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 | 86.036 | 87.594 | +1.558 |
| Tok/s | 12968.3 | 13203.2 | +234.847 |
| Avg s/step | 2.682 | 2.639 | -0.043 |
| 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: Wed Apr 22 16:01:09 UTC 2026 2026-04-22 16:01:13.888448: 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 00:00:1776873673.901246 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776873673.905024 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776873673.916717 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776873673.916737 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776873673.916739 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776873673.916741 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-22 16:01:33.334618: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0422 16:01:33.869812 137924698810176 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-22 16:01:42,910:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0422 16:01:42.910624 137924698810176 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-22 16:01:42,913:jax._src.distributed:157: Connecting to JAX distributed service on mt-05-fp8-6i6q8-slice-job-0-0.mt-05-fp8-6i6q8:8482 I0422 16:01:42.913145 137924698810176 distributed.py:157] Connecting to JAX distributed service on mt-05-fp8-6i6q8-slice-job-0-0.mt-05-fp8-6i6q8:8482 I0422 16:01:44.408121 137924698810176 max_utils.py:284] Jax distributed system initialized! I0422 16:01:51.537330 137924698810176 max_utils.py:800] System Information: Jax Version: 0.8.1 I0422 16:01:51.537437 137924698810176 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0422 16:01:51.537479 137924698810176 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709 I0422 16:01:51.537516 137924698810176 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 16:01:52.172996 137924698810176 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 16:01:52.173601 137924698810176 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 16:01:52.173788 137924698810176 checkpointing.py:688] Setting up checkpoint logger... I0422 16:01:52.173837 137924698810176 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 16:01:52.173880 137924698810176 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0422 16:01:52.174243 137924698810176 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 0x7d7091d71130>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 16:01:55.338889 137924698810176 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0422 16:01:55.339241 137924698810176 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 0x7d5c0047d8e0>}, handler_registry=None I0422 16:01:55.339480 137924698810176 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d5c0047d8e0>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0422 16:01:55.339534 137924698810176 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d5c001c6d80>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0422 16:01:55.339570 137924698810176 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 0x7d5c0047d8e0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7d5c0047d8e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d5c001c6d80>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7d5c001c6d80>}). I0422 16:01:55.340065 137924698810176 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0422 16:01:55.340155 137924698810176 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7d5be8774180> timeout: 600 secs and primary_host=0 for async checkpoint writes I0422 16:01:56.479917 137924698810176 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints I0422 16:01:56.854772 137924698810176 checkpoint_manager.py:929] [process=6][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7d6f53631a60> I0422 16:01:56.854947 137924698810176 checkpointing.py:302] Checkpoint manager created! I0422 16:01:57.937814 137924698810176 nnx_wrappers.py:455] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 16:01:57.937925 137924698810176 nnx_wrappers.py:455] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 16:01:58.307425 137924698810176 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 16:01:58.307521 137924698810176 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 16:01:58.323437 137924698810176 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 16:01:58.323498 137924698810176 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 16:01:58.383203 137924698810176 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 16:01:58.383291 137924698810176 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 16:01:58.399243 137924698810176 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 16:01:58.399304 137924698810176 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 16:01:58.415188 137924698810176 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 16:01:58.415252 137924698810176 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 16:01:58.439086 137924698810176 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 16:01:58.439168 137924698810176 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 16:01:58.499502 137924698810176 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 16:01:58.499587 137924698810176 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 16:01:58.964787 137924698810176 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0422 16:01:58.964904 137924698810176 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 16:02:01.016417 137924698810176 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: () I0422 16:02:01.016546 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.016602 137924698810176 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: () I0422 16:02:01.016641 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.016676 137924698810176 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: () I0422 16:02:01.016708 137924698810176 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: () I0422 16:02:01.016740 137924698810176 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: () I0422 16:02:01.016772 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.016802 137924698810176 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: () I0422 16:02:01.016832 