main (147168c3a) vs feat/nnx-trainstate-and-training-loop (190761531)| Metric | main 147168c3a | feat/nnx-trainstate-and-training-loop 190761531 | Diff (feat/nnx-trainstate-and-training-loop − main) |
|---|---|---|---|
| Parameters | 1.105 billion | 1.105 billion | — |
| Final loss | 6.1670 | 6.1670 | 0 |
| TFLOP/s | 85.851 | 85.959 | +0.108 |
| Tok/s | 12940.4 | 12956.6 | +16.21 |
| Avg s/step | 2.025 | 1.831 | -0.194 |
| Memory % | 1.44 | 1.44 | 0 |
| JAX | 0.9.2 | 0.9.2 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Sat Apr 18 00:17:16 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-18 00:17:41.435781: 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) I0418 00:17:41.616682 139776043521856 max_utils.py:273] Attempting to initialize the jax distributed system... I0418 00:17:50.658643 139776043521856 distributed.py:149] Starting JAX distributed service on [::]:8482 I0418 00:17:50.661011 139776043521856 distributed.py:172] Connecting to JAX distributed service on mt-05-fp8-9rggu-slice-job-0-0.mt-05-fp8-9rggu:8482 I0418 00:17:51.648971 139776043521856 max_utils.py:284] Jax distributed system initialized! I0418 00:17:56.743682 139776043521856 max_utils.py:800] System Information: Jax Version: 0.9.2 I0418 00:17:56.743784 139776043521856 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0418 00:17:56.743840 139776043521856 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0418 00:17:56.743880 139776043521856 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0418 00:17:57.451745 139776043521856 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0418 00:17:57.452049 139776043521856 checkpointing.py:677] Setting up checkpoint logger... I0418 00:17:57.452111 139776043521856 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0418 00:17:57.452156 139776043521856 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0418 00:17:57.452491 139776043521856 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7f1f93718590>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0418 00:18:02.152663 139776043521856 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0418 00:18:02.152915 139776043521856 checkpoint_manager.py:708] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f0b08306930>}, handler_registry=None I0418 00:18:02.153156 139776043521856 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f0b08306930>` 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`. I0418 00:18:02.153204 139776043521856 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f0b0830a210>` 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`. I0418 00:18:02.153239 139776043521856 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 0x7f0b08306930>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f0b08306930>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f0b0830a210>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f0b0830a210>}). I0418 00:18:02.153564 139776043521856 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0418 00:18:02.153636 139776043521856 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f0ac0697ec0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0418 00:18:03.302314 139776043521856 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_05_fp8/checkpoints I0418 00:18:03.700868 139776043521856 checkpoint_manager.py:929] [process=6][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f0b08304d70> I0418 00:18:03.701039 139776043521856 checkpointing.py:301] Checkpoint manager created! I0418 00:18:04.793488 139776043521856 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0418 00:18:04.793600 139776043521856 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 00:18:05.180224 139776043521856 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 00:18:05.180314 139776043521856 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 00:18:05.196790 139776043521856 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 00:18:05.196859 139776043521856 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 00:18:05.249473 139776043521856 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 00:18:05.249555 139776043521856 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 00:18:05.265985 139776043521856 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 00:18:05.266045 139776043521856 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 00:18:05.282385 139776043521856 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 00:18:05.282453 139776043521856 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 00:18:05.307911 139776043521856 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0418 00:18:05.307981 139776043521856 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 00:18:05.359896 139776043521856 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0418 00:18:05.359972 139776043521856 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0418 00:18:05.809650 139776043521856 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0418 00:18:05.809766 139776043521856 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0418 00:18:07.759493 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.759630 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.759679 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.759719 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.759756 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.759792 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.759906 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.759958 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.759995 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760028 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760060 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760089 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760118 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760147 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760176 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760205 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760236 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760266 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760294 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760322 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760349 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760375 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760403 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760431 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760458 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760485 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760512 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760545 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760572 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760599 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760628 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760656 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760683 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760710 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760738 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760765 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760795 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760841 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760879 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760908 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.760935 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 00:18:07.760962 139776043521856 maxtext_utils.py:1654] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: P() Physical: () I0418 00:18:07.761015 139776043521856 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0418 00:18:07.761069 139776043521856 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0418 00:18:07.761105 139776043521856 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0418 00:18:07.761153 139776043521856 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0418 00:18:07.761198 139776043521856 maxtext_utils.py:1654] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0418 00:18:07.761231 139776043521856 maxtext_utils.py:1654] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0418 00:18:07.761286 139776043521856 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0418 00:18:07.761348 139776043521856 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0418 00:18:07.761385 139776043521856 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0418 00:18:07.761419 139776043521856 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0418 00:18:07.761463 139776043521856 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0418 00:18:07.761508 139776043521856 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0418 00:18:09.406273 139776043521856 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 00:18:09.406368 139776043521856 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0418 00:18:09.421785 139776043521856 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 00:18:09.421855 139776043521856 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0418 00:18:23.224786 139776043521856 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. I0418 00:18:23.225640 139776043521856 metric_logger.py:301] number parameters: 1.105 billion I0418 00:18:38.567506 139776043521856 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0418 00:18:38.735745 139776043521856 checkpointing.py:776] Waited 0.16821956634521484 seconds for step 0 to finish before starting checkpointing. I0418 00:18:38.738163 139776043521856 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 00:18:38.740303 139776043521856 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0418 00:18:38.742039 139776043521856 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_05_fp8/checkpoints/0. I0418 00:18:39.520893 139776043521856 signaling_client.py:364] Using JaxDistributedSignalingClient I0418 00:18:39.522058 139776043521856 jax_array_handlers.py:360] Scheduling D2H of 81 prioritized jax.Array. I0418 00:18:39.522179 139776043521856 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 I0418 00:18:39.931248 139776043521856 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.410490s I0418 00:18:39.931428 139776043521856 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 3.681 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.41915273666381836 s) (per-host) I0418 00:18:39.931482 139776043521856 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.419217s (batch_requests_ready=0.003743s, total_serialization_initiated=0.415401s, others=0.000073s) I0418 00:18:39.931589 139776043521856 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.423580s (all_items=0.000017s, per_item={'items': '0.00001717'}, temp_paths=0.423563) I0418 00:18:39.932509 139776043521856 event_tracking.py:125] [process=6] [async] Finished blocking save in 1.19 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_05_fp8/checkpoints/0. I0418 00:18:39.932793 139649536980736 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 00:38:39.932769 I0418 00:18:39.941254 139776043521856 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0418 00:18:39.941560 139649024509696 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0418 00:18:39.941718 139776043521856 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776471518.738145, 'wait_for_prev_duration_secs': 6.318092346191406e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776471518.7403445, 'checkpointer_blocking_duration_secs': 1.1925342082977295, 'get_old_steps_start_time': 1776471519.932903, 'get_old_steps_duration_secs': 2.9325485229492188e-05, 'checkpoint_manager_blocking_start_time': 1776471518.7363977, 'checkpoint_manager_blocking_duration_secs': 1.2052812576293945} I0418 00:18:39.941839 139776043521856 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0418 00:18:39.941892 139776043521856 max_utils.py:750] Memstats: After params initialized: I0418 00:18:39.941944 139776043521856 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_24(process=6,(0,6,0,0)) I0418 00:18:39.941981 139776043521856 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_25(process=6,(1,6,0,0)) I0418 00:18:39.942011 139776043521856 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_28(process=6,(0,7,0,0)) I0418 00:18:39.942039 139776043521856 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_29(process=6,(1,7,0,0)) I0418 00:18:40.266412 139776043521856 metric_logger.py:196] completed step: 0, seconds: 15.342, TFLOP/s/device: 0.886, Tokens/s/device: 133.492, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52776.805 I0418 00:18:40.447726 139776043521856 metric_logger.py:196] completed step: 1, seconds: 1.697, TFLOP/s/device: 8.007, Tokens/s/device: 1206.937, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52761.840 I0418 00:18:41.271041 139776043521856 metric_logger.py:196] completed step: 2, seconds: 0.024, TFLOP/s/device: 567.289, Tokens/s/device: 85507.912, total_weights: 65536, loss: 10.816, lm_loss: 10.816, perplexity: 49832.398 I0418 00:18:41.429301 