2026-04-23 01:32:31.452898: 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 00:00:1776907951.466262 2269233 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1776907951.470346 2269233 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1776907951.481829 2269233 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776907951.481846 2269233 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776907951.481847 2269233 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1776907951.481849 2269233 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} `rope_scaling`'s factor field must be a float >= 1, got 40 `rope_scaling`'s beta_fast field must be a float, got 32 `rope_scaling`'s beta_slow field must be a float, got 1 Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} Unrecognized keys in `rope_scaling` for 'rope_type'='yarn': {'rope_theta'} 2026-04-23 01:32:35.150947: 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) I0423 01:32:35.285021 131807873025856 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag. I0423 01:33:06.990933 131807873025856 max_utils.py:800] System Information: Jax Version: 0.8.1 I0423 01:33:06.991210 131807873025856 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0423 01:33:06.991255 131807873025856 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 I0423 01:33:06.991284 131807873025856 train_utils.py:348] WARNING: 'dataset_path' might be pointing your local file system I0423 01:33:06.991457 131807873025856 train.py:703] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper. I0423 01:33:07.648252 131807873025856 maxtext_utils.py:1604] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0423 01:33:07.648518 131807873025856 checkpointing.py:677] Setting up checkpoint logger... I0423 01:33:07.648562 131807873025856 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0423 01:33:07.648597 131807873025856 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None I0423 01:33:07.649011 131807873025856 base_pytree_checkpoint_handler.py:411] 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 0x77e0a27ad460>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0423 01:33:10.116569 131807873025856 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0423 01:33:10.116951 131807873025856 checkpoint_manager.py:702] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x77decaf419a0>}, handler_registry=None I0423 01:33:10.117206 131807873025856 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x77decaf419a0>` 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`. I0423 01:33:10.117249 131807873025856 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x77df447b4a40>` 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`. I0423 01:33:10.117278 131807873025856 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 0x77decaf419a0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x77decaf419a0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x77df447b4a40>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x77df447b4a40>}). I0423 01:33:10.117558 131807873025856 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28 I0423 01:33:10.117623 131807873025856 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x77decaf54540> timeout: 600 secs and primary_host=0 for async checkpoint writes I0423 01:33:10.267956 131807873025856 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints I0423 01:33:10.268222 131807873025856 checkpoint_manager.py:921] [process=0][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_hns=False, 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), root_directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x77decaf41e80> I0423 01:33:10.268500 131807873025856 checkpointing.py:301] Checkpoint manager created! I0423 01:33:10.346879 131807873025856 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0423 01:33:10.349761 131807873025856 reader.py:262] Creating a tf.data.Dataset reading 8 files located in folders: tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0. I0423 01:33:10.409109 131807873025856 logging_logger.py:49] Constructing tf.data.Dataset __local_c4_builder for split train, from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0 I0423 01:33:10.431813 131807873025856 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2 I0423 01:33:10.431879 131807873025856 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2 I0423 01:33:11.202573 131807873025856 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0423 01:33:11.202677 131807873025856 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 01:33:11.508476 131807873025856 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 01:33:11.508564 131807873025856 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 01:33:11.521387 131807873025856 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0423 01:33:11.521440 131807873025856 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None). I0423 01:33:11.571295 131807873025856 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 01:33:11.571372 131807873025856 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 01:33:11.584228 131807873025856 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 01:33:11.584281 131807873025856 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 01:33:11.597028 131807873025856 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0423 01:33:11.597094 131807873025856 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 01:33:11.616997 131807873025856 attentions.py:1198] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0423 01:33:11.617069 131807873025856 attentions.py:1198] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None). I0423 01:33:11.666770 131807873025856 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0423 01:33:11.666841 131807873025856 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None). I0423 01:33:12.323356 131807873025856 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0423 01:33:12.323464 131807873025856 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. [DECOUPLED NO-OP] gcs_storage: using stubs. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] mldiagnostics: using stub. [DECOUPLED NO-OP] workload_monitor: using stub. [DECOUPLED NO-OP] vertex_tensorboard: using stub. fsdp: 8 I0423 01:33:13.452340 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.452453 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.452491 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.452522 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.452548 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.452574 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.452599 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.452623 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.452645 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.452668 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.452690 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.452710 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.452731 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.452752 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.452774 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.452794 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.452815 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.452835 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.452854 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.452875 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.452894 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.452914 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.452934 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.452954 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.452980 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.453001 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.453021 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.453040 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.453077 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.453098 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.453119 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.453139 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.453159 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.453178 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.453197 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.453217 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.453238 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/input_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.453258 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/input_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.453277 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/kernel_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.453296 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/kernel_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.453316 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/output_grad_amax_history Shape: float32[16,1024] Logical: PartitionSpec() Physical: () I0423 01:33:13.453335 131807873025856 maxtext_utils.py:1707] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/output_grad_scale Shape: float32[16,1] Logical: PartitionSpec() Physical: () I0423 01:33:13.453374 131807873025856 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0423 01:33:13.453414 131807873025856 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0423 01:33:13.453441 131807873025856 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0423 01:33:13.453475 131807873025856 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0423 01:33:13.453508 131807873025856 maxtext_utils.py:1707] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0423 01:33:13.453530 131807873025856 maxtext_utils.py:1707] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0423 01:33:13.453565 131807873025856 maxtext_utils.py:1707] 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) I0423 01:33:13.453600 131807873025856 maxtext_utils.py:1707] 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') I0423 01:33:13.453626 131807873025856 maxtext_utils.py:1707] 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) I0423 01:33:13.453651 131807873025856 maxtext_utils.py:1707] 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) I0423 01:33:13.453684 131807873025856 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed_vocab', 'vocab') Physical: ('fsdp', None) I0423 01:33:13.453716 131807873025856 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0423 01:33:14.926762 131807873025856 train.py:155] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 01:33:14.926880 131807873025856 train.py:155] train/xent Physical: float32[8,2048]............................................. ('fsdp', None). I0423 