 atomicity.py:790] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0`. I0419 05:09:09.981632 138010897139456 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 51.126003s. directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/0 I0419 05:09:09.983089 138010897139456 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 49.318109s. I0419 05:09:09.983306 138011417224960 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 05:09:09.983442 138011417224960 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 05:09:09.983506 138011417224960 checkpoint_manager.py:2133] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0419 05:09:09.985450 138011417224960 checkpoint_manager.py:2142] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0419 05:09:09.985660 138138457016128 checkpoint_manager.py:2036] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0419 05:09:09.985833 138138457016128 checkpoint_manager.py:1458] Waiting for previous save to complete took 26.677424 seconds. If this number is high, consider checkpointing less frequently. I0419 05:09:09.987742 138138457016128 checkpoint_manager.py:1518] [process=0] Saving checkpoint at step 10 I0419 05:09:09.991331 138138457016128 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/10. I0419 05:09:11.789844 138010897139456 atomicity.py:137] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/10 I0419 05:09:11.821889 138138457016128 jax_array_handlers.py:358] Scheduling D2H of 153 prioritized jax.Array. I0419 05:09:11.821995 138138457016128 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 I0419 05:09:11.862044 138138457016128 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.043706s I0419 05:09:11.862892 138138457016128 base_pytree_checkpoint_handler.py:129] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.096 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.7360095977783203 s) (per-host) I0419 05:09:11.862962 138138457016128 base_pytree_checkpoint_handler.py:737] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.736097s (batch_requests_ready=0.679316s, total_serialization_initiated=0.055990s, others=0.000791s) I0419 05:09:11.863052 138138457016128 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.739989s (all_items=0.000019s, per_item={'items': '0.00001907'}, temp_paths=0.739970) I0419 05:09:11.863958 138010746136320 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0419 05:09:11.864112 138138457016128 async_checkpointer.py:561] Finished blocking save. Time taken: 1.876290s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/10. I0419 05:09:11.905983 138138457016128 checkpoint_manager.py:1566] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0419 05:09:11.906301 138011417224960 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 05:09:11.906480 138138457016128 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776575323.3083599, 'wait_for_prev_duration_secs': 26.67742395401001, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776575349.9877892, 'checkpointer_blocking_duration_secs': 1.8764328956604004, 'get_old_steps_start_time': 1776575351.8642416, 'get_old_steps_duration_secs': 2.8848648071289062e-05, 'checkpoint_manager_blocking_start_time': 1776575323.306448, 'checkpoint_manager_blocking_duration_secs': 28.59999418258667} I0419 05:09:11.906600 138138457016128 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0419 05:09:11.907115 138138457016128 metric_logger.py:185] completed step: 9, seconds: 0.155, TFLOP/s/device: 87.475, Tokens/s/device: 13185.172, total_weights: 65536, loss: 10.656 I0419 05:09:11.920400 138138457016128 metric_logger.py:185] completed step: 10, seconds: 0.155, TFLOP/s/device: 87.379, Tokens/s/device: 13170.672, total_weights: 65536, loss: 10.636 I0419 05:09:12.068670 138138457016128 metric_logger.py:185] completed step: 11, seconds: 28.912, TFLOP/s/device: 0.470, Tokens/s/device: 70.836, total_weights: 65536, loss: 10.618 I0419 05:09:12.223874 138138457016128 metric_logger.py:185] completed step: 12, seconds: 0.012, TFLOP/s/device: 1116.078, Tokens/s/device: 168227.370, total_weights: 65536, loss: 10.602 I0419 05:09:12.379095 138138457016128 metric_logger.py:185] completed step: 13, seconds: 0.150, TFLOP/s/device: 90.488, Tokens/s/device: 13639.330, total_weights: 65536, loss: 10.588 I0419 05:09:12.534185 138138457016128 metric_logger.py:185] completed step: 14, seconds: 0.155, TFLOP/s/device: 87.857, Tokens/s/device: 13242.721, total_weights: 65536, loss: 10.577 I0419 05:09:12.689511 138138457016128 metric_logger.py:185] completed step: 15, seconds: 0.155, TFLOP/s/device: 87.469, Tokens/s/device: 13184.238, total_weights: 65536, loss: 10.568 I0419 05:09:12.844583 138138457016128 metric_logger.py:185] completed step: 16, seconds: 0.155, TFLOP/s/device: 87.825, Tokens/s/device: 13237.927, total_weights: 65536, loss: 10.562 I0419 05:09:12.999786 138138457016128 metric_logger.py:185] completed step: 17, seconds: 0.155, TFLOP/s/device: 87.405, Tokens/s/device: 13174.654, total_weights: 65536, loss: 10.556 I0419 05:09:13.154912 138138457016128 metric_logger.py:185] completed step: 18, seconds: 0.155, TFLOP/s/device: 87.633, Tokens/s/device: 13208.983, total_weights: 65536, loss: 10.552 I0419 05:09:13.312243 138138457016128 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0419 05:09:13.313487 138138457016128 checkpointing.py:798] Waited 0.0012524127960205078 seconds for step 19 to finish before starting checkpointing. I0419 05:09:13.315868 138138457016128 checkpoint_manager.py:2024] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 05:09:14.378499 138006589949696 atomicity.py:137] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/10/items I0419 05:09:18.822496 138010913924864 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/10/items/array_metadatas/process_0 I0419 05:09:35.748224 138010888746752 base_pytree_checkpoint_handler.py:1232] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 20.265801s (commit=19.229995s, array_metadata_write=1.035806s) I0419 05:09:35.749467 138010746136320 base_pytree_checkpoint_handler.py:129] [process=0] /jax/checkpoint/write/gbytes_per_sec: 64.160 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 24.622554540634155 s) (per-host) I0419 05:09:35.749571 138010746136320 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 23.885503s. I0419 05:09:37.536203 138010746136320 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000778s. I0419 05:09:37.811717 138010746136320 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0419 05:09:38.731520 138010746136320 base_pytree_checkpoint_handler.py:1356] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.197312s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/10/items