139776043521856 metric_logger.py:196] completed step: 3, seconds: 0.824, TFLOP/s/device: 16.485, Tokens/s/device: 2484.813, total_weights: 65536, loss: 10.431, lm_loss: 10.431, perplexity: 33901.820 I0418 00:18:41.746908 139776043521856 metric_logger.py:196] completed step: 4, seconds: 0.163, TFLOP/s/device: 83.306, Tokens/s/device: 12556.791, total_weights: 65536, loss: 9.992, lm_loss: 9.992, perplexity: 21847.639 I0418 00:18:41.753379 139776043521856 metric_logger.py:196] completed step: 5, seconds: 0.158, TFLOP/s/device: 85.894, Tokens/s/device: 12946.866, total_weights: 65536, loss: 9.549, lm_loss: 9.549, perplexity: 14035.006 I0418 00:18:43.964264 2781 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0418 00:18:47.287399 139649045477120 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_05_fp8/checkpoints/0/items/array_metadatas/process_6 I0418 00:18:58.926844 139649536980736 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 81.370 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 19.41453766822815 s) (per-host) I0418 00:18:58.926963 139649536980736 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 18.993528s. I0418 00:19:07.833441 139776043521856 metric_logger.py:196] completed step: 6, seconds: 0.318, TFLOP/s/device: 42.690, Tokens/s/device: 6434.748, total_weights: 65536, loss: 9.111, lm_loss: 9.111, perplexity: 9058.688 I0418 00:19:07.991850 139776043521856 metric_logger.py:196] completed step: 7, seconds: 25.923, TFLOP/s/device: 0.524, Tokens/s/device: 79.003, total_weights: 65536, loss: 8.685, lm_loss: 8.685, perplexity: 5914.720 I0418 00:19:08.150535 139776043521856 metric_logger.py:196] completed step: 8, seconds: 0.163, TFLOP/s/device: 83.385, Tokens/s/device: 12568.658, total_weights: 65536, loss: 8.281, lm_loss: 8.281, perplexity: 3950.010 I0418 00:19:08.155360 139776043521856 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0418 00:19:08.467204 139776043521856 checkpointing.py:776] Waited 0.31180596351623535 seconds for step 10 to finish before starting checkpointing. I0418 00:19:08.469851 139776043521856 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 00:19:14.333031 139649536980736 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 34.399579s. I0418 00:19:14.333303 139649024509696 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 00:19:14.333426 139649024509696 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0418 00:19:14.333492 139649024509696 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0418 00:19:14.336499 139649024509696 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0418 00:19:14.336686 139776043521856 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0418 00:19:14.336834 139776043521856 checkpoint_manager.py:1452] Waiting for previous save to complete took 5.866989 seconds. If this number is high, consider checkpointing less frequently. I0418 00:19:14.338575 139776043521856 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 10 I0418 00:19:14.340567 139776043521856 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_05_fp8/checkpoints/10. I0418 00:19:15.487345 139776043521856 jax_array_handlers.py:360] Scheduling D2H of 81 prioritized jax.Array. I0418 00:19:15.487438 139776043521856 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 I0418 00:19:15.525527 139776043521856 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.039259s I0418 00:19:15.525666 139776043521856 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 33.859 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04556417465209961 s) (per-host) I0418 00:19:15.525716 139776043521856 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.045623s (batch_requests_ready=0.003195s, total_serialization_initiated=0.042365s, others=0.000064s) I0418 00:19:15.525801 139776043521856 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.049784s (all_items=0.000015s, per_item={'items': '0.00001502'}, temp_paths=0.049769) I0418 00:19:15.526610 139776043521856 event_tracking.py:125] [process=6] [async] Finished blocking save in 1.19 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_05_fp8/checkpoints/10. I0418 00:19:15.526947 139649024509696 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 00:39:15.526910 I0418 00:19:15.530620 139776043521856 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0418 00:19:15.530939 139646348547840 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0418 00:19:15.531099 139776043521856 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776471548.4698038, 'wait_for_prev_duration_secs': 5.866988658905029, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776471554.338614, 'checkpointer_blocking_duration_secs': 1.188474178314209, 'get_old_steps_start_time': 1776471555.5271106, 'get_old_steps_duration_secs': 3.0279159545898438e-05, 'checkpoint_manager_blocking_start_time': 1776471548.4679968, 'checkpoint_manager_blocking_duration_secs': 7.063067436218262} I0418 00:19:15.531226 139776043521856 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0418 00:19:15.531833 139776043521856 metric_logger.py:196] completed step: 9, seconds: 0.158, TFLOP/s/device: 85.761, Tokens/s/device: 12926.845, total_weights: 65536, loss: 7.908, lm_loss: 7.908, perplexity: 2720.116 I0418 00:19:15.539123 139776043521856 metric_logger.py:196] completed step: 10, seconds: 0.159, TFLOP/s/device: 85.658, Tokens/s/device: 12911.279, total_weights: 65536, loss: 7.568, lm_loss: 7.568, perplexity: 1936.198 I0418 00:19:15.698077 139776043521856 metric_logger.py:196] completed step: 11, seconds: 7.381, TFLOP/s/device: 1.841, Tokens/s/device: 277.452, total_weights: 65536, loss: 7.288, lm_loss: 7.288, perplexity: 1463.272 I0418 00:19:15.856416 139776043521856 metric_logger.py:196] completed step: 12, seconds: 0.006, TFLOP/s/device: 2128.310, Tokens/s/device: 320802.005, total_weights: 65536, loss: 7.032, lm_loss: 7.032, perplexity: 1132.061 I0418 00:19:16.014558 139776043521856 metric_logger.py:196] completed step: 13, seconds: 0.160, TFLOP/s/device: 84.667, Tokens/s/device: 12761.953, total_weights: 65536, loss: 6.815, lm_loss: 6.815, perplexity: 911.031 I0418 00:19:16.572937 139776043521856 metric_logger.py:196] completed step: 14, seconds: 0.158, TFLOP/s/device: 86.190, Tokens/s/device: 12991.544, total_weights: 65536, loss: 6.635, lm_loss: 6.635, perplexity: 761.080 I0418 00:19:16.731409 139776043521856 metric_logger.py:196] completed step: 15, seconds: 0.553, TFLOP/s/device: 24.564, Tokens/s/device: 3702.505, total_weights: 65536, loss: 6.492, lm_loss: 6.492, perplexity: 659.915 I0418 