01:33:14.939029 131807873025856 train.py:162] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length'). I0423 01:33:14.939402 131807873025856 train.py:162] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None). I0423 01:33:16.321201 131807873025856 max_utils.py:791] Total memory size: 3.5 GB, Output size: 1.5 GB, Temp size: 1.9 GB, Argument size: 1.5 GB, Host temp size: 0.0 GB. I0423 01:33:16.321872 131807873025856 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter. I0423 01:33:16.322103 131807873025856 metric_logger.py:301] number parameters: 1.105 billion I0423 01:33:22.277236 131807873025856 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0423 01:33:22.395115 131807873025856 checkpointing.py:776] Waited 0.11786127090454102 seconds for step 0 to finish before starting checkpointing. I0423 01:33:22.400133 131807873025856 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 01:33:22.400299 131807873025856 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0 I0423 01:33:22.400695 131807873025856 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/0. I0423 01:33:22.493166 131807873025856 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient I0423 01:33:22.583170 131682579514944 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/0 I0423 01:33:22.583549 131807873025856 jax_array_handlers.py:347] Scheduling D2H of 81 prioritized jax.Array. I0423 01:33:22.583657 131807873025856 replica_slices.py:410] 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 I0423 01:33:23.333498 131682569029184 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/0/items I0423 01:33:23.524106 131682466268736 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776908003230547656, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776908003230547656, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/0/_CHECKPOINT_METADATA I0423 01:33:23.828543 2271847 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json I0423 01:33:23.828604 2271847 google_auth_provider.cc:156] Using OAuth2 AuthProvider I0423 01:33:30.072312 131807873025856 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 7.489630s I0423 01:33:30.073347 131807873025856 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.629 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 7 seconds) (per-host) I0423 01:33:30.073404 131807873025856 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 7.578555s (batch_requests_ready=0.081422s, total_serialization_initiated=7.496216s, others=0.000916s) I0423 01:33:30.073490 131807873025856 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 7.579763s (all_items=0.000021s, per_item={'items': '0.00002074'}, temp_paths=7.579742) I0423 01:33:30.081925 131682590000704 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 01:33:30.082038 131807873025856 async_checkpointer.py:561] Finished blocking save. Time taken: 7.681691s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/0. I0423 01:33:30.082287 131807873025856 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0423 01:33:30.082447 131682105558592 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 01:33:30.082525 131807873025856 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776908002.4001198, 'wait_for_prev_duration_secs': 4.887580871582031e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776908002.4003208, 'checkpointer_blocking_duration_secs': 7.68188214302063, 'get_old_steps_start_time': 1776908010.0822206, 'get_old_steps_duration_secs': 2.9325485229492188e-05, 'checkpoint_manager_blocking_start_time': 1776908002.4000173, 'checkpoint_manager_blocking_duration_secs': 7.6824791431427} I0423 01:33:30.082619 131807873025856 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0423 01:33:30.082658 131807873025856 max_utils.py:750] Memstats: After params initialized: I0423 01:33:30.082711 131807873025856 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_0(process=0,(0,0,0,0)) I0423 01:33:30.082737 131807873025856 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_1(process=0,(1,0,0,0)) I0423 01:33:30.082757 131807873025856 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_2(process=0,(0,1,0,0)) I0423 01:33:30.082776 131807873025856 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_3(process=0,(1,1,0,0)) I0423 01:33:30.082794 131807873025856 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_4(process=0,(0,2,0,0)) I0423 01:33:30.082814 131807873025856 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_5(process=0,(1,2,0,0)) I0423 01:33:30.082833 131807873025856 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_6(process=0,(0,3,0,0)) I0423 01:33:30.082851 131807873025856 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_7(process=0,(1,3,0,0)) I0423 01:33:30.464085 131807873025856 metric_logger.py:196] completed step: 0, seconds: 5.955, TFLOP/s/device: 2.282, Tokens/s/device: 343.937, total_weights: 13328, loss: 10.844, lm_loss: 10.844, perplexity: 51212.484 I0423 01:33:30.465031 131807873025856 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/tensorboard/' I0423 01:33:30.573374 131807873025856 metric_logger.py:196] completed step: 1, seconds: 8.184, TFLOP/s/device: 1.660, Tokens/s/device: 250.251, total_weights: 12332, loss: 10.871, lm_loss: 10.871, perplexity: 52634.344 I0423 01:33:30.685178 131807873025856 metric_logger.py:196] completed step: 2, seconds: 0.011, TFLOP/s/device: 