feat/nnx-set-defaults-true| Metric | Linen 73213e044 | NNX 73213e044 | Diff (NNX − Linen) |
|---|---|---|---|
| Parameters | 1.105 billion | 1.104 billion | — |
| Final loss | 6.1670 | 10.5490 | +4.382 |
| TFLOP/s | 85.813 | 87.676 | +1.863 |
| Tok/s | 12934.7 | 13215.5 | +280.864 |
| Avg s/step | 2.652 | 2.591 | -0.061 |
| 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: Sat Apr 25 15:13:54 UTC 2026 2026-04-25 15:13:58.420137: 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 00:00:1777130038.432894 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1777130038.436750 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1777130038.447826 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777130038.447845 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777130038.447848 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777130038.447850 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-25 15:14:17.746177: 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 15:14:18.269782 138956914530112 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-25 15:14:27,310:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0425 15:14:27.310603 138956914530112 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-25 15:14:27,313:jax._src.distributed:157: Connecting to JAX distributed service on mt-05-fp8-vyrqd-slice-job-0-0.mt-05-fp8-vyrqd:8482 I0425 15:14:27.313007 138956914530112 distributed.py:157] Connecting to JAX distributed service on mt-05-fp8-vyrqd-slice-job-0-0.mt-05-fp8-vyrqd:8482 I0425 15:14:28.433000 138956914530112 max_utils.py:284] Jax distributed system initialized! I0425 15:14:34.275407 138956914530112 max_utils.py:800] System Information: Jax Version: 0.8.1 I0425 15:14:34.275510 138956914530112 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0425 15:14:34.275550 138956914530112 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 15:14:34.275584 138956914530112 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0425 15:14:34.899965 138956914530112 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 15:14:34.900559 138956914530112 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 15:14:34.900761 138956914530112 checkpointing.py:688] Setting up checkpoint logger... I0425 15:14:34.900812 138956914530112 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0425 15:14:34.900855 138956914530112 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0425 15:14:34.901196 138956914530112 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 0x7e60e2378fe0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0425 15:14:37.809163 138956914530112 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0425 15:14:37.809504 138956914530112 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 0x7e4e6bd759d0>}, handler_registry=None I0425 15:14:37.809750 138956914530112 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e4e6bd759d0>` 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 15:14:37.809799 138956914530112 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e4e6bd7a870>` 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 15:14:37.809835 138956914530112 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 0x7e4e6bd759d0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e4e6bd759d0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e4e6bd7a870>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e4e6bd7a870>}). I0425 15:14:37.810336 138956914530112 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0425 15:14:37.810414 138956914530112 async_checkpointer.py:177] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7e4e6b4e0180> timeout: 600 secs and primary_host=0 for async checkpoint writes I0425 15:14:38.903812 138956914530112 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints I0425 15:14:38.914889 138956914530112 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_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e4e6bd7aae0> I0425 15:14:38.915027 138956914530112 checkpointing.py:302] Checkpoint manager created! I0425 15:14:40.022485 138956914530112 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 15:14:40.022592 138956914530112 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 15:14:40.415079 138956914530112 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 15:14:40.415183 138956914530112 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 15:14:40.431337 138956914530112 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 15:14:40.431397 138956914530112 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 15:14:40.491148 138956914530112 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 15:14:40.491235 138956914530112 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 15:14:40.507472 138956914530112 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 15:14:40.507534 138956914530112 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 15:14:40.523703 138956914530112 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 15:14:40.523760 138956914530112 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 15:14:40.553329 138956914530112 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0425 15:14:40.553432 138956914530112 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 15:14:40.614001 138956914530112 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0425 15:14:40.614092 138956914530112 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0425 15:14:41.081171 138956914530112 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0425 15:14:41.081287 138956914530112 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0425 15:14:43.135389 138956914530112 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: () I0425 15:14:43.135513 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.135562 138956914530112 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: () I0425 15:14:43.135602 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.135638 138956914530112 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: () I0425 15:14:43.135688 138956914530112 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: () I0425 15:14:43.135723 138956914530112 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: () I0425 15:14:43.135755 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.135785 138956914530112 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: () I0425 15:14:43.135815 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.135846 138956914530112 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: () I0425 15:14:43.135875 138956914530112 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: () I0425 15:14:43.135906 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0425 15:14:43.135936 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.135966 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0425 15:14:43.135994 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.136024 138956914530112 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: () I0425 15:14:43.136053 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.136081 138956914530112 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: () I0425 15:14:43.136108 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.136135 138956914530112 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: () I0425 15:14:43.136162 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.136189 138956914530112 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: () I0425 15:14:43.136216 138956914530112 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: () I0425 15:14:43.136243 138956914530112 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: () I0425 15:14:43.136277 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.136306 138956914530112 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: () I0425 15:14:43.136336 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.136363 138956914530112 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: () I0425 15:14:43.136390 138956914530112 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: () I0425 15:14:43.136416 138956914530112 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: () I0425 15:14:43.136443 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.136470 138956914530112 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: () I0425 15:14:43.136496 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.136523 138956914530112 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: () I0425 15:14:43.136549 138956914530112 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: () I0425 15:14:43.136575 138956914530112 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: () I0425 15:14:43.136603 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.136629 138956914530112 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: () I0425 15:14:43.136680 138956914530112 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0425 15:14:43.136711 138956914530112 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: () I0425 15:14:43.136739 138956914530112 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: () I0425 15:14:43.136790 138956914530112 maxtext_utils.py:1740] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0425 15:14:43.136845 138956914530112 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) I0425 15:14:43.136885 138956914530112 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) I0425 15:14:43.136934 138956914530112 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') I0425 15:14:43.136982 138956914530112 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) I0425 15:14:43.137019 138956914530112 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) I0425 15:14:43.137069 138956914530112 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) I0425 15:14:43.137119 138956914530112 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') I0425 15:14:43.137156 138956914530112 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) I0425 15:14:43.137192 138956914530112 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) I0425 15:14:43.137239 138956914530112 maxtext_utils.py:1740] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0425 15:14:43.137290 138956914530112 maxtext_utils.py:1740] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0425 15:14:44.926423 138956914530112 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 15:14:44.926519 138956914530112 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0425 15:14:44.942373 138956914530112 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0425 15:14:44.942436 138956914530112 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0425 15:14:58.355632 138956914530112 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 15:14:58.356684 138956914530112 metric_logger.py:289] number parameters: 1.105 billion I0425 15:15:12.846416 138956914530112 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0425 15:15:13.013220 138956914530112 checkpointing.py:798] Waited 0.16678428649902344 seconds for step 0 to finish before starting checkpointing. I0425 15:15:13.016000 138956914530112 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 15:15:13.017909 138956914530112 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 0 I0425 15:15:13.021117 138956914530112 async_checkpointer.py:452] [process=5] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/0. I0425 15:15:13.619620 138956914530112 signaling_client.py:364] Using JaxDistributedSignalingClient I0425 15:15:13.620719 138956914530112 jax_array_handlers.py:358] Scheduling D2H of 81 prioritized jax.Array. I0425 15:15:13.620840 138956914530112 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 15:15:13.984089 138956914530112 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.364585s I0425 15:15:13.984265 138956914530112 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.382 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.6476328372955322 s) (per-host) I0425 15:15:13.984316 138956914530112 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.647696s (batch_requests_ready=0.277736s, total_serialization_initiated=0.369888s, others=0.000072s) I0425 15:15:13.984425 138956914530112 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.651814s (all_items=0.000021s, per_item={'items': '0.00002050'}, temp_paths=0.651794) I0425 15:15:13.985357 138833609881344 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started. I0425 15:15:13.985484 