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.016863 137924698810176 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: () I0422 16:02:01.016893 137924698810176 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: () I0422 16:02:01.016925 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0422 16:02:01.016957 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.016986 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0422 16:02:01.017014 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.017044 137924698810176 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: () I0422 16:02:01.017072 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.017111 137924698810176 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: () I0422 16:02:01.017141 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.017168 137924698810176 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: () I0422 16:02:01.017195 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.017222 137924698810176 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: () I0422 16:02:01.017249 137924698810176 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: () I0422 16:02:01.017277 137924698810176 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: () I0422 16:02:01.017305 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.017332 137924698810176 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: () I0422 16:02:01.017358 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.017385 137924698810176 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: () I0422 16:02:01.017411 137924698810176 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: () I0422 16:02:01.017437 137924698810176 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: () I0422 16:02:01.017464 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.017490 137924698810176 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: () I0422 16:02:01.017516 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.017542 137924698810176 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: () I0422 16:02:01.017572 137924698810176 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: () I0422 16:02:01.017599 137924698810176 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: () I0422 16:02:01.017625 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.017651 137924698810176 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: () I0422 16:02:01.017677 137924698810176 maxtext_utils.py:1740] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0422 16:02:01.017704 137924698810176 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: () I0422 16:02:01.017729 137924698810176 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: () I0422 16:02:01.017800 137924698810176 maxtext_utils.py:1740] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0422 16:02:01.017860 137924698810176 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 16:02:01.017899 137924698810176 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0422 16:02:01.017949 137924698810176 maxtext_utils.py:1740] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0422 16:02:01.017995 137924698810176 maxtext_utils.py:1740] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0422 16:02:01.018028 137924698810176 maxtext_utils.py:1740] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0422 16:02:01.018076 137924698810176 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 16:02:01.018140 137924698810176 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: PartitionSpec('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0422 16:02:01.018181 137924698810176 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0422 16:02:01.018218 137924698810176 maxtext_utils.py:1740] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0422 16:02:01.018267 137924698810176 maxtext_utils.py:1740] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0422 16:02:01.018311 137924698810176 maxtext_utils.py:1740] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0422 16:02:02.806082 137924698810176 train.py:158] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 16:02:02.806201 137924698810176 train.py:158] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0422 16:02:02.822129 137924698810176 train.py:165] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0422 16:02:02.822188 137924698810176 train.py:165] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0422 16:02:16.079849 137924698810176 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0422 16:02:16.080946 137924698810176 metric_logger.py:289] number parameters: 1.105 billion I0422 16:02:30.824234 137924698810176 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0422 16:02:30.991511 137924698810176 checkpointing.py:798] Waited 0.1672513484954834 seconds for step 0 to finish before starting checkpointing. I0422 16:02:30.994510 137924698810176 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 16:02:30.996491 137924698810176 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 0 I0422 16:02:31.000023 137924698810176 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/0. I0422 16:02:31.638242 137924698810176 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 16:02:31.639310 137924698810176 jax_array_handlers.py:358] Scheduling D2H of 81 prioritized jax.Array. I0422 16:02:31.639427 137924698810176 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 16:02:32.004289 137924698810176 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.366176s I0422 16:02:32.004478 137924698810176 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.365 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.6523199081420898 s) (per-host) I0422 16:02:32.004532 137924698810176 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.652385s (batch_requests_ready=0.280847s, total_serialization_initiated=0.371464s, others=0.000074s) I0422 16:02:32.004660 137924698810176 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.656712s (all_items=0.000039s, per_item={'items': '0.00003886'}, temp_paths=0.656673) I0422 16:02:32.005627 137798296258304 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0422 16:02:32.005801 137924698810176 async_checkpointer.py:561] Finished blocking save. Time taken: 1.009230s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/0. I0422 16:02:32.015468 137924698810176 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 16:02:32.015796 137795637081856 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 16:02:32.015952 137924698810176 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776873750.9944823, 'wait_for_prev_duration_secs': 7.390975952148438e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776873750.996537, 'checkpointer_blocking_duration_secs': 1.0093746185302734, 'get_old_steps_start_time': 1776873752.0059333, 'get_old_steps_duration_secs': 2.6226043701171875e-05, 'checkpoint_manager_blocking_start_time': 1776873750.9923654, 'checkpoint_manager_blocking_duration_secs': 1.0235474109649658} I0422 16:02:32.016064 137924698810176 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0422 16:02:32.016126 137924698810176 max_utils.py:750] Memstats: After params initialized: I0422 16:02:32.016177 137924698810176 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_24(process=6,(0,6,0,0)) I0422 16:02:32.016209 137924698810176 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_25(process=6,(1,6,0,0)) I0422 16:02:32.016236 137924698810176 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_28(process=6,(0,7,0,0)) I0422 16:02:32.016259 137924698810176 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_29(process=6,(1,7,0,0)) I0422 16:02:32.290574 137924698810176 metric_logger.py:185] completed step: 0, seconds: 14.743, TFLOP/s/device: 0.922, Tokens/s/device: 138.912, total_weights: 65536, loss: 10.874 I0422 16:02:32.458542 137924698810176 metric_logger.py:185] completed step: 1, seconds: 1.464, TFLOP/s/device: 9.278, Tokens/s/device: 1398.529, total_weights: 65536, loss: 10.874 I0422 16:02:32.878518 137924698810176 metric_logger.py:185] completed step: 2, seconds: 0.011, TFLOP/s/device: 1258.534, Tokens/s/device: 189699.889, total_weights: 65536, loss: 10.816 I0422 16:02:33.036728 137924698810176 metric_logger.py:185] completed step: 3, seconds: 0.420, TFLOP/s/device: 32.317, Tokens/s/device: 4871.122, total_weights: 65536, loss: 10.431 I0422 16:02:33.353996 137924698810176 metric_logger.py:185] completed step: 4, seconds: 0.164, TFLOP/s/device: 82.944, Tokens/s/device: 12502.289, total_weights: 65536, loss: 9.992 I0422 16:02:33.360473 137924698810176 metric_logger.py:185] completed step: 5, seconds: 0.158, TFLOP/s/device: 85.922, Tokens/s/device: 12951.123, total_weights: 65536, loss: 9.549 I0422 16:02:54.140658 137924698810176 metric_logger.py:185] completed step: 6, seconds: 0.318, TFLOP/s/device: 42.779, Tokens/s/device: 6448.099, total_weights: 65536, loss: 9.111 I0422 16:02:54.298902 137924698810176 metric_logger.py:185] completed step: 7, seconds: 20.623, TFLOP/s/device: 0.659, Tokens/s/device: 99.305, total_weights: 65536, loss: 8.685 I0422 16:02:54.457266 137924698810176 metric_logger.py:185] completed step: 8, seconds: 0.163, TFLOP/s/device: 83.348, Tokens/s/device: 12563.184, total_weights: 65536, loss: 8.281 I0422 16:02:54.462399 137924698810176 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0422 16:02:54.773805 137924698810176 checkpointing.py:798] Waited 0.3113706111907959 seconds for step 10 to finish before starting checkpointing. I0422 16:02:54.776468 137924698810176 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 16:02:55.622547 2713 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 16:02:58.139292 137798304651008 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/0/items/array_metadatas/process_6 I0422 16:03:12.761649 137798296258304 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 38.150 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.40944290161133 s) (per-host) I0422 16:03:12.761773 137798296258304 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.756034s. I0422 16:03:20.021328 137798296258304 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 48.015574s. I0422 16:03:20.021635 137795637081856 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 16:03:20.021761 137795637081856 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 16:03:20.021811 137795637081856 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 16:03:20.023253 137795637081856 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 16:03:20.023418 137924698810176 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0422 16:03:20.023555 137924698810176 checkpoint_manager.py:1458] Waiting for previous save to complete took 25.247084 seconds. If this number is high, consider checkpointing less frequently. I0422 16:03:20.025230 137924698810176 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 10 I0422 16:03:20.028619 137924698810176 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/10. I0422 16:03:20.612952 137924698810176 jax_array_handlers.py:358] Scheduling D2H of 81 prioritized jax.Array. I0422 16:03:20.613055 137924698810176 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 16:03:20.665632 137924698810176 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.053707s I0422 16:03:20.665804 137924698810176 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.571 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.33748507499694824 s) (per-host) I0422 16:03:20.665857 137924698810176 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.337548s (batch_requests_ready=0.280277s, total_serialization_initiated=0.057199s, others=0.000072s) I0422 16:03:20.665959 137924698810176 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.341684s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=0.341667) I0422 16:03:20.666841 137795091818240 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0422 16:03:20.667002 137924698810176 async_checkpointer.py:561] Finished blocking save. Time taken: 0.641702s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/10. I0422 16:03:20.671612 137924698810176 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0422 16:03:20.671894 137795637081856 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 16:03:20.672056 137924698810176 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776873774.7764401, 'wait_for_prev_duration_secs': 25.24708366394043, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776873800.0252671, 'checkpointer_blocking_duration_secs': 0.6418600082397461, 'get_old_steps_start_time': 1776873800.6671455, 'get_old_steps_duration_secs': 2.6702880859375e-05, 'checkpoint_manager_blocking_start_time': 1776873774.7745264, 'checkpoint_manager_blocking_duration_secs': 25.897494792938232} I0422 16:03:20.672178 137924698810176 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0422 16:03:20.672575 137924698810176 metric_logger.py:185] completed step: 9, seconds: 0.158, TFLOP/s/device: 85.790, Tokens/s/device: 12931.171, total_weights: 65536, loss: 7.908 I0422 16:03:20.679444 137924698810176 metric_logger.py:185] completed step: 10, seconds: 0.158, TFLOP/s/device: 85.943, Tokens/s/device: 12954.318, total_weights: 65536, loss: 7.568 I0422 16:03:20.837710 137924698810176 metric_logger.py:185] completed step: 11, seconds: 26.215, TFLOP/s/device: 0.518, Tokens/s/device: 78.123, total_weights: 65536, loss: 7.288 I0422 16:03:20.995822 137924698810176 metric_logger.py:185] completed step: 12, seconds: 0.006, TFLOP/s/device: 2448.131, Tokens/s/device: 369009.009, total_weights: 65536, loss: 7.032 I0422 16:03:21.154003 137924698810176 metric_logger.py:185] completed step: 13, seconds: 0.159, TFLOP/s/device: 85.198, Tokens/s/device: 12841.897, total_weights: 65536, loss: 6.815 I0422 16:03:21.312234 137924698810176 metric_logger.py:185] completed step: 14, seconds: 0.158, TFLOP/s/device: 85.912, Tokens/s/device: 12949.568, total_weights: 65536, loss: 6.635 I0422 16:03:21.470268 137924698810176 metric_logger.py:185] completed step: 15, seconds: 0.158, TFLOP/s/device: 85.974, Tokens/s/device: 12958.909, total_weights: 65536, loss: 6.492 I0422 16:03:21.628700 137924698810176 metric_logger.py:185] completed step: 16, seconds: 0.158, TFLOP/s/device: 85.835, Tokens/s/device: 12937.951, total_weights: 65536, loss: 6.380 I0422 16:03:21.786669 137924698810176 metric_logger.py:185] completed step: 17, seconds: 0.159, TFLOP/s/device: 85.584, Tokens/s/device: 12900.137, total_weights: 65536, loss: 6.293 I0422 16:03:21.944964 137924698810176 metric_logger.py:185] completed step: 18, seconds: 0.158, TFLOP/s/device: 86.058, Tokens/s/device: 12971.631, total_weights: 65536, loss: 6.223 I0422 16:03:22.102722 137924698810176 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0422 16:03:22.103744 137924698810176 checkpointing.py:798] Waited 0.0010423660278320312 seconds for step 19 to finish before starting checkpointing. I0422 16:03:22.105773 137924698810176 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 16:03:26.146559 137797230909184 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/10/items/array_metadatas/process_6 I0422 16:04:03.090010 137795091818240 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 36.943 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.761651277542114 s) (per-host) I0422 16:04:03.090149 137795091818240 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.423197s. I0422 16:04:11.699445 137795091818240 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 51.032477s. I0422 16:04:11.699759 137795637081856 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 16:04:11.699876 137795637081856 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 16:04:11.699924 137795637081856 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0422 16:04:11.702588 137795637081856 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0422 16:04:11.702726 137924698810176 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. W0422 16:04:11.702793 137924698810176 checkpoint_manager.py:1458] Waiting for previous save to complete took 49.597035 seconds. If this number is high, consider checkpointing less frequently. I0422 16:04:11.704288 137924698810176 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 19 I0422 16:04:11.707723 137924698810176 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/19. I0422 16:04:12.715222 137924698810176 jax_array_handlers.py:358] Scheduling D2H of 81 prioritized jax.Array. I0422 16:04:12.715323 137924698810176 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 16:04:12.768595 137924698810176 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.054442s I0422 16:04:12.768765 137924698810176 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.608 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.33479785919189453 s) (per-host) I0422 16:04:12.768820 137924698810176 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.334863s (batch_requests_ready=0.276937s, total_serialization_initiated=0.057855s, others=0.000070s) I0422 16:04:12.768920 137924698810176 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.339166s (all_items=0.000011s, per_item={'items': '0.00001144'}, temp_paths=0.339155) I0422 16:04:12.769803 137797205731072 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0422 16:04:12.769960 137924698810176 async_checkpointer.py:561] Finished blocking save. Time taken: 1.065604s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/19. I0422 16:04:12.783905 137924698810176 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0422 16:04:12.784247 137795637081856 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 16:04:12.784409 137924698810176 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776873802.105746, 'wait_for_prev_duration_secs': 49.59703469276428, 'time_between_consecutive_saves_sec': 2.0824520587921143, 'checkpointer_blocking_start_time': 1776873851.7043278, 'checkpointer_blocking_duration_secs': 1.0657389163970947, 'get_old_steps_start_time': 1776873852.7700806, 'get_old_steps_duration_secs': 3.4809112548828125e-05, 'checkpoint_manager_blocking_start_time': 1776873802.104009, 'checkpoint_manager_blocking_duration_secs': 50.68036866188049} I0422 16:04:12.784530 137924698810176 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0422 16:04:12.784580 137924698810176 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 16:04:18.263239 137797230909184 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/linen_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/19/items/array_metadatas/process_6 I0422 16:04:54.423183 137797205731072 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 37.623 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.98917317390442 s) (per-host) I0422 16:04:54.423293 137797205731072 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.653381s. I0422 16:05:03.357180 137797205731072 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 50.587254s. I0422 16:05:03.357460 137795637081856 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 16:05:03.357600 137795637081856 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 16:05:03.357642 137795637081856 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0422 16:05:03.359491 137795637081856 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0422 16:05:03.359661 137924698810176 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. I0422 16:05:03.359803 137924698810176 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 16:05:03.360469 137924698810176 metric_logger.py:185] completed step: 19, seconds: 0.158, TFLOP/s/device: 86.036, Tokens/s/device: 12968.345, 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: Wed Apr 22 16:05:15 UTC 2026 EXIT_CODE=0
XPK Start: Wed Apr 22 17:40:42 UTC 2026 2026-04-22 17:40:46.788425: 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 00:00:1776879646.801348 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776879646.805090 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776879646.816725 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776879646.816747 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776879646.816749 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776879646.816751 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-22 17:41:06.035168: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0422 17:41:06.559869 135380875044672 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-22 17:41:15,601:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0422 17:41:15.601376 135380875044672 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-22 17:41:15,603:jax._src.distributed:157: Connecting to JAX distributed service on mt-05-fp8-dwjkn-slice-job-0-0.mt-05-fp8-dwjkn:8482 I0422 17:41:15.603694 135380875044672 distributed.py:157] Connecting to JAX distributed service on mt-05-fp8-dwjkn-slice-job-0-0.mt-05-fp8-dwjkn:8482 I0422 17:41:16.595173 135380875044672 max_utils.py:284] Jax distributed system initialized! I0422 17:41:23.350986 135380875044672 max_utils.py:800] System Information: Jax Version: 0.8.1 I0422 17:41:23.351105 135380875044672 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0422 17:41:23.351152 135380875044672 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709 I0422 17:41:23.351187 135380875044672 train_utils.py:377] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0422 17:41:23.975971 135380875044672 maxtext_utils.py:1631] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0422 17:41:24.434775 135380875044672 checkpointing.py:688] Setting up checkpoint logger... I0422 17:41:24.434903 135380875044672 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0422 17:41:24.434948 135380875044672 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0422 17:41:24.435183 135380875044672 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 0x7b2045970f80>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0422 17:41:27.167325 135380875044672 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0422 17:41:27.167673 135380875044672 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 0x7b0b8010cda0>}, handler_registry=None I0422 17:41:27.167914 135380875044672 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b0b8010cda0>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0422 17:41:27.167962 135380875044672 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b0b8038ad20>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0422 17:41:27.167999 135380875044672 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 0x7b0b8010cda0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b0b8010cda0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b0b8038ad20>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b0b8038ad20>}). I0422 17:41:27.168379 135380875044672 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0422 17:41:27.168456 135380875044672 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7b0bb82247c0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0422 17:41:28.461241 135380875044672 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints I0422 17:41:28.970199 135380875044672 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_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7b0b803891f0> I0422 17:41:28.970380 135380875044672 checkpointing.py:302] Checkpoint manager created! I0422 17:41:30.312268 135380875044672 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0422 17:41:30.312393 135380875044672 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0422 17:41:32.565977 135380875044672 maxtext_utils.py:1749] decoder/decoder_norm/scale/value Shape: float32[2048] Physical: (None,) I0422 17:41:32.566113 135380875044672 maxtext_utils.py:1749] decoder/layers/mlp/wi_0/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0422 17:41:32.566171 135380875044672 maxtext_utils.py:1749] decoder/layers/mlp/wi_1/kernel/value Shape: float32[2048,16,7168] Physical: ('fsdp', None, None) I0422 17:41:32.566216 135380875044672 maxtext_utils.py:1749] decoder/layers/mlp/wo/kernel/value Shape: float32[7168,16,2048] Physical: (None, None, 'fsdp') I0422 17:41:32.566258 135380875044672 maxtext_utils.py:1749] decoder/layers/post_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0422 17:41:32.566294 135380875044672 maxtext_utils.py:1749] decoder/layers/pre_self_attention_layer_norm/scale/value Shape: float32[2048,16] Physical: (None, None) I0422 17:41:32.566333 135380875044672 maxtext_utils.py:1749] decoder/layers/self_attention/key/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0422 17:41:32.566370 135380875044672 maxtext_utils.py:1749] decoder/layers/self_attention/out/kernel/value Shape: float32[16,16,128,2048] Physical: (None, None, None, 'fsdp') I0422 17:41:32.566405 135380875044672 maxtext_utils.py:1749] decoder/layers/self_attention/query/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0422 17:41:32.566441 135380875044672 maxtext_utils.py:1749] decoder/layers/self_attention/value/kernel/value Shape: float32[2048,16,16,128] Physical: ('fsdp', None, None, None) I0422 17:41:32.566476 135380875044672 maxtext_utils.py:1749] decoder/logits_dense/kernel/value Shape: float32[2048,32000] Physical: ('fsdp', None) I0422 17:41:32.566509 135380875044672 maxtext_utils.py:1749] token_embedder/embedding/value Shape: float32[32000,2048] Physical: (None, 'fsdp') I0422 17:41:32.976465 135380875044672 nnx_decoders.py:616] nnx_decoders/carry Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 17:41:32.976621 135380875044672 nnx_decoders.py:616] nnx_decoders/carry Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 17:41:32.982008 135380875044672 nnx_decoders.py:616] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0422 17:41:32.982164 135380875044672 nnx_decoders.py:616] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 17:41:33.144785 135380875044672 attentions.py:1084] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 17:41:33.144924 135380875044672 attentions.py:1084] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 17:41:33.296038 135380875044672 attentions.py:1085] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0422 17:41:33.296190 135380875044672 attentions.py:1085] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0422 17:41:33.493151 135380875044672 attentions.py:1150] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 17:41:33.493319 135380875044672 attentions.py:1150] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 17:41:33.646930 135380875044672 attentions.py:1151] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 17:41:33.647068 135380875044672 attentions.py:1151] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 17:41:33.797299 135380875044672 attentions.py:1152] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0422 17:41:33.797436 135380875044672 attentions.py:1152] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 17:41:33.957999 135380875044672 attentions.py:1193] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0422 17:41:33.958168 135380875044672 attentions.py:1193] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0422 17:41:34.038373 135380875044672 linears.py:541] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0422 17:41:34.038495 135380875044672 linears.py:541] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0422 17:41:47.691332 135380875044672 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0422 17:41:47.695230 135380875044672 metric_logger.py:289] number parameters: 1.104 billion I0422 17:42:01.043777 135380875044672 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0422 17:42:01.202047 135380875044672 checkpointing.py:798] Waited 0.15825700759887695 seconds for step 0 to finish before starting checkpointing. I0422 17:42:01.204306 135380875044672 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 17:42:01.206260 135380875044672 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 0 I0422 17:42:01.209192 135380875044672 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/0. I0422 17:42:01.828343 135380875044672 signaling_client.py:364] Using JaxDistributedSignalingClient I0422 17:42:01.829469 135380875044672 jax_array_handlers.py:358] Scheduling D2H of 153 prioritized jax.Array. I0422 17:42:01.829533 135380875044672 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 17:42:02.205430 135380875044672 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.379629s I0422 17:42:02.205603 135380875044672 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 2.283 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.675835371017456 s) (per-host) I0422 17:42:02.205656 135380875044672 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.675897s (batch_requests_ready=0.284281s, total_serialization_initiated=0.391546s, others=0.000070s) I0422 17:42:02.205784 135380875044672 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.680102s (all_items=0.000020s, per_item={'items': '0.00001955'}, temp_paths=0.680083) I0422 17:42:02.206760 135254055442176 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0422 17:42:02.206881 135380875044672 async_checkpointer.py:561] Finished blocking save. Time taken: 1.000542s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/0. I0422 17:42:02.208900 135380875044672 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0422 17:42:02.209233 135251900737280 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 17:42:02.209401 135380875044672 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776879721.2042873, 'wait_for_prev_duration_secs': 6.413459777832031e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776879721.2063055, 'checkpointer_blocking_duration_secs': 1.0006864070892334, 'get_old_steps_start_time': 1776879722.2070131, 'get_old_steps_duration_secs': 2.5033950805664062e-05, 'checkpoint_manager_blocking_start_time': 1776879721.2025723, 'checkpoint_manager_blocking_duration_secs': 1.006788730621338} I0422 17:42:02.209519 135380875044672 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0422 17:42:02.209607 135380875044672 max_utils.py:750] Memstats: After params initialized: I0422 17:42:02.209658 135380875044672 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_24(process=6,(0,6,0,0)) I0422 17:42:02.209690 135380875044672 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_25(process=6,(1,6,0,0)) I0422 17:42:02.209716 135380875044672 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_28(process=6,(0,7,0,0)) I0422 17:42:02.209739 135380875044672 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_29(process=6,(1,7,0,0)) I0422 17:42:02.485997 135380875044672 metric_logger.py:185] completed step: 0, seconds: 13.346, TFLOP/s/device: 1.018, Tokens/s/device: 153.449, total_weights: 65536, loss: 10.872 I0422 17:42:02.667264 135380875044672 metric_logger.py:185] completed step: 1, seconds: 1.439, TFLOP/s/device: 9.441, Tokens/s/device: 1423.005, total_weights: 65536, loss: 10.872 I0422 17:42:03.113007 135380875044672 metric_logger.py:185] completed step: 2, seconds: 0.031, TFLOP/s/device: 432.505, Tokens/s/device: 65191.787, total_weights: 65536, loss: 10.856 I0422 17:42:03.268653 135380875044672 metric_logger.py:185] completed step: 3, seconds: 0.425, TFLOP/s/device: 32.007, Tokens/s/device: 4824.443, total_weights: 65536, loss: 10.824 I0422 17:42:03.580076 135380875044672 metric_logger.py:185] completed step: 4, seconds: 0.182, TFLOP/s/device: 74.559, Tokens/s/device: 11238.298, total_weights: 65536, loss: 10.793 I0422 17:42:03.590343 135380875044672 metric_logger.py:185] completed step: 5, seconds: 0.156, TFLOP/s/device: 87.122, Tokens/s/device: 13131.909, total_weights: 65536, loss: 10.763 I0422 17:42:23.319405 135380875044672 metric_logger.py:185] completed step: 6, seconds: 0.312, TFLOP/s/device: 43.579, Tokens/s/device: 6568.734, total_weights: 65536, loss: 10.733 I0422 17:42:23.474576 135380875044672 metric_logger.py:185] completed step: 7, seconds: 19.577, TFLOP/s/device: 0.694, Tokens/s/device: 104.610, total_weights: 65536, loss: 10.706 I0422 17:42:23.629803 135380875044672 metric_logger.py:185] completed step: 8, seconds: 0.161, TFLOP/s/device: 84.323, Tokens/s/device: 12710.155, total_weights: 65536, loss: 10.680 I0422 17:42:23.638323 135380875044672 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0422 17:42:23.940758 135380875044672 checkpointing.py:798] Waited 0.30240440368652344 seconds for step 10 to finish before starting checkpointing. I0422 17:42:23.943024 135380875044672 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 17:42:24.193331 2741 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0422 17:42:26.660034 135254063834880 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_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/0/items/array_metadatas/process_6 I0422 17:42:41.131937 135254055442176 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 39.891 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 39.60211539268494 s) (per-host) I0422 17:42:41.132047 135254055442176 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 38.925219s. I0422 17:42:49.432482 135254055442176 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 47.225637s. I0422 17:42:49.432797 135251900737280 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 17:42:49.432919 135251900737280 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 17:42:49.432972 135251900737280 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0422 17:42:49.435170 135251900737280 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0422 17:42:49.435350 135380875044672 checkpoint_manager.py:2036] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0422 17:42:49.435485 135380875044672 checkpoint_manager.py:1458] Waiting for previous save to complete took 25.492465 seconds. If this number is high, consider checkpointing less frequently. I0422 17:42:49.437366 135380875044672 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 10 I0422 17:42:49.440750 135380875044672 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/10. I0422 17:42:50.016978 135380875044672 jax_array_handlers.py:358] Scheduling D2H of 153 prioritized jax.Array. I0422 17:42:50.017086 135380875044672 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 17:42:50.065940 135380875044672 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.052531s I0422 17:42:50.066088 135380875044672 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.722 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.3267338275909424 s) (per-host) I0422 17:42:50.066151 135380875044672 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.326805s (batch_requests_ready=0.264552s, total_serialization_initiated=0.062174s, others=0.000079s) I0422 17:42:50.066259 135380875044672 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.330769s (all_items=0.000015s, per_item={'items': '0.00001502'}, temp_paths=0.330754) I0422 17:42:50.067176 135249795217152 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0422 17:42:50.067337 135380875044672 async_checkpointer.py:561] Finished blocking save. Time taken: 0.629899s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/10. I0422 17:42:50.086515 135380875044672 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0422 17:42:50.086841 135251900737280 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 17:42:50.087007 135380875044672 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776879743.942996, 'wait_for_prev_duration_secs': 25.492464542388916, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776879769.437405, 'checkpointer_blocking_duration_secs': 0.6300408840179443, 'get_old_steps_start_time': 1776879770.067466, 'get_old_steps_duration_secs': 2.47955322265625e-05, 'checkpoint_manager_blocking_start_time': 1776879743.9412117, 'checkpoint_manager_blocking_duration_secs': 26.14576029777527} I0422 17:42:50.087141 135380875044672 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0422 17:42:50.087598 135380875044672 metric_logger.py:185] completed step: 9, seconds: 0.155, TFLOP/s/device: 87.485, Tokens/s/device: 13186.700, total_weights: 65536, loss: 10.656 I0422 17:42:50.099590 135380875044672 metric_logger.py:185] completed step: 10, seconds: 0.155, TFLOP/s/device: 87.654, Tokens/s/device: 13212.221, total_weights: 65536, loss: 10.636 I0422 17:42:50.249463 135380875044672 metric_logger.py:185] completed step: 11, seconds: 26.458, TFLOP/s/device: 0.514, Tokens/s/device: 77.405, total_weights: 65536, loss: 10.618 I0422 17:42:50.404525 135380875044672 metric_logger.py:185] completed step: 12, seconds: 0.010, TFLOP/s/device: 1295.246, Tokens/s/device: 195233.556, total_weights: 65536, loss: 10.602 I0422 17:42:50.559858 135380875044672 metric_logger.py:185] completed step: 13, seconds: 0.151, TFLOP/s/device: 90.107, Tokens/s/device: 13581.893, total_weights: 65536, loss: 10.588 I0422 17:42:50.714898 135380875044672 metric_logger.py:185] completed step: 14, seconds: 0.155, TFLOP/s/device: 87.553, Tokens/s/device: 13196.897, total_weights: 65536, loss: 10.577 I0422 17:42:50.869965 135380875044672 metric_logger.py:185] completed step: 15, seconds: 0.155, TFLOP/s/device: 87.503, Tokens/s/device: 13189.333, total_weights: 65536, loss: 10.568 I0422 17:42:51.025429 135380875044672 metric_logger.py:185] completed step: 16, seconds: 0.155, TFLOP/s/device: 87.622, Tokens/s/device: 13207.279, total_weights: 65536, loss: 10.562 I0422 17:42:51.180625 135380875044672 metric_logger.py:185] completed step: 17, seconds: 0.155, TFLOP/s/device: 87.695, Tokens/s/device: 13218.361, total_weights: 65536, loss: 10.556 I0422 17:42:51.335761 135380875044672 metric_logger.py:185] completed step: 18, seconds: 0.156, TFLOP/s/device: 87.358, Tokens/s/device: 13167.624, total_weights: 65536, loss: 10.552 I0422 17:42:51.493001 135380875044672 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0422 17:42:51.494146 135380875044672 checkpointing.py:798] Waited 0.0011551380157470703 seconds for step 19 to finish before starting checkpointing. I0422 17:42:51.496237 135380875044672 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 17:42:55.209668 135252446000896 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_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/10/items/array_metadatas/process_6 I0422 17:43:11.271544 135249795217152 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 73.368 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 21.53213119506836 s) (per-host) I0422 17:43:11.271653 135249795217152 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 21.204365s. I0422 17:43:19.207299 135249795217152 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 29.139994s. I0422 17:43:19.207612 135251900737280 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 17:43:19.207746 135251900737280 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 17:43:19.207805 135251900737280 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0422 17:43:19.209346 135251900737280 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0422 17:43:19.209501 135380875044672 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. W0422 17:43:19.209625 135380875044672 checkpoint_manager.py:1458] Waiting for previous save to complete took 27.713398 seconds. If this number is high, consider checkpointing less frequently. I0422 17:43:19.211298 135380875044672 checkpoint_manager.py:1518] [process=6] Saving checkpoint at step 19 I0422 17:43:19.214792 135380875044672 async_checkpointer.py:452] [process=6] Started async saving checkpoint to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/19. I0422 17:43:19.770443 135380875044672 jax_array_handlers.py:358] Scheduling D2H of 153 prioritized jax.Array. I0422 17:43:19.770550 135380875044672 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0422 17:43:19.820054 135380875044672 base_pytree_checkpoint_handler.py:153] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.053088s I0422 17:43:19.820235 135380875044672 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/blocking_gbytes_per_sec: 4.865 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.31713438034057617 s) (per-host) I0422 17:43:19.820291 135380875044672 base_pytree_checkpoint_handler.py:737] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.317200s (batch_requests_ready=0.254697s, total_serialization_initiated=0.062433s, others=0.000071s) I0422 17:43:19.820406 135380875044672 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.321190s (all_items=0.000010s, per_item={'items': '0.00001025'}, temp_paths=0.321179) I0422 17:43:19.821346 135252454393600 async_checkpointer.py:79] [process=6][thread=async_save] Background save thread started. I0422 17:43:19.821473 135380875044672 async_checkpointer.py:561] Finished blocking save. Time taken: 0.610105s. Continuing background save to gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/19. I0422 17:43:20.288053 135380875044672 checkpoint_manager.py:1566] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0422 17:43:20.288461 135251900737280 async_checkpointer.py:265] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0422 17:43:20.288638 135380875044672 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/nnx_ckpt_xpk_feat_nnx_set_defaults_true_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776879771.49621, 'wait_for_prev_duration_secs': 27.71339774131775, 'time_between_consecutive_saves_sec': 2.061000347137451, 'checkpointer_blocking_start_time': 1776879799.2113378, 'checkpointer_blocking_duration_secs': 0.6103157997131348, 'get_old_steps_start_time': 1776879799.8216677, 'get_old_steps_duration_secs': 2.1219253540039062e-05, 'checkpoint_manager_blocking_start_time': 1776879771.4944563, 'checkpoint_manager_blocking_duration_secs': 28.794148683547974} I0422 17:43:20.288759 135380875044672 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0422 17:43:20.288801 135380875044672 checkpoint_manager.py:2024] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0422 17:43:24.519730 135251909129984 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_20260422_154647/nnx_xpk_feat_nnx_set_defaults_true_20260422_154647_05_fp8/checkpoints/19/items/array_metadatas/process_6 I0422 17:43:40.599963 135252454393600 base_pytree_checkpoint_handler.py:129] [process=6] /jax/checkpoint/write/gbytes_per_sec: 74.881 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 21.09682059288025 s) (per-host) I0422 17:43:40.600086 135252454393600 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 20.778529s. I0422 17:43:49.044733 135252454393600 async_checkpointer.py:144] [process=6][thread=async_save] Background save thread done. Time taken: 29.223159s. I0422 17:43:49.045040 135251900737280 async_checkpointer.py:273] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0422 17:43:49.045183 135251900737280 async_checkpointer.py:283] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0422 17:43:49.045226 135251900737280 checkpoint_manager.py:2133] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0422 17:43:49.046592 135251900737280 checkpoint_manager.py:2142] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0422 17:43:49.046761 135380875044672 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. I0422 17:43:49.046941 135380875044672 checkpoint_manager.py:2013] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0422 17:43:49.047613 135380875044672 metric_logger.py:185] completed step: 19, seconds: 0.155, TFLOP/s/device: 87.594, Tokens/s/device: 13203.192, 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: Wed Apr 22 17:43:58 UTC 2026 EXIT_CODE=0