00:19:16.889742 139776043521856 metric_logger.py:196] completed step: 16, seconds: 0.163, TFLOP/s/device: 83.143, Tokens/s/device: 12532.202, total_weights: 65536, loss: 6.380, lm_loss: 6.380, perplexity: 589.948 I0418 00:19:17.047995 139776043521856 metric_logger.py:196] completed step: 17, seconds: 0.158, TFLOP/s/device: 85.815, Tokens/s/device: 12935.009, total_weights: 65536, loss: 6.293, lm_loss: 6.293, perplexity: 540.941 I0418 00:19:17.206132 139776043521856 metric_logger.py:196] completed step: 18, seconds: 0.158, TFLOP/s/device: 85.746, Tokens/s/device: 12924.560, total_weights: 65536, loss: 6.223, lm_loss: 6.223, perplexity: 504.108 I0418 00:19:17.363863 139776043521856 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0418 00:19:17.364865 139776043521856 checkpointing.py:776] Waited 0.0010554790496826172 seconds for step 19 to finish before starting checkpointing. I0418 00:19:17.366906 139776043521856 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 00:19:25.653768 139646314977024 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_05_fp8/checkpoints/10/items/array_metadatas/process_6 I0418 00:19:58.810022 139649024509696 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 36.459 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.32988476753235 s) (per-host) I0418 00:19:58.810128 139649024509696 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.283070s. I0418 00:20:12.278180 139649024509696 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 56.751107s. I0418 00:20:12.278461 139646348547840 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 00:20:12.278599 139646348547840 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0418 00:20:12.278650 139646348547840 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0418 00:20:12.281585 139646348547840 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0418 00:20:12.281808 139776043521856 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0418 00:20:12.281945 139776043521856 checkpoint_manager.py:1452] Waiting for previous save to complete took 54.915051 seconds. If this number is high, consider checkpointing less frequently. I0418 00:20:12.283765 139776043521856 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 19 I0418 00:20:12.285813 139776043521856 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_05_fp8/checkpoints/19. I0418 00:20:13.009186 139776043521856 jax_array_handlers.py:360] Scheduling D2H of 81 prioritized jax.Array. I0418 00:20:13.009279 139776043521856 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 I0418 00:20:13.048366 139776043521856 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.040230s I0418 00:20:13.048535 139776043521856 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 33.174 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04650425910949707 s) (per-host) I0418 00:20:13.048585 139776043521856 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.046565s (batch_requests_ready=0.003141s, total_serialization_initiated=0.043358s, others=0.000066s) I0418 00:20:13.048692 139776043521856 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.050747s (all_items=0.000012s, per_item={'items': '0.00001168'}, temp_paths=0.050736) I0418 00:20:13.049446 139776043521856 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.77 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_05_fp8/checkpoints/19. I0418 00:20:13.049734 139646348547840 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 00:40:13.049700 I0418 00:20:13.052493 139776043521856 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0418 00:20:13.052742 139644201064192 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0418 00:20:13.052900 139776043521856 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776471557.3668709, 'wait_for_prev_duration_secs': 54.9150505065918, 'time_between_consecutive_saves_sec': 3.030329465866089, 'checkpointer_blocking_start_time': 1776471612.2838047, 'checkpointer_blocking_duration_secs': 0.7660825252532959, 'get_old_steps_start_time': 1776471613.0499046, 'get_old_steps_duration_secs': 2.6941299438476562e-05, 'checkpoint_manager_blocking_start_time': 1776471557.3651414, 'checkpoint_manager_blocking_duration_secs': 55.687726974487305} I0418 00:20:13.053010 139776043521856 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0418 00:20:13.053053 139776043521856 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 00:20:19.809404 139646314977024 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260417_235819/linen_xpk_main_20260417_235819_05_fp8/checkpoints/19/items/array_metadatas/process_6 I0418 00:20:56.231195 139646348547840 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 36.544 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.229124546051025 s) (per-host) I0418 00:20:56.231318 139646348547840 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.181459s. I0418 00:21:07.337360 139646348547840 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 54.287484s. I0418 00:21:07.337656 139644201064192 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 00:21:07.337782 139644201064192 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0418 00:21:07.337841 139644201064192 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0418 00:21:07.339673 139644201064192 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0418 00:21:07.339867 139776043521856 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0418 00:21:07.340015 139776043521856 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 00:21:07.341026 139776043521856 metric_logger.py:196] completed step: 19, seconds: 0.158, TFLOP/s/device: 85.851, Tokens/s/device: 12940.403, total_weights: 65536, loss: 6.167, lm_loss: 6.167, perplexity: 476.947 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 18 00:21:18 UTC 2026 EXIT_CODE=0
190761531
· feat_nnx_trainstate_and_training_loop_20260418_060141
· full log