1216.068, Tokens/s/device: 183299.024, total_weights: 15161, loss: 10.871, lm_loss: 10.871, perplexity: 52617.816 I0423 01:33:30.796965 131807873025856 metric_logger.py:196] completed step: 3, seconds: 0.110, TFLOP/s/device: 123.025, Tokens/s/device: 18543.670, total_weights: 13327, loss: 10.857, lm_loss: 10.857, perplexity: 51888.457 I0423 01:33:30.834826 131682476754496 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/0/items/array_metadatas/process_0 I0423 01:33:51.242408 131807873025856 metric_logger.py:196] completed step: 4, seconds: 0.112, TFLOP/s/device: 121.507, Tokens/s/device: 18314.822, total_weights: 11939, loss: 10.861, lm_loss: 10.861, perplexity: 52094.562 I0423 01:33:51.262912 131807873025856 metric_logger.py:196] completed step: 5, seconds: 0.112, TFLOP/s/device: 121.607, Tokens/s/device: 18329.902, total_weights: 15502, loss: 10.882, lm_loss: 10.882, perplexity: 53230.324 I0423 01:33:51.372936 131807873025856 metric_logger.py:196] completed step: 6, seconds: 20.455, TFLOP/s/device: 0.664, Tokens/s/device: 100.124, total_weights: 13864, loss: 10.886, lm_loss: 10.886, perplexity: 53423.605 I0423 01:33:51.484886 131807873025856 metric_logger.py:196] completed step: 7, seconds: 0.009, TFLOP/s/device: 1522.367, Tokens/s/device: 229467.787, total_weights: 12988, loss: 10.829, lm_loss: 10.829, perplexity: 50459.859 I0423 01:33:51.596935 131807873025856 metric_logger.py:196] completed step: 8, seconds: 0.111, TFLOP/s/device: 122.676, Tokens/s/device: 18491.098, total_weights: 13820, loss: 10.848, lm_loss: 10.848, perplexity: 51456.266 I0423 01:33:51.605095 131807873025856 checkpointing.py:772] Waiting for step 10 to finish before checkpoint... I0423 01:33:51.821599 131807873025856 checkpointing.py:776] Waited 0.21646499633789062 seconds for step 10 to finish before starting checkpointing. I0423 01:33:51.822905 131807873025856 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 01:34:05.831820 131682116044352 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 35.750827s (commit=35.276320s, array_metadata_write=0.474507s) I0423 01:34:05.833174 131682590000704 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 291.615 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 43 seconds) (per-host) I0423 01:34:05.833300 131682590000704 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 35.751179s. I0423 01:34:06.127763 131682590000704 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776908003230547656, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/0/_CHECKPOINT_METADATA I0423 01:34:06.347656 131682590000704 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 01:34:06.563909 131682466268736 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776908003230547656, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/0/_CHECKPOINT_METADATA I0423 01:34:06.760966 131682590000704 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 01:34:06.761275 131682590000704 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.583822s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/0/items I0423 01:34:06.762115 131682590000704 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/0/items I0423 01:34:07.008233 131682590000704 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/0 I0423 01:34:07.737986 131682590000704 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/0`. I0423 01:34:07.738711 131682590000704 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 45.338375s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/0 I0423 01:34:07.738800 131682590000704 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 37.656686s. I0423 01:34:07.738956 131682105558592 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 01:34:07.739011 131682105558592 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 01:34:07.739083 131682105558592 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0423 01:34:07.739133 131682105558592 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0423 01:34:07.739251 131807873025856 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0423 01:34:07.739388 131807873025856 checkpoint_manager.py:1441] Waiting for previous save to complete took 15.916508 seconds. If this number is high, consider checkpointing less frequently. I0423 01:34:07.740889 131807873025856 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 10 I0423 01:34:07.741252 131807873025856 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/10. I0423 01:34:07.915887 131807873025856 jax_array_handlers.py:347] Scheduling D2H of 81 prioritized jax.Array. I0423 01:34:07.915989 131807873025856 replica_slices.py:410] 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 I0423 01:34:07.922714 131682590000704 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/10 I0423 01:34:08.668136 131682476754496 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/10/items I0423 01:34:10.267197 131807873025856 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 2.351762s I0423 01:34:10.275133 131807873025856 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.030 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 2 seconds) (per-host) I0423 01:34:10.275206 131807873025856 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 2.453560s (batch_requests_ready=0.087657s, total_serialization_initiated=2.358080s, others=0.007823s) I0423 01:34:10.275639 131807873025856 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 