138956914530112 async_checkpointer.py:561] Finished blocking save. Time taken: 0.967507s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/0. I0425 15:15:14.134589 138956914530112 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0425 15:15:14.134997 138831428826880 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0425 15:15:14.135161 138956914530112 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777130113.0159802, 'wait_for_prev_duration_secs': 6.747245788574219e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777130113.0179484, 'checkpointer_blocking_duration_secs': 0.967731237411499, 'get_old_steps_start_time': 1777130113.9856992, 'get_old_steps_duration_secs': 2.47955322265625e-05, 'checkpoint_manager_blocking_start_time': 1777130113.0140688, 'checkpoint_manager_blocking_duration_secs': 1.121051549911499} I0425 15:15:14.135287 138956914530112 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0425 15:15:14.135343 138956914530112 max_utils.py:750] Memstats: After params initialized: I0425 15:15:14.135402 138956914530112 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_18(process=5,(2,4,0,0)) I0425 15:15:14.135441 138956914530112 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_19(process=5,(3,4,0,0)) I0425 15:15:14.135468 138956914530112 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_22(process=5,(2,5,0,0)) I0425 15:15:14.135494 138956914530112 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_23(process=5,(3,5,0,0)) I0425 15:15:14.405003 138956914530112 metric_logger.py:185] completed step: 0, seconds: 14.490, TFLOP/s/device: 0.938, Tokens/s/device: 141.343, total_weights: 65536, loss: 10.874 I0425 15:15:14.588383 138956914530112 metric_logger.py:185] completed step: 1, seconds: 1.557, TFLOP/s/device: 8.728, Tokens/s/device: 1315.632, total_weights: 65536, loss: 10.874 I0425 15:15:15.010091 138956914530112 metric_logger.py:185] completed step: 2, seconds: 0.026, TFLOP/s/device: 523.589, Tokens/s/device: 78921.002, total_weights: 65536, loss: 10.816 I0425 15:15:15.168308 138956914530112 metric_logger.py:185] completed step: 3, seconds: 0.422, TFLOP/s/device: 32.172, Tokens/s/device: 4849.357, total_weights: 65536, loss: 10.431 I0425 15:15:15.485711 138956914530112 metric_logger.py:185] completed step: 4, seconds: 0.164, TFLOP/s/device: 83.078, Tokens/s/device: 12522.471, total_weights: 65536, loss: 9.992 I0425 15:15:15.491941 138956914530112 metric_logger.py:185] completed step: 5, seconds: 0.158, TFLOP/s/device: 85.984, Tokens/s/device: 12960.385, total_weights: 65536, loss: 9.549 I0425 15:15:35.568160 138956914530112 metric_logger.py:185] completed step: 6, seconds: 0.318, TFLOP/s/device: 42.730, Tokens/s/device: 6440.717, total_weights: 65536, loss: 9.111 I0425 15:15:35.726531 138956914530112 metric_logger.py:185] completed step: 7, seconds: 19.919, TFLOP/s/device: 0.682, Tokens/s/device: 102.815, total_weights: 65536, loss: 8.685 I0425 15:15:35.884703 138956914530112 metric_logger.py:185] completed step: 8, seconds: 0.163, TFLOP/s/device: 83.218, Tokens/s/device: 12543.486, total_weights: 65536, loss: 8.281 I0425 15:15:35.889688 138956914530112 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0425 15:15:36.201147 138956914530112 checkpointing.py:798] Waited 0.31142425537109375 seconds for step 10 to finish before starting checkpointing. I0425 15:15:36.203762 138956914530112 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 15:15:36.678382 2610 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0425 15:15:38.793344 138834096396032 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/0/items/array_metadatas/process_5 I0425 15:15:53.499108 138833609881344 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 39.334 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 40.162426710128784 s) (per-host) I0425 15:15:53.499228 138833609881344 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 39.513663s. I0425 15:16:01.078222 138833609881344 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 47.092641s. I0425 15:16:01.078533 138831428826880 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 15:16:01.078682 138831428826880 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0425 15:16:01.078734 138831428826880 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0425 15:16:01.080138 138831428826880 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0425 15:16:01.080313 138956914530112 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 15:16:01.080444 138956914530112 checkpoint_manager.py:1458] Waiting for previous save to complete took 24.876688 seconds. If this number is high, consider checkpointing less frequently. I0425 15:16:01.082438 138956914530112 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 10 I0425 15:16:01.085878 138956914530112 async_checkpointer.py:452] [process=5] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/10. I0425 15:16:01.613861 138956914530112 jax_array_handlers.py:358] Scheduling D2H of 81 prioritized jax.Array. I0425 15:16:01.613963 138956914530112 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 15:16:01.671529 138956914530112 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.058653s I0425 15:16:01.671711 138956914530112 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.942 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.31218814849853516 s) (per-host) I0425 15:16:01.671766 138956914530112 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.312254s (batch_requests_ready=0.250069s, total_serialization_initiated=0.062112s, others=0.000073s) I0425 15:16:01.671870 138956914530112 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.316143s (all_items=0.000015s, per_item={'items': '0.00001550'}, temp_paths=0.316127) I0425 15:16:01.672712 138831957305088 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started. I0425 15:16:01.672880 138956914530112 async_checkpointer.py:561] Finished blocking save. Time taken: 0.590370s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/10. I0425 15:16:02.111637 138956914530112 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0425 15:16:02.112107 138831428826880 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0425 15:16:02.112294 138956914530112 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777130136.2037332, 'wait_for_prev_duration_secs': 24.87668776512146, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777130161.0824773, 'checkpointer_blocking_duration_secs': 0.5905101299285889, 'get_old_steps_start_time': 1777130161.6730046, 'get_old_steps_duration_secs': 2.6464462280273438e-05, 'checkpoint_manager_blocking_start_time': 1777130136.2020557, 'checkpoint_manager_blocking_duration_secs': 25.910202026367188} I0425 15:16:02.112413 138956914530112 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0425 15:16:02.112870 138956914530112 metric_logger.py:185] completed step: 9, seconds: 0.158, TFLOP/s/device: 86.046, Tokens/s/device: 12969.741, total_weights: 65536, loss: 7.908 I0425 15:16:02.119641 138956914530112 metric_logger.py:185] completed step: 10, seconds: 0.158, TFLOP/s/device: 85.973, Tokens/s/device: 12958.827, total_weights: 65536, loss: 7.568 I0425 15:16:02.277332 138956914530112 metric_logger.py:185] completed step: 11, seconds: 26.228, TFLOP/s/device: 0.518, Tokens/s/device: 78.084, total_weights: 65536, loss: 7.288 I0425 15:16:02.435523 138956914530112 metric_logger.py:185] completed step: 12, seconds: 0.006, TFLOP/s/device: 2430.613, Tokens/s/device: 366368.515, total_weights: 65536, loss: 7.032 I0425 15:16:02.593778 138956914530112 metric_logger.py:185] completed step: 13, seconds: 0.159, TFLOP/s/device: 85.504, Tokens/s/device: 12888.123, total_weights: 65536, loss: 6.815 I0425 15:16:02.751893 138956914530112 metric_logger.py:185] completed step: 14, seconds: 0.158, TFLOP/s/device: 85.797, Tokens/s/device: 12932.232, total_weights: 65536, loss: 6.635 I0425 15:16:02.910072 138956914530112 metric_logger.py:185] completed step: 15, seconds: 0.158, TFLOP/s/device: 85.940, Tokens/s/device: 12953.745, total_weights: 65536, loss: 6.492 I0425 15:16:03.068120 138956914530112 metric_logger.py:185] completed step: 16, seconds: 0.158, TFLOP/s/device: 85.933, Tokens/s/device: 12952.762, total_weights: 65536, loss: 6.380 I0425 15:16:03.226390 138956914530112 metric_logger.py:185] completed step: 17, seconds: 0.159, TFLOP/s/device: 85.613, Tokens/s/device: 12904.527, total_weights: 65536, loss: 6.293 I0425 15:16:03.384541 138956914530112 metric_logger.py:185] completed step: 18, seconds: 0.158, TFLOP/s/device: 86.267, Tokens/s/device: 13003.092, total_weights: 65536, loss: 6.223 I0425 15:16:03.542484 138956914530112 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0425 15:16:03.543428 138956914530112 checkpointing.py:798] Waited 0.0009639263153076172 seconds for step 19 to finish before starting checkpointing. I0425 15:16:03.545514 138956914530112 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 15:16:07.360924 138831462397696 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/10/items/array_metadatas/process_5 I0425 15:16:43.697669 138831957305088 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 37.313 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.33808994293213 s) (per-host) I0425 15:16:43.697789 138831957305088 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.024959s. I0425 15:16:50.256399 138831957305088 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 48.583552s. I0425 15:16:50.256715 138831428826880 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 15:16:50.256829 138831428826880 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0425 15:16:50.256878 138831428826880 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0425 15:16:50.258393 138831428826880 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0425 15:16:50.258518 138956914530112 checkpoint_manager.py:2036] [process=5][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0425 15:16:50.258598 138956914530112 checkpoint_manager.py:1458] Waiting for previous save to complete took 46.713097 seconds. If this number is high, consider checkpointing less frequently. I0425 15:16:50.260156 138956914530112 checkpoint_manager.py:1518] [process=5] Saving checkpoint at step 19 I0425 15:16:50.263824 138956914530112 async_checkpointer.py:452] [process=5] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/19. I0425 15:16:51.203862 138956914530112 jax_array_handlers.py:358] Scheduling D2H of 81 prioritized jax.Array. I0425 15:16:51.203964 138956914530112 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 15:16:51.260159 138956914530112 base_pytree_checkpoint_handler.py:153] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.057338s I0425 15:16:51.260328 138956914530112 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.530 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.34053587913513184 s) (per-host) I0425 15:16:51.260378 138956914530112 base_pytree_checkpoint_handler.py:737] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.340597s (batch_requests_ready=0.279794s, total_serialization_initiated=0.060736s, others=0.000067s) I0425 15:16:51.260483 138956914530112 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.344701s (all_items=0.000010s, per_item={'items': '0.00001049'}, temp_paths=0.344691) I0425 15:16:51.261326 138831462397696 async_checkpointer.py:79] [process=5][thread=async_save] Background save thread started. I0425 15:16:51.261450 138956914530112 async_checkpointer.py:561] Finished blocking save. Time taken: 1.001229s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/19. I0425 15:16:51.379065 138956914530112 checkpoint_manager.py:1566] [process=5][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0425 15:16:51.379430 138831428826880 async_checkpointer.py:265] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0425 15:16:51.379598 138956914530112 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777130163.5454876, 'wait_for_prev_duration_secs': 46.71309733390808, 'time_between_consecutive_saves_sec': 2.4653124809265137, 'checkpointer_blocking_start_time': 1777130210.260195, 'checkpointer_blocking_duration_secs': 1.0014231204986572, 'get_old_steps_start_time': 1777130211.2616322, 'get_old_steps_duration_secs': 3.337860107421875e-05, 'checkpoint_manager_blocking_start_time': 1777130163.5437198, 'checkpoint_manager_blocking_duration_secs': 47.83584523200989} I0425 15:16:51.379721 138956914530112 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0425 15:16:51.379766 138956914530112 checkpoint_manager.py:2024] [process=5][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 15:16:55.998197 138831454004992 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/linen_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/19/items/array_metadatas/process_5 I0425 15:17:33.045205 138831462397696 base_pytree_checkpoint_handler.py:129] [process=5] /jax/checkpoint/write/gbytes_per_sec: 37.501 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.12537145614624 s) (per-host) I0425 15:17:33.045323 138831462397696 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.783793s. I0425 15:17:38.706441 138831462397696 async_checkpointer.py:144] [process=5][thread=async_save] Background save thread done. Time taken: 47.444896s. I0425 15:17:38.706742 138831428826880 async_checkpointer.py:273] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 15:17:38.706861 138831428826880 async_checkpointer.py:283] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0425 15:17:38.706908 138831428826880 checkpoint_manager.py:2133] [process=5][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0425 15:17:38.708360 138831428826880 checkpoint_manager.py:2142] [process=5][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0425 15:17:38.708524 138956914530112 checkpoint_manager.py:2036] [process=5][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0425 15:17:38.708674 138956914530112 checkpoint_manager.py:2013] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 15:17:38.709325 138956914530112 metric_logger.py:185] completed step: 19, seconds: 0.158, TFLOP/s/device: 85.813, Tokens/s/device: 12934.682, 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: Sat Apr 25 15:17:49 UTC 2026 EXIT_CODE=0
XPK Start: Sat Apr 25 16:43:51 UTC 2026 2026-04-25 16:43:55.410372: 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 00:00:1777135435.423267 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1777135435.427198 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1777135435.438876 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777135435.438900 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777135435.438902 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777135435.438904 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-25 16:44:14.803269: 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 16:44:15.333987 137469934905152 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-25 16:44:24,376:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0425 16:44:24.376873 137469934905152 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-25 16:44:24,379:jax._src.distributed:157: Connecting to JAX distributed service on mt-05-fp8-qjrkl-slice-job-0-0.mt-05-fp8-qjrkl:8482 I0425 16:44:24.379233 137469934905152 distributed.py:157] Connecting to JAX distributed service on mt-05-fp8-qjrkl-slice-job-0-0.mt-05-fp8-qjrkl:8482 I0425 16:44:25.184716 137469934905152 max_utils.py:284] Jax distributed system initialized! I0425 16:44:32.098411 137469934905152 max_utils.py:800] System Information: Jax Version: 0.8.1 I0425 16:44:32.098519 137469934905152 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0425 16:44:32.098564 137469934905152 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 16:44:32.098598 137469934905152 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0425 16:44:32.720353 137469934905152 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0425 16:44:33.177417 137469934905152 checkpointing.py:688] Setting up checkpoint logger... I0425 16:44:33.177540 137469934905152 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0425 16:44:33.177586 137469934905152 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0425 16:44:33.177814 137469934905152 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 0x7d06c6d6d2b0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0425 16:44:36.014785 137469934905152 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0425 16:44:36.015146 137469934905152 checkpoint_manager.py:709] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d0571afc200>}, handler_registry=None I0425 16:44:36.015465 137469934905152 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d0571afc200>` 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 16:44:36.015521 137469934905152 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cf26818f7d0>` 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 16:44:36.015560 137469934905152 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 0x7d0571afc200>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d0571afc200>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cf26818f7d0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7cf26818f7d0>}). I0425 16:44:36.015907 137469934905152 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0425 16:44:36.015987 137469934905152 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7cf23022c7c0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0425 16:44:36.978511 137469934905152 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/nnx_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints I0425 16:44:37.001824 137469934905152 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/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/nnx_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7cf26818c7a0> I0425 16:44:37.001958 137469934905152 checkpointing.py:302] Checkpoint manager created! I0425 16:44:38.352629 137469934905152 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0425 16:44:38.352755 137469934905152 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0425 16:44:40.596737 137469934905152 maxtext_utils.py:1749] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0425 16:44:40.596879 137469934905152 maxtext_utils.py:1749] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0425 16:44:40.596958 137469934905152 maxtext_utils.py:1749] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0425 16:44:40.597027 137469934905152 maxtext_utils.py:1749] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0425 16:44:40.597086 137469934905152 maxtext_utils.py:1749] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0425 16:44:40.597154 137469934905152 maxtext_utils.py:1749] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0425 16:44:40.597214 137469934905152 maxtext_utils.py:1749] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0425 16:44:40.597270 137469934905152 maxtext_utils.py:1749] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0425 16:44:40.597330 137469934905152 maxtext_utils.py:1749] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0425 16:44:40.597386 137469934905152 maxtext_utils.py:1749] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0425 16:44:40.597439 137469934905152 maxtext_utils.py:1749] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0425 16:44:40.597492 137469934905152 maxtext_utils.py:1749] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0425 16:44:41.013124 137469934905152 nnx_decoders.py:616] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 16:44:41.013282 137469934905152 nnx_decoders.py:616] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 16:44:41.018662 137469934905152 nnx_decoders.py:616] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0425 16:44:41.018758 137469934905152 nnx_decoders.py:616] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 16:44:41.172053 137469934905152 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 16:44:41.172211 137469934905152 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 16:44:41.326879 137469934905152 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0425 16:44:41.327010 137469934905152 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0425 16:44:41.526944 137469934905152 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 16:44:41.527087 137469934905152 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 16:44:41.682963 137469934905152 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 16:44:41.683115 137469934905152 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 16:44:41.839709 137469934905152 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0425 16:44:41.839846 137469934905152 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 16:44:42.001519 137469934905152 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0425 16:44:42.001670 137469934905152 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0425 16:44:42.084371 137469934905152 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0425 16:44:42.084498 137469934905152 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0425 16:44:55.522846 137469934905152 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 16:44:55.526716 137469934905152 metric_logger.py:289] number parameters: 1.104 billion I0425 16:45:08.684807 137469934905152 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0425 16:45:08.844067 137469934905152 checkpointing.py:798] Waited 0.15924596786499023 seconds for step 0 to finish before starting checkpointing. I0425 16:45:08.846627 137469934905152 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 16:45:08.848617 137469934905152 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 0 I0425 16:45:08.851599 137469934905152 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/nnx_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/0. I0425 16:45:09.473562 137469934905152 signaling_client.py:364] Using JaxDistributedSignalingClient I0425 16:45:09.474668 137469934905152 jax_array_handlers.py:358] Scheduling D2H of 153 prioritized jax.Array. I0425 16:45:09.474729 137469934905152 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 16:45:09.842073 137469934905152 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.371026s I0425 16:45:09.842266 137469934905152 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.319 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.6652743816375732 s) (per-host) I0425 16:45:09.842324 137469934905152 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.665337s (batch_requests_ready=0.282528s, total_serialization_initiated=0.382738s, others=0.000071s) I0425 16:45:09.842450 137469934905152 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.669493s (all_items=0.000019s, per_item={'items': '0.00001931'}, temp_paths=0.669474) I0425 16:45:09.843425 137343775790848 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0425 16:45:09.843553 137469934905152 async_checkpointer.py:561] Finished blocking save. Time taken: 0.994866s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/nnx_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/0. I0425 16:45:10.208615 137469934905152 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0425 16:45:10.209022 137341108745984 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 16:45:10.209229 137469934905152 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/nnx_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777135508.8466096, 'wait_for_prev_duration_secs': 6.365776062011719e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777135508.8486567, 'checkpointer_blocking_duration_secs': 0.9950745105743408, 'get_old_steps_start_time': 1777135509.8437538, 'get_old_steps_duration_secs': 2.47955322265625e-05, 'checkpoint_manager_blocking_start_time': 1777135508.8445766, 'checkpoint_manager_blocking_duration_secs': 1.3646113872528076} I0425 16:45:10.209359 137469934905152 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0425 16:45:10.209444 137469934905152 max_utils.py:750] Memstats: After params initialized: I0425 16:45:10.209504 137469934905152 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_24(process=6,(0,6,0,0)) I0425 16:45:10.209539 137469934905152 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_25(process=6,(1,6,0,0)) I0425 16:45:10.209568 137469934905152 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_28(process=6,(0,7,0,0)) I0425 16:45:10.209593 137469934905152 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_29(process=6,(1,7,0,0)) I0425 16:45:10.491136 137469934905152 metric_logger.py:185] completed step: 0, seconds: 13.156, TFLOP/s/device: 1.033, Tokens/s/device: 155.671, total_weights: 65536, loss: 10.872 I0425 16:45:10.652894 137469934905152 metric_logger.py:185] completed step: 1, seconds: 1.803, TFLOP/s/device: 7.536, Tokens/s/device: 1135.835, total_weights: 65536, loss: 10.872 I0425 16:45:11.099195 137469934905152 metric_logger.py:185] completed step: 2, seconds: 0.019, TFLOP/s/device: 707.774, Tokens/s/device: 106683.336, total_weights: 65536, loss: 10.856 I0425 16:45:11.254581 137469934905152 metric_logger.py:185] completed step: 3, seconds: 0.418, TFLOP/s/device: 32.523, Tokens/s/device: 4902.195, total_weights: 65536, loss: 10.824 I0425 16:45:11.565899 137469934905152 metric_logger.py:185] completed step: 4, seconds: 0.183, TFLOP/s/device: 74.439, Tokens/s/device: 11220.258, total_weights: 65536, loss: 10.793 I0425 16:45:11.575883 137469934905152 metric_logger.py:185] completed step: 5, seconds: 0.155, TFLOP/s/device: 87.618, Tokens/s/device: 13206.768, total_weights: 65536, loss: 10.763 I0425 16:45:31.147790 137469934905152 metric_logger.py:185] completed step: 6, seconds: 0.312, TFLOP/s/device: 43.536, Tokens/s/device: 6562.294, total_weights: 65536, loss: 10.733 I0425 16:45:31.302876 137469934905152 metric_logger.py:185] completed step: 7, seconds: 19.420, TFLOP/s/device: 0.700, Tokens/s/device: 105.456, total_weights: 65536, loss: 10.706 I0425 16:45:31.458024 137469934905152 metric_logger.py:185] completed step: 8, seconds: 0.161, TFLOP/s/device: 84.398, Tokens/s/device: 12721.445, total_weights: 65536, loss: 10.680 I0425 16:45:31.466558 137469934905152 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0425 16:45:31.769001 137469934905152 checkpointing.py:798] Waited 0.3024137020111084 seconds for step 10 to finish before starting checkpointing. I0425 16:45:31.771223 137469934905152 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 16:45:32.302057 2628 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0425 16:45:34.205694 137341603653376 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/nnx_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/0/items/array_metadatas/process_6 I0425 16:45:49.106936 137343775790848 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 39.563 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 39.929903745651245 s) (per-host) I0425 16:45:49.107058 137343775790848 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 39.263424s. I0425 16:45:56.149481 137343775790848 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 46.305831s. I0425 16:45:56.149748 137341108745984 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 16:45:56.149871 137341108745984 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 16:45:56.149922 137341108745984 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0425 16:45:56.151176 137341108745984 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0425 16:45:56.151384 137469934905152 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0425 16:45:56.151514 137469934905152 checkpoint_manager.py:1458] Waiting for previous save to complete took 24.380297 seconds. If this number is high, consider checkpointing less frequently. I0425 16:45:56.153352 137469934905152 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 10 I0425 16:45:56.156894 137469934905152 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/nnx_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/10. I0425 16:45:56.711032 137469934905152 jax_array_handlers.py:358] Scheduling D2H of 153 prioritized jax.Array. I0425 16:45:56.711155 137469934905152 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 16:45:56.763161 137469934905152 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.055479s I0425 16:45:56.763345 137469934905152 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.795 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3217582702636719 s) (per-host) I0425 16:45:56.763400 137469934905152 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.321825s (batch_requests_ready=0.256608s, total_serialization_initiated=0.065139s, others=0.000077s) I0425 16:45:56.763529 137469934905152 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.325642s (all_items=0.000015s, per_item={'items': '0.00001550'}, temp_paths=0.325626) I0425 16:45:56.764427 137341100353280 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0425 16:45:56.764556 137469934905152 async_checkpointer.py:561] Finished blocking save. Time taken: 0.611131s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/nnx_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/10. I0425 16:45:56.813997 137469934905152 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0425 16:45:56.814323 137341108745984 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 16:45:56.814471 137469934905152 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/nnx_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777135531.7711928, 'wait_for_prev_duration_secs': 24.38029670715332, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777135556.1533906, 'checkpointer_blocking_duration_secs': 0.6113302707672119, 'get_old_steps_start_time': 1777135556.764741, 'get_old_steps_duration_secs': 2.574920654296875e-05, 'checkpoint_manager_blocking_start_time': 1777135531.7694705, 'checkpoint_manager_blocking_duration_secs': 25.044965505599976} I0425 16:45:56.814608 137469934905152 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0425 16:45:56.815053 137469934905152 metric_logger.py:185] completed step: 9, seconds: 0.155, TFLOP/s/device: 87.622, Tokens/s/device: 13207.365, total_weights: 65536, loss: 10.656 I0425 16:45:56.825937 137469934905152 metric_logger.py:185] completed step: 10, seconds: 0.155, TFLOP/s/device: 87.631, Tokens/s/device: 13208.728, total_weights: 65536, loss: 10.636 I0425 16:45:56.977197 137469934905152 metric_logger.py:185] completed step: 11, seconds: 25.357, TFLOP/s/device: 0.536, Tokens/s/device: 80.767, total_weights: 65536, loss: 10.618 I0425 16:45:57.132188 137469934905152 metric_logger.py:185] completed step: 12, seconds: 0.010, TFLOP/s/device: 1415.473, Tokens/s/device: 213355.558, total_weights: 65536, loss: 10.602 I0425 16:45:57.287512 137469934905152 metric_logger.py:185] completed step: 13, seconds: 0.153, TFLOP/s/device: 88.968, Tokens/s/device: 13410.162, total_weights: 65536, loss: 10.588 I0425 16:45:57.442788 137469934905152 metric_logger.py:185] completed step: 14, seconds: 0.155, TFLOP/s/device: 87.554, Tokens/s/device: 13197.152, total_weights: 65536, loss: 10.577 I0425 16:45:57.597807 137469934905152 metric_logger.py:185] completed step: 15, seconds: 0.155, TFLOP/s/device: 87.591, Tokens/s/device: 13202.597, total_weights: 65536, loss: 10.568 I0425 16:45:57.752835 137469934905152 metric_logger.py:185] completed step: 16, seconds: 0.155, TFLOP/s/device: 87.494, Tokens/s/device: 13187.974, total_weights: 65536, loss: 10.562 I0425 16:45:57.907955 137469934905152 metric_logger.py:185] completed step: 17, seconds: 0.155, TFLOP/s/device: 87.683, Tokens/s/device: 13216.484, total_weights: 65536, loss: 10.556 I0425 16:45:58.063077 137469934905152 metric_logger.py:185] completed step: 18, seconds: 0.155, TFLOP/s/device: 87.590, Tokens/s/device: 13202.512, total_weights: 65536, loss: 10.552 I0425 16:45:58.220362 137469934905152 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0425 16:45:58.221522 137469934905152 checkpointing.py:798] Waited 0.0011696815490722656 seconds for step 19 to finish before starting checkpointing. I0425 16:45:58.223638 137469934905152 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 16:46:01.650923 137342719358720 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/nnx_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/10/items/array_metadatas/process_6 I0425 16:46:17.391628 137341100353280 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 75.406 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 20.94999861717224 s) (per-host) I0425 16:46:17.391735 137341100353280 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 20.627101s. I0425 16:46:25.470399 137341100353280 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 28.705751s. I0425 16:46:25.470629 137341108745984 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 16:46:25.470686 137341108745984 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 16:46:25.470737 137341108745984 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0425 16:46:25.473429 137341108745984 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0425 16:46:25.473586 137469934905152 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0425 16:46:25.473710 137469934905152 checkpoint_manager.py:1458] Waiting for previous save to complete took 27.250086 seconds. If this number is high, consider checkpointing less frequently. I0425 16:46:25.475278 137469934905152 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 19 I0425 16:46:25.478848 137469934905152 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/nnx_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/19. I0425 16:46:26.048074 137469934905152 jax_array_handlers.py:358] Scheduling D2H of 153 prioritized jax.Array. I0425 16:46:26.048198 137469934905152 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 16:46:26.101618 137469934905152 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.056964s I0425 16:46:26.101785 137469934905152 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.866 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3170735836029053 s) (per-host) I0425 16:46:26.101836 137469934905152 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.317133s (batch_requests_ready=0.250645s, total_serialization_initiated=0.066423s, others=0.000064s) I0425 16:46:26.101960 137469934905152 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.321108s (all_items=0.000010s, per_item={'items': '0.00001025'}, temp_paths=0.321098) I0425 16:46:26.102832 137343239444224 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0425 16:46:26.102989 137469934905152 async_checkpointer.py:561] Finished blocking save. Time taken: 0.627648s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/nnx_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/19. I0425 16:46:26.530925 137469934905152 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0425 16:46:26.531298 137341108745984 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0425 16:46:26.531465 137469934905152 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/nnx_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777135558.223608, 'wait_for_prev_duration_secs': 27.250085592269897, 'time_between_consecutive_saves_sec': 2.0723936557769775, 'checkpointer_blocking_start_time': 1777135585.475316, 'checkpointer_blocking_duration_secs': 0.6277883052825928, 'get_old_steps_start_time': 1777135586.103119, 'get_old_steps_duration_secs': 2.0265579223632812e-05, 'checkpoint_manager_blocking_start_time': 1777135558.2218301, 'checkpoint_manager_blocking_duration_secs': 28.309600591659546} I0425 16:46:26.531595 137469934905152 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0425 16:46:26.531639 137469934905152 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0425 16:46:30.918387 137341603653376 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 153 array_metadata.ArrayMetadata to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260425_150412/nnx_xpk_feat_nnx_set_defaults_true_20260425_150412_05_fp8/checkpoints/19/items/array_metadatas/process_6 I0425 16:46:47.405775 137343239444224 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 73.066 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 21.62101912498474 s) (per-host) I0425 16:46:47.405889 137343239444224 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 21.302943s. I0425 16:46:53.567120 137343239444224 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 27.464149s. I0425 16:46:53.567425 137341108745984 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0425 16:46:53.567548 137341108745984 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0425 16:46:53.567590 137341108745984 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0425 16:46:53.568940 137341108745984 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0425 16:46:53.569126 137469934905152 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0425 16:46:53.569302 137469934905152 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0425 16:46:53.569952 137469934905152 metric_logger.py:185] completed step: 19, seconds: 0.155, TFLOP/s/device: 87.676, Tokens/s/device: 13215.546, 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: Sat Apr 25 16:47:06 UTC 2026 EXIT_CODE=0