XPK Start: Sat Apr 18 06:18:30 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-18 06:18:54.991481: 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) I0418 06:18:55.170172 135161681655616 max_utils.py:273] Attempting to initialize the jax distributed system... I0418 06:19:04.212244 135161681655616 distributed.py:149] Starting JAX distributed service on [::]:8482 I0418 06:19:04.214720 135161681655616 distributed.py:172] Connecting to JAX distributed service on mt-05-fp8-yf4nh-slice-job-0-0.mt-05-fp8-yf4nh:8482 I0418 06:19:06.210536 135161681655616 max_utils.py:284] Jax distributed system initialized! I0418 06:19:12.304371 135161681655616 max_utils.py:800] System Information: Jax Version: 0.9.2 I0418 06:19:12.304475 135161681655616 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0418 06:19:12.304516 135161681655616 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0418 06:19:12.304550 135161681655616 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0418 06:19:13.020353 135161681655616 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0418 06:19:13.020947 135161681655616 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0418 06:19:13.021140 135161681655616 checkpointing.py:688] Setting up checkpoint logger... I0418 06:19:13.021189 135161681655616 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0418 06:19:13.021236 135161681655616 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0418 06:19:13.021563 135161681655616 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7aed2f38a9f0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0418 06:19:16.367126 135161681655616 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0418 06:19:16.367370 135161681655616 checkpoint_manager.py:708] [process=3][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ad8cc639dc0>}, handler_registry=None I0418 06:19:16.367605 135161681655616 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ad8cc639dc0>` 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`. I0418 06:19:16.367667 135161681655616 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ad8cc63e5a0>` 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`. I0418 06:19:16.367702 135161681655616 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 0x7ad8cc639dc0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7ad8cc639dc0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ad8cc63e5a0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7ad8cc63e5a0>}). I0418 06:19:16.368005 135161681655616 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0418 06:19:16.368073 135161681655616 async_checkpointer.py:192] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7ad8cc3d82c0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0418 06:19:17.911777 135161681655616 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141/linen_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141_05_fp8/checkpoints I0418 06:19:18.761210 135161681655616 checkpoint_manager.py:929] [process=3][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141/linen_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7ad8cc63a0c0> I0418 06:19:18.761392 135161681655616 checkpointing.py:302] Checkpoint manager created! I0418 06:19:19.857199 135161681655616 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0418 06:19:19.857308 135161681655616 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 06:19:20.240718 135161681655616 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 06:19:20.240809 135161681655616 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 06:19:20.257245 135161681655616 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 06:19:20.257302 135161681655616 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 06:19:20.309743 135161681655616 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 06:19:20.309822 135161681655616 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 06:19:20.326358 135161681655616 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 06:19:20.326416 135161681655616 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 06:19:20.342863 135161681655616 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 06:19:20.342923 135161681655616 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 06:19:20.368211 135161681655616 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0418 06:19:20.368281 135161681655616 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 06:19:20.419598 135161681655616 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0418 06:19:20.419685 135161681655616 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0418 06:19:20.861510 135161681655616 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0418 06:19:20.861742 135161681655616 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0418 06:19:22.775090 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.775205 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.775252 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.775291 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.775327 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.775360 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.775390 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.775420 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.775451 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.775480 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.775509 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.775538 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.775566 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.775593 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.775633 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.775666 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.775695 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.775723 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.775751 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.775778 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.775804 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.775830 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.775857 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.775884 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.775911 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.775937 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.775967 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.775995 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.776021 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.776047 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.776073 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.776101 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.776127 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.776153 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.776178 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.776204 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.776246 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.776278 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.776305 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.776332 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.776358 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: P() Physical: () I0418 06:19:22.776384 135161681655616 maxtext_utils.py:1821] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: P() Physical: () I0418 06:19:22.776432 135161681655616 maxtext_utils.py:1821] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0418 06:19:22.776484 135161681655616 maxtext_utils.py:1821] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0418 06:19:22.776520 135161681655616 maxtext_utils.py:1821] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0418 06:19:22.776565 135161681655616 maxtext_utils.py:1821] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0418 06:19:22.776610 135161681655616 maxtext_utils.py:1821] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0418 06:19:22.776656 135161681655616 maxtext_utils.py:1821] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0418 06:19:22.776704 135161681655616 maxtext_utils.py:1821] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0418 06:19:22.776751 135161681655616 maxtext_utils.py:1821] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0418 06:19:22.776785 135161681655616 maxtext_utils.py:1821] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0418 06:19:22.776817 135161681655616 maxtext_utils.py:1821] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0418 06:19:22.776860 135161681655616 maxtext_utils.py:1821] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0418 06:19:22.776901 135161681655616 maxtext_utils.py:1821] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0418 06:19:24.414138 135161681655616 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 06:19:24.414236 135161681655616 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0418 06:19:24.430007 135161681655616 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 06:19:24.430068 135161681655616 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0418 06:19:38.117869 135161681655616 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. I0418 06:19:38.118895 135161681655616 metric_logger.py:301] number parameters: 1.105 billion I0418 06:19:53.396734 135161681655616 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0418 06:19:53.567125 135161681655616 checkpointing.py:798] Waited 0.17037630081176758 seconds for step 0 to finish before starting checkpointing. I0418 06:19:53.569950 135161681655616 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 06:19:53.572006 135161681655616 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 0 I0418 06:19:53.573519 135161681655616 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141/linen_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141_05_fp8/checkpoints/0. I0418 06:19:54.316666 135161681655616 signaling_client.py:364] Using JaxDistributedSignalingClient I0418 06:19:54.317649 135161681655616 jax_array_handlers.py:360] Scheduling D2H of 81 prioritized jax.Array. I0418 06:19:54.317765 135161681655616 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 I0418 06:19:54.745546 135161681655616 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.429012s I0418 06:19:54.745759 135161681655616 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 3.529 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.437150239944458 s) (per-host) I0418 06:19:54.745815 135161681655616 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.437217s (batch_requests_ready=0.003416s, total_serialization_initiated=0.433727s, others=0.000075s) I0418 06:19:54.745924 135161681655616 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.441589s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=0.441572) I0418 06:19:54.746828 135161681655616 event_tracking.py:125] [process=3] [async] Finished blocking save in 1.17 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141/linen_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141_05_fp8/checkpoints/0. I0418 06:19:54.747157 135031985125120 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 06:39:54.747124 I0418 06:19:54.749273 135161681655616 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0418 06:19:54.749578 135031462799104 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0418 06:19:54.749752 135161681655616 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141/linen_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776493193.5699313, 'wait_for_prev_duration_secs': 6.413459777832031e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776493193.5720518, 'checkpointer_blocking_duration_secs': 1.1752653121948242, 'get_old_steps_start_time': 1776493194.7473435, 'get_old_steps_duration_secs': 3.0517578125e-05, 'checkpoint_manager_blocking_start_time': 1776493193.5677521, 'checkpoint_manager_blocking_duration_secs': 1.1819636821746826} I0418 06:19:54.749867 135161681655616 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0418 06:19:54.749919 135161681655616 max_utils.py:750] Memstats: After params initialized: I0418 06:19:54.749969 135161681655616 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_10(process=3,(2,2,0,0)) I0418 06:19:54.750003 135161681655616 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_11(process=3,(3,2,0,0)) I0418 06:19:54.750030 135161681655616 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_14(process=3,(2,3,0,0)) I0418 06:19:54.750060 135161681655616 max_utils.py:756] Using (GB) 0.45 / 31.25 (1.440000%) on TPU_15(process=3,(3,3,0,0)) I0418 06:19:55.070693 135161681655616 metric_logger.py:196] completed step: 0, seconds: 15.278, TFLOP/s/device: 0.889, Tokens/s/device: 134.052, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52776.805 I0418 06:19:55.258723 135161681655616 metric_logger.py:196] completed step: 1, seconds: 1.672, TFLOP/s/device: 8.127, Tokens/s/device: 1225.000, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52761.840 I0418 06:19:55.684917 135161681655616 metric_logger.py:196] completed step: 2, seconds: 0.030, TFLOP/s/device: 446.681, Tokens/s/device: 67328.555, total_weights: 65536, loss: 10.816, lm_loss: 10.816, perplexity: 49832.398 I0418 06:19:55.843052 135161681655616 metric_logger.py:196] completed step: 3, seconds: 0.426, TFLOP/s/device: 31.878, Tokens/s/device: 4804.951, total_weights: 65536, loss: 10.431, lm_loss: 10.431, perplexity: 33901.820 I0418 06:19:56.160627 135161681655616 metric_logger.py:196] completed step: 4, seconds: 0.165, TFLOP/s/device: 82.321, Tokens/s/device: 12408.361, total_weights: 65536, loss: 9.992, lm_loss: 9.992, perplexity: 21847.639 I0418 06:19:56.166839 135161681655616 metric_logger.py:196] completed step: 5, seconds: 0.157, TFLOP/s/device: 86.318, Tokens/s/device: 13010.857, total_weights: 65536, loss: 9.549, lm_loss: 9.549, perplexity: 14035.006 I0418 06:19:58.599222 2818 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0418 06:20:01.488866 135031471191808 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141/linen_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141_05_fp8/checkpoints/0/items/array_metadatas/process_3 I0418 06:20:15.841762 135031985125120 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 73.364 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 21.533114671707153 s) (per-host) I0418 06:20:15.841885 135031985125120 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 21.094614s. I0418 06:20:21.480381 135161681655616 metric_logger.py:196] completed step: 6, seconds: 0.318, TFLOP/s/device: 42.741, Tokens/s/device: 6442.359, total_weights: 65536, loss: 9.111, lm_loss: 9.111, perplexity: 9058.688 I0418 06:20:21.638868 135161681655616 metric_logger.py:196] completed step: 7, seconds: 25.156, TFLOP/s/device: 0.540, Tokens/s/device: 81.411, total_weights: 65536, loss: 8.685, lm_loss: 8.685, perplexity: 5914.720 I0418 06:20:21.797776 135161681655616 metric_logger.py:196] completed step: 8, seconds: 0.163, TFLOP/s/device: 83.205, Tokens/s/device: 12541.565, total_weights: 65536, loss: 8.281, lm_loss: 8.281, perplexity: 3950.010 I0418 06:20:21.802963 135161681655616 checkpointing.py:794] Waiting for step 10 to finish before checkpoint... I0418 06:20:22.114559 135161681655616 checkpointing.py:798] Waited 0.31155872344970703 seconds for step 10 to finish before starting checkpointing. I0418 06:20:22.117468 135161681655616 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 06:20:25.463678 135031985125120 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 30.716391s. I0418 06:20:25.463946 135031462799104 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 06:20:25.464086 135031462799104 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0418 06:20:25.464138 135031462799104 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0418 06:20:25.465829 135031462799104 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0418 06:20:25.466002 135161681655616 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0418 06:20:25.466144 135161681655616 checkpoint_manager.py:1452] Waiting for previous save to complete took 3.348676 seconds. If this number is high, consider checkpointing less frequently. I0418 06:20:25.467920 135161681655616 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 10 I0418 06:20:25.469985 135161681655616 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141/linen_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141_05_fp8/checkpoints/10. I0418 06:20:26.625518 135161681655616 jax_array_handlers.py:360] Scheduling D2H of 81 prioritized jax.Array. I0418 06:20:26.625628 135161681655616 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 I0418 06:20:26.679342 135161681655616 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.054748s I0418 06:20:26.679505 135161681655616 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 25.358 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.0608372688293457 s) (per-host) I0418 06:20:26.679553 135161681655616 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.060895s (batch_requests_ready=0.003041s, total_serialization_initiated=0.057789s, others=0.000065s) I0418 06:20:26.679670 135161681655616 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.065068s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.065051) I0418 06:20:26.680332 135161681655616 event_tracking.py:125] [process=3] [async] Finished blocking save in 1.21 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141/linen_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141_05_fp8/checkpoints/10. I0418 06:20:26.680696 135031462799104 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 06:40:26.680658 I0418 06:20:26.682729 135161681655616 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0418 06:20:26.683025 135026649851648 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0418 06:20:26.683188 135161681655616 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141/linen_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776493222.1174383, 'wait_for_prev_duration_secs': 3.3486762046813965, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776493225.4679606, 'checkpointer_blocking_duration_secs': 1.2128863334655762, 'get_old_steps_start_time': 1776493226.6808684, 'get_old_steps_duration_secs': 3.0517578125e-05, 'checkpoint_manager_blocking_start_time': 1776493222.11545, 'checkpoint_manager_blocking_duration_secs': 4.567705154418945} I0418 06:20:26.683300 135161681655616 checkpointing.py:409] Started an asynchronous checkpoint save for step 10 I0418 06:20:26.683932 135161681655616 metric_logger.py:196] completed step: 9, seconds: 0.159, TFLOP/s/device: 85.589, Tokens/s/device: 12900.950, total_weights: 65536, loss: 7.908, lm_loss: 7.908, perplexity: 2720.116 I0418 06:20:26.691419 135161681655616 metric_logger.py:196] completed step: 10, seconds: 0.159, TFLOP/s/device: 85.588, Tokens/s/device: 12900.787, total_weights: 65536, loss: 7.568, lm_loss: 7.568, perplexity: 1936.198 I0418 06:20:26.849731 135161681655616 metric_logger.py:196] completed step: 11, seconds: 4.886, TFLOP/s/device: 2.781, Tokens/s/device: 419.144, total_weights: 65536, loss: 7.288, lm_loss: 7.288, perplexity: 1463.272 I0418 06:20:27.008130 135161681655616 metric_logger.py:196] completed step: 12, seconds: 0.007, TFLOP/s/device: 2085.515, Tokens/s/device: 314351.497, total_weights: 65536, loss: 7.032, lm_loss: 7.032, perplexity: 1132.061 I0418 06:20:27.166672 135161681655616 metric_logger.py:196] completed step: 13, seconds: 0.159, TFLOP/s/device: 85.369, Tokens/s/device: 12867.716, total_weights: 65536, loss: 6.815, lm_loss: 6.815, perplexity: 911.031 I0418 06:20:27.715552 135161681655616 metric_logger.py:196] completed step: 14, seconds: 0.158, TFLOP/s/device: 85.775, Tokens/s/device: 12928.885, total_weights: 65536, loss: 6.635, lm_loss: 6.635, perplexity: 761.080 I0418 06:20:27.873968 135161681655616 metric_logger.py:196] completed step: 15, seconds: 0.544, TFLOP/s/device: 24.976, Tokens/s/device: 3764.685, total_weights: 65536, loss: 6.492, lm_loss: 6.492, perplexity: 659.915 I0418 06:20:28.032277 135161681655616 metric_logger.py:196] completed step: 16, seconds: 0.163, TFLOP/s/device: 83.132, Tokens/s/device: 12530.592, total_weights: 65536, loss: 6.380, lm_loss: 6.380, perplexity: 589.948 I0418 06:20:28.190550 135161681655616 metric_logger.py:196] completed step: 17, seconds: 0.158, TFLOP/s/device: 85.742, Tokens/s/device: 12923.990, total_weights: 65536, loss: 6.293, lm_loss: 6.293, perplexity: 540.941 I0418 06:20:28.348838 135161681655616 metric_logger.py:196] completed step: 18, seconds: 0.158, TFLOP/s/device: 85.748, Tokens/s/device: 12924.887, total_weights: 65536, loss: 6.223, lm_loss: 6.223, perplexity: 504.108 I0418 06:20:28.506333 135161681655616 checkpointing.py:794] Waiting for step 19 to finish before checkpoint... I0418 06:20:28.507366 135161681655616 checkpointing.py:798] Waited 0.0010519027709960938 seconds for step 19 to finish before starting checkpointing. I0418 06:20:28.509634 135161681655616 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 06:20:34.021425 135029871077120 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141/linen_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141_05_fp8/checkpoints/10/items/array_metadatas/process_3 I0418 06:21:09.706747 135031462799104 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 36.664 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.088019132614136 s) (per-host) I0418 06:21:09.706868 135031462799104 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.026053s. I0418 06:21:20.432472 135031462799104 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 53.751642s. I0418 06:21:20.432782 135026649851648 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 06:21:20.432927 135026649851648 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0418 06:21:20.432985 135026649851648 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0418 06:21:20.436120 135026649851648 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0418 06:21:20.436286 135161681655616 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0418 06:21:20.436408 135161681655616 checkpoint_manager.py:1452] Waiting for previous save to complete took 51.926797 seconds. If this number is high, consider checkpointing less frequently. I0418 06:21:20.437901 135161681655616 checkpoint_manager.py:1512] [process=3] Saving checkpoint at step 19 I0418 06:21:20.439961 135161681655616 event_tracking.py:70] [process=3] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141/linen_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141_05_fp8/checkpoints/19. I0418 06:21:21.136364 135161681655616 jax_array_handlers.py:360] Scheduling D2H of 81 prioritized jax.Array. I0418 06:21:21.136459 135161681655616 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 I0418 06:21:21.189273 135161681655616 base_pytree_checkpoint_handler.py:154] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.053900s I0418 06:21:21.189441 135161681655616 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/blocking_gbytes_per_sec: 25.671 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.06009650230407715 s) (per-host) I0418 06:21:21.189491 135161681655616 base_pytree_checkpoint_handler.py:768] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.060156s (batch_requests_ready=0.003012s, total_serialization_initiated=0.057080s, others=0.000064s) I0418 06:21:21.189591 135161681655616 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.064880s (all_items=0.000011s, per_item={'items': '0.00001073'}, temp_paths=0.064870) I0418 06:21:21.190346 135161681655616 event_tracking.py:125] [process=3] [async] Finished blocking save in 0.75 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141/linen_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141_05_fp8/checkpoints/19. I0418 06:21:21.190633 135026649851648 async_checkpointer.py:76] [process=3][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 06:41:21.190588 I0418 06:21:21.192677 135161681655616 checkpoint_manager.py:1560] [process=3][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0418 06:21:21.192956 135029871077120 async_checkpointer.py:280] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0418 06:21:21.193124 135161681655616 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141/linen_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776493228.5095882, 'wait_for_prev_duration_secs': 51.92679738998413, 'time_between_consecutive_saves_sec': 3.04371976852417, 'checkpointer_blocking_start_time': 1776493280.4379413, 'checkpointer_blocking_duration_secs': 0.7528541088104248, 'get_old_steps_start_time': 1776493281.1908126, 'get_old_steps_duration_secs': 2.5510787963867188e-05, 'checkpoint_manager_blocking_start_time': 1776493228.5076602, 'checkpoint_manager_blocking_duration_secs': 52.68543362617493} I0418 06:21:21.193231 135161681655616 checkpointing.py:409] Started an asynchronous checkpoint save for step 19 I0418 06:21:21.193273 135161681655616 checkpoint_manager.py:2020] [process=3][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 06:21:27.153745 135026633066240 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141/linen_xpk_feat_nnx_trainstate_and_training_loop_20260418_060141_05_fp8/checkpoints/19/items/array_metadatas/process_3 I0418 06:22:04.034579 135026649851648 base_pytree_checkpoint_handler.py:130] [process=3] /jax/orbax/write/gbytes_per_sec: 36.820 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.905192136764526 s) (per-host) I0418 06:22:04.034706 135026649851648 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.843942s. I0418 06:22:14.011172 135026649851648 async_checkpointer.py:160] [process=3][thread=async_save] Background save thread done. Time taken: 52.820394s. I0418 06:22:14.011442 135029871077120 async_checkpointer.py:288] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 06:22:14.011571 135029871077120 async_checkpointer.py:298] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0418 06:22:14.011633 135029871077120 checkpoint_manager.py:2137] [process=3][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0418 06:22:14.013215 135029871077120 checkpoint_manager.py:2146] [process=3][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0418 06:22:14.013383 135161681655616 checkpoint_manager.py:2032] [process=3][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0418 06:22:14.013535 135161681655616 checkpoint_manager.py:2009] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 06:22:14.014599 135161681655616 metric_logger.py:196] completed step: 19, seconds: 0.158, TFLOP/s/device: 85.959, Tokens/s/device: 12956.613, total_weights: 65536, loss: 6.167, lm_loss: 6.167, perplexity: 476.947 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 18 06:22:26 UTC 2026 EXIT_CODE=0