2.454481s (all_items=0.000026s, per_item={'items': '0.00002575'}, temp_paths=2.454455) I0423 01:34:10.277277 131682590000704 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 01:34:10.277361 131807873025856 async_checkpointer.py:561] Finished blocking save. Time taken: 2.536423s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/10. I0423 01:34:10.277565 131807873025856 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize I0423 01:34:10.277773 131682105558592 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 01:34:10.277904 131807873025856 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776908031.8228445, 'wait_for_prev_duration_secs': 15.916508436203003, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776908047.7409153, 'checkpointer_blocking_duration_secs': 2.536548614501953, 'get_old_steps_start_time': 1776908050.277484, 'get_old_steps_duration_secs': 4.076957702636719e-05, 'checkpoint_manager_blocking_start_time': 1776908031.8227549, 'checkpoint_manager_blocking_duration_secs': 18.4551203250885} I0423 01:34:10.278023 131807873025856 checkpointing.py:408] Started an asynchronous checkpoint save for step 10 I0423 01:34:10.278785 131807873025856 metric_logger.py:196] completed step: 9, seconds: 0.112, TFLOP/s/device: 121.314, Tokens/s/device: 18285.714, total_weights: 12300, loss: 10.857, lm_loss: 10.857, perplexity: 51900.086 I0423 01:34:10.290238 131807873025856 metric_logger.py:196] completed step: 10, seconds: 0.112, TFLOP/s/device: 121.192, Tokens/s/device: 18267.447, total_weights: 13436, loss: 10.876, lm_loss: 10.876, perplexity: 52905.383 I0423 01:34:10.398869 131807873025856 metric_logger.py:196] completed step: 11, seconds: 18.682, TFLOP/s/device: 0.727, Tokens/s/device: 109.624, total_weights: 13817, loss: 10.853, lm_loss: 10.853, perplexity: 51714.605 I0423 01:34:10.510804 131807873025856 metric_logger.py:196] completed step: 12, seconds: 0.010, TFLOP/s/device: 1329.985, Tokens/s/device: 200469.851, total_weights: 12076, loss: 10.852, lm_loss: 10.852, perplexity: 51622.902 I0423 01:34:10.622663 131807873025856 metric_logger.py:196] completed step: 13, seconds: 0.109, TFLOP/s/device: 124.218, Tokens/s/device: 18723.544, total_weights: 11048, loss: 10.864, lm_loss: 10.864, perplexity: 52261.953 I0423 01:34:10.734748 131807873025856 metric_logger.py:196] completed step: 14, seconds: 0.112, TFLOP/s/device: 121.473, Tokens/s/device: 18309.746, total_weights: 14267, loss: 10.834, lm_loss: 10.834, perplexity: 50738.680 I0423 01:34:10.846834 131807873025856 metric_logger.py:196] completed step: 15, seconds: 0.112, TFLOP/s/device: 121.764, Tokens/s/device: 18353.557, total_weights: 13004, loss: 10.832, lm_loss: 10.832, perplexity: 50634.074 I0423 01:34:10.958700 131807873025856 metric_logger.py:196] completed step: 16, seconds: 0.115, TFLOP/s/device: 118.585, Tokens/s/device: 17874.443, total_weights: 13811, loss: 10.867, lm_loss: 10.867, perplexity: 52422.980 I0423 01:34:11.013793 131682445297216 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/10/items/array_metadatas/process_0 I0423 01:34:11.070650 131807873025856 metric_logger.py:196] completed step: 17, seconds: 0.109, TFLOP/s/device: 124.085, Tokens/s/device: 18703.367, total_weights: 13709, loss: 10.850, lm_loss: 10.850, perplexity: 51525.906 I0423 01:34:11.187552 131807873025856 metric_logger.py:196] completed step: 18, seconds: 0.112, TFLOP/s/device: 121.530, Tokens/s/device: 18318.262, total_weights: 12831, loss: 10.839, lm_loss: 10.839, perplexity: 50989.945 I0423 01:34:11.294218 131807873025856 checkpointing.py:772] Waiting for step 19 to finish before checkpoint... I0423 01:34:11.296388 131807873025856 checkpointing.py:776] Waited 0.0021820068359375 seconds for step 19 to finish before starting checkpointing. I0423 01:34:11.296931 131807873025856 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 01:34:25.128520 131682476754496 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 14.852194s (commit=14.362157s, array_metadata_write=0.490037s) I0423 01:34:25.130386 131682590000704 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 730.159 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 17 seconds) (per-host) I0423 01:34:25.130440 131682590000704 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 14.853036s. I0423 01:34:25.601319 131682590000704 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 01:34:26.002118 131682590000704 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 01:34:26.002413 131682590000704 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.574193s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/10/items I0423 01:34:26.003188 131682590000704 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/10/items I0423 01:34:26.254459 131682590000704 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/10 I0423 01:34:26.949897 131682590000704 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/10`. I0423 01:34:26.950744 131682590000704 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 19.209812s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/10 I0423 01:34:26.950810 131682590000704 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 16.673407s. I0423 01:34:26.951020 131682105558592 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 01:34:26.951156 131682105558592 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 01:34:26.951207 131682105558592 