I0419 05:09:38.733134 138010746136320 atomicity.py:608] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/10/items I0419 05:09:39.554167 138010746136320 atomicity.py:608] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/10 I0419 05:09:41.302123 138010746136320 atomicity.py:790] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/10`. I0419 05:09:41.302837 138010746136320 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 31.315033s. directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/10 I0419 05:09:41.304229 138010746136320 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 29.440160s. I0419 05:09:41.304450 138011417224960 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 05:09:41.304580 138011417224960 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 05:09:41.304646 138011417224960 checkpoint_manager.py:2133] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0419 05:09:41.306298 138011417224960 checkpoint_manager.py:2142] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0419 05:09:41.306464 138138457016128 checkpoint_manager.py:2036] [process=0][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0419 05:09:41.306581 138138457016128 checkpoint_manager.py:1458] Waiting for previous save to complete took 27.990727 seconds. If this number is high, consider checkpointing less frequently. I0419 05:09:41.307985 138138457016128 checkpoint_manager.py:1518] [process=0] Saving checkpoint at step 19 I0419 05:09:41.311418 138138457016128 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/19. I0419 05:09:42.253546 138010746136320 atomicity.py:137] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/19 I0419 05:09:42.283346 138138457016128 jax_array_handlers.py:358] Scheduling D2H of 153 prioritized jax.Array. I0419 05:09:42.283459 138138457016128 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 I0419 05:09:42.326997 138138457016128 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.047074s I0419 05:09:42.327771 138138457016128 base_pytree_checkpoint_handler.py:129] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.864 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3171663284301758 s) (per-host) I0419 05:09:42.327867 138138457016128 base_pytree_checkpoint_handler.py:737] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.317283s (batch_requests_ready=0.257223s, total_serialization_initiated=0.059348s, others=0.000712s) I0419 05:09:42.327967 138138457016128 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.321112s (all_items=0.000013s, per_item={'items': '0.00001264'}, temp_paths=0.321100) I0419 05:09:42.328872 138010888746752 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0419 05:09:42.329025 138138457016128 async_checkpointer.py:561] Finished blocking save. Time taken: 1.020970s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/19. I0419 05:09:42.768823 138138457016128 checkpoint_manager.py:1566] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0419 05:09:42.769189 138011417224960 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0419 05:09:42.769363 138138457016128 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776575353.315839, 'wait_for_prev_duration_secs': 27.990727424621582, 'time_between_consecutive_saves_sec': 3.330348253250122, 'checkpointer_blocking_start_time': 1776575381.3080244, 'checkpointer_blocking_duration_secs': 1.0211102962493896, 'get_old_steps_start_time': 1776575382.3291478, 'get_old_steps_duration_secs': 2.09808349609375e-05, 'checkpoint_manager_blocking_start_time': 1776575353.3138573, 'checkpoint_manager_blocking_duration_secs': 29.45547127723694} I0419 05:09:42.769477 138138457016128 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0419 05:09:42.769519 138138457016128 checkpoint_manager.py:2024] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 05:09:43.985666 138006589949696 atomicity.py:137] Creating tmp directory gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/19/items I0419 05:09:46.725344 138010913924864 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/19/items/array_metadatas/process_0 I0419 05:10:06.510847 138010905532160 base_pytree_checkpoint_handler.py:1232] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 21.424646s (commit=18.966727s, array_metadata_write=2.457918s) I0419 05:10:06.512101 138010888746752 base_pytree_checkpoint_handler.py:129] [process=0] /jax/checkpoint/write/gbytes_per_sec: 64.477 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 24.501468896865845 s) (per-host) I0419 05:10:06.512222 138010888746752 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 24.183239s. I0419 05:10:08.572389 138010888746752 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0419 05:10:09.166426 138010888746752 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000907s. I0419 05:10:09.640734 138010888746752 base_pytree_checkpoint_handler.py:1356] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 1.907093s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/19/items I0419 05:10:09.642518 138010888746752 atomicity.py:608] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/19/items I0419 05:10:10.034661 138010888746752 atomicity.py:608] Finalizing gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/19 I0419 05:10:11.750792 138010888746752 atomicity.py:790] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/19`. I0419 05:10:11.751580 138010888746752 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 30.443530s. directory=gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260419_025828/nnx_xpk_feat_nnx_set_defaults_true_20260419_025828_05_fp8/checkpoints/19 I0419 05:10:11.752950 138010888746752 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 29.423965s. I0419 05:10:11.753150 138011417224960 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 05:10:11.753265 138011417224960 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0419 05:10:11.753319 138011417224960 checkpoint_manager.py:2133] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0419 05:10:11.754985 138011417224960 checkpoint_manager.py:2142] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0419 05:10:11.755168 138138457016128 checkpoint_manager.py:2036] [process=0][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0419 05:10:11.755349 138138457016128 checkpoint_manager.py:2013] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 05:10:11.756028 138138457016128 metric_logger.py:185] completed step: 19, seconds: 0.155, TFLOP/s/device: 87.557, Tokens/s/device: 13197.492, total_weights: 65536, loss: 10.549 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sun Apr 19 05:10:24 UTC 2026 EXIT_CODE=0