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts... I0423 01:34:26.951247 131682105558592 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts. I0423 01:34:26.951384 131807873025856 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=10][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=10. W0423 01:34:26.951511 131807873025856 checkpoint_manager.py:1441] Waiting for previous save to complete took 15.654594 seconds. If this number is high, consider checkpointing less frequently. I0423 01:34:26.951949 131807873025856 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 19 I0423 01:34:26.952260 131807873025856 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/19. I0423 01:34:27.116694 131682590000704 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/19 I0423 01:34:27.133619 131807873025856 jax_array_handlers.py:347] Scheduling D2H of 81 prioritized jax.Array. I0423 01:34:27.133722 131807873025856 replica_slices.py:410] 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 I0423 01:34:27.831082 131682445297216 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/19/items I0423 01:34:33.356454 131807873025856 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 6.223252s I0423 01:34:33.363587 131807873025856 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.950 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 6 seconds) (per-host) I0423 01:34:33.363664 131807873025856 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 6.328493s (batch_requests_ready=0.092535s, total_serialization_initiated=6.228935s, others=0.007023s) I0423 01:34:33.364211 131807873025856 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 6.329533s (all_items=0.000012s, per_item={'items': '0.00001240'}, temp_paths=6.329521) I0423 01:34:33.365806 131791300068928 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started. I0423 01:34:33.365891 131807873025856 async_checkpointer.py:561] Finished blocking save. Time taken: 6.413895s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/19. I0423 01:34:33.366062 131807873025856 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize I0423 01:34:33.366204 131682105558592 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0423 01:34:33.366270 131807873025856 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776908051.2969024, 'wait_for_prev_duration_secs': 15.65459394454956, 'time_between_consecutive_saves_sec': 3.557751417160034, 'checkpointer_blocking_start_time': 1776908066.951975, 'checkpointer_blocking_duration_secs': 6.414014577865601, 'get_old_steps_start_time': 1776908073.3659997, 'get_old_steps_duration_secs': 1.7642974853515625e-05, 'checkpoint_manager_blocking_start_time': 1776908051.2968583, 'checkpoint_manager_blocking_duration_secs': 22.069388389587402} I0423 01:34:33.366373 131807873025856 checkpointing.py:408] Started an asynchronous checkpoint save for step 19 I0423 01:34:33.366405 131807873025856 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0423 01:34:34.079775 131679354095168 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/19/items/array_metadatas/process_0 I0423 01:35:01.900578 131677407938112 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 28.535679s (commit=28.070045s, array_metadata_write=0.465633s) I0423 01:35:01.901719 131791300068928 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 362.471 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 34 seconds) (per-host) I0423 01:35:01.901826 131791300068928 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 28.535890s. I0423 01:35:02.368680 131791300068928 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0. I0423 01:35:02.766168 131791300068928 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389). I0423 01:35:02.766464 131791300068928 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.564204s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/19/items I0423 01:35:02.767263 131791300068928 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/19/items I0423 01:35:03.003865 131791300068928 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/19 I0423 01:35:03.699745 131791300068928 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/19`. I0423 01:35:03.700485 131791300068928 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 36.748499s. directory=gs://wanglance-maxtext/linen_ckpt_main_20260423_012248/linen_main_20260423_012248_05_fp8/checkpoints/19 I0423 01:35:03.700546 131791300068928 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 30.334612s. I0423 01:35:03.700654 131682105558592 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0423 01:35:03.700709 131682105558592 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0423 01:35:03.700748 131682105558592 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts... I0423 01:35:03.700786 131682105558592 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts. I0423 01:35:03.700901 131807873025856 checkpoint_manager.py:2006] [process=0][thread=MainThread][step=19][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=19. I0423 01:35:03.701038 131807873025856 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0423 01:35:03.701823 131807873025856 metric_logger.py:196] completed step: 19, seconds: 0.112, TFLOP/s/device: 121.357, Tokens/s/device: 18292.247, total_weights: 13717, loss: 10.851, lm_loss: 10.851, perplexity: 51591.992 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops